[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Next Article in Journal
Improved MobileNet V3-Based Identification Method for Road Adhesion Coefficient
Next Article in Special Issue
Performance of Soil Moisture Sensors at Different Salinity Levels: Comparative Analysis and Calibration
Previous Article in Journal
Development of Optical Sensors Based on Neutral Red Absorbance for Real-Time pH Measurements
Previous Article in Special Issue
Spectral Data Processing for Field-Scale Soil Organic Carbon Monitoring
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
 
Journals
Sensors
Volume 24
Issue 17
10.3390/s24175612
sensors-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Monitoring Soil Copper in Urban Land Using Visibale and Near-Infrared Spectroscopy with Spatially Nearby Samples

by
Yi Liu
1,
Tiezhu Shi
2,
Zeying Lan
3,
Kai Guo
4,
Chao Yang
2 and
Yiyun Chen
5,6,7,*
1
School of Public Administration, Guangdong University of Finance & Economics, Guangzhou 510320, China
2
State Key Laboratory of Subtropical Building and Urban Science & Guangdong–Hong Kong-Macau Joint Laboratory for Smart Cities & MNR Key Laboratory for Geo-Environmental Monitoring of Great Bay Area, Shenzhen University, Shenzhen 518060, China
3
School of Management, Guangdong University of Technology, Guangzhou 510520, China
4
School of Geography and Remote Sensing, Guangzhou University, Guangzhou 510006, China
5
School of Resource and Environmental Sciences, Wuhan University, Wuhan 430079, China
6
Key Laboratory of Geographic Information System of the Ministry of Education, Wuhan University, Wuhan 430079, China
7
Hubei Luojia Laboratory, Wuhan 430079, China
*
Author to whom correspondence should be addressed.
Sensors 2024, 24(17), 5612; https://doi.org/10.3390/s24175612
Submission received: 30 July 2024 / Revised: 28 August 2024 / Accepted: 28 August 2024 / Published: 29 August 2024
(This article belongs to the Special Issue Soil Sensing and Mapping for a Sustainable Future)
Browse Figures
Graphical abstract
"> Figure 1
<p>Location of the sampling sites.</p> "> Figure 2
<p>Equipment setup for spectral measurements.</p> "> Figure 3
<p>A validation sample with 20 and 50 spatially nearby samples used for building the Cu estimation model. The green circles denote samples that were not selected as nearby samples.</p> "> Figure 4
<p>Flowchart of using spatially nearby samples.</p> "> Figure 5
<p>Boxplot and histogram of Cu content for calibration samples (<b>a</b>) and validation samples (<b>b</b>). Repoint (·) denotes the mean value. The blue line (|) denotes the median value. Hollow circle (○) denotes the outliers. The black box denotes the interquartile range.</p> "> Figure 6
<p>Land use and sample distribution in Shenzhen city.</p> "> Figure 7
<p>Soil Cu content between predicted and measured values using spectroscopy models without considering spatially nearby samples. <math display="inline"><semantics> <mrow> <msubsup> <mrow> <mi>R</mi> </mrow> <mrow> <mi>p</mi> </mrow> <mrow> <mn>2</mn> </mrow> </msubsup> </mrow> </semantics></math> denotes coefficient of determination in prediction. RMSEP denotes the root mean square error of prediction. RPD denotes the residual predictive deviation.</p> "> Figure 8
<p>Performance of soil Cu estimation model considering different numbers of spatial nearby samples. (<b>a</b>) <math display="inline"><semantics> <mrow> <msubsup> <mrow> <mi>R</mi> </mrow> <mrow> <mi>p</mi> </mrow> <mrow> <mn>2</mn> </mrow> </msubsup> </mrow> </semantics></math>, denotes coefficient of determination in prediction. (<b>b</b>) RMSEP, denotes the root mean square error of prediction. (<b>c</b>) RPD, denotes the residual predictive deviation. The dotted blue line is the fitting line.</p> "> Figure 9
<p>Performance of soil Cu estimation model when the number of spatial nearby samples is 125. <math display="inline"><semantics> <mrow> <msubsup> <mrow> <mi>R</mi> </mrow> <mrow> <mi>p</mi> </mrow> <mrow> <mn>2</mn> </mrow> </msubsup> </mrow> </semantics></math> denotes coefficient of determination in prediction. RMSEP denotes the root mean square error of prediction. RPD denotes the residual predictive deviation. (<b>a</b>) The selected 125 nearby samples. (<b>b</b>) The performance of the Cu estimation model.</p> "> Figure 10
<p>Correlation between Cu concentration and spectral wavelengths from 350 to 2500 nm. The blue line denotes the Pearson correlation coefficient. The dotted line denotes the threshold for important wavelengths.</p> "> Figure 11
<p>Examples of spatially nearby samples and geographic subsets.</p> "> Figure 12
<p>The mean distance between validation and calibration samples when selecting different numbers of nearby samples.</p> "> Figure 13
<p>The mean distance (17 km) between the validation and calibration sample when selecting 125 nearby samples. The red circle has a radius of 17 km.</p> ">
Versions Notes

Abstract

:
Soil heavy metal contamination in urban land can affect biodiversity, ecosystem functions, and the health of city residents. Visible and near-infrared (Vis-NIR) spectroscopy is fast, inexpensive, non-destructive, and environmentally friendly compared to traditional methods of monitoring soil Cu, a common heavy metal found in urban soils. However, there has been limited research on using spatially nearby samples to build the Cu estimation model. Our study aims to investigate how spatially nearby samples influence the Cu estimation model. In our study, we collected 250 topsoil samples (0–20 cm) from China’s third-largest city and analyzed their spectra (350–2500 nm). For each unknown validation sample, we selected its spatially nearby samples to construct the Cu estimation model. The results showed that compared to the traditional method ( R p 2 = 0.75, RMSEP = 8.56, RPD = 1.73), incorporating nearby samples greatly improved the model ( R p 2 = 0.93, RMSEP = 4.02, RPD = 3.89). As the number of nearby samples increased, the performance of the Cu estimation model followed an inverted U-shaped curve—initially increasing and then declining. The optimal number of nearby samples is 125 (62.5% of the total), and the mean distance between validation and calibration samples is 17 km. Therefore, we conclude that using nearby samples significantly enhances the Cu estimation model. The optimal number of nearby samples should strike a balance, covering a moderate area without there being too few or too many.

Graphical Abstract">
Graphical Abstract

1. Introduction

Soil heavy metal contamination poses challenges to several United Nations Sustainable Development Goals (SDGs), including good health and well-being, sustainable ecosystems and cities, and climate change regulation [1,2]. Soil heavy metal contamination can directly affect biodiversity and ecosystem functions [3]. Moreover, soil heavy metal contamination in urban areas can affect the health of city residents through food chains, drinking water, and direct contact with soil in greenspaces [4]. Soil heavy metal contamination comes from both human activities and natural sources. Human activities include vehicle emissions, industrial processes, and poor waste management, while natural sources include metal-rich rock weathering [5]. Urban soil is more influenced by human activities than natural soil, especially due to rapid industrialization and urbanization, particularly in developing countries [6,7]. Thus, it is urgent to monitor soil heavy metal contamination in urban areas.
Traditional methods of measuring soil heavy metals are expensive and time-consuming [8,9]. Copper (Cu) is a common heavy metal found in urban soil [10]. The traditional method of Cu estimation is inductively coupled plasma mass spectrometer (ICP-MS) [11]. This method requires sample preparation and chemical preprocessing, making it time-consuming and costly [12,13]. In recent years, visible and near-infrared (vis-NIR) spectroscopy has been developed as an alternative method for estimating heavy metal content in soil [14,15,16,17]. This method is fast, cheap, non-destructive, and environmentally friendly [18,19,20].
When using vis-NIR spectroscopy, calibration samples were first used to build a Cu estimation model. Then, new samples were tested with this built model to predict their Cu content [21]. Covering a large area would generate many samples and increase their diversity [22]. Thus, it is challenging for the built model to be suitable for various conditions, such as different parent materials, soil types, land use, and landscapes [23]. One possible way to decrease the diversity is by considering spatial similarity [24]. In other words, the nearby samples of a validation sample can be used to build the model and predict the Cu content of that validation sample. In fact, samples that are close together are more likely to share a similar Cu estimation model than those that are far apart [25,26]. Thus, it is important to consider spatially nearby samples when building a Cu estimation model.
Early researchers have studied the spatial similarities in estimating soil properties. Viscarra Rossel et al. (2024) [27] assessed geographic similarities and pointed out that samples under similar pedo-climatic conditions benefit the local soil property estimation. Song et al. (2024) [28] considered geographical stratification and achieved the highest accuracy in soil total nitrogen estimation. Khosravi et al. (2024) [29] used geographical and texture-based stratification strategies to enhance the soil organic carbon prediction. Hong et al. (2023) [30] found that considering land use in more homogeneous classes led to better soil inorganic carbon predictions. These researchers found that similarity in position is important for estimating soil properties [31,32,33]. However, previous studies on spatial similarity did not use samples from nearby areas; instead, they used methods like geographic stratifications.
Spatially nearby samples mean using the surrounding samples to predict the Cu content of a specific sample. Instead of focusing solely on nearby samples, many researchers have explored geographic zones, geographic subsets, spatial dependence, or landscapes [34,35,36]. Shi et al. (2015) [37] used soil geographical zoning to account for similarities in soil-forming conditions. Vohland et al. (2022) [38] considered geographical subsets to improve the performance of the soil organic carbon estimation model. Geographic zones or stratifications mean dividing the samples into groups based on their location. This approach increases similarities within each area, but it does not ensure that each sample is close to its nearest samples. However, in theory, nearby samples are more suitable and preferred [27,39]. Focusing on similar spatial or close distances makes samples share more common characteristics, such as spectral shapes and environmental factors [40]. Summerauer et al. (2021) [41] found that the nearest neighbor samples are a suitable method. Dorantes et al. (2022) [42] reported that reducing the geographic extent of a calibration model may reduce the spectral feature space. Thus, our study focuses on spatially nearby samples that are located to each other instead of using geographic zones or stratifications, as many previous researchers have done. We investigated how these nearby samples influence soil Cu estimation. Moreover, how many spatially nearby samples are suitable for the soil Cu estimation model? Is it a case of “the more, the better” for performance? However, there is less research on the optimal number of spatially nearby samples required for soil Cu estimation.
To overcome this challenge, we aim to address two gaps in current soil spectroscopy research: (i) using spatially nearby samples to estimate soil Cu in urban land by vis-NIR spectroscopy; (ii) determining the optimal number of spatially nearby samples for the Cu estimation model.

2. Materials and Methods

2.1. Study Area and Sample Collection

Our study area is Shenzhen City, located in southeast China (113°46′ E to 114°37′ E and 22°27′ N to 22°52′ N), as shown in Figure 1. This city is the 3rd-largest in China and the 10th-largest in the world. In 1979, most of this city was agricultural land, with a population of 3.14 million people. However, following China’s reform and opening-up policy in 1978, the city experienced rapid industrialization and urbanization, growing to a population of 17.79 million people by 2023. The city’s GDP has soared to $482 billion, making it one of the most developed cities in the world. Previous studies have indicated that intensive human activities, such as industrial wastewater, the use of fertilizers and pesticides, vehicle emissions, and household garbage, have resulted in soil contamination by heavy metals [43,44].
This city is located near the sea and close to the Tropic of Cancer (23.5° N). It has an average temperature of 22.4 °C, making it a warm place. The average annual rainfall is 1933 mm, with most of the rain falling in summer, which may cause soil erosion. As classified by the Genetic Soil Classification of China (GSCC), the main soil types in this area are latosolic red soils, red soils, yellow soils, paddy soils, and coastal solonchaks [45]. According to the World Reference Base for Soil Resource (WRB), the main soil types are acrisols, cambisols, anthrosols, and solonchaks [46]. The city’s unique natural conditions and extensive human activities make it an ideal place to study soil heavy metal contamination.
The study area was divided into grids measuring 2 × 2 km, and a sampling site was randomly chosen from each grid. At each site, we removed surface cover or plants and collected 1.5 kg of topsoil from a depth of 0–20 cm during five sampling campaigns [43,47]. However, accessing some grids was difficult due to the city’s hilly terrain, as shown in Figure 1. We made sure to avoid artificial deposits, such as rubbles, concrete debris, and waste. In total, we collected 250 samples in November 2016, and their positions were recorded using a GPS receiver. We also noted the land use, vegetable cover, and landform at each site.

2.2. Spectral Measurement and Chemical Analysis

In the lab, the samples were air-dried and ground until they could pass through a 2 mm sieve. Then, the samples were divided into two parts: one for spectral analysis and the other for chemical analysis. The spectra were obtained using an ASD FieldSpec® 3 portable spectro-radiometer (Analytical Spectral Devices Inc., Boulder, CO, USA), which has a spectral range of 350–2500 nm [48]. The spectra scan was performed in a dark room using a halogen lamp positioned at a 45° angle above the sample (Figure 2). A fiber probe was placed 12 cm directly overhead at a 90° angle. Before measurement, the spectrometer was calibrated with a Spectralon® panel that had 99% reflectance. Each sample was scanned 10 times, and the results were averaged. The Cu content was analyzed using the diethylenetriamine penta-acetic acid method and measured with ICP-OES [11,49].

2.3. Spatially Nearby Samples

In theory, spatially nearby samples share many conditions. Therefore, our study aims to use these samples to build a Cu estimation model. For each unknown validation sample, we selected its surrounding samples based on geographic distance to construct the Cu estimation model. For example, as shown in Figure 3, we selected 20 and 50 nearby samples to predict the Cu content of the validation sample. For each validation sample, a set of nearby samples was selected to build a Cu estimation model (Figure 4). This study involved 50 validation samples, resulting in the creation of 50 Cu estimation models.
To study the influence of nearby samples on the Cu estimation model, we varied the number of nearby samples from 20 to 200 (Figure 4). Our calibration set contains a total of 200 samples. Traditionally, researchers would use all 200 calibration samples to build one model to predict the Cu content of the 50 validation samples. However, our study focuses solely on using nearby samples. The number of nearby samples varied from 20, 21, 22, …, up to 200.

2.4. Model Calibration

A total of 250 samples were divided into 200 calibration samples and 50 validation samples. The 20%/80% split is commonly used by other researchers [13,50]. The 50 validation samples were selected based on their Cu content. Samples were ordered in ascending Cu content, and every fifth sample was chosen. This method ensured that the validation sample set evenly covered the range of Cu content of this city, making it suitable for future new samples from this city.
Partial least squares regression (PLSR) was used to build the model. Although many researchers have recently used deep learning methods like random forest (RF) [51], our study focuses on spatially nearby samples and uses the most commonly used method, PLSR [52]. PLSR first projects the spectra into a low-dimensional space, where multiple regression is then performed. Based on previous research [21], the Cu estimation model does not benefit from or require spectral pretreatment, so we did not apply any.
As mentioned in Section 3.3, not all 200 calibration samples were used at once. For each validation sample, a specific number of nearby samples were selected to form the calibration set. The calibration set was then used to calibrate the PLSR model and predict the Cu content of the validation sample. The number of latent variables was determined using leave-one-out cross-validation (LOOCV). The PLSR was conducted using the PLS_toolbox (Eigenvector Research, Inc., Manson, WA, USA) within the MATLAB environment (The MathWorks, Inc., Natick, MA, USA).

2.5. Model Performance

The 50 validation samples were used to test the performance of the PLSR model. Several common indicators were used to assess the model’s performance: coefficient of determination in prediction ( R p 2 ), root mean square error of prediction (RMSEP), and residual predictive deviation (RPD). These indicators were calculated as follows:
R p 2 = 1 − ∑ i = 1 n y i − y ^ i ∑ i = 1 n y i − y ¯
R M S E P = ∑ i = 1 n y i − y ^ i n
R P D = S D R M S E P
where n is the number of validated samples, and y i and y ^ i are the measured and predicted Cu values of ith sample in the validation set. y ¯   is the average measured Cu value. SD is the standard deviation of measured Cu values.

3. Results

3.1. Descriptive Statistics of Soil Samples

The Cu content ranged from 20.45 to 103.24 mg·kg−1, with a mean value of 58.29 mg·kg−1 (Figure 5 and Table 1). The mean Cu value was three times higher than the background value (17 mg·kg−1), indicating that extensive human activities have caused significant Cu pollution in the soil. According to the pollution level (36 mg·kg−1) [20], 230 samples showed different degrees of Cu pollution, with a pollution rate of 92%. The severe pollution in Shenzhen may be attributed to the city’s rapid urbanization (Figure 6). As shown in Figure 6, much of the city’s land has been converted from cropland to construction areas. With a population of 17.79 million and industrial activities amounting to $482 billion, soil heavy metal pollution has become an urgent environmental issue.
The coefficient of variation (CV) was 0.27, indicating a medium level of variability (0.1 < CV < 1.0). The skewness (0.13) and kurtosis (0.12) were close to zero, indicating a normal distribution. For the calibration and validation sets, Levene’s test confirmed significant homogeneity at the 0.05 significance level (p = 0.99). This was also evidenced by Figure 5, where the boxplot and histogram of calibration and validation sets were very similar.

3.2. Estimation Performance of Cu Models without Considering Spatially Nearby Samples

When not considering spatially nearby samples, the performance of the Cu estimation was acceptable. The R p 2 was 0.75, RMSEP was 8.56 mg·kg−1, and RPD was 1.73 (Figure 7). Most samples were located close to the fit curve. Compared to the 1:1 line, the slope of the fit curve is less than 1 (less than a 45° angle), indicating that low Cu content was overestimated while high Cu content was underestimated. According to the histogram, the measured and predicted Cu content were similar, showing that the model did not change the Cu distribution. Given that our study area was a large city with 17.79 million people and 1997 km2, these three indicators showed that the Cu estimation model was satisfactory.

3.3. Estimation Performance of Cu Models with Spatially Nearby Samples

The R p 2 increased as more spatially nearby samples were selected (Figure 8a and Table 2). With only 20 nearby samples, the R p 2 was 0.75. It then quickly rose to 0.90 with 50 samples. Therefore, between 20 and 50 samples, nearby samples were most beneficial for improving the Cu estimation model. From 50 to 70 samples, the R p 2 decreased slowly to 0.88. When there were more than 70 samples, the R p 2 increased very slowly, reaching 0.93 at 125 samples. Beyond 125 samples, the R p 2 stabilized with minor fluctuations and tended to decrease. Thus, selecting more than 125 nearby samples did not significantly benefit the Cu estimation model and may have even degraded the model.
The RMSEP decreased as more spatially nearby samples were selected (Figure 8b). With 20 nearby samples, the RMSEP was 7.75 mg·kg−1. It then quickly dropped to 4.90 mg·kg−1 with 50 samples, showing the most significant decrease in RMSEP between 20 and 50 nearby samples. Afterward, it rose to 5.46 mg·kg−1 with 70 samples. From 70 to 125 samples, the RMSEP slowly decreased to 4.04 mg·kg−1. Beyond 125 samples, it remained stable with minor fluctuations and tended to increase. Thus, using more than 125 nearby samples did not improve the RMSEP and could even increase it.
The RPD increased in a wave-like pattern as more spatially nearby samples were selected (Figure 8c). There were five peaks in the RPD curve at 28, 43, 80, 125, and 172 samples, with RPD values of 2.81, 3.09, 3.24, 3.88, and 4.08, respectively. The RPD increased before each peak and then decreased afterward. The wave-like pattern suggested that initially, adding nearby samples improved the model, but adding too much eventually degraded it. Therefore, it is crucial to determine the right number of nearby samples—not too few and not too many.
The RPD performed similarly to R p 2 and RMESP in overall (Figure 8c). According to the fit curve (blue dotted line in Figure 8c), the RPD generally rose from 20 to 125 samples and then stabilized with a tendency to decrease. At 20 samples, the RPD was 2.01, and at 125 samples, it was 3.88.
In total, the model’s performance improved significantly as the number of samples increased from 20 ( R p 2 = 0.75, RMSEP = 7.75 mg·kg−1, RPD = 2.01) to 50 ( R p 2 = 0.90, RMSEP = 4.90 mg·kg−1, RPD = 3.14). However, its performance decreased from 50 to 70 samples ( R p 2 = 0.88, RMSEP = 5.46 mg·kg−1, RPD = 2.86). Between 70 and 125 samples, the performance increased gradually ( R p 2 = 0.93, RMSEP = 4.02 mg·kg−1, RPD = 3.89). Beyond 125 samples, the model’s performance stabilized with minor fluctuations and a tendency to degrade.
Compared to the model that did not consider spatially nearby samples ( R p 2 = 0.75, RMSEP = 8.56 mg·kg−1, and RPD = 1.73), using nearby samples significantly improved the Cu estimation model. With 20 nearby samples, the R p 2 was 0.75, RMSEP was 7.75 mg·kg−1, and RPD was 2.01, indicating similar or better performance than using all 200 samples in the traditional way. As more nearby samples were selected, R p 2 increased, RMSEP decreased, and RPD increased, resulting in superior performance than the traditional Cu estimation model. With 125 nearby samples ( R p 2 = 0.93, RMSEP = 4.02 mg·kg−1, and RPD = 3.89), the Cu estimation model greatly outperformed than traditional model ( R p 2 = 0.75, RMSEP = 8.56 mg·kg−1, and RPD = 1.73).
As shown in Figure 8, the Cu estimation model improved until it reached 125 samples. Beyond 125 samples, its performance stabilized with minor improvement and even a slight tendency to degrade. Therefore, the number of 125 nearby samples was studied specifically (Figure 9). In Figure 9b, the Cu estimation model demonstrated very high accuracy ( R p 2 = 0.93, RMSEP = 4.02 mg·kg−1, and RPD = 3.89). Most samples were located very close to the fit curve, indicating the model’s low prediction errors.
As shown in Figure 9a, the area covered by the 125 samples was very moderate—not too large to cover the entire city and not too small to cover just a tiny part of it. These findings were consistent with the “wave-like pattern” discussed in the RPD section: it is crucial to determine the right number of nearby samples—not too few and not too many. Thus, the optimal number of nearby samples should strike a balance, being neither too large nor too small, and should cover a moderate area.

4. Discussion

4.1. Estimation of Soil Cu Content in Urban Land by Vis-NIR Spectroscopy

In recent years, many researchers have used vis-NIR spectroscopy to measure soil heavy metal content. Karami et al. (2024) [53] estimated manganese (Mn), iron (Fe), nickel (Ni), zinc (Zn), and Cu using vis-NIR spectroscopy. Zhou et al. (2024) [54] estimated soil Mn with an R p 2 of 0.76. Krzebietke et al. (2023) [17] estimated soil cadmium (Cd), Cu, lead (Pb), Ni, chromium (Cr), Zn, Mn, and Fe. This shows that vis-NIR spectroscopy is a feasible method for estimating soil heavy metals. Our study successfully estimated soil heavy metal Cu in urban land with high accuracy ( R p 2 = 0.93, RMSEP = 4.02 mg·kg−1, and RPD = 3.89, Figure 9b). Similar results have been reported by others, with R p 2 values of 0.92 [17], 0.90 [55], and 0.82 [53]. However, these studies covered a very narrow range of Cu content (1.12–6.14 mg·kg−1, 0.89–6.18 mg·kg−1, 0.50–2.70 mg·kg−1). Our study covered a much broader range of Cu content (20.45 to 103.24 mg·kg−1, Figure 5) and included more samples. In contrast, some researchers obtained much worse results, such as R p 2 values of 0.5 [56], 0.26 [8], and 0.01 [57]. This may be due to the large range of Cu range (34.25–91.40 mg·kg−1, 69.3–119 mg·kg−1, 2.7–780 mg·kg−1). These studies also covered a large area, with samples located 80 to 200 km apart. In contrast, Khosravi et al. (2020) [16] focused on a small area (about 1 km2) with a broader range of Cu content (322–5099 mg·kg−1) and achieved a high R p 2 value of 0.86. Thus, when covering a large geographic area and range of Cu content, it is feasible to consider spatially nearby samples, like in our study, to improve the Cu estimation model.
The most important wavelengths for PLSR were 368–541 nm, 755–784 nm, 1156–1804 nm, 2143–2161 nm, and 2486–2498 nm, showing a high correlation coefficient (Figure 10). Specifically, wavelengths such as 368–541 nm, 755–784 nm, 1750–1804 nm, and 2143–2161 nm were identified as crucial for soil organic matter [58,59]. Previous studies have noted significant overlap in important wavelengths between soil heavy metals and soil organic matter [8,9]. This overlap is primarily due to Cu’s minimal or nonexistent response in the vis-NIR spectroscopy region, depending instead on other chemical bonds such as F e 2 O 3 , O–H, and C–H [60]. Many researchers have found the relationship between heavy metals and soil organic matter [15]. The 1400 nm was likely attributed to water. Additionally, 2486–2498 nm was relative to clay minerals and oxides [53].

4.2. The Influence of Spatially Nearby Samples on Soil Cu Estimation Model

Compared to traditional methods that did not consider spatially nearby samples, taking nearby samples into account greatly improved the Cu estimation model. The R p 2 increased from 0.75 to 0.93, the RMSEP decreased from 7.75 to 4.02 mg·kg−1, and the RPD increased from 1.73 to 3.89 (Figure 7 and Figure 8). The improvement was clear. This is because nearby samples have more in common than distant samples, making the model more reliable. In geography, distance plays an important role in the similarity of soil properties between samples [61]. Many studies suggest using nearby samples [27,40,62], but few actually do this. This may be due to the complexity of identifying each sample’s nearby samples and building a specific model for the sample. Our study took several weeks to complete the calculations, which is much longer than using a constant model that does not consider nearby samples.
When studying the spatial influence on soil properties, some researchers used geographical zones [37] or geographical subsets [38]. These methods divide samples into several areas, as shown in Figure 11, and then build a model for each subset. This approach reduces sample diversity and improves the model. It is simpler and requires less calculation time than our study. However, there were two main drawbacks: (i) a validation sample may be located on the edge of the subset, like Subset 1 and Subset 3 in Figure 11, making it less similar to other samples in that subset, and (ii) the number of samples in each subset can vary greatly, with some having too few and others having too many. Our study overcomes these limitations perfectly. Each validation sample is located centrally and surrounded by similar samples (Figure 11). Additionally, the number of nearby samples is consistent for all validation samples. Some research also used subsets based on landscape [34], soil type [23,63,64], or parental materials [65], which is similar to using geographical subsets. It is clear that using spatially nearby samples has many advantages.
The Cu estimation model’s performance varied based on the number of spatially nearby samples selected (Figure 8). Significant improvements were observed when increasing the nearby samples from 20 to 70, but the rate of improvement slowed between 70 and 125 samples and remained stable above 125. Therefore, the benefit of using spatially nearby samples is greatest when their number is small but diminishes with too many nearby samples. Other researchers have also found that increasing the calibration set size beyond a certain point does not improve the model performance [14,66] and can even lead to a decrease [37,67]. Thus, it is crucial to determine the optimal number of nearby samples.
The change in the number of nearby samples reveals a significant factor: the spatial distance between validation and calibration samples (Figure 12). Figure 12 illustrates how the mean distance between these samples varies with the number of nearby samples, increasing linearly. This trend results from our sampling strategy, which divides the study area into 2 × 2 km grids (Section 2.1). However, this linear increase in distance does not correspond to a linear increase in model performance. Instead, the model performance follows an inverted U-shaped curve—initially increasing and then decreasing. This indicates that spatial influence is strong at closer distances but diminishes over longer distances [68,69]. Therefore, it is important to investigate the limited range of spatial influence.
The optimal number of nearby samples is 125, which is 62.5% of the total calibration samples (Figure 9). The Cu estimation model with 125 nearby samples has high accuracy ( R p 2 = 0.93, RMSEP = 4.02 mg·kg−1, RPD = 3.89). Too few samples are insufficient to build a reliable model, while too many degrade the model’s performance [14,67]. When selecting 125 nearby samples, the mean distance between validation and calibration samples is 17 km (Figure 13). The red circle with a 17 km radius adequately covers the study area, being neither too big nor too small. Therefore, this distance is moderate and appropriate for the study area. In summary, the number of nearby samples should be balanced to cover a moderate area without there being too few or too many.
Our studies greatly improve the soil Cu estimation by using spatially nearby samples. However, there are still limitations that need further investigation. For example, while we applied our method in Shenzhen City, we were unsure if it would work in other cities or regions. We focused on heavy metals; it remains to be seen if this method is effective for other soil properties. Additionally, using other multivariate analysis methods, such as deep learning, may result in different performance for nearby samples compared to PLSR. These limitations present opportunities for further study and analysis.

5. Conclusions

This study explored using spatially nearby samples to estimate soil Cu in urban areas with vis-NIR spectroscopy. Our results lead to the following conclusions: (i) Using spatially nearby samples significantly improves the Cu estimation model compared to traditional methods. (ii) As the number of nearby samples increases, the performance of the Cu estimation model follows an inverted U-shaped curve, initially increasing and then declining. (iii) The optimal number of nearby samples should strike a balance, covering a moderate area without there being too few or too many. We found that using 125 nearby samples, or 62.5% of the total calibration samples, is optimal.
We successfully used spatially nearby samples to estimate soil Cu with high accuracy. However, more research is still needed on estimating soil Cu, such as using machine learning methods. While our study focuses on soil heavy metals in urban soils, we should also explore how this approach works in agricultural land and other soil properties under different environmental conditions.

Author Contributions

Conceptualization, Y.L. and Y.C.; methodology, T.S.; software, Z.L.; validation, K.G.; formal analysis, Y.L.; investigation, T.S. and K.G.; resources, Y.C.; data curation, T.S.; writing—original draft preparation, Y.C. and Y.L.; writing—review and editing, Y.C. and Z.L.; visualization, Y.L.; supervision, C.Y.; project administration, Y.L.; funding acquisition, Y.L. and Y.C. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Key Research and Development Program of China (Grant No. 2022YFB3903302), Guangzhou Science and Technology Plan Project (202102020416), Philosophy and Social Sciences Fund of the 13th Five-year Plan of Guangdong Province of China (GD20YGL11), Guangdong Basic and Applied Basic Research Foundation (2024A1515010110), and the Key R&D projects in Hubei Province (Grant No. 2021BCA220).

Data Availability Statement

The original contributions presented in the study are included in the article; further inquiries can be directed to the corresponding author.

Acknowledgments

We express our gratitude to the reviewers for offering valuable comments that have enhanced the quality of this paper. We also want to extend our significant appreciation to all the colleagues who provided essential assistance in this work.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Liu, Y.R.; van der Heijden, M.G.A.; Riedo, J.; Sanz-Lazaro, C.; Eldridge, D.J.; Bastida, F.; Moreno-Jiménez, E.; Zhou, X.-Q.; Hu, H.-W.; He, J.-Z.; et al. Soil contamination in nearby natural areas mirrors that in urban greenspaces worldwide. Nat. Commun. 2023, 14, 1706. [Google Scholar] [CrossRef]
  2. Schmidt-Traub, G.; Kroll, C.; Teksoz, K.; Durand-Delacre, D.; Sachs, J.D. National baselines for the Sustainable Development Goals assessed in the SDG Index and Dashboards. Nat. Geosci. 2017, 10, 547. [Google Scholar] [CrossRef]
  3. Hou, D.; O’Connor, D.; Igalavithana, A.D.; Alessi, D.S.; Luo, J.; Tsang, D.C.; Sparks, D.L.; Yamauchi, Y.; Rinklebe, J.; Ok, Y.S. Metal contamination and bioremediation of agricultural soils for food safety and sustainability. Nat. Rev. Earth Environ. 2020, 1, 366–381. [Google Scholar] [CrossRef]
  4. Khan, S.; Naushad, M.; Lima, E.C.; Zhang, S.X.; Shaheen, S.M.; Rinklebe, J. Global soil pollution by toxic elements: Current status and future perspectives on the risk assessment and remediation strategies—A review. J. Hazard. Mater. 2021, 417, 126039. [Google Scholar] [CrossRef]
  5. Luo, X.S.; Xue, Y.; Wang, Y.L.; Cang, L.; Xu, B.; Ding, J. Source identification and apportionment of heavy metals in urban soil profiles. Chemosphere 2015, 127, 152–157. [Google Scholar] [CrossRef]
  6. O’Riordan, R.; Davies, J.; Stevens, C.; Quinton, J.N.; Boyko, C. The ecosystem services of urban soils: A review. Geoderma 2021, 395, 115076. [Google Scholar] [CrossRef]
  7. Zhao, F.J.; Ma, Y.; Zhu, Y.G.; Tang, Z.; McGrath, S.P. Soil Contamination in China: Current Status and Mitigation Strategies. Environ. Sci. Technol. 2015, 49, 750–759. [Google Scholar] [CrossRef]
  8. Cheng, H.; Shen, R.; Chen, Y.; Wan, Q.; Shi, T.; Wang, J.; Wan, Y.; Hong, Y.; Li, X. Estimating heavy metal concentrations in suburban soils with reflectance spectroscopy. Geoderma 2019, 336, 59–67. [Google Scholar] [CrossRef]
  9. Hong, Y.; Shen, R.; Cheng, H.; Chen, S.; Chen, Y.; Guo, L.; He, J.; Liu, Y.; Yu, L.; Liu, Y. Cadmium concentration estimation in peri-urban agricultural soils: Using reflectance spectroscopy, soil auxiliary information, or a combination of both? Geoderma 2019, 354, 113875. [Google Scholar] [CrossRef]
  10. Alengebawy, A.; Abdelkhalek, S.T.; Qureshi, S.R.; Wang, M.-Q. Heavy metals and pesticides toxicity in agricultural soil and plants: Ecological risks and human health implications. Toxics 2021, 9, 42. [Google Scholar] [CrossRef]
  11. Wang, C.; Yang, Z.; Yuan, X.; Browne, P.; Chen, L.; Ji, J. The influences of soil properties on Cu and Zn availability in soil and their transfer to wheat t (Triticum aestivum L.) in the Yangtze River delta region, China. Geoderma 2013, 193, 131–139. [Google Scholar] [CrossRef]
  12. Viscarra Rossel, R.A.; Behrens, T.; Ben-Dor, E.; Brown, D.; Demattê, J.; Shepherd, K.; Shi, Z.; Stenberg, B.; Stevens, A.; Adamchuk, V.; et al. A global spectral library to characterize the world’s soil. Earth-Sci. Rev. 2016, 155, 198–230. [Google Scholar] [CrossRef]
  13. Li, S.; Viscarra Rossel, R.A.; Webster, R. The cost-effectiveness of reflectance spectroscopy for estimating soil organic carbon. Eur. J. Soil Sci. 2022, 73, e13202. [Google Scholar] [CrossRef]
  14. Ji, W.; Li, S.; Chen, S.; Shi, Z.; Viscarra Rossel, R.A.; Mouazen, A.M. Prediction of soil attributes using the Chinese soil spectral library and standardized spectra recorded at field conditions. Soil Tillage Res. 2016, 155, 492–500. [Google Scholar] [CrossRef]
  15. Gholizadeh, A.; Borůvka, L.; Saberioon, M.M.; Kozák, J.; Vašát, R.; Němeček, K. Comparing different data preprocessing methods for monitoring soil heavy metals based on soil spectral features. Soil Water Res. 2015, 10, 218–227. [Google Scholar] [CrossRef]
  16. Khosravi, V.; Ardejani, F.D.; Aryafar, A.; Yousefi, S.; Karami, S. Prediction of copper content in waste dump of Sarcheshmeh copper mine using visible and near-infrared reflectance spectroscopy. Environ. Earth Sci. 2020, 79, 165. [Google Scholar] [CrossRef]
  17. Krzebietke, S.; Daszykowski, M.; Czarnik-Matusewicz, H.; Stanimirova, I.; Pieszczek, L.; Sienkiewicz, S.; Wierzbowska, J. Monitoring the concentrations of Cd, Cu, Pb, Ni, Cr, Zn, Mn and Fe in cultivated Haplic Luvisol soils using near-infrared reflectance spectroscopy and chemometrics. Talanta 2023, 251, 123749. [Google Scholar] [CrossRef]
  18. Shi, T.; Liu, H.; Wang, J.; Chen, Y.; Fei, T.; Wu, G. Monitoring arsenic contamination in agricultural soils with reflectance spectroscopy of rice plants. Environ. Sci. Technol. 2014, 48, 6264–6272. [Google Scholar] [CrossRef]
  19. Xu, D.Y.; Chen, S.C.; Xu, H.Y.; Wang, N.; Zhou, Y.; Shi, Z. Data fusion for the measurement of potentially toxic elements in soil using portable spectrometers. Environ. Pollut. 2020, 263, 114649. [Google Scholar] [CrossRef]
  20. Wang, J.Z.; Hu, X.J.; Shi, T.Z.; He, L.; Hu, W.F.; Wu, G.F. Assessing toxic metal chromium in the soil in coal mining areas via proximal sensing: Prerequisites for land rehabilitation and sustainable development. Geoderma 2022, 405, 115399. [Google Scholar] [CrossRef]
  21. Liu, Y.; Shi, T.; Lan, Z.; Guo, K.; Zhuang, D.; Zhang, X.; Liang, X.; Qiu, T.; Zhang, S.; Chen, Y. Estimating the Soil Copper Content of Urban Land in a Megacity Using Piecewise Spectral Pretreatment. Land 2024, 13, 517. [Google Scholar] [CrossRef]
  22. Horta, A.; Malone, B.; Stockmann, U.; Minasny, B.; Bishop, T.F.A.; McBratney, A.B.; Pallasser, R.; Pozza, L. Potential of integrated field spectroscopy and spatial analysis for enhanced assessment of soil contamination: A prospective review. Geoderma 2015, 241, 180–209. [Google Scholar] [CrossRef]
  23. Liu, Y.; Shi, Z.; Zhang, G.; Chen, Y.; Li, S.; Hong, Y.; Shi, T.; Wang, J.; Liu, Y. Application of Spectrally Derived Soil Type as Ancillary Data to Improve the Estimation of Soil Organic Carbon by Using the Chinese Soil Vis-NIR Spectral Library. Remote Sens. 2018, 10, 1747. [Google Scholar] [CrossRef]
  24. Bao, Y.L.; Meng, X.T.; Ustin, S.; Wang, X.; Zhang, X.L.; Liu, H.J.; Tang, H.T. Vis-SWIR spectral prediction model for soil organic matter with different grouping strategies. Catena 2020, 195, 104703. [Google Scholar] [CrossRef]
  25. Tsakiridis, N.L.; Keramaris, K.D.; Theocharis, J.B.; Zalidis, G.C. Simultaneous prediction of soil properties from VNIR-SWIR spectra using a localized multi-channel 1-D convolutional neural network. Geoderma 2020, 367, 114208. [Google Scholar] [CrossRef]
  26. Chen, S.; Arrouays, D.; Mulder, V.L.; Poggio, L.; Minasny, B.; Roudier, P.; Libohova, Z.; Lagacherie, P.; Shi, Z.; Hannam, J.; et al. Digital mapping of GlobalSoilMap soil properties at a broad scale: A review. Geoderma 2022, 409, 115567. [Google Scholar] [CrossRef]
  27. Viscarra Rossel, R.A.; Shen, Z.F.; Lopez, L.R.; Behrens, T.; Shi, Z.; Wetterlind, J.; Sudduth, K.A.; Stenberg, B.; Guerrero, C.; Gholizadeh, A.; et al. An imperative for soil spectroscopic modelling is to think global but fit local with transfer learning. Earth-Sci. Rev. 2024, 254, 104797. [Google Scholar] [CrossRef]
  28. Song, J.H.; Shi, X.Y.; Wang, H.J.; Lv, X.; Zhang, W.X.; Wang, J.G.; Li, T.S.; Li, W.D. Combination of feature selection and geographical stratification increases the soil total nitrogen estimation accuracy based on vis-NIR and pXRF spectral fusion. Comput. Electron. Agric. 2024, 218, 108636. [Google Scholar] [CrossRef]
  29. Khosravi, V.; Gholizadeh, A.; Zizala, D.; Kodesová, R.; Saberioon, M.; Agyeman, P.C.; Vokurková, P.; Juricová, A.; Spasic, M.; Boruvka, L. On the impact of soil texture on local scale organic carbon quantification: From airborne to spaceborne sensing domains. Soil Tillage Res. 2024, 241, 106125. [Google Scholar] [CrossRef]
  30. Hong, Y.S.; Chen, Y.Y.; Chen, S.C.; Shen, R.L.; Guo, L.; Liu, Y.L.; Mouazen, A.M.; Shi, Z. Improving spectral estimation of soil inorganic carbon in urban and suburban areas by coupling continuous wavelet transform with geographical stratification. Geoderma 2023, 430, 116284. [Google Scholar] [CrossRef]
  31. Nocita, M.; Stevens, A.; Toth, G.; Panagos, P.; van Wesemael, B.; Montanarella, L. Prediction of soil organic carbon content by diffuse reflectance spectroscopy using a local partial least square regression approach. Soil Biol. Biochem. 2014, 68, 337–347. [Google Scholar] [CrossRef]
  32. Tziolas, N.; Tsakiridis, N.; Ben-Dor, E.; Theocharis, J.; Zalidis, G. A memory-based learning approach utilizing combined spectral sources and geographical proximity for improved VIS-NIR-SWIR soil properties estimation. Geoderma 2019, 340, 11–24. [Google Scholar] [CrossRef]
  33. Liu, Y.; Shi, T.; Chen, Y.; Lan, Z.; Guo, K.; Zhuang, D.; Yang, C.; Zhang, W. Monitoring the Soil Copper of Urban Land with Visible and Near-Infrared Spectroscopy: Comparing Spectral, Compositional, and Spatial Similarities. Land 2024, 13, 1279. [Google Scholar] [CrossRef]
  34. Peng, Y.; Knadel, M.; Gislum, R.; Deng, F.; Norgaard, T.; de Jonge, L.W.; Moldrup, P.; Greve, M.H. Predicting soil organic carbon at field scale using a national soil spectral library. J. Near Infrared Spectrosc. 2013, 21, 213–222. [Google Scholar] [CrossRef]
  35. Viscarra Rossel, R.A.; Webster, R. Predicting soil properties from the Australian soil visible-near infrared spectroscopic database. Eur. J. Soil Sci. 2012, 63, 848–860. [Google Scholar] [CrossRef]
  36. Udelhoven, T.; Emmerling, C.; Jarmer, T. Quantitative analysis of soil chemical properties with diffuse reflectance spectrometry and partial least-square regression: A feasibility study. Plant Soil 2003, 251, 319–329. [Google Scholar] [CrossRef]
  37. Shi, Z.; Ji, W.; Viscarra Rossel, R.A.; Chen, S.; Zhou, Y. Prediction of soil organic matter using a spatially constrained local partial least squares regression and the Chinese vis–NIR spectral library. Eur. J. Soil Sci. 2015, 66, 679–687. [Google Scholar] [CrossRef]
  38. Vohland, M.; Ludwig, B.; Seidel, M.; Hutengs, C. Quantification of soil organic carbon at regional scale: Benefits of fusing vis-NIR and MIR diffuse reflectance data are greater for in situ than for laboratory-based modelling approaches. Geoderma 2022, 405, 115426. [Google Scholar] [CrossRef]
  39. Ramirez-Lopez, L.; Behrens, T.; Schmidt, K.; Stevens, A.; Demattê, J.A.M.; Scholten, T. The spectrum-based learner: A new local approach for modeling soil vis–NIR spectra of complex datasets. Geoderma 2013, 195, 268–279. [Google Scholar] [CrossRef]
  40. Guo, L.; Zhao, C.; Zhang, H.; Chen, Y.; Linderman, M.; Zhang, Q.; Liu, Y. Comparisons of spatial and non-spatial models for predicting soil carbon content based on visible and near-infrared spectral technology. Geoderma 2017, 285, 280–292. [Google Scholar] [CrossRef]
  41. Summerauer, L.; Baumann, P.; Ramirez-Lopez, L.; Barthel, M.; Bauters, M.; Bukombe, B.; Reichenbach, M.; Boeckx, P.; Kearsley, E.; Van Oost, K.; et al. The central African soil spectral library: A new soil infrared repository and a geographical prediction analysis. Soil 2021, 7, 693–715. [Google Scholar] [CrossRef]
  42. Dorantes, M.J.; Fuentes, B.A.; Miller, D.M. Calibration set optimization and library transfer for soil carbon estimation using soil spectroscopy—A review. Soil Sci. Soc. Am. J. 2022, 86, 879–903. [Google Scholar] [CrossRef]
  43. Shi, T.; Hu, Z.; Shi, Z.; Guo, L.; Chen, Y.; Li, Q.; Wu, G. Geo-detection of factors controlling spatial patterns of heavy metals in urban topsoil using multi-source data. Sci. Total Environ. 2018, 643, 451–459. [Google Scholar] [CrossRef]
  44. Duan, D.Y.; Wang, P.; Rao, X.; Zhong, J.H.; Xiao, M.H.; Huang, F.; Xiao, R.B. Identifying interactive effects of spatial drivers in soil heavy metal pollutants using interpretable machine learning models. Sci. Total Environ. 2024, 934, 173284. [Google Scholar] [CrossRef] [PubMed]
  45. Lin, T.; Zhao, S.H.; Xi, X.P.; Yang, K.; Luo, F. Environmental Background Values of Heavy Metals and Physicochemical Properties in Different Soils in Shenzhen. Environ. Sci. 2021, 42, 3518–3526. [Google Scholar]
  46. Zhang, W.; Xu, A.; Zhang, R.; Ji, H. Review of Soil Classification and Revision of China Soil Classification System. Sci. Agric. Sin. 2014, 47, 3214–3230. [Google Scholar]
  47. Shi, T.Z.; Cui, L.J.; Wang, J.J.; Fei, T.; Chen, Y.Y.; Wu, G.F. Comparison of multivariate methods for estimating soil total nitrogen with visible/near-infrared spectroscopy. Plant Soil 2013, 366, 363–375. [Google Scholar] [CrossRef]
  48. Zhang, Z.Y.; Chen, Y.Y.; Wu, K.X.; Hong, Y.S.; Shi, T.Z.; Mouazen, A.M. On the parsimony, interpretability and predictive capability of a physically-based model in the optical domain for estimating soil moisture content. Geoderma 2024, 449, 116996. [Google Scholar] [CrossRef]
  49. Lindsay, W.L.; Norvell, W.A. Development of a DTPA Soil Test for Zinc, Iron, Manganese, and Copper1. Soil Sci. Soc. Am. J. 1978, 42, 421–428. [Google Scholar] [CrossRef]
  50. Mancini, M.; Andrade, R.; Silva, S.H.G.; Rafael, R.B.A.; Mukhopadhyay, S.; Li, B.; Chakraborty, S.; Guilherme, L.R.G.; Acree, A.; Weindorf, D.C.; et al. Multinational prediction of soil organic carbon and texture via proximal sensors. Soil Sci. Soc. Am. J. 2024, 88, 8–26. [Google Scholar] [CrossRef]
  51. Hu, B.F.; Ni, H.J.; Xie, M.D.; Li, H.Y.; Wen, Y.L.; Chen, S.C.; Zhou, Y.; Teng, H.F.; Bourennane, H.; Shi, Z. Mapping soil organic matter and identifying potential controls in the farmland of Southern China: Integration of multi-source data, machine learning and geostatistics. Land Degrad. Dev. 2023, 34, 5468–5485. [Google Scholar] [CrossRef]
  52. Guo, B.; Zhang, B.; Su, Y.; Zhang, D.; Wang, Y.; Bian, Y.; Suo, L.; Guo, X.; Bai, H. Retrieving zinc concentrations in topsoil with reflectance spectroscopy at Opencast Coal Mine sites. Sci. Rep. 2021, 11, 19909. [Google Scholar] [CrossRef]
  53. Karami, A.; Moosavi, A.A.; Pourghasemi, H.R.; Ronaghi, A.; Ghasemi-Fasaei, R.; Vidal, E.; Lado, M. Proximal sensing approach for characterization of calcareous soils using multiblock data analysis. Geoderma Reg. 2024, 36, e00752. [Google Scholar] [CrossRef]
  54. Zhou, M.; Hu, T.; Wu, M.T.; Ma, C.D.; Qi, C.C. Rapid estimation of soil Mn content by machine learning and soil spectra in large-scale. Ecol. Inform. 2024, 81, 102615. [Google Scholar] [CrossRef]
  55. Gozukara, G.; Acar, M.; Ozlu, E.; Dengiz, O.; Hartemink, A.E.; Zhang, Y. A soil quality index using Vis-NIR and pXRF spectra of a soil profile. Catena 2022, 211, 105954. [Google Scholar] [CrossRef]
  56. Nawar, S.; Mohamed, E.S.; Sayed, S.E.E.; Mohamed, W.S.; Rebouh, N.Y.; Hammam, A.A. Estimation of key potentially toxic elements in arid agricultural soils using Vis-NIR spectroscopy with variable selection and PLSR algorithms. Front. Environ. Sci. 2023, 11, 1222871. [Google Scholar] [CrossRef]
  57. Riedel, F.; Denk, M.; Müller, I.; Barth, N.; Glässer, C. Prediction of soil parameters using the spectral range between 350 and 15,000 nm: A case study based on the Permanent Soil Monitoring Program in Saxony, Germany. Geoderma 2018, 315, 188–198. [Google Scholar] [CrossRef]
  58. Viscarra Rossel, R.A.; Behrens, T. Using data mining to model and interpret soil diffuse reflectance spectra. Geoderma 2010, 158, 46–54. [Google Scholar] [CrossRef]
  59. Li, S.; Shi, Z.; Chen, S.; Ji, W.; Zhou, L.; Yu, W.; Webster, R. In situ measurements of organic carbon in soil profiles using vis-NIR spectroscopy on the Qinghai–Tibet plateau. Environ. Sci. Technol. 2015, 49, 4980–4987. [Google Scholar] [CrossRef]
  60. Shi, T.; Chen, Y.; Liu, Y.; Wu, G. Visible and near-infrared reflectance spectroscopy—An alternative for monitoring soil contamination by heavy metals. J. Hazard. Mater. 2014, 265, 166–176. [Google Scholar] [CrossRef]
  61. Li, J.K.; Xu, H.; Song, Y.P.; Tang, L.L.; Gong, Y.B.; Yu, R.L.; Shen, L.; Wu, X.L.; Liu, Y.D.; Zeng, W.M. Geography Plays a More Important Role than Soil Composition on Structuring Genetic Variation of Pseudometallophyte Commelina communis. Front. Plant Sci. 2016, 7, 1085. [Google Scholar] [CrossRef] [PubMed]
  62. Wang, Z.; Chen, S.C.; Lu, R.; Zhang, X.L.; Ma, Y.X.; Shi, Z. Non-linear memory-based learning for predicting soil properties using a regional vis-NIR spectral library. Geoderma 2024, 441, 116752. [Google Scholar] [CrossRef]
  63. McDowell, M.L.; Bruland, G.L.; Deenik, J.L.; Grunwald, S. Effects of subsetting by carbon content, soil order, and spectral classification on prediction of soil total carbon with diffuse reflectance spectroscopy. Appl. Environ. Soil Sci. 2012, 2012, 294121. [Google Scholar] [CrossRef]
  64. Vasques, G.M.; Grunwald, S.; Harris, W.G. Spectroscopic models of soil organic carbon in Florida, USA. J. Environ. Qual. 2010, 39, 923–934. [Google Scholar] [CrossRef] [PubMed]
  65. Xu, S.; Shi, X.; Wang, M.; Zhao, Y. Effects of subsetting by parent materials on prediction of soil organic matter content in a hilly area using Vis–NIR spectroscopy. PLoS ONE 2016, 11, e0151536. [Google Scholar] [CrossRef] [PubMed]
  66. Debaene, G.; Niedzwiecki, J.; Pecio, A.; Zurek, A. Effect of the number of calibration samples on the prediction of several soil properties at the farm-scale. Geoderma 2014, 214, 114–125. [Google Scholar] [CrossRef]
  67. Liu, Y.; Liu, Y.; Chen, Y.; Zhang, Y.; Shi, T.; Wang, J.; Hong, Y.; Fei, T. The Influence of Spectral Pretreatment on the Selection of Representative Calibration Samples for Soil Organic Matter Estimation Using Vis-NIR Reflectance Spectroscopy. Remote Sens. 2019, 11, 450. [Google Scholar] [CrossRef]
  68. Ettema, C.H.; Wardle, D.A. Spatial soil ecology. Trends Ecol. Evol. 2002, 17, 177–183. [Google Scholar] [CrossRef]
  69. Wei, Y.; Chen, Y.; Wang, J.; Wang, B.; Yu, P.; Hong, Y.; Zhu, L. Unveiling the explanatory power of environmental variables in soil organic carbon mapping: A global–local analysis framework. Geoderma 2024, 449, 117011. [Google Scholar] [CrossRef]
Sensors 24 05612 g001
Figure 1. Location of the sampling sites.
Figure 1. Location of the sampling sites.
Sensors 24 05612 g001
Sensors 24 05612 g002
Figure 2. Equipment setup for spectral measurements.
Figure 2. Equipment setup for spectral measurements.
Sensors 24 05612 g002
Sensors 24 05612 g003
Figure 3. A validation sample with 20 and 50 spatially nearby samples used for building the Cu estimation model. The green circles denote samples that were not selected as nearby samples.
Figure 3. A validation sample with 20 and 50 spatially nearby samples used for building the Cu estimation model. The green circles denote samples that were not selected as nearby samples.
Sensors 24 05612 g003
Sensors 24 05612 g004
Figure 4. Flowchart of using spatially nearby samples.
Figure 4. Flowchart of using spatially nearby samples.
Sensors 24 05612 g004
Sensors 24 05612 g005
Figure 5. Boxplot and histogram of Cu content for calibration samples (a) and validation samples (b). Repoint (·) denotes the mean value. The blue line (|) denotes the median value. Hollow circle (○) denotes the outliers. The black box denotes the interquartile range.
Figure 5. Boxplot and histogram of Cu content for calibration samples (a) and validation samples (b). Repoint (·) denotes the mean value. The blue line (|) denotes the median value. Hollow circle (○) denotes the outliers. The black box denotes the interquartile range.
Sensors 24 05612 g005
Sensors 24 05612 g006
Figure 6. Land use and sample distribution in Shenzhen city.
Figure 6. Land use and sample distribution in Shenzhen city.
Sensors 24 05612 g006
Sensors 24 05612 g007
Figure 7. Soil Cu content between predicted and measured values using spectroscopy models without considering spatially nearby samples. R p 2 denotes coefficient of determination in prediction. RMSEP denotes the root mean square error of prediction. RPD denotes the residual predictive deviation.
Figure 7. Soil Cu content between predicted and measured values using spectroscopy models without considering spatially nearby samples. R p 2 denotes coefficient of determination in prediction. RMSEP denotes the root mean square error of prediction. RPD denotes the residual predictive deviation.
Sensors 24 05612 g007
Sensors 24 05612 g008
Figure 8. Performance of soil Cu estimation model considering different numbers of spatial nearby samples. (a) R p 2 , denotes coefficient of determination in prediction. (b) RMSEP, denotes the root mean square error of prediction. (c) RPD, denotes the residual predictive deviation. The dotted blue line is the fitting line.
Figure 8. Performance of soil Cu estimation model considering different numbers of spatial nearby samples. (a) R p 2 , denotes coefficient of determination in prediction. (b) RMSEP, denotes the root mean square error of prediction. (c) RPD, denotes the residual predictive deviation. The dotted blue line is the fitting line.
Sensors 24 05612 g008
Sensors 24 05612 g009
Figure 9. Performance of soil Cu estimation model when the number of spatial nearby samples is 125. R p 2 denotes coefficient of determination in prediction. RMSEP denotes the root mean square error of prediction. RPD denotes the residual predictive deviation. (a) The selected 125 nearby samples. (b) The performance of the Cu estimation model.
Figure 9. Performance of soil Cu estimation model when the number of spatial nearby samples is 125. R p 2 denotes coefficient of determination in prediction. RMSEP denotes the root mean square error of prediction. RPD denotes the residual predictive deviation. (a) The selected 125 nearby samples. (b) The performance of the Cu estimation model.
Sensors 24 05612 g009
Sensors 24 05612 g010
Figure 10. Correlation between Cu concentration and spectral wavelengths from 350 to 2500 nm. The blue line denotes the Pearson correlation coefficient. The dotted line denotes the threshold for important wavelengths.
Figure 10. Correlation between Cu concentration and spectral wavelengths from 350 to 2500 nm. The blue line denotes the Pearson correlation coefficient. The dotted line denotes the threshold for important wavelengths.
Sensors 24 05612 g010
Sensors 24 05612 g011
Figure 11. Examples of spatially nearby samples and geographic subsets.
Figure 11. Examples of spatially nearby samples and geographic subsets.
Sensors 24 05612 g011
Sensors 24 05612 g012
Figure 12. The mean distance between validation and calibration samples when selecting different numbers of nearby samples.
Figure 12. The mean distance between validation and calibration samples when selecting different numbers of nearby samples.
Sensors 24 05612 g012
Sensors 24 05612 g013
Figure 13. The mean distance (17 km) between the validation and calibration sample when selecting 125 nearby samples. The red circle has a radius of 17 km.
Figure 13. The mean distance (17 km) between the validation and calibration sample when selecting 125 nearby samples. The red circle has a radius of 17 km.
Sensors 24 05612 g013
Table 1. The descriptive statistics of 250 soil samples for the calibration and validation sets.
Table 1. The descriptive statistics of 250 soil samples for the calibration and validation sets.
SampleNumberCu (mg·kg−1)
MinMaxMedianMeanStd 1CV 2SkewnessKurtosisBackgroundPollution Level
Total25020.45103.2459.4458.2915.570.270.130.1217.0036.00
Calibration20020.45103.2459.4458.2915.600.270.130.15
Validation5025.2197.0659.1858.3015.630.270.130.12
1 Std denotes standard deviation. 2 CV denotes coefficient of variation.
Table 2. Summary statistics for soil Cu estimation models using spatially nearby samples.
Table 2. Summary statistics for soil Cu estimation models using spatially nearby samples.
Number of Nearby SamplesCalibrationValidation
R c v 2 R M S E c v R p 2 RMSEPRPD
None0.649.720.758.561.83
20--0.757.752.01
50--0.904.903.14
100--0.924.563.43
150--0.934.003.90
200--0.924.343.60
Note: R c v 2 denotes coefficient of determination in cross-validation. R M S E c v denotes root mean square error in cross-validation. R p 2 denotes coefficient of determination in prediction. RMSEP denotes root mean square error of prediction. RPD denotes the residual predictive deviation. LV denotes latent variable.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Liu, Y.; Shi, T.; Lan, Z.; Guo, K.; Yang, C.; Chen, Y. Monitoring Soil Copper in Urban Land Using Visibale and Near-Infrared Spectroscopy with Spatially Nearby Samples. Sensors 2024, 24, 5612. https://doi.org/10.3390/s24175612

AMA Style

Liu Y, Shi T, Lan Z, Guo K, Yang C, Chen Y. Monitoring Soil Copper in Urban Land Using Visibale and Near-Infrared Spectroscopy with Spatially Nearby Samples. Sensors. 2024; 24(17):5612. https://doi.org/10.3390/s24175612

Chicago/Turabian Style

Liu, Yi, Tiezhu Shi, Zeying Lan, Kai Guo, Chao Yang, and Yiyun Chen. 2024. "Monitoring Soil Copper in Urban Land Using Visibale and Near-Infrared Spectroscopy with Spatially Nearby Samples" Sensors 24, no. 17: 5612. https://doi.org/10.3390/s24175612

APA Style

Liu, Y., Shi, T., Lan, Z., Guo, K., Yang, C., & Chen, Y. (2024). Monitoring Soil Copper in Urban Land Using Visibale and Near-Infrared Spectroscopy with Spatially Nearby Samples. Sensors, 24(17), 5612. https://doi.org/10.3390/s24175612

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop