Open AccessTechnical Note

Analyzing the Application of X-Band Radar for Improving Rainfall Observation and Flood Forecasting in Yeongdong, South Korea

Seong-Sim Yoon

and

Sang-Hun Lim

Korea Institute of Civil Engineering and Building Technology, Goyang-Si 10223, Korea

Author to whom correspondence should be addressed.

Remote Sens. 2022, 14(1), 43; https://doi.org/10.3390/rs14010043

Submission received: 30 October 2021 / Revised: 18 December 2021 / Accepted: 21 December 2021 / Published: 23 December 2021

(This article belongs to the Special Issue Advance of Radar Meteorology and Hydrology)

Download

Browse Figures

Figure 1
Accumulated rainfall distribution from 15 to 16 August 2019. (a) Distributed rainfall field by ground rain-gauge network; (b) Rainfall field by S-band radar network. "> Figure 2
X-band dual-polarization radar location and hydrological observation network in the Samcheok region. "> Figure 3
Prior and post application of the distributed specific differential phase estimation method, and the rainfall estimation results (30 June 2020, at 2:18): (a) KDP (JRC); (b) Surface Rainfall (SAM_SRI); (c) KDP (new); (d) Rainfall (SAM_RR). "> Figure 4
Distribution of accumulated rainfall over 1 h period distribution for maximum rainfall in each rain event: (a) 3 October 2019, 00:00; (b) 30 June 2020, 03:00; (c) 24 July 2020, 07:00; (d) 3 September 2020, 06:00; (e) 7 September 2020, 11:00. "> Figure 4 Cont.
Distribution of accumulated rainfall over 1 h period distribution for maximum rainfall in each rain event: (a) 3 October 2019, 00:00; (b) 30 June 2020, 03:00; (c) 24 July 2020, 07:00; (d) 3 September 2020, 06:00; (e) 7 September 2020, 11:00. "> Figure 5
Scatterplots of observed ground rain gauge vs. various radar rainfall (red dots: rainfall in 10 km radius, black dots: rainfall in 40-km radius): (a) Event 1 (00:00–23:00 2–3 October 2019); (b) Event 2 (00:00–23:00 30 June 2020); (c) Event 3 (00:00–23:00 24 July 2020); (d) Event 4 (00:00–23:00 2–3 September 2020); (e) Event 5 (00:00–23:00 7 September 2020). "> Figure 5 Cont.
Scatterplots of observed ground rain gauge vs. various radar rainfall (red dots: rainfall in 10 km radius, black dots: rainfall in 40-km radius): (a) Event 1 (00:00–23:00 2–3 October 2019); (b) Event 2 (00:00–23:00 30 June 2020); (c) Event 3 (00:00–23:00 24 July 2020); (d) Event 4 (00:00–23:00 2–3 September 2020); (e) Event 5 (00:00–23:00 7 September 2020). "> Figure 6
Simulated discharge and water level using rainfall scenarios for 360 min duration: (a) Rainfall scenarios; (b) Simulated discharge scenarios; (c) Simulated water level scenarios. "> Figure 7
The criteria of flood risk and flow nomograph. (a) Cross-section and criteria of flood risk; (b) Flow nomograph at station 12; (c) Stations of flow nomograph development. "> Figure 8
Mean areal precipitation of the Samcheok-osib stream (3 October 2019, at 01:50): (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR. "> Figure 9
Calculation results of the flood forecast information for October 3 2019, at 01:50: (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR. "> Figure 10
Observed water level and timeseries data of the mean real precipitation between 2 and 3 October 2019. "> Figure 11
Calculation results of the flood forecast information for 7 September 2020, at 08:30: (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR. "> Figure 12
Calculation results of the flood forecast information for 7 September 2020, at 10:30: (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR. "> Figure 13
Calculation results of the flood forecast information for 7 September 2020, at 10:50: (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR. "> Figure 14
Mean areal precipitation of the Samcheok-osib stream (7 September 2020, at 10:50): (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR. "> Figure 15
Observed water level and time-series data of the mean areal precipitation on 7 September 2020. ">

Versions Notes

Abstract

The mountainous Yeongdong region of South Korea contains mountains over 1 km. Owing to this topographic blockage, the region has a low-density rain-gauge network, and there is a low-altitude (~1.5 km) observation gap with the nearest large S-band radar. The Korean government installed an X-band dual-polarization radar in 2019 to improve rainfall observations and to prevent hydrological disasters in the Yeongdong region. The present study analyzed rainfall estimates using the newly installed X-band radar to evaluate its hydrological applicability. The rainfall was estimated using a distributed specific differential phase-based technique for a high-resolution 75 m grid. Comparison of the rainfall estimates of the X-band radar and the existing rainfall information showed that the X-band radar was less likely to underestimate rainfall compared to the S-band radar. The accuracy was particularly high within a 10 km observation radius. To evaluate the hydrological applicability of X-band radar rainfall estimates, this study developed a rain-based flood forecasting method—the flow nomograph—for the Samcheok-osib stream, which is vulnerable to heavy rain and resultant floods. This graph represents the flood risk level determined by hydrological–hydraulic modeling with various rainfall scenarios. Rainfall information (X-band radar, S-band radar, ground rain gauge) was applied as input to the flow nomograph to predict the flood level of the stream. Only the X-band radar could accurately predict the actual high-risk increase in the water level for all studied rainfall events.

Keywords:

rainfall estimation; X-band dual-polarization radar; distributed specific differential phase-based technique; flow nomograph; flood level prediction; disaster management

1. Introduction

In recent years, small X-band radars with a small coverage radius and high resolution have been employed to observe local weather phenomena in mountainous terrains and urban areas. Furthermore, X-band radar information has been used for flood forecasting [1,2,3,4,5]. The CASA radar observation network provides spatiotemporally high-resolution (0.5 km spatial and 1 min temporal resolutions) reflectivity data, which are essential for the application of nowcasting [6,7]. The CASA radar network for Houston, NETRAD, provides local quantitative precipitation estimates and can be integrated into real-time runoff models. End-users will be provided with a customized product to make informed decisions during severe weather events through the current flood alert version, the RiceU/TMC Flood Alert System (FAS2). FAS2 couples available radar data (NEXRAD) with real-time hydrologic models to predict peak flood levels with 2 to 3 h of lead time, and issues warnings via the internet to emergency personnel. It uniquely combines new urban hydrologic models with new and more accurate CASA radars (with rainfall data at the scale of a city block), which allows a notable advance in flash flood prediction in flood-prone areas such as Houston [8].

Most previous studies have used radar rainfall data with a rainfall–runoff model for flood predictions and warning systems [9,10,11]. However, rainfall–runoff models are impractical in an operational environment because of the time and expertise required, and river management authorities require a simple and practical operational method [12,13]. Bedient et al. [14] illustrated the design, operation, and performance of an advanced flood warning system developed for the Texas Medical Center, which uses the NEXRAD radar for hydrologic prediction in the Brays Bayou watershed of Houston, Texas. The basis of this flood warning system is a flow nomograph, i.e., a lookup table. A nomograph is a simple, graphical representation of the peak flows computed from observed rainfall intensities during a storm event.

The Ministry of Land, Infrastructure, and Transport and Tourism of Japan is building an X-band radar network, called XRAIN, which can make precise rainfall observations without correction from ground rain gauges. They developed a flood alert criteria nomograph for the Toga River in Kobe, Japan, and the nomograph with XRAIN rainfall provided accurate flood warning information [15]. Based on previous studies, we confirmed that a small X-band radar and a flood prediction technique based on a simple rainfall–flooding relationship are useful in areas that require rapid information because of a short hydrological response. Therefore, this study applied this concept to predict stream flooding in the Yeongdong area, which has a short concentration time (2–3 h) due to its topographical characteristics.

In South Korea, the Korea Meteorological Administration and the Flood Control Offices are currently operating 17 S-band dual-polarization radars to monitor atmospheric weather conditions and forecast floods. These radars have an observation radius of 150–240 km or greater and perform broadband observations. However, considering the complex mountainous terrain of South Korea and drastic differences in regional precipitation, the current radar observation network has limitations in providing accurate observations. In particular, the Yeongdong region, located on the eastern coast of South Korea, is surrounded by mountainous terrain. The density of the ground rain-gauge network is low in this region because of the topographic conditions. Moreover, the existing S-band radar, which is used to remedy the observational limitations of ground-gauge stations, cannot provide accurate rainfall observations in this region because of topographic shielding. Therefore, the Ministry of Environment has installed two new X-band dual-polarimetric radars on the eastern coast of South Korea. These radars can generate gridded rainfall data with a resolution of 75 m within a 40 km observation radius. The specific objectives of this study were to evaluate the rainfall measurements of the newly installed X-band radar and to evaluate its hydrological applicability for real operation.

2. Materials and Methods

2.1. Study Area and Data

This study covered Samcheok city in Gangwon Province on the eastern coast of South Korea, in the Yeongdong region. This region is adjacent to the Taebaek Mountains, which have a high altitude (~EL. 1567 m) and are topographically complex, and the East Sea. Hence, intense rainfall in summer and heavy snowfall in winter are common. Such precipitation phenomena are caused by the increased instability of air currents in the Taebaek Mountains and mechanisms such as the seeder–feeder caused by the inflow of the cold and humid air currents from the East Sea with the easterly wind [16]. Significant spatiotemporal variations in the dynamic and microphysical precipitation processes due to the topography of the Taebaek Mountains are observed in the Gangwon Province. Figure 1 shows the distribution of accumulated rainfall between 15 and 16 August 2019 based on observations from the rain-gauge station (left) and the existing composited S-band radar rainfall field (CMP_HFC) (right). This is one case that shows the limitations of rainfall observation in the Yeongdong region. Figure 1a shows the distributed rainfall field measured by the ground rain-gauge network via Barnes objective analysis in the studies by Koch et al. [17] and Lee et al. [18]. As shown in these distribution maps, the ground rain-gauge network observed over 300 mm of heavy rainfall in the East Sea region. However, the existing S-band radar network underestimated rainfall of 30 mm or less. Accurate rainfall estimation is difficult with heavy rain caused by the east wind and beam shielding. The Korean easterly inflow from the East Sea is an unusual weather phenomenon in Yeongdong, South Korea. Owing to the topographical characteristics, the east wind causes heavy rain as the air pressure and temperature decrease when the wind blows through highly mountainous areas.

The high-resolution X-band radar is useful for observing such areas in detail because a single X-band radar can cover the specific region using an independent operation. The Ministry of Environment of Korea has operated two small X-band dual-polarization radars (in Samcheok and Uljin) in the eastern coast region since 2018. These radars have a frequency of X band (9.4 GHz), an antenna size of 1.2 m, a beamwidth of 2°, and a transmitter with a low-power solid-state power amplifier. As shown in Figure 2, the X-band dual-polarization radar in Samcheok city was used to estimate rainfall. The 27 ground rain gauges within a 40 km effective observation radius of the Samcheok X-band dual-polarization radar are shown in Figure 2. The ground rain gauges refer to automatic weather stations (AWS) and the Ministry of Environment’s telemetry data (TM). Furthermore, the feasibility of flood forecasting was evaluated for the Samcheok-osib stream (purple area) located within an observation radius of 40 km. The target basin of this study is located in the east coast region, which is mountainous. The area of target basins is 394.2 km² and the stream length is 58.1 km. This area is flood-prone and has a risk of human casualties during heavy rains. The data observed by the newly installed Samcheok X-band dual-polarization radar were used to generate high-resolution gridded rainfall and mean areal precipitation. We developed a rainfall-based flood forecasting method, namely a flow nomograph, to predict the risk of stream flooding by considering the regional hydrological characteristics.

2.2. Methodology

2.2.1. Precipitation Estimation Algorithm Using X-Band Dual-Polarimetric Radar

The installed X-band dual-polarimetric radar generates dual-polarization variables (e.g., unfiltered reflectivity DZ, corrected reflectivity CZ, reflectivity at horizontal polarization transmit Z_h, differential reflectivity Z_DR, differential phase ψ_dp, specific differential phase K_DP, and correlation coefficient ρ_hv). This study used the Colorado State University (CSU) algorithm [19] for rainfall estimation among several radar rainfall estimation algorithms [20,21,22]. In the CSU–Hydrometeor Identification Rainfall Optimization (CSU-HIDRO) algorithm, precipitation echo is classified into snowfall, rainfall, and mixed particles through fuzzy logic in the hydrometeor classification process. Rainfall was estimated using different relations according to the range of the observed variables of the dual-polarization radar (e.g., Z_h, Z_DR, and K_DP). The specific differential phase, which is an important observed variable in rainfall estimation using dual-polarization variables, is not heavily affected by system deviation or hail.

The specific differential phase is conventionally calculated by applying filtering or regression analysis to the differential phase, which provides information about the shape and distribution density of the precipitation particles [23,24]. However, the peak value of the specific differential phase may be underestimated or have a negative value in the convective rainfall regions. Therefore, in this study, a specific differential phase calculation technique and a rainfall estimation technique suitable for the Samcheok X-band dual-polarization radar were developed by applying the distributed specific differential phase calculated according to the self-consistency-based method proposed by Lim et al. [25]. Horizontal reflectivity was corrected and used because the X-band radar has a relatively large signal attenuation. The process for rainfall estimation using a distributed specific differential phase is as follows: the K_DP distribution method uses the attenuation-corrected horizontal reflectivity (

{\hat{Z}}_{h})

and total ψ_dp in the rainfall region based on self-consistency. Here, the total ψ_dp refers to the difference in ψ_dp between the endpoint (r_m) and the starting point (r₀) of the rainfall region. r₀ and r_m indicate the beginning and range of precipitation echo. The rainfall region is classified using the standard deviation of ψ_dp and the correlation coefficient ρ_hv, which enables the identification of non-weather echoes. The proposed distributed K_DP can reduce the effect of ψ_dp triggered by ground clutter. This is because Z_h, which is less affected by ground clutter than ψ_dp, is used to calculate the K_DP. In addition, problems such as fluctuations of K_DP or the calculation of its negative values that may occur because of a significant change in ψ_dp can be solved because data from adjacent bins are not used [25]. Finally, the computed specific differential phase was used to calculate the rainfall based on the following R–K_DP relation (Equation (1)) [26]:

R (K_{d p}) = 23.7 K_{d p}^{0.87}

(1)

2.2.2. Flow Nomograph

A flow nomograph is based on the rainfall information required to secure the lead time to evacuation [14,15]. This nomograph contains the relationship between flood discharge and flood level from the rainfall scenarios (peak rainfall intensity and duration time) at a specific flood forecasting station, as in Equation (2). The procedure used to develop a nomograph can be classified into four parts: (1) setting up the reference flood level at the flood forecasting station; (2) setting up the rainfall hyetographs; (3) using the rainfall hyetographs to run and develop the hydrologic model and to estimate the flood discharge; and (4) developing the flow nomograph.

S t a g e_{i s o}^{k} = F l o w_{i s o}^{k} = f (P_{i n t e n s i t y}^{k}, P_{d u r a t i o n}^{k})

(2)

where k is the flood forecasting station,

P_{i n t e n s t y}^{k}

is the peak rainfall intensity, and

P_{d u r a t i o n}^{k}

is the rainfall duration in each rainfall scenario.

F l o w_{i s o}^{k}

is the isoflow of each flood forecasting reference calculated by the rainfall intensity and duration in each rainfall scenario.

S t a g e_{i s o}^{k}

is the isostage of each flood forecasting reference, which is simulated by hydraulic modeling. This stage was used to define the flood discharge range for the flood forecasting reference. This study estimated the

F l o w_{i s o}^{k}

using the rainfall–runoff model

S t a g e_{i s o}^{k}

, which was determined through hydraulic modeling with the simulated inflow from the rainfall–runoff model at each flood forecasting reference. The occurrence of a flood was determined using Equation (3):

S t a g e_{i s o}^{k} \leq S t a g e_{i s o}^{o b s} \to f l o o d f o r e c s e t

(3)

Equation (3) compares the predetermined

S t a g e_{i s o}^{k}

and

S t a g e_{i s o}^{o b s}

using the observed rainfall data (

P_{i n t e n s i t y}^{o b s}

P_{d u r a t i o n}^{o b s}

) for the flood forecasting value. Therefore, if we follow this concept, the result can be estimated by the peak rainfall intensity and duration time, which exceed the isoflow line of the nomograph. This mean is that there is a possibility of flood risk in flood forecasting.

3. Results and Discussion

3.1. Quantitative Precipitation Estimation

Figure 3 shows the precipitation estimation using a rainfall estimation technique based on a distributed specific differential phase for 30 June 2020, at 02:18 (Korea Standard Time, KST). Figure 3a,b show the specific differential phase K_DP (JRC) and surface rainfall (SRI) using the default settings of the radar manufacturer (Japan Radio Co., Ltd. Tokyo). Figure 3c,d show the newly estimated K_DP in this study and the rainfall estimated (SAM_RR) using this method. The X-band dual-polarization radar installed in Samcheok generates estimated rainfall (SAM_SRI) data through a module installed in the manufacturer’s system. However, because the estimated rainfall data were generated by applying the basic Z-R relationship (Z = 200 R^1.6) without quality control, the quantitative accuracy was low. Table 1 shows the information collected from the rain gauges and the radar data to evaluate the accuracy of this study. To reduce the influence of differences in temporal resolution, rainfall data were accumulated over a 1 h period for comparison [27].

Rainfall data within the effective radius (40 km) of the Samcheok X-band dual-polarization radar were analyzed for major heavy rainfall events in 2019 and 2020. The precipitation data (SAM_RR) were generated by the distributed specific differential phase-based rainfall estimation technique. In addition, the rainfall data provided by the manufacturer (SAM_SRI) and the rainfall data from the composite S-band radar rainfall (CMP_HFC) operated by the Ministry of Environment were shown together to compare the observation characteristics.

This study analyzed the accuracy of five rain events during 2019–2020. The events were: 2–3 October 2019 (Event 1), 30 June 2020 (Event 2), 24 July 2020 (Event 3), 2–3 September 2020 (Event 4), and 7 September 2020 (Event 5). To examine the X-band radar with the aim of improving the accuracy of estimating strong rainfall, we selected the strong rainfall events with a maximum rainfall intensity of more 40 mm/h around the Samcheok area.

To calculate the spatially distributed precipitation of each rainfall estimation technique, Barnes objective analysis was performed on the data from the automatic weather station of KMA and the Ministry of Environment’s telemetry data. The Korea Meteorological Administration (KMA) have used the Barnes scheme previously to draw a spatial distribution map of ground information for real-time analysis [21,22]. The distributed rainfall from ground rain gauges cannot be considered the actual rainfall distribution field. However, the rainfall distribution can be confirmed locally.

Figure 4 illustrates the 1 h accumulated rainfall distributions for the maximum rainfall in each rain event, generated by estimated rainfall data based on the distributed specific differential phase (SAM_RR), the estimated rainfall data from the existing SRI (SAM_SRI), and composite S-band radar rainfall (CMP_HFC). The rainfall spatial patterns between ground rain gauges and SAM_RR were intuitively similar, as shown in the fourth column of Figure 4d. As shown in the third column in Figure 4, the observed rainfall area of SAM_SRI was narrow, and the rainfall amount was notably underestimated. This is because the SAM_SRI was not calibrated by beam attenuation and also used the predetermined Z-R relation (Z = 200R^1.6). The S-band radar composite rainfall could not accurately estimate the area of severe rainfall. The results show the rainfall estimated based on the distributed specific differential phase (SAM_RR), the ground rain gauge, and the existing radar for the heavy rain event on 3 October 2019. During this heavy rainfall event, the X-band dual-polarization radar estimated the location and area of severe rainfall on the ground more accurately than the rainfall estimated by both the SRI rainfall and the composite S-band radar rainfall.

Figure 2 shows the ground rain gauges (the Korea Meteorological Administration AWS, ASOS, and the Ministry of Environment’s TM) within a 40 km effective observation radius of the Samcheok X-band dual-polarization radar. This study estimated the accumulation of rainfall over a 1 h period at 27 ground rainfall observation points within a 40 km radius and extracted the radar rainfall data corresponding to each gauge station to compare the scatterplots. The Samcheok X-band dual-polarization radar is less accurate with increasing distance, owing to the attenuation caused by rainfall. Considering this limitation, the analysis was also performed only for ground rain-gauge stations within a 10 km radius (red dots of Figure 5). Figure 5 shows the evaluation of the accumulated rainfall over a 1 h period based on the distributed specific differential phase for all events. Figure 5a shows the evaluation results for Event 1. Typhoon Mitak occurred during this event, and the maximum rainfall was 110.55 mm. The SAM_SRI and the CMP_HFC significantly underestimated the rainfall, whereas the SAM_RR accurately estimated the severe rainfall areas. In the analysis performed within a 10 km radius, the accuracy of the rainfall estimated by the X-band dual-polarization radar was higher. As shown in Figure 5, SAM_RR is more concentrated on a 1:1 line than CMP_HFC and SAM_SRI at 10 mm or less, clearly confirming that the accuracy of rainfall estimates was high. The slope of the regression line of the SAM_RR is 0.27. Figure 5b shows the evaluation results for Event 2, with a maximum rainfall of 51 mm. The correlation coefficients of SAM_RR and CMP_HFC were 0.70 and 0.69, respectively. CMP_HFC was highly correlated to the ground rain gauge; however, the rainfall was underestimated. The slope of regression of CMP_HFC (0.25) was lower than that of SAM_RR (0.47). Figure 5c shows the results for Event 3. The maximum rainfall from the ground rain gauge was 67.50 mm. Although the maximum rainfall estimated by SAM_RR was closer to the ground rainfall than the other type of rainfall (SAM_SRI and CMP_HFC), it was still underestimated by 46%. Figure 5d shows the evaluation results for Event 4, and the improvement in the rainfall accuracy by SAM_RR. Figure 5e shows the evaluation of rainfall estimation for Event 5. The rainfall estimated by the distributed specific differential phase-based Samcheok X-band radar was located closer to the 1:1 line than the rainfall estimated by the other radars. The results show that the Samcheok X-band radar significantly improved the underestimation tendency compared to the other radar rainfall estimates. The red dots are densely populated around the 1:1 line for all events, indicating that the estimation of the observed area within a 10 km radius has higher accuracy.

We computed the accuracy of the rainfall estimates using the correlation coefficient (C-CORR), slope of the regression line (SLOPE), root-mean-square error (RMSE), maximum rainfall (Max Rain), and average rainfall at the evaluation points (average rain) (Table 2). The accuracy values of the measurements of the ground rain gauge stations within a 10 km radius are shown in parentheses in Table 2. This result also confirms that the accuracy of the improved rainfall estimation technique is higher than that of the other radar rainfall estimates.

3.2. Development of Flow Nomograph in Samcheok Stream

This study developed the flow nomograph at selected points where citizens can easily access the river and property damage can occur due to flooding of parking facilities. Before deciding on the technique, we surveyed public officials in charge of disaster prevention in the target area. Considering their expertise and environment, we developed a flow nomograph that can be used immediately when rainfall information is acquired, excluding the hydrological–hydraulic modeling.

This study set up the Hydrologic Engineering Center’s Hydrologic Modeling System (HEC-HMS) and HEC’s River Analysis System (HEC-RAS) for the development of a flow nomograph [28,29]. The Samcheok-osib stream was divided into 20 sub-basins using geographical information for HEC-HMS. To improve the simulation accuracy of the rainfall–runoff model, the initial parameters (concentration time and storage coefficient) were calibrated and optimized using the observed water level flow data. Parameters of HEC-RAS are representative cross-sections for each sub-basin, including left and right bank locations, roughness coefficients, and contraction and expansion coefficients. HEC-RAS simulates the unsteady flow for the study area based on the HEC-HMS-derived hydrographs. For developing the flow nomograph, data on different rain events were required as input; however, observed rainfall data from the past cannot represent all possible rainfall situations, including very weak or extreme rainfall. Therefore, we used various rainfall events based on synthetically generated hyetographs using the Huff method [30]. The location of the peak rainfall was set as centered (at 2/4 of the time of the event). The total rainfall during a period was assumed to vary from 10 to 310 mm with 10 mm intervals. The rainfall durations were set from 60 to 360 min with 60 min intervals. Based on these conditions, 143 synthetic rainfall events were generated and used as input data for the HEC-HMS. HEC-RAS simulates the water level information for all sections of the stream using the inflow scenario produced by HEC-HMS. Figure 6 shows the simulated discharge and water level using rainfall scenarios for a 360 min duration.

The relationship between rainfall information (maximum rainfall intensity, duration) and water level rise was calculated using the simulation results. The criteria of flood risk were determined using the measured river section information, as shown in Figure 7. As shown in Figure 7c, the flow nomographs were developed at six points, which had different cross-sections in the downstream region. The advantage of a flow nomograph is that it can provide flood forecasting information based on hydrological–hydraulic modeling scenarios at the cross-section of a stream without a water level station. Point 12 has a water level observation station; therefore, this point and the observed water level were used for flood predictability evaluation.

3.3. Flood Forecasting of Samcheok Stream

The feasibility of using the newly installed X-band radar for forecasting floods was evaluated by incorporating a flow nomograph. Among the selected five rain events in the rainfall accuracy evaluation, Event 1 and Event 5 were heavy rainfall events; therefore, the rainfall events on 2 and 3 October 2019 and 7 September 2020 were selected for evaluation.

To apply the flow nomograph, we estimated the mean areal precipitation (MAP) using ground rain-gauge data, composite S-band radar rainfall (CMP_HFC), and Samcheok X-band radar rainfall (SAM_RR). The MAP was estimated using ground rain-gauge data and the Thiessen polygons method [31]. The MAP from radar rainfall was simply the sum of rainfall values at each grid cell within the predefined sub-basin divided by the number of cells within that sub-basin. Figure 8 shows the result of applying the calculated MAP every 10 min on 3 October 2019, at 01:50. The average of all 10 min rainfall intensities was recorded up to the selected duration time to be applied to the flow nomograph. At this time, the attention level was first exceeded because the flow nomograph was applied in real time from the beginning of the rainfall. As shown in Figure 9, the three types of rainfall data were marked with circles on the flow nomograph based on the cumulative rainfall data at 01:50. Only SAM_RR forecasted that the water level would rise to reach the attention level. Figure 8 shows that the ground rain-gauge station could not observe the rainfall in the upper basin of the Samcheok-osib stream at that point in time. Furthermore, since the CMP_HFC underestimated the rainfall, it failed to predict that the water level would exceed the attention level even until the rain stopped. Figure 9 illustrates the time-series data of the water level and rainfall observed at the Osibcheon bridge during Typhoon Mitak. At 01:50, the water level did not exceed the attention level. However, a rise in the water level can be forecasted by applying the flow nomograph, and this information can be used as a disaster prevention measure. As shown in Figure 10, the water level observed at the Osibcheon bridge at 02:40 increased to 2.45 m, which exceeded the attention level. Hence, the applicability of the flow nomograph is confirmed. This technique is especially advantageous because information about the rising water level can be obtained even when there is no water level observation station.

The following results were derived from predicting the ranges of water level rise by applying each set of rainfall information to the flow nomograph for the heavy rainfall event on 7 September 2020. At 08:30 on 7 September 2020, it was forecasted that the water level would exceed the attention level only when SAM_RR was applied, as shown in Figure 11. Beginning at 08:40, it was forecasted that the water level would exceed the attention level when the ground rain gauge and SAM_RR were applied. The forecast predicted the possibility of a water level rise 1 h and 50 min earlier than 10:20, at which the water level exceeded the attention level. At 10:30, it was forecasted that the water level would exceed the caution level when the ground-rain-gauge-measured rainfall was applied, as shown in Figure 12. However, when SAM_RR was applied, it was predicted that the water level would exceed the attention level. At 10:50, both the ground-rain-gauge-measured rainfall and the rainfall estimated by the X-band dual-polarization radar (SAM_RR) predicted that the water level would exceed the caution level. This forecast was made 30 min earlier than the time at which the water level exceeded the caution level (11:20). Figure 13 shows the results of predicting the stream flood level using rainfall data at 10:50 on 7 September 2020. Figure 14 shows the MAP for each set of rainfall data simultaneously. Compared with the case in 2019, the intensities of the MAP of the ground rain gauge and X-band radar were simulated similarly.

Figure 15 shows the observed water level and time-series data of each set of rainfall information for 7 September 2020. As shown in the bar graph of the rainfall information, the large composite radar underestimated the rainfall. The estimated rainfall provided by the Samcheok X-band dual-polarization radar was similar to that of the ground rain gauge. This outcome was applied to the flow nomograph, which affected the estimation of the rise in the stream’s water level.

4. Conclusions

In the present study, the estimated rainfall was analyzed using an X-band radar. This system was first installed and operated in South Korea for flood forecasting in areas with potential for rainfall observation and to supplement existing radar information in order to improve the accuracy of rainfall estimation. Furthermore, it is unique because it shows the possibility of application by applying a flood forecasting method that considers geographic characteristics that can directly use X-band radar data. It was confirmed that high-accuracy rainfall data can be obtained by applying a distributed specific differential phase to improve the accuracy of the estimated quantitative rainfall data from the X-band radar. This study also showed that the X-band radar could be useful for forecasting the flood level of streams. Based on this study, the rainfall estimation technique and flow nomograph can be successfully used for flood forecasting tasks.

Author Contributions

Conceptualization, S.-S.Y. and S.-H.L.; methodology, S.-S.Y. and S.-H.L.; software, S.-S.Y. and S.-H.L.; formal analysis, S.-S.Y.; investigation, S.-S.Y. and S.-H.L.; resources, S.-S.Y.; data curation, S.-S.Y.; writing—original draft preparation, S.-S.Y.; writing—review and editing, S.-S.Y. and S.-H.L. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by Korea Hydro and Nuclear Power Co., Ltd., grant number H21S031000, and the Ministry of Environment’s academic research funding (Establishment of the application framework and data analysis for an electromagnetic precipitation observation station).

Acknowledgments

The authors are grateful to Hyeon Gyo Jeong and Yo Han Cho (from Han River Flood Control Office, Korea Ministry of Environment) for providing the datasets used in this study and for the helpful advice.

Conflicts of Interest

The authors declare no conflict of interest.

References

WMO. Meteorological Radar at the World Radiocommunication Conference, Meteoworld March 2015. Available online: https://public.wmo.int/en/resources/meteoworld/meteorological-radar-world-radiocommunication-conference (accessed on 16 March 2021).
Anagnostou, M.N.; Kalogiros, J.; Anagnostou, E.N.; Tarolli, M.; Papadopoulos, A.; Borga, M. Performance evaluation of high-resolution rainfall estimation by X-band dual-polarization radar for flash flood applications in mountainous basins. J. Hydrol. 2010, 394, 4–16. [Google Scholar] [CrossRef]
Anagnostou, M.N.; Nikolopoulos, E.I.; Kalogiros, J.; Anagnostou, E.N.; Marra, F.; Mair, E.; Borga, M. Advancing precipitation estimation and streamflow simulations in complex terrain with X-band dual-polarization radar observations. Remote Sens. 2018, 10, 1258. [Google Scholar] [CrossRef] [Green Version]
Matrosov, S.Y.; Cifelli, R.; Gochis, D. Measurements of heavy convective rainfall in the presence of hail in flood-prone areas using an X-band polarimetric radar. J. Appl. Meteorol. Climatol. 2013, 52, 395–407. [Google Scholar] [CrossRef] [Green Version]
Paz, I.D.S.R. Quantifying the Rain Heterogeneity by X-Band Radar Measurements for Improving Flood Forecasting. Ph.D. Thesis, Université Paris-Est, Champs-sur-Marne, France, 2018. [Google Scholar]
Wang, Y.; Chandrasekar, V. Quantitative precipitation estimation in the CASA X-band dual-polarization radar network. J. Atmos. Ocean Technol. 2010, 27, 1665–1676. [Google Scholar] [CrossRef]
Junyent, F.; Chandrasekar, V.; McLaughlin, D.; Insanic, E.; Bharadwaj, N. The CASA Integrated Project 1 networked radar system. J. Atmos. Ocean Technol. 2010, 27, 61–78. [Google Scholar] [CrossRef]
Fang, Z.; Bedient, P. NEXRAD flood warning system and floodplain library for Houston, TX. Restoring Our Natural Habitat. In Proceedings of the 2007 World Environmental and Water Resources Congress, Tampa, FL, USA, 15–19 May 2007. [Google Scholar] [CrossRef] [Green Version]
Andjelkovic, I. Guidelines on non-structural measures in urban flood management. In Technical Documents in Hydrology; UNESCO: Paris, France, 2001; p. 86. [Google Scholar]
Hong, Y.; Gourley, J.J. Radar Hydrology; CRC Press: Boca Raton, FL, USA, 2015. [Google Scholar]
Sokol, Z.; Szturc, J.; Orellana-Alvear, J.; Popová, J.; Jurczyk, A.; Célleri, R. The role of weather radar in rainfall estimation and its application in meteorological and hydrological modelling—A review. Remote Sens. 2021, 13, 351. [Google Scholar] [CrossRef]
Correa, O.; García, F.; Bernal, G.; Gardona, O.D.; Rodriguez, C. Early warning system for rainfall-triggered landslides based on real-time probabilistic hazard assessment. Nat. Hazards 2020, 100, 345–361. [Google Scholar] [CrossRef] [Green Version]
Ustrnul, Z.; Wypych, A.; Henek, E.; Maciejewski, M.; Bochenek, B. Climatologically based warning system against meteorological hazards and weather extremes: The example for Poland. Nat. Hazards 2015, 77, 1711–1729. [Google Scholar] [CrossRef] [Green Version]
Bedient, P.B.; Holder, A.; Benavides, J.A.; Vieux, B.E. Radar-based flood warning system applied to tropical storm Allison. J. Hydrol. Eng. 2003, 8, 308–318. [Google Scholar] [CrossRef]
Yoon, S.-S.; Nakakita, E. Application of an X-Band multiparameter radar network for rain-based urban flood forecasting. J. Hydrol. Eng. 2017, 22, E5015005. [Google Scholar] [CrossRef]
Ji, J.B.; Lee, K.T. Analysis of precipitation in the Yeongdong region related to the east wind-focusing on the effect of the mountains. In Proceedings of the Spring Meeting of KMS, Seoul, Korea, 23–25 April 2008. [Google Scholar]
Koch, S.E.; des Jardins, M.; Kocin, P.J. An interactive Barnes objective map analysis scheme for use with satellite and conventional data. J. Clim. Appl. Meteorol. 1983, 22, 1487–1503. [Google Scholar] [CrossRef]
Lee, J.S.; Lee, Y.H.; Ha, J.C.; Lee, H.C. Objective analysis of temperature using the elevation-dependent weighting function. Atmosphere 2021, 22, 233–243. [Google Scholar] [CrossRef]
Cifelli, R.; Chandrasekar, V.; Lim, S.; Kennedy, P.C.; Wang, Y.; Rutledge, S.A. A new dual-polarization radar rainfall algorithm: Application in Colorado precipitation events. J. Atmos. Ocean Technol. 2011, 28, 352–364. [Google Scholar] [CrossRef] [Green Version]
Ryzhkov, A.V.; Giangrande, S.E.; Schuur, T.J. Rainfall estimation with a polarimetric prototype of WSR-88D. J. Appl. Meteorol. 2005, 44, 502–515. [Google Scholar] [CrossRef] [Green Version]
Thurai, M.; Mishra, K.V.; Bringi, V.N.; Krajewski, W.F. Initial results of a new composite-weighted algorithm for dual-polarized X-band rainfall estimation. J. Hydrometeorol. 2017, 18, 1081–1100. [Google Scholar] [CrossRef]
Pepler, A.S.; May, P.T. A robust error-based rain estimation method for polarimetric radar. Part II: Case study. J. Appl. Meteorol. Climatol. 2021, 51, 1702–1713. [Google Scholar] [CrossRef]
Wang, Y.; Chandrasekar, V. Algorithm for estimation of the specific differential phase. J. Atmos. Ocean Technol. 2009, 26, 2569–2582. [Google Scholar] [CrossRef]
Brandes, E.A.; Zhang, G.; Vivekanandan, J. Drop size distribution retrieval with polarimetric radar: Model and application. J. Appl. Meteorol. Climatol. 2004, 43, 461–475. [Google Scholar] [CrossRef] [Green Version]
Lim, S.; Cifelli, R.; Chandrasekar, V.; Matrosov, S.Y. Precipitation classification and quantification using X-band dual-polarization weather radar: Application in the Hydrometeorology Testbed. J. Atmos. Ocean Technol. 2013, 30, 2108–2120. [Google Scholar] [CrossRef]
Chen, H.; Lim, S.; Chandrasekar, V.; Jang, B.J. Urban hydrological applications of dual-polarization X-band radar: Case study in Korea. J. Appl. Meteorol. 2017, 22, 502–515. [Google Scholar] [CrossRef]
Miller, D.A.; Wu, S.; Kitzmiller, D. Spatial and temporal resolution considerations in evaluating and utilizing radar quantitative precipitation estimates. J. Oper. Meteor. 2013, 1, 168–184. [Google Scholar] [CrossRef]
HEC. Hydrologic Modeling System: Technical Reference Manual; US Army Corps of Engineers Hydrologic Engineering Center: Davis, CA, USA, 2000.
HEC. River Analysis System: Hydraulic Reference Manual; US Army Corps of Engineers Hydrologic Engineering Center: Davis, CA, USA, 2002.
Huff, F.A. Time distribution of rainfall in heavy storms. Water Resour. Res. 1967, 3, 1007–1019. [Google Scholar] [CrossRef]
Fiedler, F.R. Simple, practical method for determining station weights using Thiessen polygons and isohyetal maps. J. Hydrol. Eng. 2003, 8, 119–221. [Google Scholar] [CrossRef]

Figure 1. Accumulated rainfall distribution from 15 to 16 August 2019. (a) Distributed rainfall field by ground rain-gauge network; (b) Rainfall field by S-band radar network.

Figure 2. X-band dual-polarization radar location and hydrological observation network in the Samcheok region.

Figure 3. Prior and post application of the distributed specific differential phase estimation method, and the rainfall estimation results (30 June 2020, at 2:18): (a) K_DP (JRC); (b) Surface Rainfall (SAM_SRI); (c) K_DP (new); (d) Rainfall (SAM_RR).

Figure 4. Distribution of accumulated rainfall over 1 h period distribution for maximum rainfall in each rain event: (a) 3 October 2019, 00:00; (b) 30 June 2020, 03:00; (c) 24 July 2020, 07:00; (d) 3 September 2020, 06:00; (e) 7 September 2020, 11:00.

Figure 5. Scatterplots of observed ground rain gauge vs. various radar rainfall (red dots: rainfall in 10 km radius, black dots: rainfall in 40-km radius): (a) Event 1 (00:00–23:00 2–3 October 2019); (b) Event 2 (00:00–23:00 30 June 2020); (c) Event 3 (00:00–23:00 24 July 2020); (d) Event 4 (00:00–23:00 2–3 September 2020); (e) Event 5 (00:00–23:00 7 September 2020).

Figure 6. Simulated discharge and water level using rainfall scenarios for 360 min duration: (a) Rainfall scenarios; (b) Simulated discharge scenarios; (c) Simulated water level scenarios.

Figure 7. The criteria of flood risk and flow nomograph. (a) Cross-section and criteria of flood risk; (b) Flow nomograph at station 12; (c) Stations of flow nomograph development.

Figure 8. Mean areal precipitation of the Samcheok-osib stream (3 October 2019, at 01:50): (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR.

Figure 9. Calculation results of the flood forecast information for October 3 2019, at 01:50: (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR.

Figure 10. Observed water level and timeseries data of the mean real precipitation between 2 and 3 October 2019.

Figure 11. Calculation results of the flood forecast information for 7 September 2020, at 08:30: (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR.

Figure 12. Calculation results of the flood forecast information for 7 September 2020, at 10:30: (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR.

Figure 13. Calculation results of the flood forecast information for 7 September 2020, at 10:50: (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR.

Figure 14. Mean areal precipitation of the Samcheok-osib stream (7 September 2020, at 10:50): (a) Ground rain gauge; (b) CMP_HFC; (c) SAM_RR.

Figure 15. Observed water level and time-series data of the mean areal precipitation on 7 September 2020.

Table 1. The properties of rain gauge and various radar data for accuracy evaluation.

Category	Ground Rain Gauge	SAM_RR	SAM_SRI	CMP_HFC
Data	AWS, ASOS (KMA) TM (ME)	Samcheok X-band radar	Samcheok X-band radar (original)	Composite S-band radar
Spatial resolution	27 gauge stations	75 m	75 m	250 m
Temporal resolution	10 min	2.5 min	2.5 min	10 min
Rainfall estimation method	-	R (K_DP)	R (Z) (Z = 200R^1.6)	R (K_DP) (CSU-HIDRO algorithm)

Table 2. Results of accuracy analysis for 1 h accumulated rainfall from radar information.

Event	ITEM	Ground Rain Gauge	CMP_HFC	SAM_SRI	SAM_RR
1	C-CORR	-	0.58 (0.67)	0.34 (0.38)	0.62 (0.66)
	SLOPE	-	0.24	0.04	0.27
	RMSE	-	7.70 (15.50)	9.23 (17.39)	7.58 (14.11)
	Max rain (mm/h)	110.50 (84.00)	54.47 (22.07)	8.65 (8.65)	47.66 (41.86)
	Average rain (mm)	3.82 (8.07)	2.29 (2.15)	0.84 (1.03)	1.89 (2.64)
2	C-CORR	-	0.69 (0.91)	0.60 (0.93)	0.70 (0.98)
	SLOPE	-	0.25	0.13	0.47
	RMSE	-	5.16 (4.27)	5.92 (3.11)	4.60 (1.29)
	Max rain (mm/h)	51.00 (27.00)	14.96 (6.07)	11.10 (11.10)	39.15 (35.37)
	Average rain (mm)	3.29 (4.56)	1.10 (0.94)	0.50 (2.13)	1.39 (4.52)
3	C-CORR	-	0.44 (0.11)	0.64 (0.73)	0.66 (0.93)
	SLOPE	-	0.09	0.15	0.34
	RMSE	-	5.15 (12.30)	5.03 (10.11)	4.40 (6.20)
	Max rain (mm/h)	67.50 (67.50)	6.51 (4.56)	12.75 (12.75)	36.16 (36.16)
	Average rain (mm)	3.16 (5.70)	0.75 (0.84)	0.49 (2.06)	0.88 (3.60)
4	C-CORR	-	0.82 (0.81)	0.44 (0.91)	0.70 (0.95)
	SLOPE	-	0.32	0.07	0.39
	RMSE	-	5.61 (3.79)	7.53 (3.07)	5.69 (2.10)
	Max rain (mm/h)	60.50 (26.00)	23.83 (7.23)	13.70 (13.70)	38.63 (34.25)
	Average rain (mm)	3.74 (2.26)	1.55 (1.01)	0.53 (1.24)	1.62 (2.82)
5	C-CORR	-	0.82 (0.93)	0.47 (0.79)	0.72 (0.84)
	SLOPE	-	0.27	0.08	0.46
	RMSE	-	8.11 (11.58)	10.10 (10.07)	7.27 (6.50)
	Max rain (mm/h)	45.50 (45.50)	20.01 (7.20)	16.45 (16.45)	41.29 (41.29)
	Average rain (mm)	5.98 (8.48)	1.78 (1.61)	0.59 (3.06)	2.47 (7.19)

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

Share and Cite

MDPI and ACS Style

Yoon, S.-S.; Lim, S.-H. Analyzing the Application of X-Band Radar for Improving Rainfall Observation and Flood Forecasting in Yeongdong, South Korea. Remote Sens. 2022, 14, 43. https://doi.org/10.3390/rs14010043

AMA Style

Yoon S-S, Lim S-H. Analyzing the Application of X-Band Radar for Improving Rainfall Observation and Flood Forecasting in Yeongdong, South Korea. Remote Sensing. 2022; 14(1):43. https://doi.org/10.3390/rs14010043

Chicago/Turabian Style

Yoon, Seong-Sim, and Sang-Hun Lim. 2022. "Analyzing the Application of X-Band Radar for Improving Rainfall Observation and Flood Forecasting in Yeongdong, South Korea" Remote Sensing 14, no. 1: 43. https://doi.org/10.3390/rs14010043

APA Style

Yoon, S. -S., & Lim, S. -H. (2022). Analyzing the Application of X-Band Radar for Improving Rainfall Observation and Flood Forecasting in Yeongdong, South Korea. Remote Sensing, 14(1), 43. https://doi.org/10.3390/rs14010043

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

Article Menu