remote sensing

Article Drought Monitoring Utility using Satellite-Based Precipitation Products over the Xiang River Basin in

Qian Zhu 1,*, Yulin Luo 1, Dongyang Zhou 1, Yue-Ping Xu 2, Guoqing Wang 3 and Haiying Gao 1 1 School of Civil Engineering, Southeast University, Nanjing 211189, China; [email protected] (Y.L.); [email protected] (D.Z.); [email protected] (H.G.) 2 Institute of Hydrology and Water Resources, College of Civil Engineering and Architecture, Zhejiang University, Hangzhou 310058, China; [email protected] 3 State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Nanjing Hydraulic Research Institute, Nanjing 210029, China; [email protected] * Correspondence: [email protected]; Tel.: +86-151-5185-7942

 Received: 21 May 2019; Accepted: 20 June 2019; Published: 22 June 2019 

Abstract: Drought is a natural hazard disaster that can deeply affect environments, economies, and societies around the world. Therefore, accurate monitoring of patterns in drought is important. Precipitation is the key variable to define the drought index. However, the spare and uneven distribution of rain gauges limit the access of long-term and reliable in situ observations. Remote sensing techniques enrich the precipitation data at different temporal–spatial resolutions. In this study, the climate prediction center morphing (CMORPH) technique (CMORPH-CRT), the tropical rainfall measuring mission (TRMM) multi-satellite precipitation analysis (TRMM 3B42V7), and the integrated multi-satellite retrievals for global precipitation measurement (IMERG V05) were evaluated and compared with in situ observations for the drought monitoring in the Xiang River Basin, a humid region in China. A widely-used drought index, the standardized precipitation index (SPI), was chosen to evaluate the drought monitoring utility. The atmospheric water deficit (AWD) was used for comparison of the drought estimation with SPI. The results were as follows: (1) IMERG V05 precipitation products showed the highest accuracy against grid-based precipitation, followed by CMORPH-CRT, which performed better than TRMM 3B42V7; (2) IMERG V05 showed the best performance in SPI-1 (one-month SPI) estimations compared with CMORPH-CRT and TRMM 3B42V7; (3) SPI-1 was more suitable for drought monitoring than AWD in the Xiang River Basin, because its high R-values and low root mean square error (RMSE) compared with the corresponding index based on in situ observations; (4) drought conditions in 2015 were apparently more severe than that in 2016 and 2017, with the driest area mainly distributed in the southwest part of the Xiang River Basin.

Keywords: drought; TRMM 3B42V7; CMORPH-CRT; IMERG V05; SPI

1. Introduction Droughts usually occur over months to years in a region, including high and low precipitation regions, and cause damage to ecosystems and human well-being. The influence of climate change has led to an increasing trend of temperature, which is expected to significantly affect the spatial–temporal pattern of precipitation and the intensity of droughts [1]. Droughts might be further aggravated around the world, including in humid regions [2,3]. Therefore, it is essential to accurately monitor and predict drought because of its importance in risk management. Precipitation plays an important role in drought monitoring as the primary factor affecting the formation and persistence of droughts and the estimation of drought indices [4,5]. Conventionally,

Remote Sens. 2019, 11, 1483; doi:10.3390/rs11121483 www.mdpi.com/journal/remotesensing Remote Sens. 2019, 11, 1483 2 of 17 precipitation data are derived from rain gauge observations through in situ gauge networks. However, the spare and uneven distribution of rain gauges limit the access of long-term and reliable ground-based precipitation measurements [6,7]. This phenomenon is particularly common in remote regions of developing countries. Moreover, unlike air pressure and temperature, precipitation always features large spatial–temporal variation and high uncertainty [8], thus, the spatial–temporal interpolation cannot always accurately capture the dynamics of precipitation data. The development of remote sensing techniques enriches precipitation data at different spatial and temporal resolutions. In recent decades, satellite-based precipitation datasets have been widely applied for drought and flood forecasting [9–12], hydrological modelling [13,14], and water resource exploration [8]. There are now many distinct satellite-based precipitation products available, with some of them being widely used and having large-scale coverage with high spatial and temporal resolutions, such as the National Aeronautics and Space Administration (NASA) tropical rainfall measuring mission (TRMM) multi-satellite precipitation analysis (TMPA) [15], the climate prediction center morphing (CMORPH) technique [16] and integrated multi-satellite retrievals for global precipitation measurement (IMERG) [17]. The TRMM precipitation datasets have been available since November 1997 and many researchers have proved its ability for drought monitoring [18–24]. For example, Zeng et al. [18] evaluated the TRMM multi-satellite precipitation analysis (TMPA) in drought monitoring in Lancang River Basin. They discovered that one-month and three-month scale standard precipitation indexes (SPI) obtained from the version 6 monthly TMPA product from 1998 to 2009 both agreed well for most of the grid points in Lancang River Basin. Obtaining a drought index using solely CMORPH precipitation products has also been utilized for drought monitoring in many regions worldwide [25–29]. Lu et al. [26] compared the CMORPH-based SPI with the SPI estimate using in situ precipitation observations from 2221 meteorological stations across China from 1998 to 2014. They found that the CMORPH-based SPI values were generally consistent with the SPI obtained with in situ measurements, which suggested that the SPI estimate using CMORPH precipitation data products could be applied to drought assessment and monitoring. The IMERG precipitation products were characterized with high temporal and spatial resolutions (0.1 0.1 , 30 min) and its capability of real-time drought ◦ × ◦ monitoring [30–32]. Among them, Jang et al. [31] calculated the SPI based on the global precipitation measurement (GPM) IMERG data and compared them with the results obtained from the in situ observations. They confirmed that the GPM IMERG-based SPI correlated well with the SPI, based on the ground precipitation observations. province is located at the River Basin, a humid region in China with the highest incidence of drought [33]. Xiang River is the largest river in Hunan province, China, and is one of the eight of the Yangtze River. The Xiang River Basin suffered moderate drought events before the 1990s and the condition has become drier since 2003 [34]. Therefore, it is urgent and significant to investigate the spatial–temporal distribution characteristics of droughts in the Xiang River Basin. In recent years, a variety of studies have been conducted on drought monitoring in the Xiang River Basin [7,26,33,35–37]. However, most of them explored the drought monitoring with in situ precipitation observations [7,33,35,36], some of them with satellite soil moisture [37] and some of them with TMPA and CMORPH-BLD precipitation products [26]. As far as we know, there are few works evaluating the efficiency of CMORPH-CRT or IMERG V05 in drought monitoring in Xiang River Basin. Therefore, it is critical to explore the utility of satellite-based precipitation products for more effective drought monitoring and forecasting in Xiang River Basin. The aim of this study is to investigate and compare the drought monitoring utility of CMORPH-CRT, IMERG V05, and TRMM 3B42V7 precipitation estimates with the drought index, which is standard precipitation index (SPI). The remaining structure of the paper is organized as follows. The descriptions of the study area, the in situ observations, and three satellite-based precipitation products are presented in Section2; the methodology and evaluation indices are introduced in Section3; Section4 provides the results and discussions; and conclusions are drawn in Section5. Remote Sens. 2019, 11, 1483 3 of 17 Remote Sens. 2019, 11, x FOR PEER REVIEW 3 of 18

Remote Sens. 2019, 11, x FOR PEER REVIEW 3 of 18 2.2.1. Study Study Area Area and Data 2.1. Study Area 2.1. StudyThe Xiang Area river is located in the south of the Yangtze River Basin and is the largest river in HunanThe province Xiang river and isthe located Dongting in the Lake south water of systemthe Yangtze [33,37]. River As theBasin main and is the largest of the riverYangtze in The Xiang river is located in the south of the Yangtze River Basin and is the largest river in Hunan HunanRiver Basin, province the andXiang the river Dongting originates Lake from water the system mountainous [33,37]. Asarea the in main the southwest tributary ofand the flows Yangtze into province and the water system [33,37]. As the main tributary of the Yangtze River Basin, Riverthe northeast Basin, the of XiangXiang riverRiver originates Basin [36]. from The the mountainous discharge area station in the is southwest the control and station flows of into the the Xiang river originates from the mountainous area in the southwest and flows into the northeast of theXiang northeast River Basinof Xiang and River the streamflowBasin [36]. Thein this Xiangtan basin (Figuredischarge 1). stationAs shown is the in control Figure station2, the annualof the Xiang River Basin [36]. The Xiangtan discharge station is the control station of the Xiang River Basin Xiangrainfall River ranges Basin from and 1500 the streamflowto 1800 mm in and this is basin unevenly (Figure distributed, 1). As shown with in approximately Figure 2, the annual60% of and the streamflow in this basin (Figure1). As shown in Figure2, the annual rainfall ranges from 1500 rainfallannual rangesrainfall fromfalling 1500 from to April1800 mmto September. and is un evenlyThe average distributed, annual with temperature approximately is 18.4 60%°C and of to 1800 mm and is unevenly distributed, with approximately 60% of annual rainfall falling from April annualannual rainfallevaporation falling is aboutfrom April932 mm. to September. The uneven The dist averageribution annualof rainfall temperature has led to isfrequent 18.4 °C floods and to September. The average annual temperature is 18.4 C and annual evaporation is about 932 mm. annualand droughts evaporation in the is Xiang about River 932 mm.Basin The [7]. uneven distribution◦ of rainfall has led to frequent floods The uneven distribution of rainfall has led to frequent floods and droughts in the Xiang River Basin [7]. and droughts in the Xiang River Basin [7].

Figure 1. Xiang River Basin and distribution of the meteorological stations. Figure 1. Xiang River Basin and distribution of the meteorological stations. Figure 1. Xiang River Basin and distribution of the meteorological stations.

Figure 2. Ranges for annual precipitation, annual average temperature, and potential evapotranspiration. Figure 2. Ranges for annual precipitation, annual average temperature, and potential Figureevapotranspiration. 2. Ranges for annual precipitation, annual average temperature, and potential evapotranspiration. 2.2. Data sources 2.2. Data sources

Remote Sens. 2019, 11, 1483 4 of 17

2.2. Data Sources

2.2.1. In Situ Observations The daily in situ observations used in this study are collected from fourteen meteorological stations provided by the China National Meteorological Information Center (https://data.cma.cn) from 2015 to 2017. All the selected stations have complete records of daily precipitation, average daily temperature, maximum and minimum daily temperature, average pressure, average relative humidity, and solar radiation. The in situ observations are used as the reference to evaluate the accuracy of satellite-based precipitation and satellite-based drought indices at station scale.

2.2.2. Grid-Based Precipitation In order to comprehensively illustrate the accuracy and drought monitoring utility of the selected satellite-based precipitation datasets on grid scale, the dataset of daily 0.5 0.5 grid-based ◦ × ◦ precipitation over China from 2015 to 2017 is selected as the reference. It is developed by the China Meteorological Administration (CMA) and generated from the daily precipitation observations from 2472 stations using Thin Plate Spline. The dataset has been evaluated and used for exploring the spatial and temporal distributions of the precipitation over China and the results claim that the dataset can be used in meteorological monitoring and hydrological studies [38].

2.2.3. TRMM 3B42V7 The tropical rainfall measuring mission (TRMM) multi-satellite precipitation analysis (TMPA) is a satellite-based precipitation product providing the estimation of quasi-global rainfall (50◦N–50◦S) with high spatial and temporal resolution (0.25 0.25 , 3 h). The available data covers the period ◦ × ◦ from 1998 to present. The TRMM 3B42 series products are the main precipitation products obtained from TRMM merged with other satellite estimates [15]. Two main 3B42 precipitation products are the near-real-time 3B42RT and the post-real-time 3B42V7. The 3B42RT product is generated from the satellite remote sensing information, while the post-real-time 3B42V7 is corrected by the Global Precipitation Climatology Centre (GPCC) gauge precipitation [39]. This study only focuses on the post-real-time 3B42V7 product. The datasets of this product from 2015 to 2017 are available on the NASA website (https://pmm.nasa.gov/data-access/downloads/trmm).

2.2.4. CMORPH-CRT The CMORPH products, with high resolution (0.25 0.25 , 3 h) and near total global coverage ◦ × ◦ (60◦N–60◦S), are generated by the National Ocean and Atmospheric Administration (NOAA) Climate Prediction Center’s MORPHing technique [16,40]. CMORPH products estimate the global precipitation from passive microwave satellite scans [14]. CMORPH has two versions, Version 0.x and Version 1.0. The former starts from December 2002 and presents satellite-only products, while the latter starts from January 1998 and contains three individual precipitation products [16,41]. CMORPH Version 1.0 includes raw (CMORPH-RAW), bias-corrected (CMORPH-CRT), and gauge satellite-blended (CMORPH-BLD) precipitation products. In this study, 0.25◦and daily CMORPH-CRT product from 2015 to 2017 is utilized and the CMORPH-CRT datasets are freely downloaded (ftp://ftp.cpc.ncep.noaa. gov/precip/).

2.2.5. IMERG V05 The GPM mission, as a successor to the TRMM satellite, was launched on Feb 2014 by NASA and JAXA, aiming at providing an accurate and reliable global precipitation estimation with both active and passive microwave sensors [15,42,43]. There are four levels of GPM products based on different algorithms. The IMERG products are the GPM level 3 products with high spatial–temporal resolution (0.1 0.1 , 30 min) within a 60 N–60 S latitude–longitude band [15,44,45]. IMERG provides ◦ × ◦ ◦ ◦ Remote Sens. 2019, 11, 1483 5 of 17 three types of products: Near-real-time Early Run (IMERG-E, with a latency of 6 h), reprocessed near-real-time Late Run (IMERG-L, with a latency of 12 h), and gauged-adjusted Final Run (IMERG-F, Remote Sens. 2019, 11, x FOR PEER REVIEW 5 of 18 with a latency of 2.5 months). Among these three products, IMERG-F usually provides more accurate theprecipitation latest version datasets of IMERG-F compared from to 2015 the otherto 2017, two name productsly IMERG [6,15 V05,]. Therefore, is selected the for latestthis study. version This of canIMERG-F be fromobtained 2015 to 2017,from namely the IMERG Precipitatio V05, is selectedn Measurement for this study. Missions This can be obtainedwebsite fromat https://pmm.nasa.gov/data-access/downloads/gpm.the Precipitation Measurement Missions website at https: //pmm.nasa.gov/data-access/downloads/gpm. ConsideringConsidering thethe availableavailable daily daily precipitation precipitation data data of IMERG of IMERG V05 is V05 from is 2015, from while 2015, the while accessible the accessibledata of in data situ meteorologicalof in situ meteorol observationsogical observations ends at 2017, ends the at time2017, period the time of thisperiod study of this is from study 2015 is fromto 2017. 2015 to 2017.

3. Methodology 3. Methodology As mentioned above, the in situ observations and grid-based precipitation data from CMA are As mentioned above, the in situ observations and grid-based precipitation data from CMA are used as reference datasets to evaluate the accuracy of TRMM 3B42V7, CMORPH-CRT, and IMERG V05 used as reference datasets to evaluate the accuracy of TRMM 3B42V7, CMORPH-CRT, and IMERG precipitation products and their performance on drought monitoring. Generally, 1–2 month SPI and V05 precipitation products and their performance on drought monitoring. Generally, 1–2 month SPI 3–6 month SPI are suitable for exploring meteorological and agricultural droughts, respectively [46]. and 3–6 month SPI are suitable for exploring meteorological and agricultural droughts, respectively In this study, SPI-1 is selected for drought monitoring in the Xiang River Basin due to the short records [46]. In this study, SPI-1 is selected for drought monitoring in the Xiang River Basin due to the short of satellite-based precipitation data. The SPI-1 obtained from these three precipitation products are records of satellite-based precipitation data. The SPI-1 obtained from these three precipitation compared with the SPI-1 calculated from the reference datasets. For comparison, the atmospheric products are compared with the SPI-1 calculated from the reference datasets. For comparison, the water deficit (AWD) is also used to illustrate the drought conditions. Considering that potential atmospheric water deficit (AWD) is also used to illustrate the drought conditions. Considering that evapotranspiration is required to calculate AWD, the AWD is only analyzed at station scale. Because potential evapotranspiration is required to calculate AWD, the AWD is only analyzed at station the grid-based observations just provide precipitation data instead of other variables (e.g., temperature), scale. Because the grid-based observations just provide precipitation data instead of other variables there is no grid-based potential evapotranspiration data available. Based on the evaluation results, (e.g., temperature), there is no grid-based potential evapotranspiration data available. Based on the the best product among the three is then selected to analyze the drought conditions in the Xiang evaluation results, the best product among the three is then selected to analyze the drought conditionsRiver Basin. in the Xiang River Basin. TheThe analysis analysis waswas accomplishedaccomplished on on both both temporal temporal and spatialand spatial scales. scales. Figure 3Figure shows the3 shows flow-chart the flow-chartof the main of steps the main of the steps study. of Thethe st detailedudy. The steps detailed are as steps follows: are as follows: 1)(1) EvaluateEvaluate the the accuracyaccuracy ofof TRMM TRMM 3B42V7, 3B42V7, CMORPH-CRT, CMORPH-CRT, and IMERG and IMERG V05 precipitation V05 precipitation products; (2) products;Evaluate the accuracy of SPI-1 calculated from TRMM 3B42V7, CMORPH-CRT, and IMERG V05 2) Evaluateprecipitation the accuracy products; of SPI-1 calculated from TRMM 3B42V7, CMORPH-CRT, and IMERG V05 (3) precipitationCalculate the products; AWD values with in situ observations and compare the performance of AWD and 3) CalculateSPI at station the AWD scale; values with in situ observations and compare the performance of AWD and (4) SPIAnalyze at station the scale; temporal and spatial drought conditions in the Xiang River Basin. 4) Analyze the temporal and spatial drought conditions in the Xiang River Basin.

Figure 3. The flow-chart of the main steps of the study. Figure 3. The flow-chart of the main steps of the study.

Remote Sens. 2019, 11, 1483 6 of 17

3.1. Standard Precipitation Index (SPI) The SPI is a widely used drought index developed by McKee et al. [47], which is calculated with accumulated precipitation observations over a time period (e.g., one month). This index is a valuable tool for the determination of the intensity and duration of different droughts [48]. It is generally accepted that the calculation of SPI needs a minimum data-record period of 30 years [47]. However, this index has also been shown to be of use in drought monitoring on small time scales [48]. Rhee and Carbone [30] also confirmed that short data records of precipitation can be used for the estimation of SPI. Therefore, even though the IMERG data records are relatively short compared with rain gauge observations and the other two satellite-based precipitation products, the estimation of SPI based on IMERG precipitation data for drought monitoring in the Xiang River Basin is still worth exploring. SPI can be measured at different time periods, i.e., 1 month, 3 months, 6 months, or 12 months, with monthly precipitation data [49]. The SPI is defined as follows:

X X SPI = i − (1) s where Xi is the accumulated precipitation observations of a time period and X and s are the mean and standard deviation of precipitation data obtained from the time series of monthly precipitation dataset, respectively. The SPI is calculated based on the gamma probability density function, so SPI < 0 reflects drought conditions and SPI > 0 implies wet conditions [47]. According to McKee et al. [47], the levels of wet and drought conditions and their corresponding categories are listed in Table1.

Table 1. Drought categories defined for SPI values.

SPI Values Drought Category 0 to 1 Mild drought − 1 to 1.5 Moderate drought − − 1.5 to 2 Severe drought − − < = 2 Extreme drought −

3.2. Atmospheric Water Deficit (AWD) Atmospheric water deficit is the difference between precipitation (P) and potential evapotranspiration (ET0). Atmospheric water deficit is claimed to be a suitable index to reflect the drought condition related to meteorological parameters [50]. The AWD value is calculated by P minus ET0 on a weekly scale and the daily ET0 was computed using the Penman–Monteith equation:

AWD = P ET , (2) i i − i 900  0  0.408∆(Rn G) + γ T+273 u2 e ea ET0 = − − (3) ∆ + γ(1 + 0.34u2) where i represents the week of the study period, Pi and ETi denote the sum of precipitation and sum of evapotranspiration (ET0) of week i, respectively, Rn is the net radiation, ∆ is the slope of the vapor pressure curve, G is the soil heat flux, u2 is the wind speed at 2 m above ground level, γ is the psychometric constant, T is the temperature, e0 is saturation vapor pressure at air temperature, and ea is actual vapor pressure. The Penman–Monteith equation is useful for computing ET based on temperature, humidity, wind speed, and solar radiation [51]. The condition is defined as drought when AWD is lower than 0. Extreme drought occurs when AWD is lower than 50 mm [52]. As mentioned − above, no grid-based ET0 is available, therefore, the AWD values on meteorological stations are calculated and compared with the SPI-1 at station scale. Remote Sens. 2019, 11, 1483 7 of 17

3.3. Evaluation Indices Multiple statistical indices are used to evaluate the accuracy of the three satellite-based precipitation products and the performance of their SPI estimates. These indices include Pearson correlation coefficient (R-value), relative bias (BIAS), root mean square error (RMSE), Nash–Sutcliffe efficiency (NSE), the probability of detection (POD), and the false alarm ration (FAR). The six indices are defined as follows: Pn    i=1 Xi X Yi Y R = q − q − , (4) P  2 P  2 n X X n Y Y i=1 i − i=1 i − Pn 2 = (Xi Yi) NSE = 1 i 1 − (5) P  2 − n X X i=1 i − Pn i=1(Xi Yi) BIAS = Pn − (6) i=1 Xi s Pn 2 = (Xi Yi) RMSE = i 1 − (7) n A POD = (8) A + C B FAR = (9) A + B where Xi and Yi represent the in situ dataset (or grid-based dataset) and satellite-based dataset, respectively, and X and Y represent the mean values of these two datasets, respectively. For POD and FAR, A means “hits”, B means “false alarms”, and C means “misses”.

4. Results and Discussion

4.1. Evaluation of Satellite-Based Precipitation Products To investigate the efficiency of the selected three satellite-based precipitation datasets on the drought monitoring in the Xiang River Basin, the accuracy of these precipitation products needs to be evaluated against the in situ observations and grid-based precipitation firstly. The accuracy of precipitation estimates of TRMM 3B42V7, CMORPH-CRT, and IMERG V05 from January 2015 to December 2017 are compared and evaluated using in situ observations and grid-based precipitation as the references at catchment scale and grid/station scale, with the mentioned evaluation indices. At the catchment scale, the areal daily satellite-based precipitation estimates are compared with those calculated with precipitation from the fourteen meteorological stations (Figure4). The daily precipitations from CMORPH-CRT, IMERG V05, and TRMM 3B42V7 are rather consistent with gauge measurements, with high R-values, POD, and NSE values (R-values > 0.58, POD > 0.87, and NSE > 0.91), and low BIAS, FAR, and RMSE values (BIAS between 23.3% and 5.6%, FAR < 0.16, and − RMSE between 6.6 and 6.97). The results show that the CMORPH-CRT and the IMERG V05 have a comparably good performance next to TRMM 3B42V7 at the Xiang River Basin. The TRMM 3B42V7 and the CMORPH-CRT both underestimate the daily precipitation, while the IMERG V05 overestimates the daily precipitation. Figure5 shows the time series of monthly average precipitation based on three satellite-based precipitation products and in situ observations. It can be seen that heavy rains mainly occur in June and July. As for temporal variation, IMERG V05 and CMORPH-CRT show similarly good performances in capturing the dynamics of monthly precipitation and the former has a slight advantage, as CMOPRH-CRT has some minor underestimation. The TRMM 3B42V7 shows the worst performance relatively, exhibiting underestimation, especially in November 2015, June 2016, and June 2017. Remote Sens. 2019, 11, x FOR PEER REVIEW 8 of 18

CMORPH-CRTRemote Sens. 2019show, 11, x FORsimilarly PEER REVIEW good performances in capturing the dynamics of 8monthly of 18 precipitation and the former has a slight advantage, as CMOPRH-CRT has some minor Remote Sens. 2019, 11, 1483 8 of 17 underestimation.CMORPH-CRT The show TRMM similarly 3B42V7 good showsperformanc the esworst in capturing performance the dynamics relatively, of monthlyexhibiting precipitation and the former has a slight advantage, as CMOPRH-CRT has some minor underestimation, especially in November 2015, June 2016, and June 2017. underestimation. The TRMM 3B42V7 shows the worst performance relatively, exhibiting underestimation, especially in November 2015, June 2016, and June 2017.

Figure 4. Comparison of the daily precipitation data from satellite-based products with in Figure 4. Comparison of the daily precipitation data from satellite-based products with in situ observations. Figuresitu observations. 4. Comparison of the daily precipitation data from satellite-based products with in situ observations.

Figure 5. Temporal comparison of monthly average precipitation between satellite-based precipitation Figure 5. Temporal comparison of monthly average precipitation between satellite-based precipitation products products and in situ observations. and in situ observations. Figure 5. Temporal comparison of monthly average precipitation between satellite-based precipitation products Regarding the grid scale, the spatial distributions of R-value, BIAS, RMSE, and NSE based on Regarding the grid scale, the spatial distributions of R-value, BIAS, RMSE, and NSE based on andthe in satellite-based situ observations. precipitation products and daily 0.5 0.5 grid-based precipitation are shown the satellite-based precipitation products and daily 0.5° ×◦ 0.5°× grid-based◦ precipitation are shown in in FigureRegardingFigure6. 6. In In the termsterms grid of of scale,R-values, R-values, the around spatial around 40%, distributi 40%, 78%, 78%, andons 56% andof R-value,of 56% the grids of BIAS, the are grids higher RMSE, are than higherand 0.4 NSE [53] than forbased the 0.4 [on53 ] thefor satellite-based theCMORPH-CRT, CMORPH-CRT, precipitation IMERG IMERG V05, products V05, and andTRMM and TRMM daily 3B42V7 3B42V70.5° ×techniques, 0.5° techniques, grid-based respectively. respectively. precipitation The central The are centralshown and andin Figuresouthwestsouthwest 6. In terms parts parts of of R-values, theof the Xiang Xiang around River River Basin40%, Basin 78%, showshow and highhigh 56% agreement agreement of the grids with with arelarger larger higher R-values. R-values. than The0.4 [53]IMERGThe for IMERG the V05 presents the highest accuracy, followed by the TRMM 3B42V7. Regarding the results of BIAS, CMORPH-CRT,V05 presents the IMERG highest V05, accuracy, and followedTRMM 3B42V7 by the TRMM techniques, 3B42V7. respectively. Regarding theThe resultscentral of BIAS,and CMORPH-CRT presents a relatively poor performance, with underestimation among most parts of southwestCMORPH-CRT parts of presents the Xiang a relatively River Basin poor performance,show high agreement with underestimation with larger R-values. among most The parts IMERG of the basin.theThe basin. BIAS Thevalues BIAS ofvalues IMERG of IMERG V05 range V05 fromrange from20% –20% to 10%, to 10%, while while thevalues the values range range from from20% to V05 presents–20% to the20% highest for TRMM accuracy, 3B42V7. followed IMERG V05 by underestimatesthe− TRMM 3B42V7. most parts Regarding of the basin, the resultswhile TRMM of− BIAS, 20% for TRMM 3B42V7. IMERG V05 underestimates most parts of the basin, while TRMM 3B42V7 CMORPH-CRT3B42V7 shows presents the opposite a relatively pattern. poor performance, with underestimation among most parts of shows the opposite pattern. the basin. TheIn terms BIAS of RMSE,values TRMM of IMERG 3B42V7 V05 shows range the fromsmallest –20% RMSE to values10%, while compared the withvalues IMERG range V05 from –20% toandIn 20% terms CMORPH-CRT. for of TRMM RMSE, 3B42V7. TRMMSmall values 3B42V7IMERG of RMSE showsV05 areunderestimates the observ smallested in the RMSE most central valuesparts part of comparedof the basinbasin, withwith while IMERG TRMM V05 3B42V7and CMORPH-CRT.V05, shows the eastern the opposite part Small with valuespattern. TRMM of RMSE3B42V7, are and observed the southwest in the centralpart with part CMORPH-CRT. of the basin with As for IMERG NSE, V05, theIn easternit terms is obvious partof RMSE, withthat all TRMMTRMM three satellite-based 3B42V7,3B42V7 shows and theprecipitation the southwest smallest products RMSE part with valuesexhibit CMORPH-CRT. comparedgood performance with As IMERG for (NSE NSE, > V05 it is andobvious CMORPH-CRT.0.8) thatover allthe threewhole Small satellite-based basin. values Approximately of RMSE precipitation are82%, observ 90%, products anded in95% the exhibit of central grids good showed part performance of a thehigh basin value (NSE with of NSE> IMERG0.8 (>) over 0.9) for the CMORPH-CRT, IMERG V05, and TRMM 3B42V7 techniques, respectively. Higher NSE V05,the the whole eastern basin. part Approximately with TRMM 82%,3B42V7, 90%, and and the 95% southwest of grids showedpart with a high CMORPH-CRT. value of NSE As (> 0.9)for NSE, for the it CMORPH-CRT,is obvious that IMERGall three V05, satellite-based and TRMM 3B42V7precipitation techniques, products respectively. exhibit good Higher performance NSE values (NSE between > 0.8)satellite-based over the whole precipitation basin. Approximately products and 82%, in situ 90%, observations and 95% of weregrids mostly showed observed a high value in the of south NSE and(> 0.9)north for partsthe CMORPH-CRT, of the Xiang River IMERG Basin. V05, Poor and performance TRMM 3B42V7 of NSE techniques, values were respectively. mainly illustrated Higher NSE in the eastern basin with the highest elevation. Above all, these three satellite-based precipitation products are satisfactory in terms of R-value, BIAS, RMSE, and NSE for both catchment and grid scale, with acceptable accuracy and relatively good Remote Sens. 2019, 11, x FOR PEER REVIEW 9 of 18

values between satellite-based precipitation products and in situ observations were mostly observed Remote Sens.in the2019 south, 11, 1483and north parts of the Xiang River Basin. Poor performance of NSE values were mainly 9 of 17 illustrated in the eastern basin with the highest elevation. Above all, these three satellite-based precipitation products are satisfactory in terms of R-value, performanceBIAS, RMSE, in most and of NSE the for Xiang both Rivercatchment Basin. and Among grid scale, these with three acceptable products, accuracy generally, and relatively IMERG V05 performsgood the performance best based in on most the of results the Xiang of evaluationRiver Basin. Among indices. these three products, generally, IMERG V05 performs the best based on the results of evaluation indices.

FigureFigure 6. Spatial 6. Spatial distribution distribution of of thethe R-value, relative relative bias bias(BIAS), (BIAS), root mean root sq meanuare error square (RMSE), error and (RMSE), and Nash–SutcliNash–Sutcliffeffe eefficiencyfficiency (NSE) (NSE) va valueslues based based on onan anevaluation evaluation of daily of daily 0.5° × 0.5 0.5° grid-based0.5 grid-based ◦ × ◦ precipitationprecipitation data. data.

4.2. Evaluation4.2. Evaluation of the of Performance the performance of of Satellite-Based satellite-based precipitation Precipitation on drought on Drought monitoring Monitoring

4.2.1. Evaluation with Meteorological Stations The SPI-1 was calculated based on CMORPH-CRT, IMERG V05, TRMM 3B42V7 and in situ observations at catchment scale and station scale in the Xiang River Basin from 2015 to 2017. As stated above, the SPI-1 is used to explore the drought conditions. R-value, RMSE, and NSE are utilized to evaluate the performance of the satellite-based precipitation products on drought monitoring against the in situ observations. There are no BIAS values because the cumulative precipitation distribution is transformed to a normal distribution with a mean of zero [44]. The comparison of areal SPI-1 is shown in Figure7. Remote Sens. 2019, 11, x FOR PEER REVIEW 10 of 18

4.2.1. Evaluation with meteorological stations The SPI-1 was calculated based on CMORPH-CRT, IMERG V05, TRMM 3B42V7 and in situ observations at catchment scale and station scale in the Xiang River Basin from 2015 to 2017. As stated above, the SPI-1 is used to explore the drought conditions. R-value, RMSE, and NSE are utilized to evaluate the performance of the satellite-based precipitation products on drought monitoring against the in situ observations. There are no BIAS values because the cumulative precipitation distribution is transformed to a normal distribution with a mean of zero [44]. The comparison of areal SPI-1 is shown in Figure 7. Similar to the results of evaluation about the accuracy of precipitation estimates, IMERG V05 shows the best performance in SPI-1 estimations compared with CMORPH-CRT and TRMM 3B42V7, with the highest R-value (0.96) and NSE (0.8) and the lowest RMSE (0.29) (Figure 7). The SPI-1 estimates based on CMORPH-CRT and TRMM 3B42V7 are also in good agreement with those based on in situ observations. CMORPH-CRT performs better than TRMM 3B42V7, withSPI relatively higher R-value and NSE and a low RMSE. These results indicate that all the three satellite-basedRemote Sens. 2019 precipitation, 11, 1483 products are suitable for drought monitoring in the Xiang River Basin,10 of 17 while IMERG V05 performs the best on drought monitoring.

FigureFigure 7. Comparison 7. Comparison of the of theone-month one-month standardized standardized prec precipitationipitation index index (SPI-1) (SPI-1) calculated calculated from from thethe satellite-basedsatellite-based precipitation precipitation products products and and in situ in situ observations. observations.

BecauseSimilar of tothe the common results oftime evaluation period of about the select the accuracyed three precipitation of precipitation datasets estimates, and the IMERG reason V05 mentionedshows the above, best performance only SPI-1 inis SPI-1used estimations to investigate compared their withapplication CMORPH-CRT on drought and TRMMmonitoring. 3B42V7, However,with the since highest CMORPH R-value (0.96)and TRMM and NSE data (0.8) cover and the long lowester time RMSE periods, (0.29) othe (Figurer SPI7). time The SPI-1 periods estimates are alsobased utilized on CMORPH-CRTto justify whether and the TRMM results 3B42V7 based on are SPI-1 also inare good solid. agreement Therefore, with based those on basedthe available on in situ dataobservations. of in situ observations, CMORPH-CRT TRMM performs 3B42V7, better and than CMORPH-CRT, TRMM 3B42V7, the withSPI common relatively time period higher is R-valuefrom 1998and to NSE 2013. and The a low SPI-3, RMSE. SPI-6, These and results SPI-12 indicate based that on all in the situ three observations, satellite-based TRMM precipitation 3B42V7, products and CMORPH-CRTare suitable for are drought calculated monitoring at catchment in the Xiang scale River over Basin, Xiang while River IMERG Basin. V05 Figure performs 8 shows the bestthe on R-value,drought RMSE, monitoring. and NSE results based on the comparison between in situ observations and satellite-basedBecause precipitation of the common products. time period Similar of to the th selectede results threeof evaluation precipitation about datasets the SPI-1 and estimates the reason frommentioned CMORPH-CRT above, only and SPI-1TRMM is used3B42V7, to investigate the CMORPH-CRT their application performs on droughtbetter than monitoring. TRMM 3B42V7, However, especiallysince CMORPH the SPI-6 andand TRMMSPI-12. dataThe SPI-6 cover and longer SPI-12 time of periods, CMORPH-CRT other SPI shows time periodssimilar areperformance, also utilized whichto justify indicates whether that a the longer results time based scale on of SPI-1SPI based are solid. on CMORPH-CRT Therefore, based does on not the have available a significant data of in impactsitu observations,on its performance TRMM of 3B42V7, drought and monitoring. CMORPH-CRT, However, the common the longer time the period time isscales from of 1998 SPI, to the 2013. worseThe the SPI-3, performance SPI-6, and of SPI-12 TRMM based 3B42V7 on in on situ drought observations, monitoring. TRMM Short 3B42V7, time scale and of CMORPH-CRT SPI (e.g., SPI-1 are or calculatedSPI-3) might at catchmentmore suitable scale for over TRMM Xiang 3B42V7 River Basin.for drought Figure estimations8 shows the at R-value, catchment RMSE, scale and over NSE Xiangresults River based Basin. on the comparison between in situ observations and satellite-based precipitation products. Similar to the results of evaluation about the SPI-1 estimates from CMORPH-CRT and TRMM 3B42V7, the CMORPH-CRT performs better than TRMM 3B42V7, especially the SPI-6 and SPI-12. The SPI-6 and SPI-12 of CMORPH-CRT shows similar performance, which indicates that a longer time scale of SPI based on CMORPH-CRT does not have a significant impact on its performance of drought monitoring. However, the longer the time scales of SPI, the worse the performance of TRMM 3B42V7 on drought monitoring. Short time scale of SPI (e.g., SPI-1 or SPI-3) might more suitable for TRMM 3B42V7 for drought estimations at catchment scale over Xiang River Basin. The short-term records of precipitation from IMERG limited its investigation for long-term drought monitoring, however, in this study, the high resolution but short-term IMERG precipitation product

shows the best performance in both precipitation detection and drought monitoring over Xiang River Basin. Thus, the use of the IMERG precipitation product for further application in drought monitoring is still worth exploring. Based on the above analysis, it can be seen that the SPI-1 has comparable performance with other SPI time periods. Therefore, considering the short available time period of IMERG, the following analysis is just based on SPI-1. Time series of SPI-1 values from the three satellite-based precipitation products and in situ observations are shown in Figure9. This illustrates that SPI-1 based on IMERG V05 captures the variations well, indicating that IMERG V05 can be an alternative precipitation source instead of in situ observations for drought monitoring. Compared with IMERG V05 and CMORPH-CRT, TRMM 3B42V7 is less suitable for drought monitoring over the Xiang River Basin, because of the disparities along the time. With regard to the capability of drought detection at station scale, 10 meteorological stations within the Xiang River Basin were selected for statistical metric calculation (Table2). NSE values for the SPI-1 Remote Sens. 2019, 11, 1483 11 of 17 based on IMERG V05 are mainly between 0.42 and 0.78. However, the values of NSE for SPI-1 based on CMORPH-CRT are between 0.11 and 0.51 and for SPI-1 based on TRMM 3B42V7, the values are between 0.06 and 0.39. These results indicate that IMERG V05 apparently has better capability in drought monitoring than those of CMORPH-CRT and TRMM 3B42V7. Moreover, SPI based on the IMERG V05 shows a higher correlation with in situ observations than TRMM 3B42V7 and CMORPH-CRT, with the R-values ranging from 0.58 to 0.95. As for RMSE, similar to the above results, the RMSE values of IMERG V05 are lower than those of CMORPH-CRT and TRNN 3B42V7 in most meteorological stations. Remote Sens. 2019, 11, x FOR PEER REVIEW 11 of 18

Figure 8. Comparison of the SPI-3, SPI-6, and SPI-12 calculated from the satellite-based precipitation Remoteproducts Sens. 2019 and, 11 in, x situFOR observations.PEER REVIEW 12 of 18 Figure 8. Comparison of the SPI-3, SPI-6, and SPI-12 calculated from the satellite-based precipitation products and in situ observations.

The short-term records of precipitation from IMERG limited its investigation for long-term drought monitoring, however, in this study, the high resolution but short-term IMERG precipitation product shows the best performance in both precipitation detection and drought monitoring over Xiang River Basin. Thus, the use of the IMERG precipitation product for further application in drought monitoring is still worth exploring. Based on the above analysis, it can be seen that the SPI-1 has comparable performance with other SPI time periods. Therefore, considering the short available time period of IMERG, the following analysis is just based on SPI-1. Time series of SPI-1 values from the three satellite-based precipitation products and in situ observations are shown in Figure 9. This illustrates that SPI-1 based on IMERG V05 captures the variations well, indicating that IMERG V05 can be an alternative precipitation source instead of in situ observations for drought monitoring. Compared with IMERG V05 and CMORPH-CRT, TRMM 3B42V7 is less suitable for drought monitoring over the Xiang River Basin, because of the disparities along the time. FigureFigure 9. 9.Temporal Temporal comparisoncomparison of of 1-month 1-month SPI SPI values values between between satellite-based satellite-based precipitation precipitation products products andand in in situ situ observations. observations.

With regard to the capability of drought detection at station scale, 10 meteorological stations within the Xiang River Basin were selected for statistical metric calculation (Table 2). NSE values for the SPI-1 based on IMERG V05 are mainly between 0.42 and 0.78. However, the values of NSE for SPI-1 based on CMORPH-CRT are between 0.11 and 0.51 and for SPI-1 based on TRMM 3B42V7, the values are between 0.06 and 0.39. These results indicate that IMERG V05 apparently has better capability in drought monitoring than those of CMORPH-CRT and TRMM 3B42V7. Moreover, SPI based on the IMERG V05 shows a higher correlation with in situ observations than TRMM 3B42V7 and CMORPH-CRT, with the R-values ranging from 0.58 to 0.95. As for RMSE, similar to the above results, the RMSE values of IMERG V05 are lower than those of CMORPH-CRT and TRNN 3B42V7 in most meteorological stations.

Table 2. SPI-1 comparison between satellites and in situ in each meteorological station.

Station CMORPH IMERG TRMM Station CMORPH IMERG TRMM

R-value 0.73 0.95 0.49 R-value 0.70 0.76 0.57 Malingpo RMSE 0.73 0.32 1.00 RMSE 0.77 0.69 0.93

NSE 0.50 0.78 0.24 NSE 0.51 0.49 0.27

R-value 0.39 0.66 0.61 R-value 0.61 0.76 0.48 RMSE 1.10 0.82 0.88 Changning RMSE 0.88 0.69 1.01

NSE 0.16 0.43 0.39 NSE 0.36 0.48 0.20

R-value 0.25 0.69 0.49 R-value 0.53 0.70 0.57 Shuangfeng RMSE 1.22 0.79 1.02 Daoxian RMSE 0.97 0.78 0.93

NSE 0.11 0.51 0.27 NSE 0.32 0.42 0.30

R-value 0.43 0.71 0.24 R-value 0.54 0.58 0.51 RMSE 1.06 0.77 1.23 RMSE 0.97 0.93 1.00

NSE 0.25 0.49 0.06 NSE 0.35 0.42 0.23

R-value 0.63 0.78 0.52 R-value 0.71 0.62 0.34 RMSE 0.85 0.67 0.97 Lianxian RMSE 0.75 0.88 1.15

NSE 0.43 0.59 0.37 NSE 0.44 0.48 0.11

Remote Sens. 2019, 11, 1483 12 of 17

Table 2. SPI-1 comparison between satellites and in situ in each meteorological station.

Station CMORPH IMERG TRMM Station CMORPH IMERG TRMM R-value 0.73 0.95 0.49 R-value 0.70 0.76 0.57 Malingpo RMSE 0.73 0.32 1.00 Hengyang RMSE 0.77 0.69 0.93 NSE 0.50 0.78 0.24 NSE 0.51 0.49 0.27 R-value 0.39 0.66 0.61 R-value 0.61 0.76 0.48 Shaoyang RMSE 1.10 0.82 0.88 Changning RMSE 0.88 0.69 1.01 Remote Sens. 2019, NSE11, x FOR PEER 0.16 REVIEW 0.43 0.39 NSE 0.36 0.48 0.20 13 of 18 R-value 0.25 0.69 0.49 R-value 0.53 0.70 0.57 Shuangfeng RMSE 1.22 0.79 1.02 Daoxian RMSE 0.97 0.78 0.93 NSE 0.11 0.51 0.27 NSE 0.32 0.42 0.30 For comparisonR-value between 0.43 two 0.71different 0.24 drought indices, R-value the atmospheric 0.54 water 0.58 deficit 0.51 (AWD) Nanyue RMSE 1.06 0.77 1.23 Chenzhou RMSE 0.97 0.93 1.00 was also used toNSE estimate 0.25 the drought 0.49 conditions 0.06 at catchment NSE(Figure 10) 0.35 and station 0.42 scale (Table 0.23 3). Figure 10 showsR-value the comparison 0.63 re 0.78sults of 0.52 the one-week AWD R-value calculated 0.71 from the 0.62 satellite-based 0.34 Yongzhou RMSE 0.85 0.67 0.97 Lianxian RMSE 0.75 0.88 1.15 precipitation productsNSE and 0.43 in situ observations. 0.59 0.37 CMORPH-CRT-based NSE AWD 0.44 presents 0.48 similar 0.11 values of R-value and RMSE with those of IMERG V05, with a high R-value of nearly 0.9 and a low RMSE belowFor 6.1. comparison BIAS of both between CMORPH-CRT two different and drought IMERG indices,V05 based the AWD atmospheric are lower water than deficitTRMM (AWD) 3B42V7 was(–50.6%), also used with to bias estimate values the between drought –2% conditions and 12%. at Apparently, catchment (Figure AWD based10) and on station the IMERG scale V05 (Table shows3). a higher correlation with in situ observations than TRMM 3B42V7 and CMORPH-CRT, with the Figure 10 shows the comparison results of the one-week AWD calculated from the satellite-based R-value ranging from 0.7 to 0.99 (Table 3). As for BIAS and RMSE, the TRMM 3B42V7-based AWD precipitation products and in situ observations. CMORPH-CRT-based AWD presents similar values shows very high bias values in most meteorological stations. The values of BIAS and RMSE for of R-value and RMSE with those of IMERG V05, with a high R-value of nearly 0.9 and a low RMSE IMERG V05-based AWD is relatively low compared to those of CMORPH-CRT in most below 6.1. BIAS of both CMORPH-CRT and IMERG V05 based AWD are lower than TRMM 3B42V7 meteorological stations. These results indicate that the IMERG V05 is superior to TRMM3B42V7 and ( 50.6%), with bias values between 2% and 12%. Apparently, AWD based on the IMERG V05 −CMORPH-CRT in AWD estimates. − shows a higher correlation with in situ observations than TRMM 3B42V7 and CMORPH-CRT, with the Above all, SPI-1 based on satellite-based precipitation products have good performance in both R-value ranging from 0.7 to 0.99 (Table3). As for BIAS and RMSE, the TRMM 3B42V7-based AWD temporal and spatial analyses with acceptable accuracy and capability for drought monitoring in the shows very high bias values in most meteorological stations. The values of BIAS and RMSE for Xiang River Basin. Even though the R-value of satellite-based AWD is not much higher than IMERG V05-based AWD is relatively low compared to those of CMORPH-CRT in most meteorological satellite-based SPI, the RMSE values of satellite-based AWD are much larger than those of SPI-1 stations. These results indicate that the IMERG V05 is superior to TRMM3B42V7 and CMORPH-CRT based on satellite-based precipitation products. Therefore, SPI-1 might be more suitable for drought in AWD estimates. monitoring in the Xiang River Basin.

FigureFigure 10. 10.Comparison Comparison of the of one-week the one-week AWD calculatedAWD calculat fromed the from satellite-based the satellite-based precipitation precipitation products andproducts in situ observations.and in situ observations.

AboveTable all,3. SPI-1One-week based AWD on satellite-based comparison precipitationbetween satellites products and havein situ good observations performance at ineach both temporalmeteorological and spatial station. analyses with acceptable accuracy and capability for drought monitoring in the Xiang River Basin. Even though the R-value of satellite-based AWD is not much higher than satellite-basedStation SPI, theCMORPH RMSE valuesIMERG of satellite-based TRMM Station AWD are much largerCMORPH than IMERG those of SPI-1 TRMM based on satellite-based precipitation products. Therefore, SPI-1 might be more suitable for drought R-value 0.53 0.90 0.96 R-value 0.47 0.69 0.51 monitoring in the Xiang River Basin. Malingpo Bias 24.55 37.02 12.13 Hengyang Bias –44.28 –25.55 –110.89

RMSE 15.55 7.49 5.29 RMSE 19.34 15.44 20.00

R-value 0.71 0.81 0.71 R-value 0.97 0.99 0.97 Shaoyang Bias 33.12 37.38 33.12 Changning Bias -25.24 21.86 -50.23

RMSE 9.25 7.54 9.25 RMSE 13.76 12.97 11.35

R-value 0.56 0.71 0.62 R-value 0.96 0.98 0.98

Shuangfeng Bias –28.46 –26.75 –83.04 Daoxian Bias –18.77 –22.78 –44.85

RMSE 16.02 12.88 16.09 RMSE 20.84 13.58 10.50

R-value 0.65 0.73 0.77 R-value 0.97 0.98 0.93 Nanyue Bias 128.59 128.60 –124.50 Chenzhou Bias 64.18 53.12 –59.03

RMSE 12.42 10.23 8.00 RMSE 15.54 10.86 10.25

Remote Sens. 2019, 11, 1483 13 of 17

TableRemote 3. Sens.One-week 2019, 11, x FOR AWD PEER REVIEW comparison between satellites and in situ observations 14 at of 18 each meteorological station.

R-value 0.71 0.82 0.74 R-value 0.96 0.98 0.93 Station CMORPH IMERG TRMM Station CMORPH IMERG TRMM R-value 0.53 0.90 0.96 R-value 0.47 0.69 0.51 Yongzhou Bias –4.97 16.02 –112.60 Lianxian Bias –52.17 7.03 –65.48 Malingpo Bias 24.55 37.02 12.13 Hengyang Bias 44.28 25.55 110.89 − − − RMSERMSE 15.559.14 7.28 7.49 8.59 5.29 RMSE RMSE 10.59 19.3410.75 15.44 10.61 20.00 R-value 0.71 0.81 0.71 R-value 0.97 0.99 0.97 Shaoyang Bias 33.12 37.38 33.12 Changning Bias 25.24 21.86 50.23 4.2.2. Evaluation with grid-based precipitation − − RMSE 9.25 7.54 9.25 RMSE 13.76 12.97 11.35 To furtherR-value evaluate 0.56 the accuracy 0.71 of three 0.62 satellite-based precipitation R-value products 0.96 on0.98 drought 0.98 Shuangfeng Bias 28.46 26.75 83.04 Daoxian Bias 18.77 22.78 44.85 − − − − − − monitoring,RMSE the spatial 16.02 patterns of 12.88 R-value, 16.09RMSE, and NSE of SPI-1 RMSE based on 20.84 these products 13.58 are 10.50 conductedR-value and shown 0.65in Figure 11. 0.73 0.77 R-value 0.97 0.98 0.93 Nanyue Bias 128.59 128.60 124.50 Chenzhou Bias 64.18 53.12 59.03 It is obvious that the spatial distributions− of R-value, RMSE, and NSE of SPI-1 based on TRMM− RMSE 12.42 10.23 8.00 RMSE 15.54 10.86 10.25 3B42V7 are similar to those of IMERG V05, with high R-values and NSE values in the south basin R-value 0.71 0.82 0.74 R-value 0.96 0.98 0.93 Yongzhouand relatively Bias low R-values4.97 in the 16.02 west basin.112.60 Poor performances Lianxian Biasof R-value, 52.17NSE, and RMSE 7.03 are 65.48 − − − − observed inRMSE CMORPH-CRT 9.14 product, 7.28 with low 8.59 R-values and NSE RMSEvalues and 10.59high RMSE 10.75values in 10.61 most of the basin. These results indicate that IMERG V05 and TRMM 3B42V7 are more suitable for 4.2.2. Evaluationdrought monitoring with Grid-Based with SPI-1 than Precipitation CMORPH-CRT at grid scale. However, these satellite-based SPI values all show good performance in the southern Xiang River Basin, which is located in a region Towith further relatively evaluate low altitude the accuracy compared of to threesurrounding satellite-based regions. The precipitation values of R-value products and NSE on are drought monitoring,high and the the spatial RMSE patternsvalues are of low R-value, in the southern RMSE, Xiang and River NSE Basin, of SPI-1 implying based that on satellite-based these products are conductedprecipitation and shown products in Figure can be 11 used. to monitor drought in the southern basin, while uncertainties are much larger in the western basin.

Figure 11. Spatial distribution of the R-values and RMSE values based on an evaluation of daily 0.5 0.5 grid-based precipitation data. ◦ × ◦ It is obvious that the spatial distributions of R-value, RMSE, and NSE of SPI-1 based on TRMM 3B42V7 are similar to those of IMERG V05, with high R-values and NSE values in the south basin Remote Sens. 2019, 11, 1483 14 of 17 and relatively low R-values in the west basin. Poor performances of R-value, NSE, and RMSE are observed in CMORPH-CRT product, with low R-values and NSE values and high RMSE values in most of the basin. These results indicate that IMERG V05 and TRMM 3B42V7 are more suitable for drought monitoring with SPI-1 than CMORPH-CRT at grid scale. However, these satellite-based SPI values all show good performance in the southern Xiang River Basin, which is located in a region with relativelyRemote Sens. low2019, altitude 11, x FOR compared PEER REVIEW tosurrounding regions. The values of R-value and NSE are high 15 of and 18 the RMSE values are low in the southern Xiang River Basin, implying that satellite-based precipitation productsFigure 11. can Spatial be used distribution to monitor of the drought R-values in theand southernRMSE valu basin,es based while onuncertainties an evaluationare of daily much 0.5° larger × 0.5° in thegrid-based western precipitation basin. data.

4.3. Analysis of drought Drought conditions Conditions in in the the Xiang Xiang River River Basin Basin The SPI-1 based on IMERG V05 are selected for catchment drought studies from 2015 to 2017. AsAs ZhuZhu etet al.al. [[34]34] mentioned,mentioned, AugustAugust basically represents the driest period in the Xiang River Basin. Therefore,Therefore, thethe spatialspatial distributionsdistributions ofof SPI-1SPI-1 inin AugustAugust fromfrom 20152015 toto 20172017 areare selectedselected toto bebe compared.compared. As shown in Figure 1212,, August of 2015 susufferedffered the most severe and wide range of droughts, withwith thethe driestdriest areaarea mainlymainly inin thethe southwestsouthwest partpart ofof thethe XiangXiang RiverRiver Basin.Basin. However, thethe driestdriest partpart of the basin in 20152015 tendedtended toto becomebecome wet inin 2016,2016, butbut thethe northernnorthern basinbasin stillstill remainedremained under mild drought in AugustAugust 2016.2016. During During August August of of 2017, th thee severity of drought in the north part of was relieved comparedcompared to to 2015 2015 and and 2016, 2016, but thebut moderate the moderate drought drought shifted fromshifted the from north the to thenorth southwest to the partsouthwest of the basin.part of As the a basin. whole, As drought a whole, conditions drought in conditions 2015 were in apparently 2015 were more apparently severe thanmore insevere 2016 andthan 2017 in 2016 and and the severity2017 and and the range severity of droughts and range in theof droughts Xiang River in Basinthe Xiang present River a decreasing Basin present trend a fromdecreasing 2015 to trend 2017. from Even 2015 though to the2017. severity Even ofthough drought the was severity reduced of indrought August was 2017, reduced there still in remainsAugust a2017, small there area still with remains moderate a small drought area inwith the moderate southwest drought basin. in the southwest basin.

Figure 12. The distributions of IMERG V05 SPI-1 in August from 2015 to 2017. Figure 12. The distributions of IMERG V05 SPI-1 in August from 2015 to 2017. 5. Conclusions 5. Conclusions In this study, the accuracy of three satellite-based precipitation products, namely CMORPH-CRT, IMERGIn V05,this andstudy, TRMM the 3B42V7, accuracy was evaluatedof three with satellite-based the in situ observations precipitation and theproducts, datasets ofnamely daily 0.5CMORPH-CRT,0.5 grid-based IMERG precipitation V05, and TRMM over China 3B42V7, obtained was evaluated from CMA. with Their the applicationsin situ observations on drought and ◦ × ◦ monitoringthe datasets were of daily also evaluated0.5° × 0.5°with grid-based grid-based precipitation SPI-1. In addition,over China the obtained efficiency from of satellite-based CMA. Their SPI-1applications values wereon drought explored monitoring by comparing were with also the ev AWDaluated based with on grid-based meteorological SPI-1. variables In addition, at station the scale.efficiency Moreover, of satellite-based the drought conditionsSPI-1 values in Augustwere explored from 2015 by to comparing 2017 were analyzedwith the basedAWD on based IMERG on V05meteorological SPI-1 values. variables The main at conclusionsstation scale. are Moreover, as follows: the drought conditions in August from 2015 to 2017 were analyzed based on IMERG V05 SPI-1 values. The main conclusions are as follows: (1) IMERG V05 precipitation product shows the highest accuracy in the Xiang River Basin. 1) IMERG V05 precipitation product shows the highest accuracy in the Xiang River Basin. CMORPH-CRT performs better than the TRMM 3B42V7 precipitation product at both catchment CMORPH-CRT performs better than the TRMM 3B42V7 precipitation product at both catchment and grid scales. and grid scales. (2)2) BasedBased on evaluationevaluation results results of SPI-1,of SPI-1, IMERG IMERG V05 shows V05 theshows best the performance best performance in SPI-1 estimations in SPI-1 estimationsat both station at both and gridstation scales. and Thegrid CMORPH-CRT scales. The CMORPH-CRT has a better performance has a better than performance TRMM 3B42V7 than TRMMon drought 3B42V7 monitoring on drought at stationmonitoring scale, at but station shows scale, the worstbut shows performance the worst at performance grid scale. at grid scale. 3) The drought conditions of the Xiang River Basin in 2015 were apparently more severe than in 2016 and 2017, with the driest area mainly in the southwest part of the basin. The severity and range of droughts in the Xiang River Basin present a decreasing trend from 2015 to 2017. In summary, IMERG V05 precipitation product is more suitable for drought monitoring in the Xiang River Basin. The performances of CMORPH-CRT and TRMM 3B42V7 are acceptable but the application for drought estimation of these two products still needs further determination.

Remote Sens. 2019, 11, 1483 15 of 17

(3) The drought conditions of the Xiang River Basin in 2015 were apparently more severe than in 2016 and 2017, with the driest area mainly in the southwest part of the basin. The severity and range of droughts in the Xiang River Basin present a decreasing trend from 2015 to 2017.

In summary, IMERG V05 precipitation product is more suitable for drought monitoring in the Xiang River Basin. The performances of CMORPH-CRT and TRMM 3B42V7 are acceptable but the application for drought estimation of these two products still needs further determination.

Author Contributions: Conceptualization, Q.Z. and Y.L.; methodology, Y.L.; software, D.Y.; validation, Q.Z., Y.-P.X. and Y.L.; formal analysis, Y.L.; investigation, Q.Z.; resources, G.W.; data curation, D.Y.; writing—original draft preparation, Y.L.; writing—review and editing, Q.Z.; visualization, Y.L.; supervision, Q.Z.; project administration, Q.Z.; funding acquisition, Q.Z. Funding: This study was financially supported by the Natural Science Foundation of Jiangsu Province (BK20180403) and the National Key Research and Development Programs of China (2016YFA0601501). Acknowledgments: Thank the National Meteorological Information Center of China Meteorological Administration for archiving the observed climate data (http://data.cma.cn/). Thank the National Aeronautics Space Agency (NASA) and the National Ocean and Atmospheric Administration (NOAA)for providing the satellite-based precipitation data (https://pmm.nasa.gov/data-access/downloads/trmm; https://pmm.nasa.gov/ data-access/downloads/gpm; ftp://ftp.cpc.ncep.noaa.gov/precip/). Authors have great thanks to Jie Wang for technical supporting and the anonymous reviewers for providing so valuable comments. Conflicts of Interest: The authors declare no conflict of interest.

References

1. Edenhofer, O.; Pichs-Madruga, R.; Sokona, Y.; Farahani, E.; Kadner, S.; Seyboth, K.; Adler, A.; Baum, I.; Brunner, S.; Eickemeier, P.; et al. (Eds.) IPCC, 2014: Climate Change 2014: Mitigation of Climate Change. Contribution of Working Group III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change; Cambridge University Press: Cambridge, UK, 2014. 2. Dai, A. Increasing drought under global warming in observations and models. Nat. Clim. Chang. 2013, 3, 52. [CrossRef] 3. Dai, A. Drought under global warming: A review. Wiley Interdisciplinary Reviews. Clim. Chang. 2011, 2, 45–65. 4. Tao, H.; Fischer, T.; Zeng, Y.; Fraedrich, K. Evaluation of TRMM 3B43 precipitation data for drought monitoring in Jiangsu Province, China. Water 2016, 8, 221. [CrossRef] 5. Katsanos, D.; Retalis, A.; Tymvios, F.; Michaelides, S. Study of extreme wet and dry periods in Cyprus using climatic indices. Atmos. Res. 2018, 208, 88–93. [CrossRef] 6. Tan, M.; Tan, K.; Chua, V.; Chan, N. Evaluation of TRMM Product for Monitoring Drought in the Kelantan River Basin, Malaysia. Water 2017, 9, 57. [CrossRef] 7. Tian, Y.; Xu, Y.-P.; Wang, G. Agricultural drought prediction using climate indices based on Support Vector Regression in Xiangjiang River basin. Sci. Total Environ. 2018, 622, 710–720. [CrossRef][PubMed] 8. Zhong, R.; Chen, X.; Lai, C.; Wang, Z.; Lian, Y.; Yu, H.; Wu, X. Drought monitoring utility of satellite-based precipitation products across mainland China. J. Hydrol. 2019, 568, 343–359. [CrossRef] 9. Toté, C.; Patricio, D.; Boogaard, H.; van der Wijngaart, R.; Tarnavsky, E.; Funk, C. Evaluation of Satellite Rainfall Estimates for Drought and Flood Monitoring in Mozambique. Remote Sens. 2015, 7, 1758–1776. [CrossRef] 10. Fan, F.M.; Collischonn, W.; Quiroz, K.; Sorribas, M.; Buarque, D.; Siqueira, V. Flood forecasting on the Tocantins River using ensemble rainfall forecasts and real-time satellite rainfall estimates. J. Flood Risk Manag. 2016, 9, 278–288. [CrossRef] 11. Zambrano, F.; Wardlow, B.; Tadesse, T.; Lillo-Saavedra, M.; Lagos, O. Evaluating satellite-derived long-term historical precipitation datasets for drought monitoring in Chile. Atmos. Res. 2017, 186, 26–42. [CrossRef] 12. Gao, F.; Zhang, Y.; Ren, X.; Yao, Y.; Hao, Z.; Cai, W. Evaluation of CHIRPS and its application for drought monitoring over the Haihe River Basin, China. Nat. Hazard 2018, 92, 155–172. [CrossRef] 13. Tobin, K.J.; Bennett, M.E. Adjusting Satellite Precipitation Data to Facilitate Hydrologic Modeling. J. Hydrometeorol. 2010, 11, 966–978. [CrossRef] Remote Sens. 2019, 11, 1483 16 of 17

14. Zhao, H.; Yang, S.; Wang, Z.; Zhou, X.; Luo, Y.; Wu, L. Evaluating the suitability of TRMM satellite rainfall data for hydrological simulation using a distributed hydrological model in the Weihe River catchment in China. J. Geogr. Sci. 2015, 25, 177–195. [CrossRef] 15. Huffman, G.J.; Bolvin, D.T.; Nelkin, E.J.; Wolff, D.B.; Adler, R.F.; Gu, G.; Hong, Y.; Bowman, K.P.; Stocker, E.F. The TRMM multisatellite precipitation analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeorol. 2007, 8, 38–55. [CrossRef] 16. Joyce, R.J.; Janowiak, J.E.; Arkin, P.A.; Xie, P. CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeorol. 2004, 5, 487–503. [CrossRef] 17. Huffman, G.; Bolvin, D.; Braithwaite, D.; Hsu, K.; Joyce, R.; Kidd, C.; Sorooshian, S.; Xie, P.; Yoo, S.-H. Developing the Integrated Multi-Satellite Retrievals for GPM (IMERG). Available online: http://meetingorganizer.copernicus.org/EGU2012/EGU2012-6921.pdf (accessed on 27 April 2012). 18. Zeng, H.; Li, L.; Li, J. The evaluation of TRMM Multisatellite Precipitation Analysis (TMPA) in drought monitoring in the Lancang River Basin. J. Geogr. Sci. 2012, 22, 273–282. [CrossRef] 19. Lingtong, D.; Qingjiu, T.; Yan, H.; Jun, L. Drought monitoring based on TRMM data and its reliability validation in Shandong province. Trans. Chin. Soc. Agric. Eng. 2012, 28, 121–126. 20. Naumann, G.; Barbosa, P.; Carrao, H.; Singleton, A.; Vogt, J. Monitoring drought conditions and their uncertainties in Africa using TRMM data. J. Appl. Meteorol. Climatol. 2012, 51, 1867–1874. [CrossRef] 21. De Jesús, A.; Breña-Naranjo, J.; Pedrozo-Acuña, A.; Alcocer Yamanaka, V. The use of TRMM 3B42 product for drought monitoring in Mexico. Water 2016, 8, 325. [CrossRef] 22. Santos, C.A.G.; Neto, R.M.B.; de Araújo Passos, J.S.; da Silva, R.M. Drought assessment using a TRMM-derived standardized precipitation index for the upper São Francisco River basin, Brazil. Environ. Monit. Assess. 2017, 189, 250. [CrossRef] 23. Zhao, Q.; Chen, Q.; Jiao, M.; Wu, P.; Gao, X.; Ma, M.; Hong, Y. The temporal-spatial characteristics of drought in the Loess Plateau using the remote-sensed TRMM precipitation data from 1998 to 2014. Remote Sens. 2018, 10, 838. [CrossRef] 24. Xia, L.; Zhao, F.; Mao, K.; Yuan, Z.; Zuo, Z.; Xu, T. SPI-Based Analyses of Drought Changes over the Past 60 Years in China’s Major Crop-Growing Areas. Remote Sens. 2018, 10, 171. [CrossRef] 25. Lessel, J.; Sweeney, A.; Ceccato, P. An agricultural drought severity index using quasi-climatological anomalies of remotely sensed data. Int. J. Remote Sens. 2016, 37, 913–925. [CrossRef] 26. Lu, J.; Jia, L.; Menenti, M.; Yan, Y.; Zheng, C.; Zhou, J. Performance of the Standardized Precipitation Index Based on the TMPA and CMORPH Precipitation Products for Drought Monitoring in China. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2018, 11, 1387–1396. [CrossRef] 27. Monteleone, B.; Martina, M. Remote-sensing based model for drought identification. Available online: http://adsabs.harvard.edu/abs/2018EGUGA.2019658M (accessed on 23 April 2018). 28. Prat, O.; Leeper, R.; Bell, J.; Nelson, B.; Adams, J.; Ansari, S. Toward Earlier Drought Detection Using Remotely Sensed Precipitation Data from the Reference Environmental Data Record (REDR) CMORPH. Available online: http://adsabs.harvard.edu/abs/2018EGUGA.2011468P (accessed on 13 April 2018). 29. Wang, F.; Yang, H.; Wang, Z.; Zhang, Z.; Li, Z. Drought Evaluation with CMORPH Satellite Precipitation Data in the Basin by Using Gridded Standardized Precipitation Evapotranspiration Index. Remote Sens. 2019, 11, 485. [CrossRef] 30. Rhee, J.; Carbone, G.J. Estimating drought conditions for regions with limited precipitation data. J. Appl. Meteorol. Clim. 2011, 50, 548–559. [CrossRef] 31. Jang, S.; Center, A.C.; Rhee, J.; Center, A.C.; Yoon, S.; Center, A.C.; Lee, T.; Park, K.; Center, A.C. Evaluation of GPM IMERG Applicability Using SPI based Satellite Precipitation. J. Korean Soc. Agric. Eng. 2018, 59, 29–39. 32. Rushi, B.R.; Mishra, V.; Ellenburg, W.L.; Qamer, F.M.; Limaye, A.S.; Irwin, D. Application of Satellite Remote Sensing in Drought Monitoring in Bangladesh Using a User-Friendly Tool. Available online: http://adsabs.harvard.edu/abs/2018AGUFMGC31K1381R (accessed on 21 June 2019). 33. Du, J.; Fang, J.; Xu, W.; Shi, P. Analysis of dry/wet conditions using the standardized precipitation index and its potential usefulness for drought/flood monitoring in Hunan Province, China. Stoch. Environ. Res. Risk Assess. 2013, 27, 377–387. [CrossRef] 34. Zhang, T.; Zhang, X.; Wu, H.; Shen, L. Analysis of Pan Evaporation Trend and Its Influence Factors in Xiangjiang River Basin. Progress. Inquis. Mutat. Clim. 2013, 9, 35–42. Remote Sens. 2019, 11, 1483 17 of 17

35. Ma, C.; Pan, S.; Wang, G.; Liao, Y.; Xu, Y.-P. Changes in precipitation and temperature in Xiangjiang River Basin, China. Theor. Appl. Climatol. 2016, 123, 859–871. [CrossRef] 36. Zhao, L.; Wu, J.; Fang, J. Robust response of streamflow drought to different timescales of meteorological drought in Xiangjiang River Basin of China. Adv. Meteorol. 2016.[CrossRef] 37. Zhu, Q.; Luo, Y.; Xu, Y.-P.; Tian, Y.; Yang, T. Satellite Soil Moisture for Agricultural Drought Monitoring: Assessment of SMAP-Derived Soil Water Deficit Index in Xiang River Basin, China. Remote Sens. 2019, 11, 362. [CrossRef] 38. Yufei, Z.; Jiang, Z. Assessing quality of grid daily precipitation datasets in China in recent 50 years. Plateau Meteorol. 2015, 34, 50–258. 39. Huffman, G.J.; Bolvin, D.T. Real-Time TRMM Multi-Satellite Precipitation Analysis Data Set Documentation. Available online: https://pmm.nasa.gov/sites/default/files/document_files/3B4XRT_doc_V7_180426.pdf (accessed on 26 April 2018). 40. Wei, G.; Lü, H.; Crow, W.T.; Zhu, Y.; Wang, J.; Su, J. Comprehensive Evaluation of GPM-IMERG, CMORPH, and TMPA Precipitation Products with Gauged Rainfall over Mainland China. Adv. Meteorol. 2018, 2018, 1–18. [CrossRef] 41. Jiang, S.-h.; Zhou, M.; Ren, L.-l.; Cheng, X.-r.; Zhang, P.-j. Evaluation of latest TMPA and CMORPH satellite precipitation products over Yellow River Basin. Water Sci. Eng. 2016, 9, 87–96. [CrossRef] 42. Li, N.; Tang, G.; Zhao, P.; Hong, Y.; Gou, Y.; Yang, K. Statistical assessment and hydrological utility of the latest multi-satellite precipitation analysis IMERG in Ganjiang River basin. Atmos. Res. 2017, 183, 212–223. [CrossRef] 43. Wang, Z.; Zhong, R.; Lai, C.; Chen, J. Evaluation of the GPM IMERG satellite-based precipitation products and the hydrological utility. Atmos. Res. 2017, 196, 151–163. [CrossRef] 44. Sharifi, E.; Steinacker, R.; Saghafian, B. Assessment of GPM-IMERG and Other Precipitation Products against Gauge Data under Different Topographic and Climatic Conditions in Iran: Preliminary Results. Remote Sens. 2016, 8, 135. [CrossRef] 45. Chen, F.; Li, X. Evaluation of IMERG and TRMM 3B43 Monthly Precipitation Products over Mainland China. Remote Sens. 2016, 8, 472. [CrossRef] 46. Tan, M.L.; Chua, V.P.; Tan, K.C.; Brindha, K. Evaluation of TMPA 3B43 and NCEP-CFSR precipitation products in drought monitoring over Singapore. Int. J. Remote Sens. 2018, 39, 2089–2104. [CrossRef] 47. McKee, T.B.; Doesken, N.J.; Kleist, J. The relationship of drought frequency and duration to time scales. In Proceedings of the 8th Conference on Applied Climatology, American Meteorological Society, Boston, MA, USA, 17–22 January 1993; 17, pp. 179–183. 48. Livada, I.; Assimakopoulos, V. Spatial and temporal analysis of drought in Greece using the Standardized Precipitation Index (SPI). Theor. Appl. Climatol. 2007, 89, 143–153. [CrossRef] 49. Sönmez, F.K.; Koemuescue, A.U.; Erkan, A.; Turgu, E. An analysis of spatial and temporal dimension of drought vulnerability in Turkey using the standardized precipitation index. Nat. Hazard 2005, 35, 243–264. [CrossRef] 50. Torres, G.M.; Lollato, R.P.; Ochsner, T.E. Comparison of Drought Probability Assessments Based on Atmospheric Water Deficit and Soil Water Deficit. Agron. J. 2013, 105, 428–436. [CrossRef] 51. Moran, M.; Rahman, A.; Washburne, J.; Goodrich, D.; Weltz, M.; Kustas, W. Combining the Penman-Monteith equation with measurements of surface temperature and reflectance to estimate evaporation rates of semiarid grassland. Agric. For. Meteorol. 1996, 80, 87–109. [CrossRef] 52. Martínez-Fernández, J.; Sánchez, N.; González-Zamora, A.; Gumuzzio-Such, A.; Herrero-Jiménez, C.M. Feasibility of the SMOS soil moisture for agricultural drought monitoring: Assessment with the soil water deficit index. IEEE Int. Geosci. Remote Sens. Symp. (IGARSS) 2015, 11, 976–979. 53. Albergel, C.; De Rosnay, P.; Gruhier, C.; Muñoz-Sabater, J.; Hasenauer, S.; Isaksen, L.; Kerr, Y.; Wagner, W. Evaluation of remotely sensed and modelled soil moisture products using global ground-based in situ observations. Remote Sens. Environ. 2012, 118, 215–226. [CrossRef]

© 2019 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 (http://creativecommons.org/licenses/by/4.0/).