[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Next Issue
Volume 13, October-1
Previous Issue
Volume 13, September-1
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
8.3
4.2
Journals
Remote Sensing
Volume 13
Issue 18
remotesensing-logo

Journal Menu

► Journal Menu

Journal Browser

► Journal Browser
share announcement

Remote Sens., Volume 13, Issue 18 (September-2 2021) – 243 articles

Cover Story (view full-size image): Remote sensing products are important for estimating landscape scale ecosystem services, such as carbon sequestration. Such efforts often use time series data of plant greenness, which can provide a proxy of plant biomass, coverage, and photosynthetic activity. However, this is difficult within tidal wetlands due to frequent flooding that leads to attenuations of the vegetation signal. This uncertainty contributes to gaps in upscaling and budgeting of these important “blue carbon” ecosystems and projecting their future. This study developed and investigated a strategy to model annual trajectories of a greenness indicator in a tidally affected estuarine setting by combining per-pixel historical multi-year phenological variability with relevant landscape properties, and current-year climatic characteristics from open access data. View this paper.
  • Issues are regarded as officially published after their release is announced to the table of contents alert mailing list.
  • You may sign up for e-mail alerts to receive table of contents of newly released issues.
  • PDF is the official format for papers published in both, html and pdf forms. To view the papers in pdf format, click on the "PDF Full-text" link, and use the free Adobe Reader to open them.
Order results
Result details
Section
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
21 pages, 3036 KiB  
Article
Indoor Activity and Vital Sign Monitoring for Moving People with Multiple Radar Data Fusion
by Xiuzhu Yang, Xinyue Zhang, Yi Ding and Lin Zhang
Remote Sens. 2021, 13(18), 3791; https://doi.org/10.3390/rs13183791 - 21 Sep 2021
Cited by 23 | Viewed by 5290
Abstract
The monitoring of human activity and vital signs plays a significant role in remote health-care. Radar provides a non-contact monitoring approach without privacy and illumination concerns. However, multiple people in a narrow indoor environment bring dense multipaths for activity monitoring, and the received [...] Read more.
The monitoring of human activity and vital signs plays a significant role in remote health-care. Radar provides a non-contact monitoring approach without privacy and illumination concerns. However, multiple people in a narrow indoor environment bring dense multipaths for activity monitoring, and the received vital sign signals are heavily distorted with body movements. This paper proposes a framework based on Frequency Modulated Continuous Wave (FMCW) and Impulse Radio Ultra-Wideband (IR-UWB) radars to address these challenges, designing intelligent spatial-temporal information fusion for activity and vital sign monitoring. First, a local binary pattern (LBP) and energy features are extracted from FMCW radar, combined with the wavelet packet transform (WPT) features on IR-UWB radar for activity monitoring. Then the additional information guided fusing network (A-FuseNet) is proposed with a modified generative and adversarial structure for vital sign monitoring. A Cascaded Convolutional Neural Network (CCNN) module and a Long Short Term Memory (LSTM) module are designed as the fusion sub-network for vital sign information extraction and multisensory data fusion, while a discrimination sub-network is constructed to optimize the fused heartbeat signal. In addition, the activity and movement characteristics are introduced as additional information to guide the fusion and optimization. A multi-radar dataset with an FMCW and two IR-UWB radars in a cotton tent, a small room and a wide lobby is constructed, and the accuracies of activity and vital sign monitoring achieve 99.9% and 92.3% respectively. Experimental results demonstrate the superiority and robustness of the proposed framework. Full article
(This article belongs to the Special Issue Radar Signal Processing and System Design for Urban Health)
►▼ Show Figures

Figure 1

Figure 1
<p>Flowchart of the proposed framework, composed by the feature extraction and fusion for activity monitoring and the A-FuseNet for vital sign monitoring.</p> Full article ">Figure 2
<p>Experimental setup.</p> Full article ">Figure 3
<p>Dataset generation scenarios.</p> Full article ">Figure 4
<p>Schematic diagram of proposed feature extraction and fusion for activity monitoring.</p> Full article ">Figure 5
<p>Structure of the proposed A-FuseNet for vital sign monitoring.</p> Full article ">Figure 6
<p>Doppler velocity for different body movements.</p> Full article ">Figure 7
<p>Classification performance with different proportions of training samples for activity monitoring.</p> Full article ">Figure 8
<p>Classification performance comparison of different methods for activity monitoring.</p> Full article ">Figure 9
<p>A generated heartbeat signal from A-FuseNet.</p> Full article ">
18 pages, 3837 KiB  
Article
Temporal and Spatial Autocorrelation as Determinants of Regional AOD-PM2.5 Model Performance in the Middle East
by Khang Chau, Meredith Franklin, Huikyo Lee, Michael Garay and Olga Kalashnikova
Remote Sens. 2021, 13(18), 3790; https://doi.org/10.3390/rs13183790 - 21 Sep 2021
Cited by 9 | Viewed by 2986
Abstract
Exposure to fine particulate matter (PM2.5) air pollution has been shown in numerous studies to be associated with detrimental health effects. However, the ability to conduct epidemiological assessments can be limited due to challenges in generating reliable PM2.5 estimates, particularly [...] Read more.
Exposure to fine particulate matter (PM2.5) air pollution has been shown in numerous studies to be associated with detrimental health effects. However, the ability to conduct epidemiological assessments can be limited due to challenges in generating reliable PM2.5 estimates, particularly in parts of the world such as the Middle East where measurements are scarce and extreme meteorological events such as sandstorms are frequent. In order to supplement exposure modeling efforts under such conditions, satellite-retrieved aerosol optical depth (AOD) has proven to be useful due to its global coverage. By using AODs from the Multiangle Implementation of Atmospheric Correction (MAIAC) of the MODerate Resolution Imaging Spectroradiometer (MODIS) and the Multiangle Imaging Spectroradiometer (MISR) combined with meteorological and assimilated aerosol information from the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2), we constructed machine learning models to predict PM2.5 in the area surrounding the Persian Gulf, including Kuwait, Bahrain, and the United Arab Emirates (U.A.E). Our models showed regional differences in predictive performance, with better results in the U.A.E. (median test R2 = 0.66) than Kuwait (median test R2 = 0.51). Variable importance also differed by region, where satellite-retrieved AOD variables were more important for predicting PM2.5 in Kuwait than in the U.A.E. Divergent trends in the temporal and spatial autocorrelations of PM2.5 and AOD in the two regions offered possible explanations for differences in predictive performance and variable importance. In a test of model transferability, we found that models trained in one region and applied to another did not predict PM2.5 well, even if the transferred model had better performance. Overall the results of our study suggest that models developed over large geographic areas could generate PM2.5 estimates with greater uncertainty than could be obtained by taking a regional modeling approach. Furthermore, development of methods to better incorporate spatial and temporal autocorrelations in machine learning models warrants further examination. Full article
(This article belongs to the Special Issue Spatial and Spatio-Temporal Statistics: Methods and Applications in Remote Sensing)
►▼ Show Figures

Figure 1

Figure 1
<p>Map of the study region with all PM<math display="inline"><semantics> <msub> <mrow/> <mrow> <mn>2.5</mn> </mrow> </msub> </semantics></math> monitors (<b>left</b>) and only those in Kuwait (<b>right</b>).</p> Full article ">Figure 2
<p>Areas within each country where temporal and spatial autocorrelations for MAIAC AOD were evaluated (shaded pink) and the PM<math display="inline"><semantics> <msub> <mrow/> <mrow> <mn>2.5</mn> </mrow> </msub> </semantics></math> monitors (yellow circles) in each country.</p> Full article ">Figure 3
<p>Validation of AERONET AOD interpolated to 550 nm versus MAIAC AOD (<b>left</b>), MISR AOD (<b>center</b>), and MISR “raw” AOD (<b>right</b>) with one-to-one line (dotted) and correlation line (red).</p> Full article ">Figure 4
<p>The top 10 most important variables for the random forest overall models using MAIAC AOD (<b>left</b>) and MISR AOD (<b>right</b>), as measured by increase in MSE when excluded.</p> Full article ">Figure 5
<p>Test <math display="inline"><semantics> <msup> <mi>R</mi> <mn>2</mn> </msup> </semantics></math> across different settings: overall and regional models, MISR and MAIAC models, and models with and without AOD variables.</p> Full article ">Figure 6
<p>The top 10 most important variables for the random forest models based on monitors in Kuwait using MAIAC AOD (<b>top left</b>) and MISR AOD (<b>top right</b>) and in the U.A.E. using MAIAC AOD (<b>bottom left</b>) and MISR AOD (<b>bottom right</b>), as measured by increase in MSE when excluded.</p> Full article ">Figure 7
<p>Autocorrelation functions for PM<math display="inline"><semantics> <msub> <mrow/> <mrow> <mn>2.5</mn> </mrow> </msub> </semantics></math> (<b>left</b>) and median ACFs for MAIAC AOD (<b>right</b>) at different sites in Kuwait and the U.A.E.</p> Full article ">Figure 8
<p>Daily semivariograms for MAIAC AOD in Kuwait and the U.A.E. (grey lines in top row) and median semivariograms (red lines in both top and bottom rows).</p> Full article ">Figure A1
<p>Out-of-region PM<math display="inline"><semantics> <msub> <mrow/> <mrow> <mn>2.5</mn> </mrow> </msub> </semantics></math> predictions for Kuwait-trained models over the U.A.E. (<b>left column</b>) and out-of-region PM<math display="inline"><semantics> <msub> <mrow/> <mrow> <mn>2.5</mn> </mrow> </msub> </semantics></math> predictions for U.A.E.-trained models over Kuwait (<b>right column</b>) by MAIAC and MISR AODs (top and bottom rows, respective).</p> Full article ">Figure A2
<p>Autocorrelation functions for ERA5 meteorological variables with high importance in the random forests models. The Kuwait City and Central monitors shared the same MERRA-2 pixel and, thus, the same ACF.</p> Full article ">Figure A3
<p>Autocorrelation functions for MERRA-2 assimilated aerosol extinction variables, including dust, sulfate, black carbon, and organic carbon. The Kuwait City and Central monitors shared the same MERRA-2 pixel and, thus, the same ACF. Organic carbon was not found to be useful in any models and was excluded in the final PM<math display="inline"><semantics> <msub> <mrow/> <mrow> <mn>2.5</mn> </mrow> </msub> </semantics></math> models.</p> Full article ">
27 pages, 6309 KiB  
Article
Optimizing the Sowing Date to Improve Water Management and Wheat Yield in a Large Irrigation Scheme, through a Remote Sensing and an Evolution Strategy-Based Approach
by Salwa Belaqziz, Saïd Khabba, Mohamed Hakim Kharrou, El Houssaine Bouras, Salah Er-Raki and Abdelghani Chehbouni
Remote Sens. 2021, 13(18), 3789; https://doi.org/10.3390/rs13183789 - 21 Sep 2021
Cited by 13 | Viewed by 3663
Abstract
This study aims to investigate the effects of an optimized sowing calendar for wheat over a surface irrigation scheme in the semi-arid region of Haouz (Morocco) on irrigation water requirements, crop growth and development and on yield. For that, a scenario-based simulation approach [...] Read more.
This study aims to investigate the effects of an optimized sowing calendar for wheat over a surface irrigation scheme in the semi-arid region of Haouz (Morocco) on irrigation water requirements, crop growth and development and on yield. For that, a scenario-based simulation approach based on the covariance matrix adaptation–evolution strategy (CMA-ES) was proposed to optimize both the spatiotemporal distribution of sowing dates and the irrigation schedules, and then evaluate wheat crop using the 2011–2012 growing season dataset. Six sowing scenarios were simulated and compared to identify the most optimal spatiotemporal sowing calendar. The obtained results showed that with reference to the existing sowing patterns, early sowing of wheat leads to higher yields compared to late sowing (from 7.40 to 5.32 t/ha). Compared with actual conditions in the study area, the spatial heterogeneity is highly reduced, which increased equity between farmers. The results also showed that the proportion of plots irrigated in time can be increased (from 40% to 82%) compared to both the actual irrigation schedules and to previous results of irrigation optimization, which did not take into consideration sowing dates optimization. Furthermore, considerable reduction of more than 40% of applied irrigation water can be achieved by optimizing sowing dates. Thus, the proposed approach in this study is relevant for irrigation managers and farmers since it provides an insight on the consequences of their agricultural practices regarding the wheat sowing calendar and irrigation scheduling and can be implemented to recommend the best practices to adopt. Full article
(This article belongs to the Special Issue Remote Sensing in Irrigated Crop Water Stress Assessment)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>(<b>a</b>) Main irrigated sectors in the Tensift El Haouz basin; (<b>b</b>) the gravity irrigation network in the R3 study area.</p> Full article ">Figure 2
<p>Flowchart of the proposed approach for yield and water resources optimization based on spatiotemporal sowing date optimization.</p> Full article ">Figure 3
<p>Reference evapotranspiration (ET<sub>0</sub>), rainfall amounts and air mean temperature during 2011–2012 agricultural season in the study area (R3).</p> Full article ">Figure 4
<p>Distribution of the sowing dates over the plots for each sowing scenario. Day number 1 is equal to the start date which change from one scenario to another.</p> Full article ">Figure 5
<p>Convergence of the objective function for each sowing scenario during the optimization process.</p> Full article ">Figure 6
<p>Optimal sowing date distribution simulated with the first sowing scenario.</p> Full article ">Figure 7
<p>Example of the simulated NDVI profiles for one single plot.</p> Full article ">Figure 8
<p>Stress coefficient (Ks) map of the first sowing scenario (<b>a</b>) and the sixth sowing scenario (<b>b</b>) at the beginning of the third irrigation round.</p> Full article ">Figure 9
<p>Variation of Ks coefficient and day of irrigation at the third irrigation round (IR), from one sowing scenario to another.</p> Full article ">Figure 10
<p>Distribution of the optimized IPI index at plot level for each sowing scenario (IRi represents the irrigation round number in the growing season).</p> Full article ">Figure 11
<p>IPI index distribution with sowing date optimization (for each sowing date) and without sowing date optimization (from previous study) at the third irrigation round of the 2011–2012 agricultural season.</p> Full article ">Figure 12
<p>Comparison of the variation of irrigation amount (mm), between simulated scenarios and the real case (observed irrigation at the 2011–2012 agricultural season). The statistical Student’s <span class="html-italic">t</span>-test shows that there are two groups of scenarios (S1, S5 and S6) and (S2, S3 and S4) at <span class="html-italic">p</span> &gt; 0.05.</p> Full article ">Figure 13
<p>Variation of grain yield (t/ha), between simulated scenarios and the real case (2011–2012 agricultural season) based on the NDVI profiles (1st approach). The statistical Student’s <span class="html-italic">t</span>-test shows that S1 and S2 are identical (<span class="html-italic">p</span> &gt; 0.05), and the other scenarios are significantly different (<span class="html-italic">p</span> ≤ 0.05).</p> Full article ">Figure 14
<p>Variation of grain yield (t/ha), between simulated scenarios and the real case (2011–2012 agricultural season) based on the AquaCrop model (second approach). The statistical Student’s <span class="html-italic">t</span>-test shows that S1 and S2 are identical (<span class="html-italic">p</span> &gt; 0.05), and the other scenarios are significantly different (<span class="html-italic">p</span> ≤ 0.05).</p> Full article ">Figure 15
<p>Variation of the biomass (t/ha), between simulated scenarios and the real case (2011–2012 agricultural season) based on the AquaCrop model. The statistical Student’s <span class="html-italic">t</span>-test shows that S1 and S2 are identical (<span class="html-italic">p</span> &gt; 0.05), and the other scenarios are significantly different (<span class="html-italic">p</span> ≤ 0.05).</p> Full article ">
19 pages, 9026 KiB  
Article
A Model for the Relationship between Rainfall, GNSS-Derived Integrated Water Vapour, and CAPE in the Eastern Central Andes
by Maryam Ramezani Ziarani, Bodo Bookhagen, Torsten Schmidt, Jens Wickert, Alejandro de la Torre, Zhiguo Deng and Andrea Calori
Remote Sens. 2021, 13(18), 3788; https://doi.org/10.3390/rs13183788 - 21 Sep 2021
Cited by 10 | Viewed by 2903
Abstract
Atmospheric water vapour content is a key variable that controls the development of deep convective storms and rainfall extremes over the central Andes. Direct measurements of water vapour are challenging; however, recent developments in microwave processing allow the use of phase delays from [...] Read more.
Atmospheric water vapour content is a key variable that controls the development of deep convective storms and rainfall extremes over the central Andes. Direct measurements of water vapour are challenging; however, recent developments in microwave processing allow the use of phase delays from L-band radar to measure the water vapour content throughout the atmosphere: Global Navigation Satellite System (GNSS)-based integrated water vapour (IWV) monitoring shows promising results to measure vertically integrated water vapour at high temporal resolutions. Previous works also identified convective available potential energy (CAPE) as a key climatic variable for the formation of deep convective storms and rainfall in the central Andes. Our analysis relies on GNSS data from the Argentine Continuous Satellite Monitoring Network, Red Argentina de Monitoreo Satelital Continuo (RAMSAC) network from 1999 to 2013. CAPE is derived from version 2.0 of the ECMWF’s (European Centre for Medium-Range Weather Forecasts) Re-Analysis (ERA-interim) and rainfall from the TRMM (Tropical Rainfall Measuring Mission) product. In this study, we first analyse the rainfall characteristics of two GNSS-IWV stations by comparing their complementary cumulative distribution function (CCDF). Second, we separately derive the relation between rainfall vs. CAPE and GNSS-IWV. Based on our distribution fitting analysis, we observe an exponential relation of rainfall to GNSS-IWV. In contrast, we report a power-law relationship between the daily mean value of rainfall and CAPE at the GNSS-IWV station locations in the eastern central Andes that is close to the theoretical relationship based on parcel theory. Third, we generate a joint regression model through a multivariable regression analysis using CAPE and GNSS-IWV to explain the contribution of both variables in the presence of each other to extreme rainfall during the austral summer season. We found that rainfall can be characterised with a higher statistical significance for higher rainfall quantiles, e.g., the 0.9 quantile based on goodness-of-fit criterion for quantile regression. We observed different contributions of CAPE and GNSS-IWV to rainfall for each station for the 0.9 quantile. Fourth, we identify the temporal relation between extreme rainfall (the 90th, 95th, and 99th percentiles) and both GNSS-IWV and CAPE at 6 h time steps. We observed an increase before the rainfall event and at the time of peak rainfall—both for GNSS-integrated water vapour and CAPE. We show higher values of CAPE and GNSS-IWV for higher rainfall percentiles (99th and 95th percentiles) compared to the 90th percentile at a 6-h temporal scale. Based on our correlation analyses and the dynamics of the time series, we show that both GNSS-IWV and CAPE had comparable magnitudes, and we argue to consider both climatic variables when investigating their effect on rainfall extremes. Full article
(This article belongs to the Section Remote Sensing in Geology, Geomorphology and Hydrology)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>(<b>A</b>) Topographic overview (from [<a href="#B45-remotesensing-13-03788" class="html-bibr">45</a>] data) of the study region in the central Andes in north-western Argentina with the outline of the internally-drained central Andes in white (see inset for location in South America with the internally-drained central Andes shown in red). Black lines are international borders. White stars show the GNSS stations used in this study: San Miguel de Tucumán (TUCU, <span class="html-italic">n</span> = 15 years) and San Fernando del Valle de Catamarca (CATA, <span class="html-italic">n</span> = 6 years). (<b>B</b>) Annual mean rainfall derived from TRMM 3B42 (1999–2013) shows rainfall distribution and the location of the GNSS station locations (white stars).</p> Full article ">Figure 2
<p>Logarithmically binned rainfall data (<b>A</b>) and GNSS-IWV (<b>B</b>) for TUCU station (blue dots) and for CATA station (red dots). The fitting parameters of the lognormal distribution (see Equation (<a href="#FD11-remotesensing-13-03788" class="html-disp-formula">11</a>)) show the differences between two stations (2008–2013). Note that the tail of the distribution exhibits power-law behaviour starting at <math display="inline"><semantics> <msub> <mi>x</mi> <mrow> <mi>m</mi> <mi>i</mi> <mi>n</mi> </mrow> </msub> </semantics></math> = 12.9 and with the estimated exponent <math display="inline"><semantics> <mi>α</mi> </semantics></math> = 2.5 for the TUCU station and <math display="inline"><semantics> <msub> <mi>x</mi> <mrow> <mi>m</mi> <mi>i</mi> <mi>n</mi> </mrow> </msub> </semantics></math> = 28 and the estimated exponent <math display="inline"><semantics> <mi>α</mi> </semantics></math> = 2.9 for the CATA station.</p> Full article ">Figure 3
<p>The power-law behaviour using a maximum likelihood approach of log-binned data for the independent variable CAPE (<b>A</b>) for TUCU and (<b>B</b>) for CATA following methods described in [<a href="#B50-remotesensing-13-03788" class="html-bibr">50</a>]. We identify a power-law like behaviour for CAPE values above 2500 J/kg (for TUCU) and 2900 J/kg (for CATA). The <span class="html-italic">p</span>-value greater than 0.1 (TUCU = 0.9, CATA = 0.4) confirms that a power law is a plausible hypothesis for the data.</p> Full article ">Figure 4
<p>The daily mean GNSS-integrated water vapour (orange line) vs. daily mean rainfall from TRMM data (blue line) (<b>A</b>) for TUCU and (<b>C</b>) for CATA stations and daily mean CAPE (red line) from ERA-interim vs. daily mean rainfall from TRMM data (blue line) (<b>B</b>) for TUCU and (<b>D</b>) for CATA stations for (2010–2013). Both datasets represent a high seasonal agreement with rainfall in both stations.</p> Full article ">Figure 5
<p>Squared wavelet coherence between the CAPE and rainfall (2008–2013) for TUCU (<b>A</b>) and CATA (<b>B</b>) stations and between the GNSS-IWV and rainfall (2008–2013) for TUCU (<b>C</b>) and CATA (<b>D</b>) stations. The arrows indicate the lag phase relation between rainfall and CAPE and rainfall and GNSS-IWV.</p> Full article ">Figure 6
<p>Quantile–Quantile plot on a <math display="inline"><semantics> <mrow> <mi>l</mi> <mi>o</mi> <msub> <mi>g</mi> <mi>e</mi> </msub> </mrow> </semantics></math> scale of rainfall vs. GNSS-integrated water vapour shows that most of the data are well correlated when the <math display="inline"><semantics> <mrow> <mi>l</mi> <mi>o</mi> <msub> <mi>g</mi> <mi>e</mi> </msub> </mrow> </semantics></math>(TRMM rainfall) vs. GNSS-IWV is considered in both stations (<b>A</b>) for TUCU and (<b>B</b>) for CATA. Below the 10th percentile, rainfall and GNSS-IWV do not follow an identical distribution at both stations. Black dashed lines indicate the rainfall percentiles (the 10th and 90th percentiles).</p> Full article ">Figure 7
<p>Quantile–Quantile plot on <math display="inline"><semantics> <mrow> <mi>l</mi> <mi>o</mi> <msub> <mi>g</mi> <mi>e</mi> </msub> </mrow> </semantics></math> scale of rainfall vs. <math display="inline"><semantics> <mrow> <mi>l</mi> <mi>o</mi> <msub> <mi>g</mi> <mi>e</mi> </msub> </mrow> </semantics></math> scale of CAPE essentially shows that most of the data are well behaved within the assumed relation for both stations (<b>A</b>) for TUCU and (<b>B</b>) for CATA. Below the 10th percentile, rainfall and CAPE do not follow an identical distribution at both stations. Black dashed lines indicate the rainfall percentiles (the 10th and 90th percentiles).</p> Full article ">Figure 8
<p>The logarithmically binned CAPE vs. median rainfall (<b>A</b>) for TUCU and (<b>B</b>) for CATA. The exponent <math display="inline"><semantics> <mi>β</mi> </semantics></math> in (Equation (<a href="#FD4-remotesensing-13-03788" class="html-disp-formula">4</a>)) is 0.38 for TUCU and 0.3 for CATA.</p> Full article ">Figure 9
<p>The coefficients of quantile regression analysis with their 90% confidence bounds (gray shading), Equations (<a href="#FD7-remotesensing-13-03788" class="html-disp-formula">7</a>)–(<a href="#FD9-remotesensing-13-03788" class="html-disp-formula">9</a>) for 0.75, 0.8, 0.85, 0.9 quantiles and for TUCU and CATA stations. The least-squares regression coefficients (red solid line) with their 90% confidence bounds (dashed lines) are represented.</p> Full article ">Figure 10
<p>The goodness-of-fit criterion, Equations (<a href="#FD7-remotesensing-13-03788" class="html-disp-formula">7</a>)–(<a href="#FD10-remotesensing-13-03788" class="html-disp-formula">10</a>) for 0.75, 0.8, 0.85, and 0.9 quantiles and for TUCU and CATA stations. An improvement for higher quantiles is represented. The values are in percent.</p> Full article ">Figure 11
<p>Averaged GNSS-integrated water vapour (blue line) and CAPE (red line) for the TUCU and CATA stations for the 90th percentile rainfall (<b>A</b>,<b>D</b>), for the 95th percentile rainfall (<b>B</b>,<b>E</b>), and for the 99th percentile rainfall (<b>C</b>,<b>F</b>). We selected all times with rainfall above the 90th, 95th, and 99th percentiles, respectively and their corresponding GNSS-integrated water vapour and CAPE amounts. We then show the correlation for 72 h (event day plus day before and day after). Note that the GNSS-integrated water vapour and CAPE generally increase during the day before the event and that peak values—both for GNSS-integrated water vapour and CAPE—are observed at the day of the 90th, 95th, and 99th event rainfall.</p> Full article ">
17 pages, 6834 KiB  
Article
Aerial and UAV Images for Photogrammetric Analysis of Belvedere Glacier Evolution in the Period 1977–2019
by Carlo Iapige De Gaetani, Francesco Ioli and Livio Pinto
Remote Sens. 2021, 13(18), 3787; https://doi.org/10.3390/rs13183787 - 21 Sep 2021
Cited by 7 | Viewed by 3391
Abstract
Alpine glaciers are strongly suffering the consequences of the temperature rising and monitoring them over long periods is of particular interest for climate change tracking. A wide range of techniques can be successfully applied to survey and monitor glaciers with different spatial and [...] Read more.
Alpine glaciers are strongly suffering the consequences of the temperature rising and monitoring them over long periods is of particular interest for climate change tracking. A wide range of techniques can be successfully applied to survey and monitor glaciers with different spatial and temporal resolutions. However, going back in time to retrace the evolution of a glacier is still a challenging task. Historical aerial images, e.g., those acquired for regional cartographic purposes, are extremely valuable resources for studying the evolution and movement of a glacier in the past. This work analyzed the evolution of the Belvedere Glacier by means of structure from motion techniques applied to digitalized historical aerial images combined with more recent digital surveys, either from aerial platforms or UAVs. This allowed the monitoring of an Alpine glacier with high resolution and geometrical accuracy over a long span of time, covering the period 1977–2019. In this context, digital surface models of the area at different epochs were computed and jointly analyzed, retrieving the morphological dynamics of the Belvedere Glacier. The integration of datasets dating back to earlier times with those referring to surveys carried out with more modern technologies exploits at its full potential the information that at first glance could be thought obsolete, proving how historical photogrammetric datasets are a remarkable heritage for glaciological studies. Full article
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>(<b>a</b>) Location of Belvedere Glacier, base map (source: Swisstopo <a href="http://www.geo.admin.ch" target="_blank">www.geo.admin.ch</a>, accessed on 10 August 2021); (<b>b</b>) view of Monte Rosa from the Belvedere Glacier surface (photograph by Francesco Ioli).</p> Full article ">Figure 2
<p>(<b>a</b>) Sketch of the 1977 aerial acquisitions; (<b>b</b>) sample image; (<b>c</b>) fiducial mark example.</p> Full article ">Figure 3
<p>(<b>a</b>) Sketch of the 1991 aerial acquisitions; (<b>b</b>) sample image; (<b>c</b>) fiducial mark example.</p> Full article ">Figure 4
<p>(<b>a</b>) Sketch of the 2001 aerial acquisitions; (<b>b</b>) sample image; (<b>c</b>) fiducial mark example.</p> Full article ">Figure 5
<p>(<b>a</b>) Sketch of the 2009 aerial acquisitions; (<b>b</b>) sample image.</p> Full article ">Figure 6
<p>(<b>a</b>) Sketch of the 2019 UAV acquisitions; (<b>b</b>) sample image.</p> Full article ">Figure 7
<p>Workflow for the reconstruction of models and their spatial alignment in a common reference frame.</p> Full article ">Figure 8
<p>Sketches of the GCP and CP locations: (<b>a</b>) 1977, 1991 and 2001 surveys; (<b>b</b>) 2009 survey; (<b>c</b>) 2019 survey.</p> Full article ">Figure 9
<p>Examples of points chosen as GCPs or CPs. (<b>a</b>–<b>c</b>), artificial features in 1977, 1991 and 2001 aerial images, respectively; (<b>d</b>–<b>f</b>), natural features in 1977, 1991 and 2001 aerial images, respectively.</p> Full article ">Figure 10
<p>Model accuracy comparison in terms of RMS error on CPs.</p> Full article ">Figure 11
<p>Initial rasterized point clouds: (<b>a</b>) 1977; (<b>b</b>) 1991; (<b>c</b>) 2001; (<b>d</b>) 2009; (<b>e</b>) 2019.</p> Full article ">Figure 12
<p>Binary masks defining the survey coverage: (<b>a</b>) 1977; (<b>b</b>) 1991; (<b>c</b>) 2001; (<b>d</b>) 2009; (<b>e</b>) 2019.</p> Full article ">Figure 13
<p>Interpolated gridded point clouds: (<b>a</b>) 1977; (<b>b</b>) 1991; (<b>c</b>) 2001; (<b>d</b>) 2009; (<b>e</b>) 2019.</p> Full article ">Figure 14
<p>Final mask for defining the glacier boundaries: (<b>a</b>) binary mask overlaps; (<b>b</b>) common area; (<b>c</b>) high-variability area; (<b>d</b>) filtered high-variability area; (<b>e</b>) buffered high-variability area, i.e., considered glacier shape.</p> Full article ">Figure 15
<p>Glacier contours at different epochs: (<b>a</b>) 1977; (<b>b</b>) 1991; (<b>c</b>) 2001; (<b>d</b>) 2009; (<b>e</b>) 2019.</p> Full article ">Figure 16
<p>Binned glacier altitude variations in different periods with bin size set to 25 m: (<b>a</b>) 1977–1991; (<b>b</b>) 1991–2001; (<b>c</b>) 2001–2009; (<b>d</b>) 2009–2019.</p> Full article ">Figure 16 Cont.
<p>Binned glacier altitude variations in different periods with bin size set to 25 m: (<b>a</b>) 1977–1991; (<b>b</b>) 1991–2001; (<b>c</b>) 2001–2009; (<b>d</b>) 2009–2019.</p> Full article ">
18 pages, 4455 KiB  
Article
Influence of Scale Effect of Canopy Projection on Understory Microclimate in Three Subtropical Urban Broad-Leaved Forests
by Xueyan Gao, Chong Li, Yue Cai, Lei Ye, Longdong Xiao, Guomo Zhou and Yufeng Zhou
Remote Sens. 2021, 13(18), 3786; https://doi.org/10.3390/rs13183786 - 21 Sep 2021
Cited by 7 | Viewed by 2783
Abstract
The canopy is the direct receiver and receptor of external environmental variations, and affects the microclimate and energy exchange between the understory and external environment. After autumn leaf fall, the canopy structure of different forests shows remarkable variation, causes changes in the microclimate [...] Read more.
The canopy is the direct receiver and receptor of external environmental variations, and affects the microclimate and energy exchange between the understory and external environment. After autumn leaf fall, the canopy structure of different forests shows remarkable variation, causes changes in the microclimate and is essential for understory vegetation growth. Moreover, the microclimate is influenced by the scale effect of the canopy. However, the difference in influence between different forests remains unclear on a small scale. In this study, we aimed to analyze the influence of the scale effect of canopy projection on understory microclimate in three subtropical broad-leaved forests. Three urban forests: evergreen broad-leaved forest (EBF), deciduous broad-leaved forest (DBF), and mixed evergreen and deciduous broad-leaved forest (MBF) were selected for this study. Sensors for environmental monitoring were used to capture the microclimate data (temperature (T), relative humidity (RH), and light intensity (LI)) for each forest. Terrestrial laser scanning was employed to obtain the canopy projection intensity (CPI) at each sensor location. The results indicate that the influence range of canopy projection on the microclimate was different from stand to stand (5.5, 5, and 3 m). Moreover, there was a strong negative correlation between T and RH, and the time for T and LI to reach a significant correlation in different urban forests was different, as well as the time for RH and LI during the day. Finally, the correlation between CPI and the microclimate showed that canopy projection had the greatest effect on T and RH in MBF, followed by DBF and EBF. In conclusion, our findings confirm that canopy projection can significantly affect understory microclimate. This study provides a reference for the conservation of environmentally sensitive organisms for urban forest management. Full article
(This article belongs to the Special Issue Remote Sensing of Urban Forest Structure)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Study area and location of measuring points where microclimate factors, scanning locations of TLS, and canopy structural parameters were measured.</p> Full article ">Figure 2
<p>Least square linear fitting method was adopted to calibrate the temperature data of each sensor, with one sensor as the standard. The same method was used for relative humidity and light intensity. (<b>a</b>) Temperature (T) measurement data of sensors at the same place, (<b>b</b>) T data of sensors after calibration.</p> Full article ">Figure 3
<p>Interpolated graphs of microclimate data in mixed evergreen deciduous broad-leaved forest at 12:00. All results were divided into 10 levels. A deeper blue color indicates the lowest value, whereas a deeper red color indicates the highest value.</p> Full article ">Figure 4
<p>Extraction process of tree characteristics. (<b>a1</b>) Measurement of transverse diameter. (<b>a2</b>) Measurement of longitudinal diameter. (<b>b</b>) Measurement of tree height. (<b>c</b>) Measurement of canopy thickness. (<b>d</b>) Measurement of canopy cover area. (<b>e</b>) Vertical slices with 1 mm intervals were used to calculate canopy volume.</p> Full article ">Figure 5
<p>Relationship between the area represented by each point cloud and the distance from the scanner to the target.</p> Full article ">Figure 6
<p>The process of CPI computation.</p> Full article ">Figure 7
<p>Correlation of canopy projection intensity to temperature (T) and relative humidity (RH) with scales at 12:00 in the three forests.</p> Full article ">Figure 8
<p>Diurnal variation in the average canopy projection intensity (CPI) within the three broad-leaved forests. EBF, evergreen broad-leaved forest; DBF, deciduous broad-leaved forest; MBF, mixed evergreen, and deciduous broad-leaved forest.</p> Full article ">Figure 9
<p>Curves of hourly variation in microclimate factors in the three forests over a day. (<b>a</b>) The average temperature (T), (<b>b</b>) average relative humidity (RH), and (<b>c</b>) average light intensity (LI).</p> Full article ">Figure 10
<p>Hourly correlation among understory microclimate factors over a day. (<b>a</b>) T and RH, (<b>b</b>) T and LI, (<b>c</b>) RH and LI. LI, light intensity; RH, relative humidity; T, temperature.</p> Full article ">Figure 11
<p>Hourly correlation between CPI and microclimate factors over a day. (<b>a</b>) CPI and T, (<b>b</b>) CPI and RH, (<b>c</b>) CPI and LI. CPI, canopy projection intensity; LI, light intensity; RH, relative humidity; T, temperature.</p> Full article ">Figure 12
<p>Interpolation results of temperature (°C) during the day in mixed evergreen and deciduous broad-leaved forest and hourly superposition effect of canopy projection.</p> Full article ">
24 pages, 4447 KiB  
Article
Remote Sensing of Turbidity in the Tennessee River Using Landsat 8 Satellite
by A. K. M. Azad Hossain, Caleb Mathias and Richard Blanton
Remote Sens. 2021, 13(18), 3785; https://doi.org/10.3390/rs13183785 - 21 Sep 2021
Cited by 36 | Viewed by 6969
Abstract
The Tennessee River in the United States is one of the most ecologically distinct rivers in the world and serves as a great resource for local residents. However, it is also one of the most polluted rivers in the world, and a leading [...] Read more.
The Tennessee River in the United States is one of the most ecologically distinct rivers in the world and serves as a great resource for local residents. However, it is also one of the most polluted rivers in the world, and a leading cause of this pollution is storm water runoff. Satellite remote sensing technology, which has been used successfully to study surface water quality parameters for many years, could be very useful to study and monitor the quality of water in the Tennessee River. This study developed a numerical turbidity estimation model for the Tennessee River and its tributaries in Southeast Tennessee using Landsat 8 satellite imagery coupled with near real-time in situ measurements. The obtained results suggest that a nonlinear regression-based numerical model can be developed using Band 4 (red) surface reflectance values of the Landsat 8 OLI sensor to estimate turbidity in these water bodies with the potential of high accuracy. The accuracy assessment of the estimated turbidity achieved a coefficient of determination (R2) value and root mean square error (RMSE) as high as 0.97 and 1.41 NTU, respectively. The model was also tested on imagery acquired on a different date to assess its potential for routine remote estimation of turbidity and produced encouraging results with R2 value of 0.94 and relatively high RMSE. Full article
(This article belongs to the Special Issue Estimating Inland Water Quality from Remote Sensing Data)
►▼ Show Figures

Figure 1

Figure 1
<p>Remote sensing of water quality (modified from [<a href="#B8-remotesensing-13-03785" class="html-bibr">8</a>]).</p> Full article ">Figure 2
<p>Map showing the location and extent of the study site, the Tennessee River and its surrounding tributaries.</p> Full article ">Figure 3
<p>Landsat 8 OLI images acquired over the study site on 2 December 2018, 13 February 2019, and 15 August 2019.</p> Full article ">Figure 4
<p>Location of in situ turbidity samples collected on each of the three dates for which Landsat 8 satellite images were acquired.</p> Full article ">Figure 5
<p>Flowchart illustrates the workflow followed in development of the model.</p> Full article ">Figure 6
<p>Map showing the locations of in situ turbidity samples collected on each of the three dates, separated into the sample points used for training and testing the model.</p> Full article ">Figure 7
<p>Regression scatterplots for all bands: linear regressions (<b>left</b>); nonlinear exponential regressions (<b>middle</b>); nonlinear power regressions (<b>right</b>).</p> Full article ">Figure 8
<p>Turbidity estimated on 2 December 2018 by the turbidity estimation model. The insets highlight areas of high turbidity variability due to the convergence of lakes and streams. Site A details the confluence of the Tennessee and Hiwassee Rivers, Site B highlights the increased turbidity of South Chickamauga Creek, and Site C details the Harrison Bay area.</p> Full article ">Figure 9
<p>Turbidity estimated on 13 February 2019 by the turbidity estimation model. The insets highlight areas of high turbidity variability due to the convergence of lakes and streams. Site A details the confluence of the Tennessee and Hiwassee Rivers, Site B highlights the increased turbidity of South Chickamauga Creek, and Site C details the Harrison Bay area.</p> Full article ">Figure 10
<p>Scatterplots illustrating the relationship between the predicted and observed turbidity values for 2 December 2018 (<b>left</b>) and 13 February 2019 (<b>right</b>).</p> Full article ">Figure 11
<p>Turbidity estimated on 15 August 2019 by the turbidity estimation model. The insets highlight areas of high turbidity variability due to the convergence of lakes and streams. Site A details the confluence of the Tennessee and Hiwassee Rivers, Site B highlights the increased turbidity of South Chickamauga Creek, and Site C details the Harrison Bay area.</p> Full article ">Figure 12
<p>Scatterplot between the predicted and observed turbidity values for the 15 August 2019 image acquisition.</p> Full article ">
23 pages, 7517 KiB  
Article
Reduction of Air Pollution in Poland in Spring 2020 during the Lockdown Caused by the COVID-19 Pandemic
by Patryk Tadeusz Grzybowski, Krzysztof Mirosław Markowicz and Jan Paweł Musiał
Remote Sens. 2021, 13(18), 3784; https://doi.org/10.3390/rs13183784 - 21 Sep 2021
Cited by 9 | Viewed by 3047
Abstract
The COVID-19 pandemic has affected many aspects of human well-being including air quality. The present study aims at quantifying this effect by means of ground-level concentrations of NO2, PM2.5, as well as aerosol optical depth (AOD) measurements and tropospheric [...] Read more.
The COVID-19 pandemic has affected many aspects of human well-being including air quality. The present study aims at quantifying this effect by means of ground-level concentrations of NO2, PM2.5, as well as aerosol optical depth (AOD) measurements and tropospheric NO2 column number density (NO2 TVCD), during the imposed governmental restrictions in spring 2020. The analyses were performed for both urban and non-built-up areas across the whole of Poland accompanied by Warsaw (urban site) and Strzyzow (a background site). The results revealed that mean PM2.5 concentrations in spring 2020 for urban and non-built-up areas across Poland and for Warsaw were 20%, 23%, 15% lower than the 10-year average, respectively. Analogous mean NO2 concentrations were lower by 20%, 18%, 30% and NO2 TVCD revealed 9%, 4%, 9% reductions in 2020 as compared to 2019. Regarding mean AOD, retrieved from MERRA-2 reanalysis, it was found that for the whole of Poland during spring 2020 the reduction in AOD as compared to the 10-year average was 15%. The contribution of the lockdown within total air pollution reduction is not easily assessable due to anomalous weather conditions in 2020 which resulted in advection of clean air masses identified from MERRA-2 reanalysis and Strzyzow observatory. Full article
(This article belongs to the Special Issue Ground-Based Remote Sensing of the Atmosphere during the COVID-19 Lockdown)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Location of GIOS stations measuring NO<sub>2</sub>, PM<sub>2.5</sub> and the AERONET station. Red squares correspond to urban stations where PM<sub>2.5</sub> and NO<sub>2</sub> are measured, while green squares correspond to non-built-up stations. Red triangles correspond to PM<sub>2.5</sub> urban stations. Red circles to NO<sub>2</sub> urban stations, while green circles are for NO<sub>2</sub> non-built-up stations. A green cross gives the location of the AERONET background station in Strzyzow where aerosol optical properties were measured.</p> Full article ">Figure 2
<p>Relative frequency of 96 h back-trajectories at 500 m over Warsaw obtained from HYSPLIT simulation for (<b>a</b>) 2019, (<b>b</b>) 2020, and (<b>c</b>) for 2010–2019. Different colors show the direction of air mass transport.</p> Full article ">Figure 3
<p>Frequency of surface inversions (temperature gradient greater than 3 °C/100 m) obtained from radiosonde launches at Legionowo weather station 25 km from Warsaw. The blue, red, and orange bars correspond to the springs of 2019, 2020, and to the reference period 2010–2019.</p> Full article ">Figure 4
<p>Temporal variability of 5-day running means of total AOD at 550 nm over Poland obtained from MERRA-2 reanalysis. Blue, green and red lines show data for 2019, 2020 and 2010–2019, respectively. The pink shadow indicates standard deviations of AOD for the 2010–2019 period.</p> Full article ">Figure 5
<p>Temporal variability of 5-day running means (<b>a</b>) AOD at 500 nm obtained from the CIMEL sun photometer; (<b>b</b>) equivalent black carbon concentration (ng/m<sup>3</sup>) from the AE-31 aethalometer; and (<b>c</b>) the aerosol scattering coefficient at 525 nm from the Aurora 4000 nephelometer, all at the Strzyzow station. Blue, green, and red lines correspond to 2019, 2020, and 2010–2019. The pink shadow indicates standard deviations for the 2010–2019 period.</p> Full article ">Figure 6
<p>Mean decadal ground-based PM<sub>2.5</sub> concentrations (µg/m<sup>3</sup>) from GIOS air quality stations in Poland in 2019 (blue line), 2020 (green line) and the multi-annual average (dashed red line). Standard deviations corresponding to multi-annual averages are marked in pink. Panel (<b>a</b>) depicts the average from 15 urban stations with respect to a 10-year multi-annual mean. Panel (<b>b</b>) depicts a single non-built-up station with respect to an 8-year multi-annual mean.</p> Full article ">Figure 7
<p>Mean decadal ground-based PM<sub>2.5</sub> concentrations (µg/m<sup>3</sup>) from two GIOS air quality stations located in Warsaw in 2019 (blue line), 2020 (green line) and the 10-year multi-annual averages (dashed red line). Standard deviations corresponding to multi-annual averages are marked in pink.</p> Full article ">Figure 8
<p>March to June mean ground-based PM<sub>2.5</sub> concentrations (µg/m<sup>3</sup>) (black line: (<b>a</b>–<b>c</b>)), linear regression (orange line: (<b>a</b>,<b>c</b>)) and anomalies (red bars: (<b>a</b>–<b>c</b>)) with respect to trends over: (<b>a</b>) urban areas (<b>b</b>) non-built-up areas; (<b>c</b>) Warsaw. For statistically insignificant trends, anomalies with respect to averages were calculated (purple line: (<b>b</b>)).</p> Full article ">Figure 9
<p>Mean decadal ground-based NO<sub>2</sub> concentrations (µg/m<sup>3</sup>) from GIOS air quality stations in Poland in 2019 (blue line), 2020 (green line), and the 10-year multi-annual average (dashed red line). Standard deviations corresponding to multi-annual averages are marked in pink. Panel (<b>a</b>) depicts the average from all 78 stations in Poland; panel (<b>b</b>) depicts the average from 68 urban stations; panel (<b>c</b>) depicts the average from 10 non-built-up stations.</p> Full article ">Figure 10
<p>Mean decadal ground-based NO<sub>2.5</sub> concentrations (µg/m<sup>3</sup>) from three GIOS air quality stations located in Warsaw in 2019 (blue line), 2020 (green line) and the 10-year multi-annual average (dashed red line). Standard deviations corresponding to multi-annual averages are marked in pink.</p> Full article ">Figure 11
<p>Mean ground-based NO<sub>2</sub> concentrations (µg/m<sup>3</sup>) every year (black line: (<b>a</b>–<b>d</b>)), trends (orange line: (<b>a</b>,<b>b</b>)) and anomalies (red bars: (<b>a</b>–<b>d</b>)) with respect to trends across Poland (<b>a</b>); urban areas (<b>b</b>); non-built-up areas (<b>c</b>); and Warsaw (<b>d</b>). For statistically insignificant trends, anomalies were calculated with respect to averages (purple line: (<b>c</b>,<b>d</b>)).</p> Full article ">Figure 12
<p>Decade variability of median tropospheric NO<sub>2</sub> column number density (molec/cm<sup>2</sup>·10<sup>15</sup>) over Poland (<b>a</b>); urban areas in Poland (<b>b</b>); and non-built-up areas in Poland (<b>c</b>); in 2020 (green line) and 2019 (blue line). The X axis corresponds to a decade of the year.</p> Full article ">Figure 13
<p>NO<sub>2</sub> TVCD (molec/cm<sup>2</sup>) over Poland during 11th decade of the year: 2019 (<b>a</b>) and 2020 (<b>b</b>). Changes in NO<sub>2</sub> TVCD (%) over Poland during the 11th decade (<b>c</b>) (difference of NO<sub>2</sub> TVCD during the 11th decade in 2020 (<b>b</b>) with respect to 2019 (<b>a</b>).</p> Full article ">Figure 14
<p>Decade variability of median NO<sub>2</sub> TVCD (molec/cm<sup>2</sup>·10<sup>15</sup>) in Warsaw in 2020 (green) and 2019 (blue).</p> Full article ">
13 pages, 4655 KiB  
Communication
An Application of Sea Ice Tracking Algorithm for Fast Ice and Stamukhas Detection in the Arctic
by Valeria Selyuzhenok and Denis Demchev
Remote Sens. 2021, 13(18), 3783; https://doi.org/10.3390/rs13183783 - 21 Sep 2021
Cited by 6 | Viewed by 2621
Abstract
For regional environmental studies it is important to know the location of the fast ice edge which affects the coastal processes in the Arctic. The aim of this study is to develop a new automated method for fast ice delineation from SAR imagery. [...] Read more.
For regional environmental studies it is important to know the location of the fast ice edge which affects the coastal processes in the Arctic. The aim of this study is to develop a new automated method for fast ice delineation from SAR imagery. The method is based on a fine resolution hybrid sea ice tracking algorithm utilizing advantages of feature tracking and cross-correlation approaches. The developed method consists of three main steps: drift field retrieval at sub-kilometer scale, selection of motionless features and edge delineation. The method was tested on a time series of C-band co-polarized (HH) ENVISAT ASAR and Sentinel-1 imagery in the Laptev and East Siberian Seas. The comparison of the retrieved edges with the operational ice charts produced by the Arctic and Antarctic Research Institute (Russia) showed a good agreement between the data sets with a mean distance between the edges of <15 km. Thanks to the high density of the ice drift product, the method allows for detailed fast ice edge delineation. In addition, large stamukhas with horizontal size of tens of kilometers can be detected. The proposed method can be applied for regional fast ice mapping and large stamukhas detection to aid coastal research. Additionally, the method can serve as a tool for operational sea ice mapping. Full article
(This article belongs to the Section Remote Sensing Communications)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Flowchart of the fast ice detection algorithm.</p> Full article ">Figure 2
<p>A guide vector by the feature tracking algorithm (solid black arrow) and an associated search area W defined around the end of the vector (dotted black line). The size of searching zone W is set empirically and is 32 × 32 pixels. The red arrows illustrate potential ice drift vectors for a closest computational grid cell (red circle) by the pattern matching algorithm. The black outlined circles correspond to computational grid points used in the pattern matching algorithm.</p> Full article ">Figure 3
<p>An example of methods steps output. The yellow arrows show the sea ice drift field calculated from the ENVISAT SAR image pair taken on 19 (the background image) and 22 December 2007 (step 1). The filtered near-zero displacement points are shown in red (step 2). The blue line shows the resultant fast ice polygons (step 3).</p> Full article ">Figure 4
<p>The fast ice conditions near the Lena Delta for 06 December 2007. The dark blue line shows the fast ice detected by the proposed algorithm. The light blue corresponds to the fast ice from AARI map issued on 5 December 2007. For the cross-comparison, distances between the fast ice edges were measured along the black/red lines (automatically drawn Euclidean distances). The red lines correspond to the negative distances, while the black lines—to the positive ones.</p> Full article ">Figure 5
<p>The retrieved ENVISAT-based fast ice polygons and the corresponding AARI fast ice data for the Laptev Sea region for 3–6 Demember 2007 (<b>a</b>), 6–12 December 2007 (<b>b</b>), 19–22 December 2007 (<b>c</b>), 22–26 December 2007 (<b>d</b>). The dates in the legend correspond to the timing of the first and the second SAR images and the issue date of the AARI ice charts.</p> Full article ">Figure 6
<p>The fast ice polygons derived from Sentinel-1 data and the corresponding AARI fast ice zones over the East Siberian Sea region for 21–31 January 2016 (<b>a</b>), 31 January–2 February 2016 (<b>b</b>), 31 January–1 February 2016 (<b>c</b>), 14–22 February 2016 (<b>d</b>). The dates in the legend correspond to the timing of the first and second SAR images and the issue date of the AARI ice charts.</p> Full article ">Figure 7
<p>The distribution of the measured distances between the AARI and the SAR-derived fast ice edge from ENVISAT ASAR and Sentinel-1 data. The positive values correspond to the cases when the AARI edge is closer to the shore compared to the retrieved edge. Negative values correspond to the cases when the ice edge by the AARI experts is more advanced compared to the retrieved one.</p> Full article ">Figure A1
<p>Algorithm sensitivity to the varying displacement threshold tested with Sentinel-1 (<b>a</b>–<b>c</b>) and ENVISAT data (<b>d</b>). The red lines indicate the overlap of the processed SAR images. The dashed yellow and dark blue lines show the derived fast ice polygons. The fast ice from the AARI chart is shown in light blue.</p> Full article ">
20 pages, 4899 KiB  
Article
Improving the Robustness of the MTI-Estimated Mining-Induced 3D Time-Series Displacements with a Logistic Model
by Jiancun Shi, Zefa Yang, Lixin Wu and Siyu Qiao
Remote Sens. 2021, 13(18), 3782; https://doi.org/10.3390/rs13183782 - 21 Sep 2021
Cited by 1 | Viewed by 1839
Abstract
The previous multi-track InSAR (MTI) method can be used to retrieve mining-induced three-dimensional (3D) surface displacements with high spatial–temporal resolution by incorporating multi-track interferometric synthetic aperture radar (InSAR) observations with a prior model. However, due to the track-by-track strategy used in the previous [...] Read more.
The previous multi-track InSAR (MTI) method can be used to retrieve mining-induced three-dimensional (3D) surface displacements with high spatial–temporal resolution by incorporating multi-track interferometric synthetic aperture radar (InSAR) observations with a prior model. However, due to the track-by-track strategy used in the previous MTI method, no redundant observations are provided to estimate 3D displacements, causing poor robustness and further degrading the accuracy of the 3D displacement estimation. This study presents an improved MTI method to significantly improve the robustness of the 3D mining displacements derived by the previous MTI method. In this new method, a fused-track strategy, instead of the previous track-by-track one, is proposed to process the multi-track InSAR measurements by introducing a logistic model. In doing so, redundant observations are generated and further incorporated into the prior model to solve 3D displacements. The improved MTI method was tested on the Datong coal mining area, China, with Sentinel-1 InSAR datasets from three tracks. The results show that the 3D mining displacements estimated by the improved MTI method had the same spatial–temporal resolution as those estimated by the previous MTI method and about 33.5% better accuracy. The more accurate 3D displacements retrieved from the improved MTI method can offer better data for scientifically understanding the mechanism of mining deformation and assessing mining-related geohazards. Full article
(This article belongs to the Special Issue EO for Mapping Natural Resources and Geohazards)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Flow chart of the improved MTI method for retrieving time-series 3D mining displacements. TS, time-series.</p> Full article ">Figure 2
<p>Geographic location of the region of interest (marked by the yellow star) and the footprints of the collected Sentinel-1 SAR images.</p> Full article ">Figure 3
<p>Cumulative LOS displacements over the AOI (marked by the yellow star in <a href="#remotesensing-13-03782-f002" class="html-fig">Figure 2</a>) from (<b>a</b>) track 040, (<b>b</b>) track 113 and (<b>c</b>) track 120 Sentinel-1 SAR images in the period from January 2018 to May 2019. The red triangle represents the location of a continuously deployed GPS receiver named GPS01. The yellow triangle shows a point P near the mining center of the AOI.</p> Full article ">Figure 4
<p>The QL+GN-estimated parameters <b><span class="html-italic">a</span></b> (<b>a</b>,<b>d</b>,<b>g</b>), <b><span class="html-italic">b</span></b> (<b>b</b>,<b>e</b>,<b>h</b>) and <b><span class="html-italic">c</span></b> (<b>c</b>,<b>f</b>,<b>i</b>) of the logistic model for tracks 040 (<b>a</b>–<b>c</b>), 113 (<b>d</b>–<b>f</b>) and 120 (<b>g</b>–<b>i</b>). The blue dashed rectangular boxes represent the area of deformation due to underground mining. Note that the parameters <b><span class="html-italic">a</span></b> (<b>a</b>,<b>d</b>,<b>g</b>) have no units, and the units of <b><span class="html-italic">b</span></b> (<b>b</b>,<b>e</b>,<b>h</b>) are the reciprocal of the time units.</p> Full article ">Figure 5
<p>(<b>a</b>–<b>c</b>) Estimated annual rates of 3D displacement over the AOI between January 2018 and May 2019 in the vertical, easting and northing directions, respectively. (<b>d</b>–<b>i</b>) The estimated time-series 3D displacements along the profiles AA′ and BB′ (marked by dashed lines in (<b>a</b>–<b>c</b>)).</p> Full article ">Figure 6
<p>(<b>a</b>–<b>c</b>) Comparison of the improved MTI method results and the original MTI method for surface point GPS01 (marked by the red triangle in <a href="#remotesensing-13-03782-f003" class="html-fig">Figure 3</a>) in the region of interest between January 2018 and May 2019.</p> Full article ">Figure 7
<p>The average 3D displacements rates in the vertical, easting and northing directions generated by the simulation procedure (<b>a</b>–<b>c</b>), the improved MTI method (<b>d</b>–<b>f</b>) and the MTI (<b>g</b>–<b>i</b>) methods. (<b>j</b>–<b>l</b>) Three-dimensional RMSE statistical histograms.</p> Full article ">Figure 8
<p>Fitting RMSEs (<b>a</b>–<b>c</b>) and the corresponding histograms (<b>d</b>–<b>f</b>) of the time-series LOS displacements for tracks 040, 113 and 120 using the logistic model. (<b>g</b>–<b>i</b>) An example of the fit of the time-series LOS displacements at the GPS01 point using the logistic model.</p> Full article ">Figure 9
<p>Comparison of the time-series LOS displacement fitting for track 113 at the GPS01 point using logistic, linear and cubic polynomials.</p> Full article ">Figure 10
<p>Comparison of the logistic fit of the LOS displacements of a pixel of the AOI (marked by the yellow triangle in <a href="#remotesensing-13-03782-f003" class="html-fig">Figure 3</a>) using the previous GA+SA and the proposed QL+GN algorithms. The light gray curves indicate the fit of the GA+SA from 50 independent iterations, and the red curve is the mean fit. The dashed blue curve indicates the average fit of 50 iterations of the QL+GN algorithm (almost complete overlap).</p> Full article ">Figure A1
<p>Annual average 3D deformation rates estimated by proposed new method and the MTI-based method. (<b>a</b>–<b>c</b>) The average deformation rates in the vertical, easting and northing directions, respectively, derived by the proposed method. (<b>d</b>–<b>f</b>) The results derived by the MTI-based method.</p> Full article ">
15 pages, 36979 KiB  
Article
Recovery of Tropical Cyclone Induced SST Cooling Observed by Satellite in the Northwestern Pacific Ocean
by Zheng Ling, Zhifeng Chen, Guihua Wang, Hailun He and Changlin Chen
Remote Sens. 2021, 13(18), 3781; https://doi.org/10.3390/rs13183781 - 21 Sep 2021
Cited by 5 | Viewed by 3796
Abstract
Based on the satellite observed sea surface temperature (SST), the recovery of SST cooling induced by the tropical cyclones (TCs) over the northwestern Pacific Ocean is investigated. The results show that the passage of a TC induces a mean maximum cooling in the [...] Read more.
Based on the satellite observed sea surface temperature (SST), the recovery of SST cooling induced by the tropical cyclones (TCs) over the northwestern Pacific Ocean is investigated. The results show that the passage of a TC induces a mean maximum cooling in the SST of roughly −1.25 °C. It was also found that most of this cooling (~87%) is typically erased within 30 days of TC passage. This recovery time depends upon the degree of cooling, with stronger (weaker) SST cooling corresponding to longer (shorter) recovery time. Further analyses show that the mixed layer depth (MLD) and the upper layer thermocline temperature gradient (UTTG) also play an important role in the SST response to TCs. The maximum cooling increases ~0.1 °C for every 7 m decrease in the MLD or every 0.04 °C/m increase in the UTTG. The combined effects of MLD and TC intensity and translation speed on the SST response are also discussed. Full article
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Comparison between (<b>a</b>) the MLD (m) and (<b>b</b>) UTTG (°C/m) calculated from the IAP dataset (blue lines) and the Argo dataset (red lines). The thin and weighted lines represent the original and 21-point smoothed data, respectively.</p> Full article ">Figure 2
<p>(<b>a</b>) Observed frequency (%) of the day of maximum cooling occurrence and (<b>b</b>) the corresponding mean maximum cooling for each day. D0 denotes the time of TC passage.</p> Full article ">Figure 3
<p>(<b>a</b>) Observed frequency (%) of the TC-induced SST cooling occurrence in each 0.5 °C bin and (<b>b</b>) the corresponding recovery time for each bin. The thin and bold error bars represent the standard deviation of the individual value and the mean, respectively. The standard deviation of the mean is multiplied by 10 in the plot.</p> Full article ">Figure 4
<p>The composite time series of TC-induced SST cooling relative to (<b>a</b>) the day of TC passage and (<b>b</b>) the day of maximum cooling based on different recovery time.</p> Full article ">Figure 5
<p>(<b>a</b>) Mean maximum cooling (°C) on each day in different MLD bins and (<b>b</b>) the average and standard deviation of maximum cooling (°C) and recovery time (days) in different MLD bins. D0 denotes the day of TC passage. The thin and bold error bars represent the standard deviation of the individual value and the mean, respectively. The standard deviation of the mean is multiplied by 10 in the plot.</p> Full article ">Figure 6
<p>(<b>a</b>) Mean maximum cooling (°C) on each day in each UTTG bin and (<b>b</b>) the average and standard deviation of maximum cooling (°C) and recovery time (days) in each UTTG bin. The thin and bold error bars represent the standard deviation of the individual value and the mean, respectively. The standard deviation of the mean is multiplied by 10 in the plot. D0 in (<b>a</b>) denotes the day of TC passage.</p> Full article ">Figure 7
<p>The average and standard deviation of maximum cooling and recovery time for each MLD bin for different (<b>a</b>) TC intensity and (<b>b</b>) translation speed (unit: m/s). The sizes of the circles in (<b>a</b>) represent the TD (smallest), TS (next to smallest), STS (next to largest), and TY (largest). The sizes of circles in (<b>b</b>) represent the translation speed of the TCs with &lt;2.5 m/s (smallest), 2.5–5 m/s (next to smallest), 5–7.5 m/s (next to largest) and &gt;7.5 m/s (largest). The thin and bold error bars represent the standard deviation of the individual value and the mean, respectively. The standard deviation of the mean is multiplied by 5 in the plot.</p> Full article ">Figure 8
<p>Spatial distribution of (<b>a</b>) MLD (unit: m), (<b>b</b>) UTTG (unit: 10<sup>−2</sup> °C/m), (<b>c</b>) TC intensity (unit: m/s), (<b>d</b>) TC translation speeds (unit: m/s). All panels are observed/computed on a 1° × 1° spatial grid.</p> Full article ">Figure 9
<p>Spatial distribution of (<b>a</b>) observed maximum cooling (unit: °C), (<b>b</b>) observed recovery time (unit: day), and (<b>c</b>) regressed maximum cooling (unit: °C) calculated from the regression function on MLD, UTTG, TC intensity, and TC translation speeds and (<b>d</b>) regressed recovery time (unit: day) calculated from the regression function on maximum cooling. Contours in (<b>c</b>,<b>d</b>) represent the difference between observation and regressed values (regressed value minus observation) on a 3° × 3° spatial grid. All panels are observed/computed on a 1° × 1° spatial grid.</p> Full article ">Figure 10
<p>Comparison of maximum cooling (<b>a</b>) and recovery time (<b>b</b>) between observation and the regressed value in each 0.5 °C bin of observed maximum cooling.</p> Full article ">
24 pages, 13180 KiB  
Article
A Physics-Driven CNN Model for Real-Time Sea Waves 3D Reconstruction
by Mara Pistellato, Filippo Bergamasco, Andrea Torsello, Francesco Barbariol, Jeseon Yoo, Jin-Yong Jeong and Alvise Benetazzo
Remote Sens. 2021, 13(18), 3780; https://doi.org/10.3390/rs13183780 - 21 Sep 2021
Cited by 12 | Viewed by 4543
Abstract
One of the most promising techniques for the analysis of Spatio-Temporal ocean wave fields is stereo vision. Indeed, the reconstruction accuracy and resolution typically outperform other approaches like radars, satellites, etc. However, it is computationally expensive so its application is typically restricted to [...] Read more.
One of the most promising techniques for the analysis of Spatio-Temporal ocean wave fields is stereo vision. Indeed, the reconstruction accuracy and resolution typically outperform other approaches like radars, satellites, etc. However, it is computationally expensive so its application is typically restricted to the analysis of short pre-recorded sequences. What prevents such methodology from being truly real-time is the final 3D surface estimation from a scattered, non-equispaced point cloud. Recently, we studied a novel approach exploiting the temporal dependence of subsequent frames to iteratively update the wave spectrum over time. Albeit substantially faster, the unpredictable convergence time of the optimization involved still prevents its usage as a continuously running remote sensing infrastructure. In this work, we build upon the same idea, but investigating the feasibility of a fully data-driven Machine Learning (ML) approach. We designed a novel Convolutional Neural Network that learns how to produce an accurate surface from the scattered elevation data of three subsequent frames. The key idea is to embed the linear dispersion relation into the model itself to physically relate the sparse points observed at different times. Assuming that the scattered data are uniformly distributed in the spatial domain, this has the same effect of increasing the sample density of each single frame. Experiments demonstrate how the proposed technique, even if trained with purely synthetic data, can produce accurate and physically consistent surfaces at five frames per second on a modern PC. Full article
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>The WASSfast reconstruction pipeline. Input stereo frames are analyzed to extract a sparse set of corresponding feature points for triangulation. This create a sparse 3D point cloud from which a gridded 3D surface is estimated. The original approach described in [<a href="#B15-remotesensing-13-03780" class="html-bibr">15</a>] is shown at the top with the name “PU Mode”. At the bottom, the CNN mode described in this paper uses a CNN to directly reconstruct the surface with a learning-based approach.</p> Full article ">Figure 2
<p>The point discretization process used to prepare data for the subsequent WASSfast CNN. <b>Top-left</b>: initially, points are defined in the left (or right) camera reference system. <b>Top-right</b>: points are transformed to the mean sea-plane reference system spanning the <span class="html-italic">x</span>–<span class="html-italic">y</span> axis with the z oriented upward. <b>Bottom-right</b>: points are parallel projected into the regular grid defined on the mean sea-plane. <b>Bottom-left</b>: A closeup of what happens if multiple points (a, b, c) falls on the same grid cell. A random point is chosen and its <span class="html-italic">x</span>–<span class="html-italic">y</span> coordinates are approximated to the coordinate of the grid cell center. This way, the entire grid cell takes the elevation value of the randomly chosen point.</p> Full article ">Figure 3
<p>The WASSfast surface reconstruction CNN. Input is composed by 3 frames taken at time <math display="inline"><semantics> <mrow> <mo>(</mo> <mi>t</mi> <mo>−</mo> <mn>1</mn> <mo>)</mo> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <mo>(</mo> <mi>t</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> </semantics></math>. Each frame is a 2-channel <math display="inline"><semantics> <mrow> <mn>256</mn> <mo>×</mo> <mn>256</mn> <mo>×</mo> <mn>2</mn> </mrow> </semantics></math> tensor containing the sparse elevation data and the validity mask. The phase rotation matrices <math display="inline"><semantics> <msub> <mi mathvariant="script">P</mi> <mrow> <msub> <mo>Δ</mo> <mi>t</mi> </msub> <mo>−</mo> <mn>1</mn> </mrow> </msub> </semantics></math> and <math display="inline"><semantics> <msub> <mi mathvariant="script">P</mi> <mrow> <msub> <mo>Δ</mo> <mi>t</mi> </msub> <mo>+</mo> <mn>1</mn> </mrow> </msub> </semantics></math> are assumed to be known according to the current sequence frame rate, wave propagation direction, etc. Frames <math display="inline"><semantics> <msub> <mi mathvariant="script">I</mi> <mrow> <mi>t</mi> <mo>−</mo> <mn>1</mn> </mrow> </msub> </semantics></math> and <math display="inline"><semantics> <msub> <mi mathvariant="script">I</mi> <mrow> <mi>t</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </semantics></math> are processed in parallel by 2 depth completion blocks with shared weights. Then, the temporal combiner transports the surfaces <math display="inline"><semantics> <msub> <mi mathvariant="script">S</mi> <mrow> <mi>t</mi> <mo>−</mo> <mn>1</mn> </mrow> </msub> </semantics></math> and <math display="inline"><semantics> <msub> <mi mathvariant="script">S</mi> <mrow> <mi>t</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </semantics></math> to time <span class="html-italic">t</span>. The predicted surfaces are multiplied by their original masks (<math display="inline"><semantics> <msub> <mi mathvariant="script">M</mi> <mrow> <mi>t</mi> <mo>−</mo> <mn>1</mn> </mrow> </msub> </semantics></math>, <math display="inline"><semantics> <msub> <mi mathvariant="script">M</mi> <mrow> <mi>t</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </semantics></math>) and merged with <math display="inline"><semantics> <msub> <mi mathvariant="script">I</mi> <mi>t</mi> </msub> </semantics></math>, creating new denser data <math display="inline"><semantics> <mrow> <mo>(</mo> <mover accent="true"> <mi mathvariant="script">I</mi> <mo>¯</mo> </mover> <mo>,</mo> <mover accent="true"> <mi mathvariant="script">M</mi> <mo>¯</mo> </mover> <mo>)</mo> </mrow> </semantics></math>. The result is then processed by the surface reconstruction block to produce the final surface <math display="inline"><semantics> <msub> <mi mathvariant="script">O</mi> <mi>t</mi> </msub> </semantics></math>.</p> Full article ">Figure 4
<p>The depth completion block involves a sequence of sparse convolution layers (see <a href="#remotesensing-13-03780-f005" class="html-fig">Figure 5</a>), interleaved by ReLU activations.</p> Full article ">Figure 5
<p>The sparse convolution operation takes an input tensor composed by sparse data (in white) and the associated validity mask (in yellow). Data are convolved and then normalized to account only the valid points encompassed by the kernel. Mask is dilated by the max pooling operation and finally concatenated to the output.</p> Full article ">Figure 6
<p>Example of one synthetically generated scenario with network input at different densities <math display="inline"><semantics> <mrow> <mi>d</mi> <mo>=</mo> <mn>0.1</mn> <mo>,</mo> <mn>0.2</mn> </mrow> </semantics></math> and the corresponding <math display="inline"><semantics> <msub> <mover accent="true"> <mi mathvariant="script">O</mi> <mo>¯</mo> </mover> <mi>t</mi> </msub> </semantics></math> (<b>right</b>).</p> Full article ">Figure 7
<p>Top row: comparison of the mean absolute error (<b>left</b>) and peak signal to noise ratio (<b>right</b>) of the surface reconstructed with WASSfast CNN, SparseCnn and IDW varying the sample density. Bottom row: frequency spectra of timeseries extracted from a grid center when reconstructing a synthetic sequence at different sampling densities.</p> Full article ">Figure 8
<p>Qualitative result of our CNN for sea waves’ surface reconstruction. From <b>left</b> to <b>right</b>: sparse input data, IDW interpolation, output of depth completion, WASSfast CNN output <math display="inline"><semantics> <msub> <mi mathvariant="script">O</mi> <mi>t</mi> </msub> </semantics></math>, ground truth output <math display="inline"><semantics> <msub> <mover accent="true"> <mi mathvariant="script">O</mi> <mo>¯</mo> </mover> <mi>t</mi> </msub> </semantics></math>. Each row shows a different scenario with an increasing sampling. Note how the full network output (with temporal combiner and an additional feed-forward CNN step) improves the reconstruction of the depth completion block alone (SparseCNN), especially at high frequencies. Colorbar is in meters.</p> Full article ">Figure 9
<p>Surface reconstruction errors (in meters) when reconstructing the synthetically generated data. From <b>left</b> to <b>right</b>: sparse input data, ground truth, IDW interpolation, sparseCNN, WASSfast CNN. Each row shows a different scenario with an increasing sampling.</p> Full article ">Figure 10
<p>Time series comparison between WASS, WASSfast PU and WASSfast CNN on the three sequences at the Gageocho ORS. Pearson’s correlation between each WASSfast mode and standard WASS is reported in the legends.</p> Full article ">Figure 11
<p>Frequency-spectra comparison between WASS, WASSfast PU, and WASSfast CNN. <b>Bottom</b>-<b>right</b>: The reconstructed area (red polygon) with the grid point used to extract the elevation timeseries.</p> Full article ">Figure 12
<p>Directional spectra sliced from the 3D spectrum <math display="inline"><semantics> <mrow> <mi mathvariant="script">S</mi> <mo>(</mo> <msub> <mi>K</mi> <mi>x</mi> </msub> <mo>,</mo> <msub> <mi>K</mi> <mi>y</mi> </msub> <mo>,</mo> <msub> <mi>ω</mi> <mi>a</mi> </msub> <mo>)</mo> </mrow> </semantics></math> at <math display="inline"><semantics> <mrow> <mi>ω</mi> <mo>=</mo> <mn>0.3</mn> <mo>,</mo> <mn>0.4</mn> <mo>,</mo> <mn>0.5</mn> </mrow> </semantics></math> Hz for record G20200916T01000. <b>Top row</b>: WASSfast PU mode; <b>Bottom row</b>: WASSfast CNN mode.</p> Full article ">Figure 13
<p>Qualitative comparison of the surface grid reconstructed by WASS (<b>Top</b>) and WASSfast PU (<b>Mid</b>) and WASSfast CNN (<b>Bottom</b>) for record G201810061400.</p> Full article ">
27 pages, 4803 KiB  
Article
Satellite-Derived Barrier Response and Recovery Following Natural and Anthropogenic Perturbations, Northern Chandeleur Islands, Louisiana
by Julie C. Bernier, Jennifer L. Miselis and Nathaniel G. Plant
Remote Sens. 2021, 13(18), 3779; https://doi.org/10.3390/rs13183779 - 21 Sep 2021
Cited by 4 | Viewed by 2863
Abstract
The magnitude and frequency of storm events, relative sea-level rise (RSLR), sediment supply, and anthropogenic alterations drive the morphologic evolution of barrier island systems, although the relative importance of any one driver will vary with the spatial and temporal scales considered. To explore [...] Read more.
The magnitude and frequency of storm events, relative sea-level rise (RSLR), sediment supply, and anthropogenic alterations drive the morphologic evolution of barrier island systems, although the relative importance of any one driver will vary with the spatial and temporal scales considered. To explore the relative contributions of storms and human alterations to sediment supply on decadal changes in barrier landscapes, we applied Otsu’s thresholding method to multiple satellite-derived spectral indices for coastal land-cover classification and analyzed Landsat satellite imagery to quantify changes to the northern Chandeleur Islands barrier system since 1984. This high temporal-resolution dataset shows decadal-scale land-cover oscillations related to storm–recovery cycles, suggesting that shorter and (or) less resolved time series are biased toward storm impacts and may significantly overpredict land-loss rates and the timing of barrier morphologic state changes. We demonstrate that, historically, vegetation extent and persistence were the dominant controls on alongshore-variable landscape response and recovery following storms, and are even more important than human-mediated sediment input. As a result of extensive vegetation losses over the past few decades, however, the northern Chandeleur Islands are transitioning to a new morphologic state in which the landscape is dominated by intertidal environments, indicating reduced resilience to future storms and possibly rapid transitions in morphologic state with increasing rates of RSLR. Full article
(This article belongs to the Special Issue New Insights into Ecosystem Monitoring Using Geospatial Techniques)
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) Regional map showing locations of tropical cyclones that passed within 200 km of the study area since 1984. Hurricane Frederic (1979) is also shown. Inset satellite image indicates study-area extent shown in panels (<b>b</b>,<b>c</b>). (<b>b</b>) Landsat 5 satellite image acquired 25-March-1984 and (<b>c</b>) Landsat 8 satellite image acquired 10-January-2019 show subaerial configuration of the northern Chandeleur Islands at the beginning and end of the analysis period, respectively. Imagery is overlaid with study-area subdivisions and maximum as-constructed berm extent. False-color images use bands 4, 5, 3 (Landsat 5) or 5, 6, 3 (Landsat 8). [Abbreviations: HP, Hewes Point; LST, approximate latitude of longshore transport node; PI, Palos Island; SC, Smack Channel Cut; WI, Western Islands].</p> Full article ">Figure 2
<p>(<b>a</b>) Land-cover classification workflow, showing successive thresholding and masking of modified normalized difference water index (mNDWI), normalized difference bare land index (NBLI), and normalized difference vegetation index (NDVI) spectral indices to extract (<b>b</b>) water, (<b>c</b>) sand and vegetated, and (<b>d</b>) intertidal subclasses defined in this study. (<b>e</b>) Vector shoreline, sand, and vegetated extents were extracted by contouring the mNDWI, NBLI, and NDVI images using the calculated Otsu thresholds. Extent shown in panels (<b>b</b>–<b>e</b>) corresponds to subarea 2 (<a href="#remotesensing-13-03779-f001" class="html-fig">Figure 1</a>; <a href="#remotesensing-13-03779-t001" class="html-table">Table 1</a>) and uses Landsat 5 image acquired 31-January-1986 as example.</p> Full article ">Figure 3
<p>Temporal changes in (<b>a</b>,<b>b</b>) land-cover area, (<b>c</b>) land-cover extent as a percent of total barrier-platform (sand plus vegetated plus intertidal) area, and (<b>d</b>) ratio of intertidal to total island (sand plus vegetated) extents across the entire study area. After 2010, the barrier platform was dominated by intertidal areas, indicated by values in (<b>d</b>) that are greater than 1 (horizontal dashed line). The timing of berm construction (brown vertical lines) and tropical cyclones (gray vertical lines) that passed within 200 km of the northern Chandeleur Islands are shown. Significant decreases in sand, vegetated, total island (sand plus vegetated), and barrier platform (sand plus vegetated plus intertidal) extents were observed after Hurricanes Georges (28-September-1998) and Katrina (29-August-2005) (black vertical lines); long-term changes show somewhat cyclical trends, indicated by dashed lines in (<b>b</b>), that can be related to storm–recovery cycles.</p> Full article ">Figure 4
<p>Temporal changes in land-cover extents for (<b>a</b>) subset area 1, (<b>b</b>) subset area 2, (<b>c</b>) subset area 3, (<b>d</b>) subset area 4, and (<b>e</b>) subset area 5 illustrates alongshore-variable response of the northern Chandeleur Islands to perturbations such as tropical cyclones (gray vertical lines), including Hurricanes Georges (28-September-1998) and Katrina (29-August-2005) (black vertical lines), and berm construction (brown vertical lines). See <a href="#remotesensing-13-03779-t001" class="html-table">Table 1</a> for description of study area subdivisions.</p> Full article ">Figure 5
<p>Temporal changes in land-cover extents as a percent of total barrier-platform (sand plus vegetated plus intertidal) area at the time of image acquisition for (<b>a</b>) subset area 1, (<b>b</b>) subset area 2, (<b>c</b>) subset area 3, (<b>d</b>) subset area 4, and (<b>e</b>) subset area 5 illustrates alongshore-variable response of the northern Chandeleur Islands to perturbations such as tropical cyclones (gray vertical lines), including Hurricanes Georges (28-September-1998) and Katrina (29-August-2005) (black vertical lines), and berm construction (brown vertical lines). See <a href="#remotesensing-13-03779-t001" class="html-table">Table 1</a> for description of study area subdivisions.</p> Full article ">Figure 6
<p>Boxplot showing distribution of (<b>a</b>) back-barrier platform and (<b>b</b>) sea-shoreline positions along cross-shore transects spaced 300 m alongshore relative to offshore baseline (<a href="#app1-remotesensing-13-03779" class="html-app">Figure S1</a>) at the northern Chandeleur Islands for 25-March-1984 to 22-April-1998 (pre-Hurricane Georges), 10-January-1999 to 24-March-2005 (pre-Hurricane Katrina), 18-October-2005 to 18-Febuary-2010 (post-Hurricane Katrina), and 3-December-2010 to 19-January-2019 (post-berm construction). 25-March-1984 barrier-platform extent is delineated by gray line; dashed horizontal lines indicate study area subdivision boundaries.</p> Full article ">Figure 7
<p>Plot showing barrier-platform (colored bars) and vegetated (gray bars) widths averaged along cross-shore transects spaced 300 m alongshore (<a href="#app1-remotesensing-13-03779" class="html-app">Figure S1</a>) at the northern Chandeleur Islands for (<b>a</b>) 25-March-1984 to 22-April-1998 (pre-Hurricane Georges), (<b>b</b>) 10-January-1999 to 24-March-2005 (pre-Hurricane Katrina), (<b>c</b>) 18-October-2005 to 18-Febuary-2010 (post Hurricane Katrina), and (<b>d</b>) 3-December-2010 to 19-January-2019 (post-berm construction). Bars represent average feature width based a minimum of 3 observations per transect per period; no bar indicates 0 (not present), 1, or 2 observations per transect per time period. Dashed horizontal lines indicate study area subdivision boundaries. Feature widths were not calculated along transect 57, where the oblique nature of the back-barrier marsh island, coupled with a change in baseline orientation, caused inconsistencies in delineating feature extents.</p> Full article ">Figure 8
<p>Linear regression plots of barrier-platform versus vegetated width along cross-shore transects spaced 300 m alongshore (<a href="#app1-remotesensing-13-03779" class="html-app">Figure S1</a>) at the northern Chandeleur Islands for (<b>a</b>) 25-March-1984 to 22-April-1998 (pre-Hurricane Georges), (<b>b</b>) 10-January-1999 to 24-March-2005 (pre-Hurricane Katrina), (<b>c</b>) 18-October-2005 to 18-Febuary-2010 (post Hurricane Katrina), and (<b>d</b>) 3-December-2010 to 19-January-2019 (post-berm construction). Individual observations (light gray crosses) are overlaid with values averaged along cross-shore transects for each period; dashed lines indicate regression 95% confidence bounds.</p> Full article ">Figure 9
<p>Vegetation persistence plot for the northern Chandeleur Islands represents the percent of classed land-cover images from this study for which each pixel was classed as vegetated for the entire analysis period (25-March-1984 to 19-January-2019; N = 75). Imagery is overlaid with study-area subdivisions (<a href="#remotesensing-13-03779-t001" class="html-table">Table 1</a>); 25-March-1984 barrier-platform extent is shown for reference.</p> Full article ">Figure 10
<p>Vegetation persistence plots for the northern Chandeleur Islands represent the percent of classed land-cover images from this study for which each pixel was classed as vegetated for (<b>a</b>) 25-March-1984 to 22-April-1998 (pre-Hurricane Georges; N = 27), (<b>b</b>) 10-January-1999 to 24-March-2005 (pre-Hurricane Katrina; N = 23), (<b>c</b>) 18-October-2005 to 18-Febuary-2010 (post Hurricane Katrina; N = 8), and (<b>d</b>) 3-December-2010 to 19-January-2019 (post-berm construction; N = 17). Imagery is overlaid with study-area subdivisions (<a href="#remotesensing-13-03779-t001" class="html-table">Table 1</a>); 25-March-1984 barrier-platform extent is shown for reference.</p> Full article ">
21 pages, 4414 KiB  
Review
Progress and Trends in the Application of Google Earth and Google Earth Engine
by Qiang Zhao, Le Yu, Xuecao Li, Dailiang Peng, Yongguang Zhang and Peng Gong
Remote Sens. 2021, 13(18), 3778; https://doi.org/10.3390/rs13183778 - 21 Sep 2021
Cited by 122 | Viewed by 16488
Abstract
Earth system science has changed rapidly due to global environmental changes and the advent of Earth observation technology. Therefore, new tools are required to monitor, measure, analyze, evaluate, and model Earth observation data. Google Earth (GE) was officially launched by Google in 2005 [...] Read more.
Earth system science has changed rapidly due to global environmental changes and the advent of Earth observation technology. Therefore, new tools are required to monitor, measure, analyze, evaluate, and model Earth observation data. Google Earth (GE) was officially launched by Google in 2005 as a ”geobrowser”, and Google Earth Engine (GEE) was released in 2010 as a cloud computing platform with substantial computational capabilities. The use of these two tools or platforms in various applications, particularly as used by the remote sensing community, has developed rapidly. In this paper, we reviewed the applications and trends in the use of GE and GEE by analyzing peer-reviewed articles, dating up to January 2021, in the Web of Science (WoS) core collection using scientometric analysis (i.e., by using CiteSpace) and meta-analysis. We found the following: (1) the number of articles describing the use of GE or GEE increased substantially from two in 2006 to 530 in 2020. The number of GEE articles increased much faster than those concerned with the use of GE. (2) Both GE and GEE were extensively used by the remote sensing community as multidisciplinary tools. GE articles covered a broader range of research areas (e.g., biology, education, disease and health, economic, and information science) and appeared in a broader range of journals than those concerned with the use of GEE. (3) GE and GEE shared similar keywords (e.g., “land cover”, “water”, “model”, “vegetation”, and “forest”), which indicates that their application is of great importance in certain research areas. The main difference was that articles describing the use of GE emphasized its use as a visual display platform, while those concerned with GEE placed more emphasis on big data and time-series analysis. (4) Most applications of GE and GEE were undertaken in countries, such as the United States, China, and the United Kingdom. (5) GEE is an important tool for analysis, whereas GE is used as an auxiliary tool for visualization. Finally, in this paper, the merits and limitations of GE and GEE, and recommendations for further improvements, are summarized from an Earth system science perspective. Full article
(This article belongs to the Special Issue Feature Papers for Remote Sensing Image Processing Section)
►▼ Show Figures

Figure 1

Figure 1
<p>Change in the number of publications and citations relevant to GE and GEE (dated up to January 2021).</p> Full article ">Figure 2
<p>Journals in which papers related to (<b>a</b>) GE and (<b>b</b>) GEE were published.</p> Full article ">Figure 3
<p>Research disciplines in which (<b>a</b>) GE and (<b>b</b>) GEE were applied.</p> Full article ">Figure 4
<p>Keywords related to (<b>a</b>) GE and (<b>b</b>) GEE.</p> Full article ">Figure 5
<p>Co-occurrence keywords were used in relation to (<b>a</b>) GE and (<b>b</b>) GEE. The size of the node represents the frequency of the occurrence of the keyword, the connecting lines indicate the co-occurrence relationships for the keyword, and a purple outer circle indicates that the node is a key node (betweenness centrality &gt; 0.1).</p> Full article ">Figure 6
<p>Numbers of published GE and GEE papers shown by (<b>a</b>) journal and (<b>b</b>) research area.</p> Full article ">
24 pages, 6805 KiB  
Article
Mapping Canopy Heights in Dense Tropical Forests Using Low-Cost UAV-Derived Photogrammetric Point Clouds and Machine Learning Approaches
by He Zhang, Marijn Bauters, Pascal Boeckx and Kristof Van Oost
Remote Sens. 2021, 13(18), 3777; https://doi.org/10.3390/rs13183777 - 20 Sep 2021
Cited by 14 | Viewed by 4132
Abstract
Tropical forests are a key component of the global carbon cycle and climate change mitigation. Field- or LiDAR-based approaches enable reliable measurements of the structure and above-ground biomass (AGB) of tropical forests. Data derived from digital aerial photogrammetry (DAP) on the [...] Read more.
Tropical forests are a key component of the global carbon cycle and climate change mitigation. Field- or LiDAR-based approaches enable reliable measurements of the structure and above-ground biomass (AGB) of tropical forests. Data derived from digital aerial photogrammetry (DAP) on the unmanned aerial vehicle (UAV) platform offer several advantages over field- and LiDAR-based approaches in terms of scale and efficiency, and DAP has been presented as a viable and economical alternative in boreal or deciduous forests. However, detecting with DAP the ground in dense tropical forests, which is required for the estimation of canopy height, is currently considered highly challenging. To address this issue, we present a generally applicable method that is based on machine learning methods to identify the forest floor in DAP-derived point clouds of dense tropical forests. We capitalize on the DAP-derived high-resolution vertical forest structure to inform ground detection. We conducted UAV-DAP surveys combined with field inventories in the tropical forest of the Congo Basin. Using airborne LiDAR (ALS) for ground truthing, we present a canopy height model (CHM) generation workflow that constitutes the detection, classification and interpolation of ground points using a combination of local minima filters, supervised machine learning algorithms and TIN densification for classifying ground points using spectral and geometrical features from the UAV-based 3D data. We demonstrate that our DAP-based method provides estimates of tree heights that are identical to LiDAR-based approaches (conservatively estimated NSE = 0.88, RMSE = 1.6 m). An external validation shows that our method is capable of providing accurate and precise estimates of tree heights and AGB in dense tropical forests (DAP vs. field inventories of old forest: r2 = 0.913, RMSE = 31.93 Mg ha−1). Overall, this study demonstrates that the application of cheap and easily deployable UAV-DAP platforms can be deployed without expert knowledge to generate biophysical information and advance the study and monitoring of dense tropical forests. Full article
(This article belongs to the Special Issue UAV Applications for Forest Management: Wood Volume, Biomass, Mapping)
►▼ Show Figures

Figure 1

Figure 1
<p>Illustration of DAP and ALS data. Transect (thickness = 0.5 m) showing the comparison between the DAP and ALS point clouds for a dense tropical forest (from the Yangambi datasets, see text below). The black and red dash lines show that the ALS- and DAP-measured digital surface models (DSM) are well aligned. The red solid line shows the limited capability of DAP to detect the ground. The black line is the ALS-derived digital terrain model (DTM).</p> Full article ">Figure 2
<p>Flowchart showing the structure of the study.</p> Full article ">Figure 3
<p>Description of the study sites. (<b>a</b>) Location of the study site. (<b>b</b>) Satellite image of the study sites. (<b>c</b>) Illustration of the remote sensing data sources in the Yangambi site. The gray-scale background layer indicates the ALS-based DEM. (<b>d</b>) Location of inventory plots in the Yoko site. The orthomosaic depicts the survey area covered by DAP. The gray-scale background layer represents SRTM altimetry.</p> Full article ">Figure 4
<p>Setup of the UAV survey. (<b>a</b>) Mavic drone with a L1D-20c camera model. (<b>b</b>) GoPro Hero 3 camera mounted on a Phantom 3 drone with the PPK-GPS system. (<b>c</b>) Reach RS base station placed in an open area providing positioning correction for the PPK solution. (<b>d</b>) Flight plan for UAV photogrammetry of the Yangambi site. (<b>e</b>) Images of the Yoko site processed in the Pix4D Mapper by applying the “fusion” approach.</p> Full article ">Figure 5
<p>(<b>a</b>) Digital surface model (DSM) of the Yangambi site derived from DAP (framed in red outline) and ALS (outside the red frame). (<b>b</b>) Comparison of the elevation values of each grid (0.5 m × 0.5 m) between DAP- and ALS-derived DSMs. (<b>c</b>) Cumulative probability of the absolute error between DAP- and ALS-derived DSMs.</p> Full article ">Figure 6
<p>(<b>a</b>) Normalized point cloud of the intersection area in the Yangambi site derived from ALS (top) and DAP (bottom). (<b>b</b>) Horizontal mirrored histogram showing elevation profiles of DAP- and ALS-derived point clouds. (<b>c</b>) Demonstration of ground point distribution using a threshold of 2 m from the DAP point cloud (n = 8547 pts).</p> Full article ">Figure 7
<p>(<b>a</b>) 3D view of the DAP-derived digital elevation model (DEM) and local minimum points. The DEM was the bottom layer of the raster, which was rasterized from the DAP point cloud using the minimum elevation value at 0.5 m resolution. Local minima points reflect candidate points detected by a moving window (r = 20 m) based on the DEM. (<b>b</b>) Histogram showing the shift of elevation values of the candidate points (r = 20 m) between ALS and DAP measurements. (<b>c</b>) Variation of ground/non-ground points with different sizes of the moving window. The accuracy indicates the ratio of ground points to the total number of candidate points.</p> Full article ">Figure 8
<p>Comparison of DAP-based outputs and the ALS reference. (<b>a</b>) DTM generated using DAP-based methods. The black marks depict the points that were classified as ground. (<b>b</b>) The reference ALS DTM. (<b>c</b>) Comparison of elevation values between DAP-derived DTMs and the reference ALS DTM of each grid. (<b>d</b>) Difference map between DAP-derived DTMs and the reference ALS DTM.</p> Full article ">Figure 9
<p>(<b>a</b>) CHM generated using the DAP approach. (<b>b</b>) CHM of the ALS reference. (<b>c</b>) Standard error map of co-kriging in DTM generation. The black marks refer to the points that were classified as ground. We qualified the DTM/CHM spatially by an SE of 1.5 m as the threshold for the following figures. (<b>d</b>) CHM comparison at grid level (0.5 m × 0.5 m). Black points show high-quality observations, and red points show low-quality observations. The “overall” statistics includes all observations, and “high quality” refers to statistics that include only high-quality observations—the same for the remainder. (<b>e</b>) CHM comparison at tree level. The observations were treetop heights. (<b>f</b>) Segmentation of tree crowns. Points within the polygon show the treetops. Blue outlines show high-quality observations, and red outlines show low-quality observations. (<b>g</b>) Plot-level (40 m × 40 m) comparison of mean canopy height. (<b>h</b>) Plot-level comparison of the 75th percentile of canopy height.</p> Full article ">Figure 10
<p>Application of DAP-derived CHM and <span class="html-italic">AGB</span> estimations for the Yoko site (external validation). (<b>a</b>) Comparison between CHM-derived mean canopy height (<span class="html-italic">H</span><sub>mean</sub>) and tree height measured from field inventory. The error bar on X shows the standard error of sampled trees, and the error bar on Y shows the standard error in DTM generation. (<b>b</b>) <span class="html-italic">AGB</span> estimation using CHM-derived mean canopy height (<span class="html-italic">H</span><sub>mean</sub>). The CHM-<span class="html-italic">AGB</span> model was calibrated using field inventory plots with different stand age classes. (<b>c</b>) <span class="html-italic">AGB</span> estimation using CHM-derived 75th percentile of canopy height (<span class="html-italic">H</span><sub>75</sub>). (<b>d</b>) <span class="html-italic">AGB</span> estimation map in the Yoko site.</p> Full article ">Figure A1
<p>DRC LiDAR sampling distribution over the forest-type classification map. The study sites (Yangambi and Yoko) were located within the same type of forest. Note: the LiDAR transects (~1.5 km × 11 km) are highlighted in bold for display. Data source of forest-type classification map: Réjou-Méchain et al. (2021) [<a href="#B30-remotesensing-13-03777" class="html-bibr">30</a>].</p> Full article ">Figure A2
<p>Precision maps of the Mavic camera derived from Monte Carlo simulation. (<b>a</b>) Precision map using the single GPS of the Mavic drone without the assistance of GoPro PPK-GPS. (<b>b</b>) Precision map with the assistance of GoPro PPK-GPS georeferencing.</p> Full article ">Figure A3
<p>(<b>a</b>) Demonstration of spectral and structural features extracted from the UAV-DAP point cloud for ground point classification. (<b>b</b>) Importance of features in the RF classification model. The candidate points were derived from a moving window of 5 m.</p> Full article ">Figure A4
<p>Relation between resampled coarse DAP DSM and reference ALS DTM. Note: the black auxiliary line is the 1:1 reference line, and the red auxiliary line is the 1:1 reference line with mean tree height as intercept.</p> Full article ">Figure A5
<p>DTM generation in the Yoko site. (<b>a</b>) Raw orthomosaic. (<b>b</b>) Ground points filtered by CSF. (<b>c</b>) Rasterized DTM based on CSF and the ground points filtered by RF+TIN workflow. (<b>d</b>) DTM generated by the “stack and minimum” method.</p> Full article ">
19 pages, 2881 KiB  
Article
Improving YOLOv5 with Attention Mechanism for Detecting Boulders from Planetary Images
by Linlin Zhu, Xun Geng, Zheng Li and Chun Liu
Remote Sens. 2021, 13(18), 3776; https://doi.org/10.3390/rs13183776 - 20 Sep 2021
Cited by 135 | Viewed by 20701
Abstract
It is of great significance to apply the object detection methods to automatically detect boulders from planetary images and analyze their distribution. This contributes to the selection of candidate landing sites and the understanding of the geological processes. This paper improves the state-of-the-art [...] Read more.
It is of great significance to apply the object detection methods to automatically detect boulders from planetary images and analyze their distribution. This contributes to the selection of candidate landing sites and the understanding of the geological processes. This paper improves the state-of-the-art object detection method of YOLOv5 with attention mechanism and designs a pyramid based approach to detect boulders from planetary images. A new feature fusion layer has been designed to capture more shallow features of the small boulders. The attention modules implemented by combining the convolutional block attention module (CBAM) and efficient channel attention network (ECA-Net) are also added into YOLOv5 to highlight the information that contribute to boulder detection. Based on the Pascal Visual Object Classes 2007 (VOC2007) dataset which is widely used for object detection evaluations and the boulder dataset that we constructed from the images of Bennu asteroid, the evaluation results have shown that the improvements have increased the performance of YOLOv5 by 3.4% in precision. With the improved YOLOv5 detection method, the pyramid based approach extracts several layers of images with different resolutions from the large planetary images and detects boulders of different scales from different layers. We have also applied the proposed approach to detect the boulders on Bennu asteroid. The distribution of the boulders on Bennu asteroid has been analyzed and presented. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>The architecture of the YOLOv5 method. The network consists of three main parts: backbone, neck, and output. Backbone part focuses on extracting feature information from input images, neck part fuses the extracted feature information and generates three scales of feature maps, and the output part detects the objects from these generated feature maps.</p> Full article ">Figure 2
<p>The architecture of the improved YOLOv5 method. Compared to original YOLOv5 method, there are three improvements in the architecture. First, a new fusion layer is added which generates a large scale of feature map with the size of 152 × 152 × 255. Second, new connections represented by red lines have been added to bring feature information from backbone into feature fusion layers. Third, the attention modules are added into feature fusion layers.</p> Full article ">Figure 3
<p>The structure of the attention modules. The attention modules are implemented by combining the ECA and the CBAM attention mechanisms. The above part is the ECA attention module which implements the channel attention, and the blow part is the CBAM spatial attention module.</p> Full article ">Figure 4
<p>The framework of the pyramid based approach.</p> Full article ">Figure 5
<p>The number and proportion of the boulders of different scales.</p> Full article ">Figure 6
<p>The number and proportion of boulders in each region.</p> Full article ">Figure 7
<p>The distribution of the detected boulders whose diameters are less than 5 m in each region.</p> Full article ">Figure 8
<p>The distribution of the detected boulders whose diameters are between 5 m and 10 m in each region.</p> Full article ">Figure 9
<p>The distribution of the detected boulders whose diameters are between 10 m and 30 m in each region.</p> Full article ">Figure 10
<p>The distribution of the detected boulders whose diameters are more than 30 m in each region.</p> Full article ">Figure 10 Cont.
<p>The distribution of the detected boulders whose diameters are more than 30 m in each region.</p> Full article ">
32 pages, 19658 KiB  
Article
Quantitative Assessment of Riverbed Planform Adjustments, Channelization, and Associated Land Use/Land Cover Changes: The Ingauna Alluvial-Coastal Plain Case (Liguria, Italy)
by Andrea Mandarino, Giacomo Pepe, Andrea Cevasco and Pierluigi Brandolini
Remote Sens. 2021, 13(18), 3775; https://doi.org/10.3390/rs13183775 - 20 Sep 2021
Cited by 13 | Viewed by 3125
Abstract
The active-channel planform adjustments that have occurred along the Centa, lower Arroscia and lower Neva rivers since 1930, along with the riverbed channelization processes and the land-use and land-cover changes in disconnected riverine areas, were investigated through a multitemporal analysis based on remote [...] Read more.
The active-channel planform adjustments that have occurred along the Centa, lower Arroscia and lower Neva rivers since 1930, along with the riverbed channelization processes and the land-use and land-cover changes in disconnected riverine areas, were investigated through a multitemporal analysis based on remote sensing and geographical information systems (GIS). These watercourses flow through the largest Ligurian alluvial-coastal plain in a completely anthropogenic landscape. This research is based on the integrated use of consolidated and innovative metrics for riverbed planform analysis. Specific indices were introduced to assess active-channel lateral migration in relation to the active-channel area abandonment and formation processes. The Arroscia and Neva riverbeds experienced narrowing, progressive stabilization, and braiding phenomena disappearance from 1930 to the early 1970s, and then slight narrowing up to the late 1980s. Subsequently, generalized stability was observed. Conversely, the Centa was not affected by relevant planform changes. Recently, all rivers underwent a slight to very slight width increase triggered by the November 2016 high-magnitude flood. The active-channel adjustments outlined in this paper reflect the relevant role in conditioning the river morphology and dynamics played by channelization works built from the 1920s to the early 1970s. They (i) narrowed, straightened, and stabilized the riverbed and (ii) reduced the floodable surface over the valley-floor. Thus, large disconnected riverine areas were occupied by human activities and infrastructures, resulting in a progressive increase in vulnerable elements exposed to hydrogeomorphic hazards. The outlined morphological dynamics (i) display significant differences in terms of chronology, type, and magnitude of active-channel planform adjustments with respect to the medium- and short-term morphological evolution of most Italian rivers and (ii) reflect the widespread urbanization of Ligurian major valley floors that occurred over the 20th century. The outcomes from this study represent an essential knowledge base from a river management perspective; the novel metrics enlarge the spectrum of available GIS tools for active-channel planform analysis. Full article
(This article belongs to the Special Issue Geomorphological Mapping and Process Monitoring Using Remote Sensing)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Location of the study reach. (<b>a</b>) The Centa River catchment. The Liguria Region is depicted in red; the red arrow indicates the Centa catchment location. (<b>b</b>) The investigated reaches of the Arroscia, Neva and Centa rivers.</p> Full article ">Figure 2
<p>Schematization of the methodological framework. Hexagon (rectangle) indicates input/output data (process). Blue rectangle indicates that field surveys and, for historical features, archival data supported and validated the process.</p> Full article ">Figure 3
<p>Active channel and macrochannel definition: (<b>a</b>) Google Earth image dated back to 2020 showing the Centa River upstream of Albenga and (<b>b</b>) the corresponding sketch of bank protections, active channel, and macrochannel. The blue arrow indicates the flow direction.</p> Full article ">Figure 4
<p>(<b>a</b>) Overlapping of a couple of active channels referring to two different times and (<b>b</b>) definition of stable (SACA), abandoned (AACA) and newly formed (NACA) active channel area.</p> Full article ">Figure 5
<p>(<b>a</b>) Active channel width (CW) and channel width variation index (∆W) referring to the active channel width in (<b>b</b>) 1930 and (<b>c</b>) 1954 at the reach scale.</p> Full article ">Figure 6
<p>(<b>a</b>) Active-channel centerline length (CL), (<b>b</b>) sinuosity index (SI), (<b>c</b>) braiding index (BI), and (<b>d</b>) channel pattern at the reach scale over the period 1930–2020. NC: no change; C: change; W: wandering; B: braided; SAB: sinuous with alternate bars; ST: straight.</p> Full article ">Figure 7
<p>The channel abandonment index (CAI) and channel formation index (CFI) referred to each investigated period and to the periods 1930–2020 and 1954–2020. Letters on bars indicate the qualitative class for metrics: VL: very low (≤15%); L: low (&gt;15% and ≤35%); M: moderate (&gt;35% and ≤50%); H: high (&gt;50% and ≤75%); VH: very high (&gt;75%).</p> Full article ">Figure 8
<p>Abandoned (AACA) and newly formed (NACA) active channel areas included within the macrochannel referred to each period investigated and to the periods 1930–2020 and 1954–2020. The macrochannel was mapped in reaches A2, A3, N2, C1, and C2. A1 and N1 were excluded due to the lack of continuous or almost continuous longitudinal bank protections or numerous groynes that allow the macrochannel to be mapped.</p> Full article ">Figure 9
<p>(<b>a</b>) Longitudinal bank protection extent and (<b>b</b>) groyne density over time at the reach scale. The lack of data referring to the 1954 longitudinal bank protection extent in reaches A1 and N1 is associated with the impossibility of mapping them with a certain degree of accuracy.</p> Full article ">Figure 10
<p>Current LULC in the abandoned areas of the (<b>a</b>) 1930, (<b>b</b>) 1954 and (<b>c</b>) 1973 active-channels. S: sea; N: natural and seminatural area; A: agricultural area; M: abandoned man-made area; Q: quarry area; T: transport area; I: industrial area; U: urban area.</p> Full article ">Figure 11
<p>Active-channel planform migration in (<b>a</b>) stable and (<b>b</b>) changing active-channel width conditions, referring to the period t1-t2. Black arrows indicate the active-channel centerline horizontal displacement.</p> Full article ">Figure 12
<p>Time chart of reach-scale active channel morphological adjustments and related magnitudes. The terms “contiguous” and “noncontiguous” pattern refer to [<a href="#B52-remotesensing-13-03775" class="html-bibr">52</a>,<a href="#B148-remotesensing-13-03775" class="html-bibr">148</a>]. The active-channel planform migration magnitude was defined according to the combined analysis of CAI and CFI (see <a href="#remotesensing-13-03775-t002" class="html-table">Table 2</a>).</p> Full article ">Figure 13
<p>Channelization structures depicted on the 1930 map along the (<b>a</b>) Arroscia (reach A3) and (<b>b</b>) Neva (reaches N1 and N2) rivers. The present-day channelization structures along reaches (<b>c</b>) A1, (<b>d</b>) A2, (<b>e</b>) A3, (<b>f</b>) A3, (<b>g</b>) N1, and (<b>h</b>) C2. The yellow arrows marked with 1 and 2 indicate groynes and longitudinal bank protections, respectively. In (<b>f</b>), the dotted line represents a schematic topographic section of the most common channelization structure in reaches A2, A3, N2, and C1. 1: ground level of cultivated fields; 2: embankment for flood protection; 3 gabions for erosion protection; 4: active channel. Photographs: Andrea Mandarino.</p> Full article ">Figure 14
<p>LULC evolution in reach A3: (<b>a</b>) 1954, (<b>b</b>) 1973, (<b>c</b>) 1988, (<b>d</b>) 2020. The flow is from left to right. The progressive occupation by human activities and structures of the left side of the valley floor is evident. The white arrow in (<b>c</b>) indicates an approximatively 500 m long reach where active-channel forms were obliterated by bulldozers. The white arrow in (<b>d</b>) pinpoints the after-2016 bank retreat process that involved some areas out of the macrochannel. The white dotted line in (<b>d</b>) represents the active channel in 1930.</p> Full article ">
21 pages, 67535 KiB  
Article
IFRAD: A Fast Feature Descriptor for Remote Sensing Images
by Qinping Feng, Shuping Tao, Chunyu Liu, Hongsong Qu and Wei Xu
Remote Sens. 2021, 13(18), 3774; https://doi.org/10.3390/rs13183774 - 20 Sep 2021
Cited by 3 | Viewed by 2200
Abstract
Feature description is a necessary process for implementing feature-based remote sensing applications. Due to the limited resources in satellite platforms and the considerable amount of image data, feature description—which is a process before feature matching—has to be fast and reliable. Currently, the state-of-the-art [...] Read more.
Feature description is a necessary process for implementing feature-based remote sensing applications. Due to the limited resources in satellite platforms and the considerable amount of image data, feature description—which is a process before feature matching—has to be fast and reliable. Currently, the state-of-the-art feature description methods are time-consuming as they need to quantitatively describe the detected features according to the surrounding gradients or pixels. Here, we propose a novel feature descriptor called Inter-Feature Relative Azimuth and Distance (IFRAD), which will describe a feature according to its relation to other features in an image. The IFRAD will be utilized after detecting some FAST-alike features: it first selects some stable features according to criteria, then calculates their relationships, such as their relative distances and azimuths, followed by describing the relationships according to some regulations, making them distinguishable while keeping affine-invariance to some extent. Finally, a special feature-similarity evaluator is designed to match features in two images. Compared with other state-of-the-art algorithms, the proposed method has significant improvements in computational efficiency at the expense of reasonable reductions in scale invariance. Full article
(This article belongs to the Section Remote Sensing Image Processing)
►▼ Show Figures

Figure 1

Figure 1
<p>Flowchart of IFRAD-based Registration: In this paper, we use the classical FAST detector [<a href="#B10-remotesensing-13-03774" class="html-bibr">10</a>] in the Feature Detection module.</p> Full article ">Figure 2
<p>Reference and Sensed Images: (<b>a</b>) Reference image <math display="inline"><semantics> <mrow> <mi mathvariant="bold">R</mi> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </semantics></math>; (<b>b</b>) Sensed image <math display="inline"><semantics> <mrow> <mi mathvariant="bold">S</mi> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </semantics></math>; (<b>c</b>) Parallax difference; In (<b>c</b>), red channel shows <math display="inline"><semantics> <mrow> <mi mathvariant="bold">R</mi> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </semantics></math>, while <math display="inline"><semantics> <mrow> <mi mathvariant="bold">S</mi> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </semantics></math> is shown in cyan channel, the parallax is caused by about 20° yaw, and about 30° pitch.</p> Full article ">Figure 3
<p>Illustration of Secondary Features: (<b>a</b>) All detected FAST-features in the reference image <math display="inline"><semantics> <mrow> <mi mathvariant="bold">R</mi> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </semantics></math> shown in <a href="#remotesensing-13-03774-f002" class="html-fig">Figure 2</a>a; (<b>b</b>) Secondary features marked with * in yellow; Both of these images are Gaussian smoothed before the feature detection process; In (<b>a</b>), all the features are marked with blobs in different colors; the size of each blob represents the modulated response strength of the corresponding feature.</p> Full article ">Figure 4
<p>The difference of the detected secondary features in the similar pattern but in different parallax: (<b>a</b>,<b>b</b>) are local areas from <a href="#remotesensing-13-03774-f002" class="html-fig">Figure 2</a>a,b, respectively. All the blobs are the secondary features; the size and the color of the blobs both vary by the magnitude of responses. Note that some features are shown as secondary features in one image, but do not appear in the other.</p> Full article ">Figure 5
<p>Determination of Primary Features: Assuming that there exist 14 secondary features (presented as bold points A–N) in an image, the number represents the response intensity of each feature. The circles (only four are shown for clarity) represent the domains of the corresponding features. With the determined domain radius <span class="html-italic">R</span>, under criterion (<a href="#FD8-remotesensing-13-03774" class="html-disp-formula">8</a>), features A, B, C, F, J, K, M, N are determined as primary features. Under criterion (<a href="#FD21-remotesensing-13-03774" class="html-disp-formula">21</a>), features D and G are also primary features.</p> Full article ">Figure 6
<p>Result of determining primary features from secondary features: in this figure, only secondary features are labeled with a star in red or yellow, and the primary features are marked with a red star.</p> Full article ">Figure 7
<p>Calculation of Azimuth, Distance and Relation-strength: (<b>a</b>) Illustration of RADs calculation, only four are presented; (<b>b</b>) List of these RADs, they are sorted in ascending order of azimuth.</p> Full article ">Figure 8
<p>Demonstration of Feature Relations: (<b>a</b>) shows one of the primary features to all other secondary features, the reference and dominant orientations are also shown; (<b>b</b>) A bar graph of Azimuth vs. Relation strength; (<b>c</b>) A bar graph of remapped azimuth vs. relation strength, with the dominant orientation set as 0.</p> Full article ">Figure 9
<p>Illustration of OIH: (<b>a</b>) An image is divided into 10 fan-shaped regions with a start of dominant orientation; (<b>b</b>) A 10-bin-OIH calculated from (<b>a</b>) according to Formula (<a href="#FD15-remotesensing-13-03774" class="html-disp-formula">15</a>).</p> Full article ">Figure 10
<p>An Example of Unstable Feature Description: With another primary-feature-pair as an example, (<b>a</b>,<b>d</b>) is the bar graphs of azimuth-vs-relation strength of the same primary feature in reference and sensed images, respectively, and the dominant orientation is determined according to Equation (<a href="#FD13-remotesensing-13-03774" class="html-disp-formula">13</a>); (<b>b</b>,<b>e</b>) are the remapped bar graphs according to (<b>a</b>,<b>d</b>); (<b>c</b>,<b>f</b>) are 30-bin-OIHs obtained from (<b>b</b>,<b>e</b>), according to (<a href="#FD18-remotesensing-13-03774" class="html-disp-formula">18</a>), the cosine distance between them is 0.3293.</p> Full article ">Figure 11
<p>An Example of Solving the Unstable Issue: (<b>a</b>,<b>d</b>) are the same bar graphs shown in <a href="#remotesensing-13-03774-f010" class="html-fig">Figure 10</a>a,d, but different in dominant orientations, and are determined according to Equation (<a href="#FD19-remotesensing-13-03774" class="html-disp-formula">19</a>). In this example, <math display="inline"><semantics> <mrow> <mi>α</mi> <mo>=</mo> <mn>0.7</mn> </mrow> </semantics></math>; (<b>b</b>,<b>e</b>) are the remapped bar graphs according to (<b>a</b>,<b>d</b>); (<b>c</b>,<b>f</b>) are 30-bin-OIHs obtained from (<b>b</b>,<b>e</b>), according to (<a href="#FD18-remotesensing-13-03774" class="html-disp-formula">18</a>) the cosine distance between them is 0.0254.</p> Full article ">Figure 12
<p>Image Registration Result: With parameters of <math display="inline"><semantics> <mrow> <mi>α</mi> <mo>=</mo> <mn>0.7</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>n</mi> <mo>=</mo> <mn>30</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>t</mi> <mo>=</mo> <mn>0.8</mn> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <mi>R</mi> <mo>=</mo> <mn>1</mn> <mo>/</mo> <mn>100</mn> <mo>·</mo> <mo movablelimits="true" form="prefix">min</mo> <mo>(</mo> <mi>M</mi> <mo>,</mo> <mi>N</mi> <mo>)</mo> </mrow> </semantics></math>. (<b>a</b>) shows correctly matched feature-pairs; (<b>b</b>) shows registered image; (<b>c</b>) shows some magnified views of local images.</p> Full article ">Figure 13
<p>Five Groups of Experiments on Remote Sensing Images: For each group, the ones on the top are reference images, and the ones on the bottom are the corresponding sensed images; (<b>a</b>–<b>c</b>) The two images in each group are captured by the same sensor but in different views; (<b>d</b>,<b>e</b>) Color images captured by the same sensor but in different views; Image Dimensions: (<b>a</b>,<b>b</b>) 3042-by-2048; (<b>c</b>) 3072-by-2304; (<b>d</b>) 3644-by-3644; (<b>e</b>) 3366-by-1936. Image Sources: (<b>a</b>,<b>b</b>) captured from our laboratory image boards; (<b>c</b>) captured by the Zhuhai-1 satellite; (<b>d</b>,<b>e</b>) obtained from Google Earth.</p> Full article ">Figure 14
<p>Registration Results Under Different ET, the intensity of images are rescaled to 0–1 for visibility.</p> Full article ">Figure 15
<p>Changes of OIH of a Feature Caused By Translational Shift: First column: original images with different scales (they are cropped from the image shown in <a href="#remotesensing-13-03774-f013" class="html-fig">Figure 13</a>b by altering the translational shift. Compared to the first row, the percentages of overlap area of the second and the third row are: 64.75% and 37.13% respectively). Second column: The relation graph of the same primary feature to the other secondary features. Third column: bar graph of remapped azimuth-vs-relation strength according to the second column. Last column: 50-bin-OIHs according to the third column; the distance of the second and the last OIH to the first OIH are 0.2099 and 0.1653, respectively.</p> Full article ">Figure 16
<p>Demonstration of the Limitation Caused by Scale Changes: First column: original images with different scales (they are cropped from the image shown in <a href="#remotesensing-13-03774-f013" class="html-fig">Figure 13</a>d by altering the scale; from top to bottom are: 0.8×, 1.1×, 1.4×, respectively). Second column: The relation graph of the same primary feature to other secondary features—note that the number of secondary features may be increased or reduced by changing the scale. Third column: bar graph of remapped azimuth-vs.-relation strength according to the second column; it can be seen that the relation strengths are also affected by altering scales. Last column: 50-bin-OIHs according to the third column.</p> Full article ">
24 pages, 105003 KiB  
Article
Combined GRACE and MT-InSAR to Assess the Relationship between Groundwater Storage Change and Land Subsidence in the Beijing-Tianjin-Hebei Region
by Wen Yu, Huili Gong, Beibei Chen, Chaofan Zhou and Qingquan Zhang
Remote Sens. 2021, 13(18), 3773; https://doi.org/10.3390/rs13183773 - 20 Sep 2021
Cited by 13 | Viewed by 3704
Abstract
Beijing-Tianjin-Hebei (BTH) has been suffering from severe groundwater storage (GWS) consumption and land subsidence (LS) for a long period. The overexploitation of groundwater brings about severe land subsidence, which affects the safety and development of BTH. In this paper, we utilized multi-frame synthetic [...] Read more.
Beijing-Tianjin-Hebei (BTH) has been suffering from severe groundwater storage (GWS) consumption and land subsidence (LS) for a long period. The overexploitation of groundwater brings about severe land subsidence, which affects the safety and development of BTH. In this paper, we utilized multi-frame synthetic aperture radar datasets obtained by the Rardarsat-2 satellite to monitor land subsidence’s temporal and spatial distribution in the BTH from 2012 to 2016 based on multi-temporal interferometric synthetic aperture radar (MT-InSAR). In addition, we also employed the Gravity Recovery and Climate Experiment (GRACE) mascon datasets acquired by the Center for Space Research (CSR) and Jet Propulsion Laboratory (JPL) to obtain the GWS anomalies (GWSA) of BTH from 2003 to 2016. Then we evaluate the accuracy of the results obtained. Furthermore, we explored the relationship between the regional GWSA and the average cumulative subsidence in the BTH. The total volume change of subsidence is 59.46% of the total volume change of groundwater storage. Moreover, the long-term decreasing trend of the GWSA (14.221 mm/year) and average cumulative subsidence (17.382 mm/year) show a relatively high consistency. Finally, we analyze the heterogeneity of GWS change (GWSC) and LS change (LSC) in the four typical areas by the Lorenz curve model. The implementation of the South-to-North Water Diversion Project (MSWDP) affects the heterogeneity of GWSC and LSC. It can be seen that the largest heterogeneity of LSC lags behind the GWSC in the Tianjin-Langfang-Hengshui-Baoding area. The largest uneven subsidence in Beijing and Tianjin occurred in 2015, and the largest uneven subsidence in Hengshui-Baoding occurred in 2014. After that, the heterogeneity of subsidence gradually tends to stable. Full article
(This article belongs to the Special Issue Application of Remote Sensing in Hydrogeology: Landslides, Land Subsidence and Uplift)
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) The location of the NCP in China. (<b>b</b>) The location of BTH in NCP. (<b>c</b>) The location of BTH area.</p> Full article ">Figure 2
<p>Schematic diagram of Lorenz curve.</p> Full article ">Figure 3
<p>The flowchart of this study (5.1, 5.2, and 5.3 refer to the chapters where the corresponding content appears in this paper).</p> Full article ">Figure 4
<p>The average land subsidence rate from 2012 to 2016.</p> Full article ">Figure 5
<p>The accumulated subsidence rate from 2012 to 2016.</p> Full article ">Figure 6
<p>The time-series subsidence changes and area of subsidence exceeding 50 mm from 2012 to 2016 in BTH.</p> Full article ">Figure 7
<p>Comparison between subsidence derived by InSAR and leveling results from 2015 to 2016, whose locations are shown in <a href="#remotesensing-13-03773-f001" class="html-fig">Figure 1</a>.</p> Full article ">Figure 8
<p>Time series of total water storage anomalies (TWSA) estimated from the CSR and JPL.</p> Full article ">Figure 9
<p>Time series of the SMSA and SWESA from GLDAS.</p> Full article ">Figure 10
<p>Time series of the groundwater storage anomaly (GWSA) from GRACE and GLDAS.</p> Full article ">Figure 11
<p>Comparison between GWSA derived by GRACE and in situ groundwater level results from 2003 to 2016.</p> Full article ">Figure 12
<p>Spatial trend map of GWSA and land subsidence in BTH from 2012 to 2016. (<b>A</b>) The panel A is the spatial trend map of GWSA; (<b>B</b>) The panel B is the spatial trend map of land subsidence.</p> Full article ">Figure 13
<p>The relationship between groundwater storage consumption and subsidence volume in the cities of BTH.</p> Full article ">Figure 14
<p>Long-term time series trend of abnormal of groundwater storage and accumulated land subsidence.</p> Full article ">Figure 15
<p>Typical Areas in BTH (<b>a</b>) Chaoyang District in Beijing; (<b>b</b>) Wuqing City in Tianjin and Bazhou City in Langfang; (<b>c</b>) Jing County in Hengshui and Dongguang County in Cangzhou; (<b>d</b>) Gaoyang County in Baoding.</p> Full article ">Figure 16
<p>Chaoyang District in Beijing.</p> Full article ">Figure 17
<p>Lorenz curve of Chaoyang District in Beijing.</p> Full article ">Figure 18
<p>Wuqing City in Tianjin and Bazhou City in Langfang.</p> Full article ">Figure 19
<p>Lorenz curve of Wuqing City in Tianjin and Bazhou City in Langfang.</p> Full article ">Figure 20
<p>Jing County in Hengshui and Dongguang County in Cangzhou.</p> Full article ">Figure 21
<p>Lorenz curve of Jing County in Hengshui and Dongguang County in Cangzhou.</p> Full article ">Figure 22
<p>Gaoyang County in Baoding.</p> Full article ">Figure 23
<p>Lorenz curve of Gaoyang County in Baoding.</p> Full article ">
17 pages, 831 KiB  
Article
Tensor-Based Reduced-Dimension MUSIC Method for Parameter Estimation in Monostatic FDA-MIMO Radar
by Tengxian Xu, Xianpeng Wang, Mengxing Huang, Xiang Lan and Lu Sun
Remote Sens. 2021, 13(18), 3772; https://doi.org/10.3390/rs13183772 - 20 Sep 2021
Cited by 16 | Viewed by 2483
Abstract
Frequency diverse array (FDA) radar has attracted much attention due to the angle and range dependence of the beam pattern. Multiple-input-multiple-output (MIMO) radar has high degrees of freedom (DOF) and spatial resolution. The FDA-MIMO radar, a hybrid of FDA and MIMO radar, can [...] Read more.
Frequency diverse array (FDA) radar has attracted much attention due to the angle and range dependence of the beam pattern. Multiple-input-multiple-output (MIMO) radar has high degrees of freedom (DOF) and spatial resolution. The FDA-MIMO radar, a hybrid of FDA and MIMO radar, can be used for target parameter estimation. This paper investigates a tensor-based reduced-dimension multiple signal classification (MUSIC) method, which is used for target parameter estimation in the FDA-MIMO radar. The existing subspace methods deteriorate quickly in performance with small samples and a low signal-to-noise ratio (SNR). To deal with the deterioration difficulty, the sparse estimation method is then proposed. However, the sparse algorithm has high computation complexity and poor stability, making it difficult to apply in practice. Therefore, we use tensor to capture the multi-dimensional structure of the received signal, which can optimize the effectiveness and stability of parameter estimation, reduce computation complexity and overcome performance degradation in small samples or low SNR simultaneously. In our work, we first obtain the tensor-based subspace by the high-order-singular value decomposition (HOSVD) and establish a two-dimensional spectrum function. Then the Lagrange multiplier method is applied to realize a one-dimensional spectrum function, estimate the direction of arrival (DOA) and reduce computation complexity. The transmitting steering vector is obtained by the partial derivative of the Lagrange function, and automatic pairing of target parameters is then realized. Finally, the range can be obtained by using the least square method to process the phase of transmitting steering vector. Method analysis and simulation results prove the superiority and reliability of the proposed method. Full article
(This article belongs to the Special Issue Advances in Joint Radar-Communication Systems, Multi-Carrier Radars, Passive Radar Networks, and Waveform Design)
►▼ Show Figures

Figure 1

Figure 1
<p>Schematic diagram of monostatic FDA-MIMO radar.</p> Full article ">Figure 2
<p>The spatial spectrum of angle dimension.</p> Full article ">Figure 3
<p>Partial view of the spatial spectrum of angle dimension.</p> Full article ">Figure 4
<p>2D point cloud of estimated targets.</p> Full article ">Figure 5
<p>RMSE of DOA estimation versus SNR.</p> Full article ">Figure 6
<p>RMSE of range estimation versus SNR.</p> Full article ">Figure 7
<p>RMSE of DOA estimation versus <span class="html-italic">L</span>.</p> Full article ">Figure 8
<p>RMSE of range estimation versus <span class="html-italic">L</span>.</p> Full article ">Figure 9
<p>PSD of DOA estimation versus SNR.</p> Full article ">Figure 10
<p>PSD of range estimation versus SNR.</p> Full article ">Figure 11
<p>The simulation running time comparison.</p> Full article ">
18 pages, 56883 KiB  
Article
Multi-Modality and Multi-Scale Attention Fusion Network for Land Cover Classification from VHR Remote Sensing Images
by Tao Lei, Linze Li, Zhiyong Lv, Mingzhe Zhu, Xiaogang Du and Asoke K. Nandi
Remote Sens. 2021, 13(18), 3771; https://doi.org/10.3390/rs13183771 - 20 Sep 2021
Cited by 21 | Viewed by 3257
Abstract
Land cover classification from very high-resolution (VHR) remote sensing images is a challenging task due to the complexity of geography scenes and the varying shape and size of ground targets. It is difficult to utilize the spectral data directly, or to use traditional [...] Read more.
Land cover classification from very high-resolution (VHR) remote sensing images is a challenging task due to the complexity of geography scenes and the varying shape and size of ground targets. It is difficult to utilize the spectral data directly, or to use traditional multi-scale feature extraction methods, to improve VHR remote sensing image classification results. To address the problem, we proposed a multi-modality and multi-scale attention fusion network for land cover classification from VHR remote sensing images. First, based on the encoding-decoding network, we designed a multi-modality fusion module that can simultaneously fuse more useful features and avoid redundant features. This addresses the problem of low classification accuracy for some objects caused by the weak ability of feature representation from single modality data. Second, a novel multi-scale spatial context enhancement module was introduced to improve feature fusion, which solves the problem of a large-scale variation of objects in remote sensing images, and captures long-range spatial relationships between objects. The proposed network and comparative networks were evaluated on two public datasets—the Vaihingen and the Potsdam datasets. It was observed that the proposed network achieves better classification results, with a mean F1-score of 88.6% for the Vaihingen dataset and 92.3% for the Potsdam dataset. Experimental results show that our model is superior to the state-of-the-art network models. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>The top row is the spectral image of VHR remote sensing images, and the bottom row is the corresponding DSM image. The first two columns are buildings and roads in shadows. The latter two columns are trees and low vegetation with extremely similar spectral characteristics. The digital surface model avoids the interference of shadow, occlusion, and other factors.</p> Full article ">Figure 2
<p>The overall architecture of MMAFNet.</p> Full article ">Figure 3
<p>Multi-modality fusion module (MFM). (<b>a</b>) MFM-0. (<b>b</b>) MFM-<span class="html-italic">n</span> (<span class="html-italic">n</span> ∈ [1, 3]).</p> Full article ">Figure 4
<p>Multi-scale spatial context enhancement module.</p> Full article ">Figure 5
<p>The architecture of the residual skip connection.</p> Full article ">Figure 6
<p>Sample images of Potsdam and Vaihingen datasets, digital surface models, and their corresponding labels. (<b>a</b>) Potsdam TOP, (<b>b</b>) Potsdam DSM, and (<b>c</b>) Potsdam GT, (<b>d</b>) Vaihingen TOP, (<b>e</b>) Vaihingen DSM, and (<b>f</b>) Vaihingen GT.</p> Full article ">Figure 7
<p>Comparison of experimental results for five images in the Potsdam dataset.</p> Full article ">Figure 8
<p>The results of MMAFNet and other comparative methods on the Potsdam dataset.</p> Full article ">Figure 9
<p>Experimental results of MMAFNet and other comparative methods for five images in the Vaihingen dataset.</p> Full article ">Figure 10
<p>The results of MMAFNet on the Vaihingen dataset are shown and compared.</p> Full article ">Figure 11
<p>Comparison of ablation experiment results of different modules. The first row is a visual comparison between the baseline and the result of adding the multi-modality fusion module. The second row shows that the results verify the benefits of the multi-scale spatial context module. The third row shows the visualization results of the baseline using the residual skip connection strategy.</p> Full article ">
27 pages, 13305 KiB  
Article
Using Convolutional Neural Networks for Detection and Morphometric Analysis of Carolina Bays from Publicly Available Digital Elevation Models
by Mark A. Lundine and Arthur C. Trembanis
Remote Sens. 2021, 13(18), 3770; https://doi.org/10.3390/rs13183770 - 20 Sep 2021
Cited by 2 | Viewed by 3791
Abstract
Carolina Bays are oriented and sandy-rimmed depressions that are ubiquitous throughout the Atlantic Coastal Plain (ACP). Their origin has been a highly debated topic since the 1800s and remains unsolved. Past population estimates of Carolina Bays have varied vastly, ranging between as few [...] Read more.
Carolina Bays are oriented and sandy-rimmed depressions that are ubiquitous throughout the Atlantic Coastal Plain (ACP). Their origin has been a highly debated topic since the 1800s and remains unsolved. Past population estimates of Carolina Bays have varied vastly, ranging between as few as 10,000 to as many as 500,000. With such a large uncertainty around the actual population size, mapping these enigmatic features is a problem that requires an automated detection scheme. Using publicly available LiDAR-derived digital elevation models (DEMs) of the ACP as training images, various types of convolutional neural networks (CNNs) were trained to detect Carolina bays. The detection results were assessed for accuracy and scalability, as well as analyzed for various morphologic, land-use and land cover, and hydrologic characteristics. Overall, the detector found over 23,000 Carolina Bays from southern New Jersey to northern Florida, with highest densities along interfluves. Carolina Bays in Delmarva were found to be smaller and shallower than Bays in the southeastern ACP. At least a third of Carolina Bays have been converted to agricultural lands and almost half of all Carolina Bays are forested. Few Carolina Bays are classified as open water basins, yet almost all of the detected Bays were within 2 km of a water body. In addition, field investigations based upon detection results were performed to describe the sedimentology of Carolina Bays. Sedimentological investigations showed that Bays typically have 1.5 m to 2.5 m thick sand rims that show a gradient in texture, with coarser sand at the bottom and finer sand and silt towards the top. Their basins were found to be 0.5 m to 2 m thick and showed a mix of clayey, silty, and sandy deposits. Last, the results compiled during this study were compared to similar depressional features (i.e., playa-lunette systems) to pinpoint any similarities in origin processes. Altogether, this study shows that CNNs are valuable tools for automated geomorphic feature detection and can lead to new insights when coupled with various forms of remotely sensed and field-based datasets. Full article
(This article belongs to the Special Issue How the Combination of Satellite Remote Sensing with Artificial Intelligence Can Solve Coastal Issues)
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) Jones Lake, a Carolina Bay in North Carolina. (<b>b</b>) A flooded Carolina Bay in Delaware. (<b>c</b>) A Carolina Bay in Delaware with bald cypress trees. (<b>d</b>) LiDAR DEM of Carolina Bays on the eastern shore of Virginia. (<b>e</b>) LiDAR DEM of Carolina Bays in North Carolina.</p> Full article ">Figure 2
<p>Carolina Bays in central Delaware. (<b>a</b>) LiDAR DEM gridded at 10 m. (<b>b</b>) Aerial imagery from ESRI World Imagery basemap.</p> Full article ">Figure 3
<p>Traditional computer vision algorithms tested for Carolina Bay detection. (<b>a</b>) Input DEM. (<b>b</b>) Local minima detector (each white point is a local minima). (<b>c</b>) Laplacian of Gaussians blob detector (detected blobs and their radii are shown in yellow). (<b>d</b>) Scale invariant feature transform (SIFT; keypoints and their radii are shown in blue).</p> Full article ">Figure 4
<p>(<b>a</b>) Input DEM. (<b>b</b>) Mask annotation. (<b>c</b>) K-means unsupervised two-class classifier. (<b>d</b>) GaussianNB supervised classifier. (<b>e</b>) Decision Tree supervised classifier. (<b>f</b>) Random Forest supervised classifier. (<b>g</b>) Quadratic Discriminant supervised classifier. (<b>h</b>) MLP supervised classifier. (<b>i</b>) AdaBoost supervised classifier.</p> Full article ">Figure 5
<p>(<b>a</b>) National Map Elevation 10-m DEM availability by production method. (<b>b</b>) Boundary for mosaic of 10-m elevation data used for this study.</p> Full article ">Figure 6
<p>Sediment sample locations across Delmarva.</p> Full article ">Figure 7
<p>The general workflow for getting from the ACP DEM to Carolina Bay detections with information on spatial distribution, morphology, land-use and land-cover, and surface hydrology.</p> Full article ">Figure 8
<p>Precision and recall curves for Faster R-CNN and Mask R-CNN Carolina Bay detectors.</p> Full article ">Figure 9
<p>Kernel density estimation for (<b>a</b>) area (<b>b</b>) perimeter, and (<b>c</b>) maximum relief for the Delaware training and test annotations, Delaware Faster R-CNN detection at 60%, Delaware Mask R-CNN detections at 30%, and existing data from DGS.</p> Full article ">Figure 10
<p>Comparing annotated bounding boxes with detection bounding boxes. (<b>a</b>) Bounding box area. (<b>b</b>) Bounding box perimeter. (<b>c</b>) Bounding box easting centroid. (<b>d</b>) Bounding box northing centroid.</p> Full article ">Figure 11
<p>Delaware’s Carolina Bay detection counts at 60% threshold for various DEM tile sizes and overlap amounts compared to the annotation count.</p> Full article ">Figure 12
<p>(<b>a</b>) Carolina Bay detections across the ACP. (<b>b</b>) Heat map of Carolina Bay detections showing higher vs lower density areas.</p> Full article ">Figure 13
<p>Comparing detection results with manually annotated polygons. Each plot shows an OLS fit with the x-axis quantity as the independent variable and the y-axis quantity as the dependent variable. One-to-one fits are plotted on the area and perimeter plots. (<b>a</b>) Area. (<b>b</b>) Perimeter. (<b>c</b>) Centroid longitude. (<b>d</b>) Centroid latitude.</p> Full article ">Figure 14
<p>(<b>a</b>) Distribution of maximum relief from Carolina Bay detections across the ACP. (<b>b</b>) Distribution of area from Carolina Bay detections across the ACP.</p> Full article ">Figure 15
<p>Polar plot of kernel density estimation of Carolina Bay detection major axis orientations.</p> Full article ">Figure 16
<p>General land-use and land-cover type of Carolina Bay detections by fraction of total area and fraction of total count.</p> Full article ">Figure 17
<p>KDEs for fraction of Bay covered by lake/pond water and swamp/marsh water.</p> Full article ">Figure 18
<p>Closest bodies of water to Carolina Bay detections by fraction of total detection count.</p> Full article ">Figure 19
<p>EDCF for distance to water body for Carolina Bay detections that did not intersect any NHD water bodies.</p> Full article ">Figure 20
<p>Grain size cumulative mass fractions with standard errors by sample type. Top: Grain size cumulative mass fractions for samples with more than 10% silt/clay. Bottom: Grain size cumulative mass fractions for samples with less than 10% silt/clay.</p> Full article ">Figure 21
<p>Profile of hand auger samples with descriptions taken at a Carolina Bay in Delaware. (<b>a</b>) Broader geographic location. (<b>b</b>) Location of samples plotted on a DEM. (<b>c</b>) Cross-section with sediment descriptions.</p> Full article ">Figure 22
<p>Principal component analysis of various topographic metrics within Carolina Bay detections.</p> Full article ">Figure 23
<p>(<b>a</b>) Carolina Bay detections at various image footprint scales. (<b>b</b>) Aggregation of overlapping detections. (<b>c</b>) PAEK-smoothed polygons.</p> Full article ">Figure 23 Cont.
<p>(<b>a</b>) Carolina Bay detections at various image footprint scales. (<b>b</b>) Aggregation of overlapping detections. (<b>c</b>) PAEK-smoothed polygons.</p> Full article ">Figure 24
<p>Smooth, aggregated, multi-scale detections in (<b>a</b>) Central Delaware, (<b>b</b>) Southern Delaware, (<b>c</b>) Virginia, (<b>d</b>) North Carolina, (<b>e</b>) South Carolina, and (<b>f</b>) Georgia.</p> Full article ">Figure 24 Cont.
<p>Smooth, aggregated, multi-scale detections in (<b>a</b>) Central Delaware, (<b>b</b>) Southern Delaware, (<b>c</b>) Virginia, (<b>d</b>) North Carolina, (<b>e</b>) South Carolina, and (<b>f</b>) Georgia.</p> Full article ">Figure 25
<p>Comparing a Carolina Bay elevation profile with a hypothetical impact crater elevation profile. The sediment deposit thickness curve is plotted to show how much basin sediment fill has occurred in the Carolina Bay since its formation.</p> Full article ">
21 pages, 10458 KiB  
Article
A High Latitude Model for the E Layer Dominated Ionosphere
by Sumon Kamal, Norbert Jakowski, Mohammed Mainul Hoque and Jens Wickert
Remote Sens. 2021, 13(18), 3769; https://doi.org/10.3390/rs13183769 - 20 Sep 2021
Cited by 2 | Viewed by 2279
Abstract
Under certain conditions, the ionization of the E layer can dominate over that of the F2 layer. This phenomenon is called the E layer dominated ionosphere (ELDI) and occurs mainly in the auroral regions. In the present work, we model the variation of [...] Read more.
Under certain conditions, the ionization of the E layer can dominate over that of the F2 layer. This phenomenon is called the E layer dominated ionosphere (ELDI) and occurs mainly in the auroral regions. In the present work, we model the variation of the ELDI for the Northern and Southern Hemispheres. Our proposed Neustrelitz ELDI Event Model (NEEM) is an empirical, climatological model that describes ELDI characteristics by means of four submodels for selected model observables, considering the dependencies on appropriate model drivers. The observables include the occurrence probability of ELDI events and typical E layer parameters that are important to describe the propagation medium for High Frequency (HF) radio waves. The model drivers are the geomagnetic latitude, local time, day of year, solar activity and the convection electric field. During our investigation, we found clear trends for the model observables depending on the drivers, which can be well represented by parametric functions. In this regard, the submodel NEEM-N characterizes the peak electron density NmE of the E layer, while the submodels NEEM-H and NEEM-W describe the corresponding peak height hmE and the vertical width wvE of the E layer electron density profile, respectively. Furthermore, the submodel NEEM-P specifies the ELDI occurrence probability %ELDI. The dataset underlying our studies contains more than two million vertical electron density profiles covering a period of almost 13 years. These profiles were derived from ionospheric GPS radio occultation observations on board the six COSMIC/FORMOSAT-3 satellites (Constellation Observing System for Meteorology, Ionosphere and Climate/Formosa Satellite Mission 3). We divided the dataset into a modeling dataset for determining the model coefficients and a test dataset for subsequent model validation. The normalized root mean square deviation (NRMS) between the original and the predicted model observables yields similar values across both datasets and both hemispheres. For NEEM-N, we obtain an NRMS varying between 36.1% and 47.1% and for NEEM-H, between 6.1% and 6.3%. In the case of NEEM-W, the NRMS varies between 38.5% and 41.1%, while it varies between 56.5% and 60.3% for NEEM-P. In summary, the proposed NEEM utilizes primary relationships with geophysical and solar wind observables, which are useful for describing ELDI occurrences and the associated changes of the E layer properties. In this manner, the NEEM paves the way for future prediction of the ELDI and of its characteristics in technical applications, especially from the fields of telecommunications and navigation. Full article
(This article belongs to the Special Issue Radio Occultations for Numerical Weather Prediction, Ionosphere, and Space Weather)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Flowchart of the data pre-processing steps.</p> Full article ">Figure 2
<p>Observed (red) and predicted (black) NmE trends for the modeling dataset. Left column: Northern Hemisphere. Right column: Southern Hemisphere. Last row: root mean square deviation (RMS) and normalized RMS (NRMS) between observation and prediction.</p> Full article ">Figure 3
<p>Observed (red) and predicted (black) hmE trends for the modeling dataset. Left column: Northern Hemisphere. Right column: Southern Hemisphere. Last row: RMS and NRMS deviation between observation and prediction.</p> Full article ">Figure 4
<p>Observed (red) and predicted (black) wvE trends for the modeling dataset. Left column: Northern Hemisphere. Right column: Southern Hemisphere. Last row: RMS and NRMS deviation between observation and prediction.</p> Full article ">Figure 5
<p>Observed (red) and predicted (black) %ELDI trends for the modeling dataset. (<b>Left</b>) Northern Hemisphere. (<b>Right</b>) Southern Hemisphere. Last row: RMS and NRMS deviation between observation and prediction.</p> Full article ">Figure 6
<p>Observed (red) and predicted (black) NmE trends for the test dataset. (<b>Left</b>) Northern Hemisphere. (<b>Right</b>) Southern Hemisphere. Last row: RMS and NRMS deviation between observation and prediction.</p> Full article ">Figure 7
<p>Observed (red) and predicted (black) hmE trends for the test dataset. (<b>Left</b>) Northern Hemisphere. (<b>Right</b>) Southern Hemisphere. Last row: RMS and NRMS deviation between observation and prediction.</p> Full article ">Figure 8
<p>Observed (red) and predicted (black) wvE trends for the test dataset. (<b>Left</b>) Northern Hemisphere. (<b>Right</b>) Southern Hemisphere. Last row: RMS and NRMS deviation between observation and prediction.</p> Full article ">Figure 9
<p>Observed (red) and predicted (black) %ELDI trends for the test dataset. (<b>Left</b>) Northern Hemisphere. (<b>Right</b>) Southern Hemisphere. Last row: RMS and NRMS deviation between observation and prediction.</p> Full article ">
19 pages, 4615 KiB  
Article
Approaching Global Instantaneous Precise Positioning with the Dual- and Triple-Frequency Multi-GNSS Decoupled Clock Model
by Nacer Naciri and Sunil Bisnath
Remote Sens. 2021, 13(18), 3768; https://doi.org/10.3390/rs13183768 - 20 Sep 2021
Cited by 8 | Viewed by 2608
Abstract
Precise Point Positioning (PPP), as a global precise positioning technique, suffers from relatively long convergence times, hindering its ability to be the default precise positioning technique. Reducing the PPP convergence time is a must to reach global precise positions, and doing so in [...] Read more.
Precise Point Positioning (PPP), as a global precise positioning technique, suffers from relatively long convergence times, hindering its ability to be the default precise positioning technique. Reducing the PPP convergence time is a must to reach global precise positions, and doing so in a few minutes to seconds can be achieved thanks to the additional frequencies that are being broadcast by the modernized GNSS constellations. Due to discrepancies in the number of signals broadcast by each satellite/constellation, it is necessary to have a model that can process a mix of signals, depending on availability, and perform ambiguity resolution (AR), a technique that proved necessary for rapid convergence. This manuscript does so by expanding the uncombined Decoupled Clock Model to process and fix ambiguities on up to three frequencies depending on availability for GPS, Galileo, and BeiDou. GLONASS is included as well, without carrier-phase ambiguity fixing. Results show the possibility of consistent quasi-instantaneous global precise positioning through an assessment of the algorithm on a network of global stations, as the 67th percentile solution converges below 10 cm horizontal error within 2 min, compared to 8 min with a triple-frequency solution, showing the importance of having a flexible PPP-AR model frequency-wise. In terms of individual datasets, 14% of datasets converge instantaneously when mixing dual- and triple-frequency measurements, compared to just 0.1% in that of dual-frequency mode without ambiguity resolution. Two kinematic car datasets were also processed, and it was shown that instantaneous centimetre-level positioning with a moving receiver is possible. These results are promising as they only rely on ultra-rapid global satellite products, allowing for instantaneous real-time precise positioning without the need for any local infrastructure or prior knowledge of the receiver’s environment. Full article
(This article belongs to the Special Issue GNSS Precise Point Positioning: Towards Global Instantaneous cm-Level Accuracy)
►▼ Show Figures

Figure 1

Figure 1
<p>Reference satellite choice strategy when processing a combination of dual- and triple-frequency satellites.</p> Full article ">Figure 2
<p>IGS stations used in processing.</p> Full article ">Figure 3
<p>Ground tracks of datasets used in kinematic processing.</p> Full article ">Figure 4
<p>Time series comparison of 100th, 95th, and 67th percentile horizontal and vertical errors for dual-frequency, triple-frequency, and mixed dual-/triple-frequency combinations both with and without ambiguity resolution. Horizontal dashed lines represent convergence threshold.</p> Full article ">Figure 5
<p>Horizontal convergence time for dual-frequency, triple-frequency, and mixed dual-/triple-frequency results with and without AR at 100th, 95th, and 67th percentiles. Convergence time statistics correspond to results in <a href="#remotesensing-13-03768-f004" class="html-fig">Figure 4</a>.</p> Full article ">Figure 6
<p>Histogram of number of satellites being processed in dual-frequency, triple-frequency, and mixed dual-/triple-frequency processing. Graph is based on all epochs from all datasets. “dual” and “dual/triple” histograms are overlapping.</p> Full article ">Figure 7
<p>100th, 95th, and 67th percentile horizontal and vertical convergence times and rms for solution combining dual-frequency and triple-frequency satellites, with and without AR.</p> Full article ">Figure 8
<p>Float, dual-frequency fixed, and mixed dual-/triple-frequency fixed histogram of horizontal convergence time of all processed datasets.</p> Full article ">Figure 9
<p>(<b>a</b>) Dual-frequency, triple-frequency, and mixed dual-/triple-frequency 3D errors for station GODS, on DOY 72, 2021 between UTC hours 5 and 6, and (<b>b</b>) number of satellites with two and three frequencies at station GODS on DOY 72, 2021.</p> Full article ">Figure 10
<p>Float and fixed dual-frequency and dual-/triple-frequency horizontal (<b>a</b>) and vertical (<b>b</b>) errors, number of satellites per frequency (<b>c</b>), and ambiguity success rate (<b>d</b>) for the first kinematic dataset collected near York University on DOY 337, 2020.</p> Full article ">Figure 11
<p>Float and fixed dual-frequency and dual-/triple-frequency horizontal (<b>a</b>) and vertical (<b>b</b>) errors for open sky kinematic dataset collected near York University on DOY 151, 2021. Car is static in the first 7 min, and kinematic in the rest of dataset.</p> Full article ">
16 pages, 2069 KiB  
Article
Monitoring Large-Scale Restoration Interventions from Land Preparation to Biomass Growth in the Sahel
by Moctar Sacande, Antonio Martucci and Andreas Vollrath
Remote Sens. 2021, 13(18), 3767; https://doi.org/10.3390/rs13183767 - 20 Sep 2021
Cited by 9 | Viewed by 3515
Abstract
In this work we demonstrate that restoration interventions in arid to semi-arid landscapes can be independently assessed by remote sensing methods throughout all phases. For early verification, we use Sentinel-1 radar imagery that is sensitive to changes in soil roughness and thus able [...] Read more.
In this work we demonstrate that restoration interventions in arid to semi-arid landscapes can be independently assessed by remote sensing methods throughout all phases. For early verification, we use Sentinel-1 radar imagery that is sensitive to changes in soil roughness and thus able to rapidly detect disturbances due to mechanised ploughing, including identification of the time of occurrence and the surface area prepared for planting. Subsequently, time series of the normalized difference vegetation index (NDVI) derived from high-resolution imagery enabled tracking and verifying of the increase in biomass and the long-term impact of restoration interventions. We assessed 111 plots within the Great Green Wall area in Burkina Faso, Niger, Nigeria and Senegal. For 58 plots, the interventions were successfully verified, corresponding to an area of more than 7000 ha of degraded land. Comparatively, these computerised data were matched with field data and high-resolution imagery, for which the NDVI was used as an indicator of subsequent biomass growth in the plots. The trends were polynomial and presented clear vegetation gains for the monthly aggregates over the last 2 years (2018–2020). The qualitative data on planted species also showed an increase in biodiversity as direct sown seeds of a minimum of 10 native Sahel species (six woody mixed with four fodder herbaceous species) were planted per hectare. This innovative and standardised monitoring method provides an objective and timely assessment of restoration interventions and will likely appeal more actors to confidently invest in restoration as a part of zero-net climate mitigation. Full article
(This article belongs to the Special Issue Analysis of Land Cover Change within Semiarid Environments Using Satellite Imagery and GIS)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>(<b>A</b>) Map of the 111 large-scale restoration intervention sites in the Great Green Wall in Burkina Faso (with three sites highlighted and further described in sections below), Niger, Nigeria and Senegal, which were assessed using radar detections. (<b>B</b>) The techniques used for large-scale restoration interventions on the ground from mechanised land preparation for soil permeability and rainwater harvesting and beginning of seedlings/biomass growth in the field.</p> Full article ">Figure 2
<p>Radar detection with the time series profile of mean decibel value of restoration plots ploughed in dry and wet seasons in northern Burkina Faso, combined with total monthly precipitations. (<b>A</b>) At the onset of the dry season (October 2016) dates and pattern on the ground were clearly detected in Sampelga village (87.2 ha). Panels in GEE application (see GEE code link at <a href="https://code.earthengine.google.com/0dd6c148e4a8a56b8c3a5f59849faea7" target="_blank">https://code.earthengine.google.com/0dd6c148e4a8a56b8c3a5f59849faea7</a>, accessed on 6 September 2021) for selecting and displaying images of the GPS delineated plot (in red line). (<b>B</b>) Similar detection was obtained for the site of Sibe (250.4 ha) mid-dry season (January 2017). (<b>C</b>) No detection was made for the site of Cisse (152 ha), because it was ploughed in the wet season (July 2016) when surface roughness also increases by new vegetation.</p> Full article ">Figure 2 Cont.
<p>Radar detection with the time series profile of mean decibel value of restoration plots ploughed in dry and wet seasons in northern Burkina Faso, combined with total monthly precipitations. (<b>A</b>) At the onset of the dry season (October 2016) dates and pattern on the ground were clearly detected in Sampelga village (87.2 ha). Panels in GEE application (see GEE code link at <a href="https://code.earthengine.google.com/0dd6c148e4a8a56b8c3a5f59849faea7" target="_blank">https://code.earthengine.google.com/0dd6c148e4a8a56b8c3a5f59849faea7</a>, accessed on 6 September 2021) for selecting and displaying images of the GPS delineated plot (in red line). (<b>B</b>) Similar detection was obtained for the site of Sibe (250.4 ha) mid-dry season (January 2017). (<b>C</b>) No detection was made for the site of Cisse (152 ha), because it was ploughed in the wet season (July 2016) when surface roughness also increases by new vegetation.</p> Full article ">Figure 3
<p>Multi-temporal RGB composites of Sentinel-1 SAR, VV backscatter. (<b>A</b>) Sampelga site: composite of images of 23 October 2016 (Red) and 16 November 2016 (Green/Blue). (<b>B</b>) Sibe site: composite of images of 8 January 2017 (Red) and 13 February 2017 (Green/Blue). (<b>C</b>) Cisse site: composite of images of 6 June 2016 (Red) and 6 August 2016 (Green/Blue), coinciding with the rainy season.</p> Full article ">Figure 4
<p>Digital monitoring of restoration plots in Sampelga (<b>A1</b>), Sibe (<b>B1</b>), and Cisse (<b>C1</b>), Burkina Faso, with NDVI−Landsat (30 m) variations comparing the monthly and annual averages for the period before interventions (2010–2014) and after interventions (2015–2020). The linear and polynomial trendlines show clear gains toward the end of the data series with higher peaks in months of latest years (<b>A2</b>,<b>B2</b>,<b>C2</b>), demonstrating increase in vegetation, with values of 0.15 in 2017 up to 0.2 in 2020 for the Cisse site (<b>C2</b>). The recurrent detection of negative deviations from average in rainy months is due to lack of data (cloudy images out of 2 images per month).</p> Full article ">Figure 4 Cont.
<p>Digital monitoring of restoration plots in Sampelga (<b>A1</b>), Sibe (<b>B1</b>), and Cisse (<b>C1</b>), Burkina Faso, with NDVI−Landsat (30 m) variations comparing the monthly and annual averages for the period before interventions (2010–2014) and after interventions (2015–2020). The linear and polynomial trendlines show clear gains toward the end of the data series with higher peaks in months of latest years (<b>A2</b>,<b>B2</b>,<b>C2</b>), demonstrating increase in vegetation, with values of 0.15 in 2017 up to 0.2 in 2020 for the Cisse site (<b>C2</b>). The recurrent detection of negative deviations from average in rainy months is due to lack of data (cloudy images out of 2 images per month).</p> Full article ">Figure 5
<p>Visual representation of restoration plots in Sampelga, Sibe and Cisse, Burkina Faso, with NDVI−Sentinel-2/Copernicus (10 m). The green patches are the vegetation in the plots comparatively showing the increase from May 2020 and 2021, during the dry season.</p> Full article ">
21 pages, 5851 KiB  
Article
Building Extraction from Airborne LiDAR Data Based on Multi-Constraints Graph Segmentation
by Zhenyang Hui, Zhuoxuan Li, Penggen Cheng, Yao Yevenyo Ziggah and JunLin Fan
Remote Sens. 2021, 13(18), 3766; https://doi.org/10.3390/rs13183766 - 20 Sep 2021
Cited by 14 | Viewed by 4280
Abstract
Building extraction from airborne Light Detection and Ranging (LiDAR) point clouds is a significant step in the process of digital urban construction. Although the existing building extraction methods perform well in simple urban environments, when encountering complicated city environments with irregular building shapes [...] Read more.
Building extraction from airborne Light Detection and Ranging (LiDAR) point clouds is a significant step in the process of digital urban construction. Although the existing building extraction methods perform well in simple urban environments, when encountering complicated city environments with irregular building shapes or varying building sizes, these methods cannot achieve satisfactory building extraction results. To address these challenges, a building extraction method from airborne LiDAR data based on multi-constraints graph segmentation was proposed in this paper. The proposed method mainly converted point-based building extraction into object-based building extraction through multi-constraints graph segmentation. The initial extracted building points were derived according to the spatial geometric features of different object primitives. Finally, a multi-scale progressive growth optimization method was proposed to recover some omitted building points and improve the completeness of building extraction. The proposed method was tested and validated using three datasets provided by the International Society for Photogrammetry and Remote Sensing (ISPRS). Experimental results show that the proposed method can achieve the best building extraction results. It was also found that no matter the average quality or the average F1 score, the proposed method outperformed ten other investigated building extraction methods. Full article
(This article belongs to the Special Issue Remote Sensing Based Building Extraction II)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Flowchart of the proposed method. Filtering is first applied for removing the ground points. Then, the proposed multi-constrains graph segmentation is adopted to achieve the segmentation results. According to the geometric features, the initial building points can be obtained. Finally, an optimization step is applied to obtain the final building extraction results.</p> Full article ">Figure 2
<p>Graph segmentation based on multi-constraints. (<b>a</b>) The result of graph segmentation based on multi-constraints; (<b>b</b>) enlarged version of the area I in (<b>a</b>); (<b>c</b>) enlarged version of the area II in (<b>a</b>). The colors are assigned randomly according to different segmentation objects. Different colors represent different segmentation objects.</p> Full article ">Figure 3
<p>Building points optimization.</p> Full article ">Figure 4
<p>The tested three datasets provided by the ISPRS. (<b>a</b>) Area1; (<b>b</b>) Area2; (<b>c</b>) Area3.</p> Full article ">Figure 5
<p>Building extraction results of the proposed method. (<b>a</b>) Extraction results for Area1; (<b>b</b>) extraction results for Area2; (<b>c</b>) extraction results for Area3; (<b>d</b>) the extraction results of Area1 integrating with the corresponding orthophoto image; (<b>e</b>) the extraction results of Area2 integrating with the corresponding orthophoto image; (<b>f</b>) the extraction results of Area3 integrating with the corresponding orthophoto image. Yellow represents correctly extracted buildings (<math display="inline"><semantics> <mrow> <mi>T</mi> <mi>P</mi> </mrow> </semantics></math>), red represents wrongly extracted buildings (<math display="inline"><semantics> <mrow> <mi>F</mi> <mi>P</mi> </mrow> </semantics></math>), and blue represents omitted buildings (<math display="inline"><semantics> <mrow> <mi>F</mi> <mi>N</mi> </mrow> </semantics></math>).</p> Full article ">Figure 6
<p>Average quality comparison between the proposed method and ten other methods. (<b>a</b>) Per-area level; (<b>b</b>) per-object level. The average qualities of the ten methods are obtained from the corresponding references. The methods proposed by Maltezos et al. (2019), Doulamis et al. (2003) and Protopapadakis et al. (2016) do not provide the quality at per-object level.</p> Full article ">Figure 7
<p>Average F1 score comparison between the proposed method and ten other methods. (<b>a</b>) Per-area level; (<b>b</b>) per-object level. The F1 score of the ten methods are obtained from the corresponding references. The methods proposed by Maltezos et al. (2019), Doulamis et al. (2003) and Protopapadakis et al. (2016) do not provide the quality at per-object level.</p> Full article ">Figure 8
<p>Graph segmentation results with different <math display="inline"><semantics> <mi>ς</mi> </semantics></math>. (<b>a</b>) <math display="inline"><semantics> <mrow> <mi>ς</mi> <msup> <mrow> <mrow> <mo>=</mo> <mn>1</mn> </mrow> </mrow> <mo>∘</mo> </msup> </mrow> </semantics></math>; (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>ς</mi> <msup> <mrow> <mrow> <mo>=</mo> <mn>10</mn> </mrow> </mrow> <mo>∘</mo> </msup> </mrow> </semantics></math>; (<b>c</b>) <math display="inline"><semantics> <mrow> <mi>ς</mi> <msup> <mrow> <mrow> <mo>=</mo> <mn>15</mn> </mrow> </mrow> <mo>∘</mo> </msup> </mrow> </semantics></math>; (<b>d</b>) the reference segmentation results. Different object primitives are colored with different colors.</p> Full article ">Figure 9
<p>Initial building points extraction results with different roughness threshold. (<b>a</b>) The threshold is 0.02; (<b>b</b>) the threshold is 0.04; (<b>c</b>) the threshold is 0.06. Yellow represents correctly extracted buildings (<math display="inline"><semantics> <mrow> <mi>T</mi> <mi>P</mi> </mrow> </semantics></math>), red represents wrongly extracted buildings (<math display="inline"><semantics> <mrow> <mi>F</mi> <mi>P</mi> </mrow> </semantics></math>), and blue represents omitted buildings (<math display="inline"><semantics> <mrow> <mi>F</mi> <mi>N</mi> </mrow> </semantics></math>).</p> Full article ">Figure 10
<p>The study areas of OpenTopography datasets. (<b>a</b>) The true-color image of S1; (<b>b</b>) the point cloud data of S1, which is colored according to different labels; (<b>c</b>) the true-color image of S2; (<b>d</b>) the point cloud data of S2, which is colored according to different labels.</p> Full article ">Figure 11
<p>Building extraction results of OpenTopography datasets. (<b>a</b>) The building extraction result for S1; (<b>b</b>) the reference building extraction results of S1 from orthophoto image; (<b>c</b>) the building extraction result for S2; (<b>d</b>) the reference building extraction results of S2 from orthophoto image. Yellow represents correctly extracted buildings (<math display="inline"><semantics> <mrow> <mi>T</mi> <mi>P</mi> </mrow> </semantics></math>), red represents wrongly extracted buildings (<math display="inline"><semantics> <mrow> <mi>F</mi> <mi>P</mi> </mrow> </semantics></math>), and blue represents omitted buildings (<math display="inline"><semantics> <mrow> <mi>F</mi> <mi>N</mi> </mrow> </semantics></math>).</p> Full article ">
24 pages, 6823 KiB  
Article
Potential Impacts of Assimilating Every-10-Minute Himawari-8 Satellite Radiance with the POD-4DEnVar Method
by Jingnan Wang, Lifeng Zhang, Jiping Guan, Xiaodong Wang, Mingyang Zhang and Yuan Wang
Remote Sens. 2021, 13(18), 3765; https://doi.org/10.3390/rs13183765 - 20 Sep 2021
Cited by 1 | Viewed by 2691
Abstract
The Advanced Himawari Imager (AHI) onboard the Himawari-8 geostationary satellite provides continuous observations every 10 min. This study investigates the assimilation of every-10-min radiance from the AHI with the POD-4DEnVar method. Cloud detection is conducted in the AHI quality control procedure to remove [...] Read more.
The Advanced Himawari Imager (AHI) onboard the Himawari-8 geostationary satellite provides continuous observations every 10 min. This study investigates the assimilation of every-10-min radiance from the AHI with the POD-4DEnVar method. Cloud detection is conducted in the AHI quality control procedure to remove cloudy and precipitation-affected observations. Historical samples and physical ensembles are combined to construct four-dimensional ensembles according to the observed frequency of the Himawari-8 satellite. The purpose of this study was to test the potential impacts of assimilating high temporal resolution observations with POD-4DEnVar in a numerical weather prediction (NWP) system. Two parallel experiments were performed with and without Himawari-8 radiance assimilation during the entire month of July 2020. The results of the experiment with radiance assimilation show that it improves the analysis and forecast accuracy of geopotential, horizontal wind field and relative humidity compared to the experiment without radiance assimilation. Moreover, the equitable threat score (ETS) of 24-h accumulated precipitation shows that assimilating Himawari-8 radiance improves the rainfall forecast accuracy. Improvements were found in the structure, amplitude and location of the precipitation. In addition, the ETS of hourly accumulated precipitation indicates that assimilating high temporal resolution Himawari-8 radiance can improve the prediction of rapidly developed rainfall. Overall, assimilating every-10-min AHI radiance from Himawari-8 with POD-4DEnVar has positive impacts on NWP. Full article
(This article belongs to the Section Atmospheric Remote Sensing)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Design of the assimilation window and its sub-windows.</p> Full article ">Figure 2
<p>The flowchart of four-dimensional ensemble samples’ construction.</p> Full article ">Figure 3
<p>The Himawari-8 brightness temperature (K) at 0000 UTC 20 July 2020 for (<b>a</b>) observations, (<b>b</b>) observations minus simulations from background (OMB) and (<b>c</b>) observations minus simulations from analysis (OMA).</p> Full article ">Figure 4
<p>The scatterplots of observations against CRTM simulated brightness temperature (K) from (<b>a</b>) background without BC, (<b>b</b>) background with BC and (<b>c</b>) analysis with BC.</p> Full article ">Figure 5
<p>Two-nested domain design and observations covered in the domains. Color dots represent the brightness temperature from AHI channel 2 (units: K) at 0000 UTC 20 July 2020. The numbers represent the main provinces covered in domain 2 (1: Henan; 2: Hubei; 3: Anhui; 4: Jiangsu; 5: Hunan; 6: Jiangxi; 7: Zhejiang; 8: Fujian).</p> Full article ">Figure 6
<p>Schematic configuration of data assimilation and forecast experiments.</p> Full article ">Figure 7
<p>The vertical profiles of average bias (dash lines) and RMSE (solid lines) in CTRL (blue line) and HIM8 (red line) experiments for (<b>a</b>) geopotential, (<b>b</b>) U, (<b>c</b>) V, (<b>d</b>) temperature, (<b>e</b>) water vapor mixing ratio and (<b>f</b>) relative humidity from analysis fields.</p> Full article ">Figure 8
<p>The vertical profiles of average bias (dash lines) and RMSE (solid lines) in CTRL (blue line) and HIM8 (red line) experiments for (<b>a</b>) geopotential, (<b>b</b>) U, (<b>c</b>) V, (<b>d</b>) temperature, (<b>e</b>) water vapor mixing ratio and (<b>f</b>) relative humidity from 24-h forecast fields.</p> Full article ">Figure 9
<p>The evolution of hourly precipitation (box, left Y-axis with unit of mm) and growth rate of precipitation every 3 h (solid line, right Y-axis) at different times (X-axis) on (<b>a</b>) 5 July (<b>b</b>) 6 July and (<b>c</b>) 18 July 2020.</p> Full article ">Figure 10
<p>The ETS of 24-h accumulated precipitation initialized at (<b>a</b>) 0000 UTC on 5 July (<b>b</b>) 0000 UTC on 6 July and (<b>c</b>) 0000 UTC on 18 July 2020 for different thresholds.</p> Full article ">Figure 11
<p>The SAL evaluation on 24-h accumulated precipitation initialized at (<b>a</b>,<b>d</b>,<b>g</b>) 0000 UTC on 5 July (<b>b</b>,<b>e</b>,<b>h</b>) 0000 UTC on 6 July and (<b>c</b>,<b>f</b>,<b>i</b>) 0000 UTC on 18 July 2020 for different thresholds.</p> Full article ">Figure 12
<p>The distribution of 24-h accumulated precipitation initialized at (<b>a</b>,<b>d</b>,<b>g</b>) 0000 UTC on 5 July (<b>b</b>,<b>e</b>,<b>h</b>) 0000 UTC on 6 July and (<b>c</b>,<b>f</b>,<b>i</b>) 0000 UTC on 18 July 2020 from observations (first row) and CTRL (second row) and HIM8 (third row) experiments.</p> Full article ">Figure 13
<p>The ETS evolution of hourly precipitation on (<b>a</b>,<b>d</b>,<b>g</b>) 5 July (<b>b</b>,<b>e</b>,<b>h</b>) 6 July and (<b>c</b>,<b>f</b>,<b>i</b>) 18 July 2020 for different thresholds.</p> Full article ">Figure 14
<p>The increments in (<b>a</b>,<b>d</b>) water vapor mixing ratio (shaded; g/kg), (<b>b</b>,<b>e</b>) temperature (shaded; K) and (<b>c</b>,<b>f</b>) vertical velocity (shaded; m/s) at 0000 UTC on 18 July 2020. (<b>a</b>–<b>c</b>) 500 hPa and (<b>d</b>–<b>f</b>) 700 hPa. The vectors show the direction and magnitude of the horizontal wind field from analysis.</p> Full article ">Figure 15
<p>(<b>a</b>) The 24-h accumulated precipitation ETS on 18 July for different thresholds. (<b>b</b>–<b>d</b>) The hourly accumulated precipitation ETS for 0.1, 5 and 10 mm on 18 July. The green lines represent the HIM8-1h experiment and red represents HIM8.</p> Full article ">
25 pages, 6050 KiB  
Article
Modeling the Near-Surface Energies and Water Vapor Fluxes Behavior in Response to Summer Canopy Density across Yanqi Endorheic Basin, Northwestern China
by Patient Mindje Kayumba, Gonghuan Fang, Yaning Chen, Richard Mind’je, Yanan Hu, Sikandar Ali and Mapendo Mindje
Remote Sens. 2021, 13(18), 3764; https://doi.org/10.3390/rs13183764 - 20 Sep 2021
Cited by 1 | Viewed by 2390
Abstract
The Yanqi basin is the main irrigated and active agroecosystem in semi-arid Xinjiang, northwestern China, which further seeks responses to the profound local water-related drawbacks in relation to the unceasing landscape desiccation and scant precipitation. Yet, it comes as an astonishment that a [...] Read more.
The Yanqi basin is the main irrigated and active agroecosystem in semi-arid Xinjiang, northwestern China, which further seeks responses to the profound local water-related drawbacks in relation to the unceasing landscape desiccation and scant precipitation. Yet, it comes as an astonishment that a few reported near-surface items and water vapor fluxes as so far required for water resources decision support, particularly in a scarce observation data region. As a contributive effort, here we adjusted the sensible heat flux (H) calibration mechanism of Surface Energy Balance Algorithm for Land (SEBAL) to high-resolution satellite dataset coupled with in-situ observation, through a wise guided “anchor” pixel assortment from surface reflectance-α, Leaf area index-LAI, vegetation index-NDVI, and surface temperature (Pcold, Phot) to model the robustness of energy fluxes and Evapotranspiration-ETa over the basin. Results reasonably reflected ETa which returned low RMSE (0.6 mm d−1), MAE (0.48 mm d−1) compared to in-situ recordings, indicating the competence of SEBAL to predict vapor fluxes in this region. The adjustment unveiled the estimates of the land-use contribution to evapotranspiration with an average ranging from 3 to 4.69 mm d−1, reaching a maximum of 5.5 mm d−1. Furthermore, findings showed a high striking energy dissipation (LE/Rn) across grasslands and wetlands. The vegetated surfaces with a great evaporative fraction were associated with the highest LE/Rn (70–90%), and water bodies varying between 20% and 60%, while the desert ecosystem dissipated the least energy with a low evaporative fraction. Still, besides high portrayed evaporation in water, grasslands and wetlands varied interchangeably in accounting for the highest ETa followed by cropland. Finally, a substantial nexus between available energy (Rn-G) and ETa informed the available energy, influenced by NDVI to be the primary driver of these oases’ transpiration. This study provides essentials of near-surface energy fluxes and the likelihood of ETa with considerable baseline inferences for Yanqi that may be beneficial for long-term investigations that will attend in agrometeorological services and sustainable management of water resources in semi-arid regions. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>Location of Yanqi ecosystem in the large Bosten Lake Basin, China. The map is from Chinese Standard Map (<a href="http://bzdt.ch.mnr.gov.cn/" target="_blank">http://bzdt.ch.mnr.gov.cn/</a>, GS (2019)1652).</p> Full article ">Figure 2
<p>Meteorological condition during the study time satellite overpass in the study area (Xinjiang, local time).</p> Full article ">Figure 3
<p>Flow chart of the net surface radiation computation.</p> Full article ">Figure 4
<p>Model performance prototype of bivariate scatter plots, and histograms showing anchor pixels selection through NDVI, Ts (cold-dry), and surface albedo.</p> Full article ">Figure 5
<p>Spatial mean variation of surface energy components across Yanqi basin during the summer season (Rn: Net Radiation, G: Ground heat fluxes, H: Sensible heat flux, and LE: Latent heat flux).</p> Full article ">Figure 6
<p>SEBAL-derived available energy distribution across the Yanqi land use types during the understudied period. On each boxplot, the central mark represents the median, while the upper and lower edge determines the first and the third quartiles, respectively. Lately, the whiskers extend to a minimum up to a maximum value.</p> Full article ">Figure 7
<p>Magnitude variation of Yanqi vegetated canopy density (NDVI) related with available energy under the studied period.</p> Full article ">Figure 8
<p>Near-surface energy components proportion over Yanqi basin where: (<b>a</b>) The partition of main near-surface energies for LE of evapotranspiration during summertime (Rn, G, and H); (<b>b</b>) Dissipation of net radiation amongst latent heat flux of evapotranspiration (LE/Rn) during the understudied period (DOY), (<b>c</b>) Evaporative fraction displaying the ratio between the latent heat flux and the available energy.</p> Full article ">Figure 9
<p>Spatial distribution of SEBAL derived daily <math display="inline"><semantics> <mrow> <msub> <mrow> <mi>ET</mi> </mrow> <mi mathvariant="normal">a</mi> </msub> </mrow> </semantics></math> across Yanqi basin.</p> Full article ">Figure 10
<p>Summer season daily mean ETa over Yanqi land cover canopy.</p> Full article ">Figure 11
<p>Trends of daily ETa across every single land-use type. (The red dotted line indicates the 3rd order polynomial trend).</p> Full article ">Figure 12
<p>Monthly and season spatial distribution of SEBAL derived ETa variability across the Yanqi basin.</p> Full article ">Figure 13
<p>Nexus between available energy and <math display="inline"><semantics> <mrow> <mi>ETa</mi> </mrow> </semantics></math> over Yanqi during the summer season 2019, Day of the Year (DOY). (The black dotted line indicates a reasonable agreement based on the fitting linear model while the orange dots represent the evapotranspiration-ETa and green representing the available energy Rn-G).</p> Full article ">Figure 14
<p>The trend of mean <math display="inline"><semantics> <mrow> <mrow> <mi>ETa</mi> </mrow> </mrow> </semantics></math> estimates of vegetated patches against the average latent heat of evapotranspiration over time. (The black dotted line shows the peak of summer warmth, while the dotted red line represents the polynomial trend of LET).</p> Full article ">
14 pages, 3050 KiB  
Technical Note
Development and Validation of Machine-Learning Clear-Sky Detection Method Using 1-Min Irradiance Data and Sky Imagers at a Polluted Suburban Site, Xianghe
by Mengqi Liu, Xiangao Xia, Disong Fu and Jinqiang Zhang
Remote Sens. 2021, 13(18), 3763; https://doi.org/10.3390/rs13183763 - 20 Sep 2021
Cited by 6 | Viewed by 2719
Abstract
Clear-sky detection (CSD) is of critical importance in solar energy applications and surface radiative budget studies. Existing CSD methods are not sufficiently validated due to the lack of high-temporal resolution and long-term CSD ground observations, especially at polluted sites. Using five-year high resolution [...] Read more.
Clear-sky detection (CSD) is of critical importance in solar energy applications and surface radiative budget studies. Existing CSD methods are not sufficiently validated due to the lack of high-temporal resolution and long-term CSD ground observations, especially at polluted sites. Using five-year high resolution ground-based solar radiation data and visual inspected Total Sky Imager (TSI) measurements at polluted Xianghe, a suburban site, this study validated 17 existing CSD methods and developed a new CSD model based on a machine-learning algorithm (Random Forest: RF). The propagation of systematic errors from input data to the calculated global horizontal irradiance (GHI) is confirmed with Mean Absolute Error (MAE) increased by 99.7% (from 20.00 to 39.93 W·m−2). Through qualitative evaluation, the novel Bright-Sun method outperforms the other traditional CSD methods at Xianghe site, with high accuracy score 0.73 and 0.92 under clear and cloudy conditions, respectively. The RF CSD model developed by one-year irradiance and TSI data shows more robust performance, with clear/cloudy-sky accuracy score of 0.78/0.88. Overall, the Bright-Sun and RF CSD models perform satisfactorily at heavy polluted sites. Further analysis shows the RF CSD model built with only GHI-related parameters can still achieve a mean accuracy score of 0.81, which indicates RF CSD models have the potential in dealing with sites only providing GHI observations. Full article
(This article belongs to the Special Issue Artificial Intelligence in Remote Sensing of Atmospheric Environment)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Validation of MERRA-2 against daily AERONET during 2005–2009. (<b>a</b>) AOD, (<b>b</b>) PWV, and (<b>c</b>) AE.</p> Full article ">Figure 2
<p>Colored scatter plots of measured clear-sky GHI and REST2-calculated GHI<sub>cs</sub> by (<b>a</b>) AERONET data and (<b>b</b>) MERRA-2 data. The color bar indicates the frequency of match, and the dashed red line indicates the 1:1 line.</p> Full article ">Figure 3
<p>The histogram of samples (year of 2005) against different sky conditions (blue for clear and orange for cloudy).</p> Full article ">Figure 4
<p>The flowchart in RF CSD model training and testing.</p> Full article ">Figure 5
<p>The accuracy score of CSD<sub>sky</sub> methods and the RF CSD model under clear (blue bar) and cloudy (orange bar) conditions.</p> Full article ">Figure 6
<p>Four-day typical examples from TSI images: (<b>a</b>) clear and clean skies with AOD of 0.08 on 14 January 2009; (<b>b</b>) clear but polluted skies with AOD of 0.52 on 15 January 2009; (<b>c</b>) cloudy (cirrus) with AOD of 0.69 on 16 January; (<b>d</b>,<b>e</b>) cloudy in the morning but clear in the afternoon on 24 March 2009 (AOD = 0.14). CSD detection results for the four typical examples are attached. The panels in the eight upper rows show determined clear periods determined by the conventional CSD<sub>sky</sub> methods (blue line) that overlap the GHI data (black line). The panels in the lower two rows demonstrate Bright-Sun (green line), the RF CSD model (orange line), and TSI (red line) identified clear periods overlapping the GHI data (black line).</p> Full article ">Figure 7
<p>Accuracy score of the Bright-Sun (the green line) and RF CSD model (the red line) under clear-sky (<b>a</b>) and cloudy-sky (<b>b</b>) conditions, and the frequency of occurrence for each AOD background.</p> Full article ">
19 pages, 3508 KiB  
Article
Landscape Structure and Seasonality: Effects on Wildlife Species Richness and Occupancy in a Fragmented Dry Forest in Coastal Ecuador
by Xavier Haro-Carrión, Jon Johnston and María Juliana Bedoya-Durán
Remote Sens. 2021, 13(18), 3762; https://doi.org/10.3390/rs13183762 - 19 Sep 2021
Cited by 5 | Viewed by 3147
Abstract
Despite high fragmentation and deforestation, little is known about wildlife species richness and occurrence probabilities in tropical dry forest (TDF) landscapes. To fill this gap in knowledge, we used a Sentinel-2-derived land-cover map, Normalized Difference Vegetation Index (NDVI) data and a multi-species occupancy [...] Read more.
Despite high fragmentation and deforestation, little is known about wildlife species richness and occurrence probabilities in tropical dry forest (TDF) landscapes. To fill this gap in knowledge, we used a Sentinel-2-derived land-cover map, Normalized Difference Vegetation Index (NDVI) data and a multi-species occupancy model to correct for detectability to assess the effect of landscape characteristics on medium and large mammal occurrence and richness in three TDF areas that differ in disturbance and seasonality in Ecuador. We recorded 15 species of medium and large mammals, distributed in 12 families; 1 species is critically Endangered, and 2 are Near-Threatened. The results indicate that species occupancy is related to low forest cover and high vegetation seasonality (i.e., high difference in NDVI between the wet and dry seasons). We believe that the apparent negative effect of forest cover is an indicator of species tolerance for disturbance. The three sampling areas varied from 98% to 40% forest cover, yet species richness and occupancy were not significantly different among them. Vegetation seasonality indicates that more seasonal forests (i.e., those where most tree species lose their leaves during the dry season) tend to have higher mammal species occupancy compared to less seasonal, semi-deciduous forests. Overall, occupancy did not vary between the dry and wet seasons, but species-specific data indicate that some species exhibit higher occupancy during the wet season. This research offers a good understanding of mammal species’ responses to habitat disturbance and fragmentation in TDFs and provides insights to promote their conservation. Full article
(This article belongs to the Special Issue Remote Sensing for Applied Wildlife Ecology)
►▼ Show Figures

Figure 1

Figure 1
<p><b>Study landscape.</b> (<b>a</b>) Location of Cantón Sucre within Ecuador; (<b>b</b>) Section of the study landscape where camera traps were used; (<b>c</b>) The three camera trap sampling areas.</p> Full article ">Figure 2
<p>Landscape characteristics of the study landscape. (<b>a</b>) Land-cover map; (<b>b</b>) Mean Annual NDVI; (<b>c</b>) Seasonality: Wet season NDVI—Dry season NDVI. Areas with high values indicate high seasonality; (<b>d</b>) NDVI metrics used as predictors in occupancy models; (<b>e</b>) Seasonality across sampling areas. Boxes detail wildlife sampling area. Red dots indicate sampling sites.</p> Full article ">Figure 3
<p>Richness estimates: (<b>a</b>) Camera-level and (<b>b</b>) Sampling-area estimates using two estimators Shannon and Simpson indexes in the Bahia forest of southwestern Ecuador.</p> Full article ">Figure 4
<p>Estimated beta coefficients for the mean occupancy estimation (Posterior means with 95% and 50% Bayesian credible intervals). Black bars and the red star indicate 95% CI no overlapping with zero. Gray bars with closed circles and open red circles indicate 50% CI not overlapping with zero. Gray bars with open circles indicate that the 50% CI overlapping zero.</p> Full article ">Figure 5
<p>Comparison of the mean occupancy estimates per species per season in the three sampling areas.</p> Full article ">Figure A1
<p>Estimated beta coefficients for mean detection estimation in the study landscape (Posterior means with 95% and 50% Bayesian credible intervals). Black bars indicate that the 95% CI does not overlap zero, and gray bars with closed circles indicate that the 50% CI does not overlap zero, but the 95% CI and gray bars with open circles indicate that the 50% CI overlap zero.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-243
Previous Issue
Next Issue
Back to TopTop