1. Introduction
As one of the fundamental components of the global hydrologic cycle, precipitation connects the atmosphere with land surface by providing a key input that drive a range of land surface processes, and thus has a significant importance to the sustainable development of economy, ecology, agriculture and other fields (Li et al. 2018). Knowledge of the timing, intensity and distribution of precipitation is essential to understand the hydrological cycle and manage water resource effectively (Wake 2013). Ground measurement networks (rain gauge or ground-based radar) are either sparse in both time and space or do not exist at all due to the climatic conditions, terrain and other limited conditions, especially in remote mountainous areas (Li et al. 2015; Zhang et al. 2019). Satellite remote sensing, having more complete coverage, particularly over high altitudes and remote regions, is an effective way to obtain the timely precipitation information from regional to global scales and provides a possible alternative for ground-based measurements.
Many global precipitation datasets have been produced with improvements of remote sensing science, such as Climate Prediction Center morphing technique (CMORPH) (Joyce et al. 2004), Global Satellite Mapping of Precipitation (GSMaP) (Kubota et al. 2007), Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis (TMPA) (Huffman et al. 2007), Integrated Multisatellite Retrievals for Global Precipitation Measurement (IMERG) (Hou et al. 2014), Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks–Climate Data Record (PERSIANN-CDR) (Ashouri et al. 2015), and Climate Hazards Group Infrared Precipitation with Station Data (CHIRPS) (Funk et al. 2015). These satellite-based precipitation products have been widely used in various fields in recent years, such as hydrological modeling (Yong et al. 2010), drought and flood disasters (Chen and Zhao 2016), El Niño analysis (Hamid et al. 2001) and extreme rainfall detection (Katiraie-Boroujerdy et al. 2017). However, satellite-based remote sensing is an indirect way to measure precipitation, containing inherent systematic and random errors over a large range of spatiotemporal scales (Shen et al. 2014). This uncertainty of satellite precipitation products needs to be evaluated when they are used for practical application purposes.
Numerous studies have been carried out to evaluate the spatiotemporal pattern of errors in satellite precipitation products all over the world (Yong et al. 2015; Tan and Duan 2017; Prakash et al. 2018). It is shown that different satellite precipitation products have different performances since they vary in data sources, spatial and temporal resolutions, and retrieval algorithms. Sorooshian et al. (2011) pointed out that the error in a satellite precipitation estimate is closely related to rainfall intensity, season, land cover, geographical location and topography. The error of satellite precipitation estimates usually shows a decreasing trend when scaling up the spatial and temporal scales, and shows an increasing trend with increasing rainfall intensity (Kuang et al. 2012). The performance of a satellite precipitation product often shows great spatial variability (Ebrahimi et al. 2017), although generally satellite precipitation estimates are relatively reliable over flat, homogenous regions (Huffman et al. 2007; Hu et al. 2014). However, due to the poor water conservation capacity, limited rainfall amount and strong evapotranspiration, precipitation regime in arid mountainous areas is usually characterized by low intensity of rain rate but high spatial heterogeneity, which is difficult for spaceborne satellite sensors to capture (Tang et al. 2016a; Huang et al. 2016). Nevertheless, accurate and reliable precipitation estimation plays a vital role in ecoenvironment protection and agricultural production over arid mountainous areas (Lu et al. 2018). Almazroui (2011) studied the short-term rainfall climatology regime over an arid mountainous region (Saudi Arabia) using TMPA data for the period 1998–2009. It was found that TMPA data tend to underestimate rainfall in the wet season. Lu et al. (2018) evaluated the accuracy of two widely used satellite precipitation products (TMPA 3B43 and GPM IMERG) from 2014 to 2017 in Xinjiang, China, which is characterized by complex topography and dry climate. The results indicated that both TMPA 3B43 and GPM IMERG overestimate monthly precipitation overall, but GPM IMERG performs better than TMPA 3B43. Similar work had also been carried out by Ji and Chen (2012). However, there is still a lack of comprehensive comparison of multiple satellite-based precipitation estimates over these arid mountainous areas in order to evaluate their applicability for specific purposes over such regions. Moreover, the frequency of extreme events is increasing under climate change, which leads to a significant impact on production and livelihood especially for the arid mountainous areas (Xia et al. 2012). So far, only a few studies have focused on the evaluation of extreme rainfall detection applicability of satellite precipitation products (Jiang et al. 2017; Katiraie-Boroujerdy et al. 2017; Liu et al. 2017). Consequently, the comparative evaluation of various satellite precipitation products in the arid mountainous areas, especially for the extreme rainfall detection capability, needs further region-specific investigations.
To address the above mentioned knowledge gaps, this study assessed three satellite-based precipitation products (TMPA, CHIRPS, and PERSIANN-CDR) over a typical arid mountainous river basin (Kai-Kong River basin) of China with a long-term ground-based validation records from 2000 to 2015. The main objectives of the present work were to 1) evaluate the accuracy in spatiotemporal distribution of multiple satellite precipitation estimates over Kai-Kong River basin, 2) investigate the extreme rainfall detection capability of multiple satellite estimates, and 3) understand the spatial heterogeneity of the relationship between satellite precipitation estimates and its geographical control factors.
2. Materials and methods
a. Study area
The Kai-Kong River basin is located in Xinjiang Uygur Autonomous Region in the northwest of China, with a latitude of 40°25′–43°21′N and a longitude of 82°57′−90°39′E (Fig. 1a). It is on the southern slope of the Tianshan Mountains and the northern edge of the Tarim basin. The Kai-Kong River basin is composed of Kaidu River, Bosten Lake, and Kongque River (Fig. 1b) with a catchment area of 104 395 km2. The Kai-Kong River basin has a very complex landform, with the topography slanted downward from the northwest to the southeast. Far from the ocean, this region is dominated by a continental climate, with high levels of solar radiation, low rainfall, and high evaporation. The average annual precipitation of Kai-Kong River basin is only about 180 mm, but has obvious seasonal variations within a year: most of the precipitation falls in the summer (June–August), while the precipitation amount in winter (December–February) is very limited.
Study area of the Kai-Kong River basin: (a) elevation and the distribution of rain gauges and (b) land-cover types.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Study area of the Kai-Kong River basin: (a) elevation and the distribution of rain gauges and (b) land-cover types.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Study area of the Kai-Kong River basin: (a) elevation and the distribution of rain gauges and (b) land-cover types.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Desert and grassland are the two main land-cover types of the Kai-Kong River basin, occupying the majority of the total watershed area (87%). The upper mountain areas are mainly covered by grassland (30.1% of the whole basin) and glacier (5.5% of the whole basin). The desert is distributed in the downstream region of the basin, accounting for 56.9% of the total watershed area. The arable land is mainly distributed around Bosten Lake and rivers, where the water resources can support the growth of crops.
b. Datasets
1) Rain gauge data
The rain gauge data from 2000 to 2015 were obtained from the China Meteorological Administration, which collects hourly precipitation records from about 2400 automatic weather stations over China since January 1951. The dataset is the best ground-based rainfall observation of the study region (Shen et al. 2010; Zhang et al. 2011; Ma et al. 2016) and was subjected to a series of rigorous quality controls, including examining for extreme values and internal consistency checks, as well as removal of questionable data (Chen and Li 2016). Daily precipitation was accumulated over 24 h from 0800 China local time, which is equivalent to 0000 coordinated universal time (UTC). Therefore, the daily precipitation of gauges was directly compatible with the satellite precipitation estimates since the satellite precipitation estimates measure daily precipitation at 0000 UTC. However, restricted by terrain, the distribution of rain gauges is extremely sparse with only 37 rain gauges available within the Kai-Kong River basin and its surrounding area. This emphasizes the need of reliable alternative satellite precipitation datasets to enhance the understanding of the spatiotemporal pattern of precipitation over the entire basin. All of the satellite precipitation products evaluated here incorporate rain gauge data from a variety of sources using different schemes (Duan et al. 2016), including the Global Precipitation Climatology Centre (GPCC) monthly rain gauge analysis (used for TMPA), Global Precipitation Climatology Project (GPCP) monthly rain gauge data (used for PERSIANN-CDR) and multiple sourced rain gauge data (used for CHIRPS). The satellite precipitation products (TMPA, CHIRPS, and PERSIANN-CDR) used in this study are essentially satellite-gauge blended products, which had been adjusted in time and space using gauge data. Therefore, the 37 rain gauges in the study area were selected to evaluate the performance of the gauge blended satellite precipitation estimates in the study, and their locations are shown in Fig. 1a.
2) TMPA
TRMM launched in 1997 and is jointly developed by the U.S. National Aeronautics and Space Administration (NASA) and the Japan Aerospace Exploration Agency (JAXA) with the aim of studying rainfall for weather and climate research in tropical as well as subtropical regions. Five satellite sensors are onboard the TRMM satellite, including the Precipitation Radar (PR), TRMM Microwave Imager (TMI), Visible and Infrared Scanner (VIRS), Clouds and the Earth’s Radiant Energy System (CERES), and the Lightning Imaging Sensor (LIS) (Huffman et al. 2007). Among them, the PR, TMI and VIRS are the primary precipitation measuring instruments for the TRMM satellite. The TRMM mission offers three different levels of data processing, of which the level-3 products are widely used in hydrometeorological research. To provide the best satellite precipitation estimates, the level-3 TMPA products are designed to combine precipitation estimates from different satellite systems as well as rain gauges (Huffman et al. 2007; Tang et al. 2016a). The TMPA dataset contains two products with quasi-global coverage starting in January 1998: 3-hourly combined passive microwave–infrared radar (PMW-IR) near-real-time estimates without gauge-based adjustments (TMPA 3B42 RT) and combined PMW-IR-gauge post-real-time estimates of precipitation 2-month latency (TMPA 3B42). The post-real-time TMPA 3B42, version 7 (abbreviated as TMPA 3B42 v7), 3-hourly data cover 50°S–50°N, 180°–180° at 0.25° × 0.25° spatial resolution. The original TMPA 3B42 v7 data from 2000 to 2015 are available online (https://doi.org/10.5067/TRMM/TMPA/3H/7). Since the TMPA products represent a 3-h precipitation total centered on a specified time (0000, 0300, 0600, 0900, 1200, 1500, 1800, 2100 UTC), the daily precipitation is generated by calculating the mean rain rate from half the 0000 UTC and next-day 0000 UTC estimates and the seven intermediate 3-h estimates. This processing procedure is the common method at present to obtain the TMPA daily precipitation (Tan and Santo 2018).
3) CHIRPS
CHIRPS is a quasi-global rainfall dataset from 1981 to the near-present, which incorporates 0.05° resolution satellite imagery with in situ station data to create gridded rainfall time series for trend analysis and hydrologic forecasts. The CHIRPS data build on the “smart interpolation” techniques (Funk et al. 2015) and high resolution and long period of record precipitation estimates based on IR cold cloud duration (CCD) observations. The main datasets used to build CHIRPS products include the monthly precipitation climatology [Climate Hazards Group Precipitation climatology (CHPclim)], the CCD information based on thermal infrared data archived from National Oceanic and Atmospheric Administration’s (NOAA’s) Climate Prediction Center dataset (CPC) and NOAA National Climate Data Center, the version 2 atmospheric model rainfall field from the NOAA Climate Forecast System (CFS), the TMPA 3B42 v7 data, and the multisource rain gauge stations data (Duan et al. 2016). Three main components are involved in the CHIRPS process: CHPclim, the satellite-only Climate Hazards Group Infrared Precipitation (CHIRP), and the station blending procedure that produces the CHIRPS (Funk et al. 2015). On 12 February 2015, version 2.0 of CHIRPS (abbreviated as CHIRPS v2) was released and is available to the public (at
4) PERSIANN-CDR
PERSIANN is a satellite-based precipitation retrieval algorithm that is mainly based on combined IR data from geostationary satellites with estimates generated by an artificial neural network technique to convert the IR information into rain rates (Sorooshian et al. 2000). PERSIANN-CDR provides daily precipitation estimates at 0.25° spatial resolution for the quasi-global coverage of 60°N–60°S from 1983 to present, aiming at producing a consistent, long-term, high-resolution near-global dataset for hydrometeorology research (Ashouri et al. 2015; Tan and Santo 2018). The main datasets for the construction of PERSIANN-CDR product include the gridded satellite infrared data (GridSat-BI), National Centers for Environmental Prediction (NCEP) Stage IV radar data, and the GPCP monthly precipitation data at 2.5°. The PERSIANN-CDR product is first generated from the PERSIANN algorithm using GridSat-BI IR data as the main input satellite data. Second, the NCEP stage IV radar data are used to create the nonlinear regression parameters of the artificial neural network model. The GPCP monthly product is then used to remove the biases of the rain rate estimates. Last, the 3-hourly adjusted PERSIANN precipitation data are accumulated to the daily scale to produce the PERSIANN-CDR product. More details about the PERSIANN-CDR data can be found in Ashouri et al. (2015). The PERSIANN-CDR daily precipitation data from 2000 to 2015, which are freely available to the public (at https://www.ncei.noaa.gov/data/precipitation-persiann/access/) (Sorooshian et al. 2014) were used in the study.
5) Other geographical datasets
The Système Probatoire d’Observation de la Terre (SPOT) VEGETATION (VGT) normalized difference vegetation index (NDVI) (abbreviated as SPOT VGT NDVI) products are used to analyze the impact of vegetation cover on the performance of satellite precipitation. The SPOT VGT NDVI product is a 10-day composite dataset with 1-km spatial resolution from 1999 to present. Geometric processing, radiometric correction, atmospheric corrections, and synthesis composition of the SPOT VGT NDVI product are conducted in the VEGETATION image processing center to guarantee the product quality. It has been evaluated to be reliable in arid and semiarid area in northwest China (An et al. 2015). In this study, the 10-day SPOT VGT NDVI product from 2000 to 2015 at 1-km spatial resolution, which is freely available (at http://www.vito-eodata.be/collections/srv/eng/main.home) (Maisongrande et al. 2004), were used. The average annual NDVI data were acquired by averaging the values of the SPOT VGT NDVI product from 2000 to 2015. The original 1-km spatial resolution data were aggregated to 0.25° spatial resolution through averaging to make it consistent with the satellite precipitation data.
The Shuttle Radar Topography Mission Digital Elevation Model (SRTM DEM) data at 90-m spatial resolution were obtained from the NASA website (at http://www2.jpl.nasa.gov/srtm/) (Rabus et al. 2003) to analyze the impact of elevation on the performance of satellite precipitation. The original 90-m SRTM DEM data were aggregated to 0.25° spatial resolution. Furthermore, the longitude data and latitude data at 0.25° spatial resolutions were extracted from the projection information of the DEM data.
c. Methodology
1) Evaluation metrics
Because the rain gauge stations in the study area are very sparse, spatial interpolation based on gauge observations would result in serious errors (Lu et al. 2018). All statistical analysis between point-based rain gauge data and grid-based satellite precipitation estimates was conducted in a direct comparison way using the nearest neighbor (NN) method. In the NN method, the gauge station at the grid scale was extracted and matched to the nearest satellite pixel (Ma et al. 2016; Chen et al. 2018). In addition, another grid-to-point extraction technique [bilinear weighted interpolation (BWI)] was adopted to compare with the NN method to assess the spatial mismatch between (the point based) rain gauge data and (the grid based) satellite precipitation estimates (Ebrahimi et al. 2017). Seven commonly used indicators were selected to evaluate the performance of multiple satellite precipitation products (TMPA, CHIRPS, and PERSIANN-CDR) in the Kai-Kong River basin, including Pearson correlation coefficients (CC), relative bias (RB), relative root-mean-square error (RRMSE), probability of detection (POD), false alarm ratio (FAR) and critical success index (CSI), as shown in Table 1.
Statistical metrics used in evaluation and comparison. Notation: n represents the number of rain gauge stations; Si represents satellite precipitation estimates; Gi represents observed precipitation by rain gauge; H (hits) represents precipitation observed both by the rain gauge and satellite, M (misses) represents precipitation observed only by the rain gauge, and F (false alarms) represents precipitation observed only by the satellite.
The CC, RB, and RRMSE were selected to describe the degree of linear correlation, systematic bias and random error between satellite precipitation products and gauge observations (Yong et al. 2010; Li et al. 2013; Chen et al. 2018). All the analysis was carried out at multiple time scales (daily, monthly, seasonal, and annual), while the data records at monthly, seasonal, and annual scales were obtained by accumulating daily precipitation records. For the Kai-Kong River basin, the four seasons were defined as spring (March–May), summer (June–August), autumn (September–November), and winter (December–February).
The POD, FAR, and CSI were also calculated to quantitatively investigate the detection ability of satellite precipitation products by identifying rain/no rain events through thresholding at multiple values. POD indicates the fraction of precipitation occurrences correctly detected by the satellite products to the total number of actual precipitation events; FAR represents the fraction of precipitation occurrences falsely detected among all the events detected by the satellite estimates; CSI represents the overall fraction of precipitation events correctly detected by the satellite estimates. The satellite precipitation data show perfect agreement with rain gauge observations when POD = 1, FAR = 0, and CSI = 1. Considering relatively low daily precipitation in the study area, the thresholds at different rainfall rates (0.1, 0.2, 0.5, 0.7, 1, 2, 5, 10, 15, 20, 25, 30, 35, 40, 45, and 50 mm day−1) were set to evaluate the precipitation detection capability of multiple satellite estimates, especially for extreme precipitation.
2) Extreme rainfall indices
Four standard extreme indices introduced by the World Climate Research Programme (WRCP) were selected to evaluate how well the satellite precipitation products can detect and monitor the extreme precipitation events (Katiraie-Boroujerdy et al. 2017): two intensity indices (Rx1day and Rx5day) and two continuity indices [consecutive dry days (CDD) and consecutive wet days (CWD)]. Rx1day means annual maximum 1-day precipitation amount; Rx5day means annual maximum 5-day consecutive precipitation; CDD represents annual largest number of consecutive days with daily precipitation less than 1.0 mm; CWD represents annual largest number of consecutive days with daily precipitation greater than 1.0 mm (Miao et al. 2015). The four indices were calculated at each rain gauge once a year, and then the mean extreme rainfall indices were obtained by averaging all gauges in each year. In addition, the moving average method with a 5-yr window was used to smooth the mean extreme rainfall indices on time series of 16 years for checking satellite precipitation detection capability on interannual variation.
3) Seasonal trend analysis
4) Residual error analysis
3. Results
a. Spatiotemporal pattern of error
Table 2 shows the linear regression results of TMPA, CHIRPS, and PERSIANN-CDR against the rain gauge data from 2000 to 2015 over the Kai-Kong River basin at the daily, monthly, seasonal and annual scales based on the NN and BWI methods. The results based on the NN method indicated that the three products had low accuracy at daily scale with small CC (0.219, 0.252, and 0.227 for TMPA, CHIRPS, and PERSIANN-CDR, respectively) and large RRMSE (4.711, 5.196, and 4.368 for TMPA, CHIRPS, and PERSIANN-CDR, respectively). A significant improvement of CC values (above 0.6) and a significant reduction of RRMSE values (below 1.0) were seen as time scale increased beyond the daily scale. The best performance occurred at annual scale, and the largest improvement of performance was from daily to monthly scales. In terms of comparisons between different satellite precipitation products, the performance of the results of three satellite precipitation estimates at daily scale had only slight discrepancies. However, there were significant differences in accuracy among monthly, seasonal, and annual scales. PERSIANN-CDR had the worst performance compared to the other two satellite precipitation estimates. The CC values of CHIRPS were the largest at monthly, seasonal and annual scales. The positive RB values of CHIRPS at multiple time scales indicated that CHIRPS tended to overestimate the gauge precipitation in the Kai-Kong River basin, while the negative RB values of TMPA and PERSIANN-CDR indicated an underestimation. The results based on the BWI method were similar to those of the NN method at daily scale. At monthly, seasonal and annual scales, the BWI method had better performance than the NN method, especially for the TMPA product (mean CC increased from 0.737 to 0.788).
Evaluation results of TMPA, CHIRPS, and PERSIANN-CDR at multiple time scales over the Kai-Kong River basin.
Figure 2 shows the spatial distribution of mean seasonal precipitation of rain gauge observations and satellite estimates (TMPA, CHIRPS, and PERSIANN-CDR) from 2000 to 2015 in the Kai-Kong River basin. The spatial distributions of all three satellite precipitation products were consistent with the rain gauge observations, with precipitation decreasing from the northwest to the southeast. The highest precipitation was mainly distributed around the Tianshan Mountains where the elevations are very high (~4000 m). However, there was an obvious seasonal cycle of precipitation, which was captured by all the three datasets. The precipitation was mainly concentrated in summer, while the amount in winter was extremely small (about 10 mm). It can be seen from Fig. 2 that the spatial distribution of the seasonal PERSIANN-CDR precipitation had a very smooth descending gradient from northwest to southeast, compared with the TMPA and CHIRPS. In particular, the seasonal precipitation of PERSIANN-CDR was very small, and the maximum precipitation did not exceed 200 mm. Some abnormally large precipitation amounts were also found in the TMPA product, especially in Figs. 2b, 2f, and 2j, where the land-cover types are dominated by water and glaciers (Fig. 1b). Overall, compared with the TMPA and PERSIANN-CDR, the seasonal precipitation of CHIRPS was the most consistent with rain gauge observations, especially over the mountainous areas in the northwest of the Kai-Kong River basin.
Spatial distribution of the mean seasonal precipitation estimated by rain gauges, TMPA, CHIRPS, and PERSIANN-CDR over the Kai-Kong River basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Spatial distribution of the mean seasonal precipitation estimated by rain gauges, TMPA, CHIRPS, and PERSIANN-CDR over the Kai-Kong River basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Spatial distribution of the mean seasonal precipitation estimated by rain gauges, TMPA, CHIRPS, and PERSIANN-CDR over the Kai-Kong River basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Figure 3 shows the trend rates (slope) of the rain gauge, TMPA, CHIRPS, and PERSIANN-CDR estimated seasonal precipitation from 2000 to 2015 in the Kai-Kong River basin. Although the spatial distributions of the seasonal precipitation of rain gauge and three satellite estimates were consistent (Fig. 2), their trend rates varied greatly in space. TMPA and CHIRPS had similar slope trends of the seasonal precipitation with gauge observations, presenting increasing trends in spring and autumn and minimum variation trend in winter. In summer, the estimated precipitation decreased in the northern part of the basin but increased in the southern part. In contrast, PERSIANN-CDR showed an overall increasing trend in four seasons, failing to capture the decrease of the gauge observations.
Trend rates (slope) of the seasonal precipitation by rain gauges, TMPA, CHIRPS, and PERSIANN-CDR from 2000 to 2015.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Trend rates (slope) of the seasonal precipitation by rain gauges, TMPA, CHIRPS, and PERSIANN-CDR from 2000 to 2015.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Trend rates (slope) of the seasonal precipitation by rain gauges, TMPA, CHIRPS, and PERSIANN-CDR from 2000 to 2015.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Figure 4 presents the interannual variations of mean annual precipitation of the rain gauge, TMPA, CHIRPS, and PERSIANN-CDR from 2000 to 2015 using a moving average. The mean annual precipitation of rain gauge data showed a reversing interannual variation between the periods 2000–11 and 2012–15. The annual precipitation of rain gauge observations showed a slight decline during 2000–12 and an increase after 2012. The mean annual precipitation of CHIRPS showed the consistent interannual variation with rain gauge data. PERSIANN-CDR and TMPA also had an approximate interannual variation with rain gauge data; however, they both significantly underestimated rainfall compared against rain gauges, suggesting PERSIANN-CDR and TMPA may not be a good alternative for rain gauge observations in the study area.
Interannual variations of the mean annual precipitation of rain gauges, TMPA, CHIRPS and PERSIANN-CDR from 2000 to 2015.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Interannual variations of the mean annual precipitation of rain gauges, TMPA, CHIRPS and PERSIANN-CDR from 2000 to 2015.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Interannual variations of the mean annual precipitation of rain gauges, TMPA, CHIRPS and PERSIANN-CDR from 2000 to 2015.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
b. Performance with different rain event thresholds
Figure 5 shows the cumulative distribution functions (CDF) derived from the daily precipitation for the three satellite precipitation products (TMPA, CHIRPS, and PERSIANN-CDR) and the rain gauge data from 2000 to 2015 over the Kai-Kong River basin. The CHIRPS and rain gauge data showed good agreement between their CDF curves. The CDF curves of TMPA and PERSIANN-CDR were below the CDF curve of the rain gauge data when the rainfall was approximately less than 1.0 and 0.5 mm day−1, respectively, indicating the underestimation of TMPA and PERSIANN-CDR when the rainfall was small.
The cumulative distribution functions (CDF) of the TMPA, CHIRPS, and PERSIANN-CDR precipitation products and the rain gauge data.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
The cumulative distribution functions (CDF) of the TMPA, CHIRPS, and PERSIANN-CDR precipitation products and the rain gauge data.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
The cumulative distribution functions (CDF) of the TMPA, CHIRPS, and PERSIANN-CDR precipitation products and the rain gauge data.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Figures 6a–c present the results of three detection skill metrics (FAR, POD, and CSI) of the daily precipitation at different rainfall thresholds for TMPA, CHIRPS, and PERSIANN-CDR. As rainfall rate increases, the POD and CSI decreased and the FAR increased significantly, which indicated the limited capability of satellite sensors for detecting large rainfall events. PERSIANN-CDR had better scores when rainfall rates were less than 1.0 mm day−1, with high POD and CSI and low FAR scores. Compared with TMPA and PERSIANN-CDR, CHIRPS had higher POD and CSI values and lower FAR values when the rainfall rates were greater than 1.0 mm day−1, implying better precipitation detection capability over the study area. Figure 6d shows the number of observations for rain gauges and satellite estimates at different rainfall thresholds; the number of observations decreased with increasing in rainfall intensity, particularly above 25 mm day−1.
(a) FAR, (b) POD, and (c) CSI of the daily precipitation for satellite estimates at different rainfall thresholds; (d) number of observations for rain gauges and satellite estimates at different rainfall thresholds.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
(a) FAR, (b) POD, and (c) CSI of the daily precipitation for satellite estimates at different rainfall thresholds; (d) number of observations for rain gauges and satellite estimates at different rainfall thresholds.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
(a) FAR, (b) POD, and (c) CSI of the daily precipitation for satellite estimates at different rainfall thresholds; (d) number of observations for rain gauges and satellite estimates at different rainfall thresholds.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
c. Extreme rainfall detection skills
Figure 7 shows the scatterplots and box plots of four extreme rainfall indices between the rain gauge and TMPA, CHIRPS, and PERSIANN-CDR from 2000 to 2015. It could be seen that in the four extreme rainfall indices, CHIRPS was closest to those of gauge observations, showing the best detection capability of extreme rainfall. The values of Rx1day and Rx5day were underestimated by TMPA and PERSIANN-CDR. Since the satellite precipitation estimates were usually greater than zero when the gauges showed no rainfall events, TMPA and PERSIANN-CDR tended to underestimate CDD but overestimate CWD. Compared to PERSIANN-CDR and TMPA, CHIRPS had the smallest underestimation of CDD and overestimation of CWD.
Scatterplots and box chart of four extreme rainfall indices between rain gauges and TMPA, CHIRPS, and PERSIANN-CDR.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Scatterplots and box chart of four extreme rainfall indices between rain gauges and TMPA, CHIRPS, and PERSIANN-CDR.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Scatterplots and box chart of four extreme rainfall indices between rain gauges and TMPA, CHIRPS, and PERSIANN-CDR.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Figure 8 shows the interannual variations of four extreme rainfall indices of the rain gauge, TMPA, CHIRPS and PERSIANN-CDR from 2000 to 2015 using moving average. The interannual variations of Rx1day and Rx5day of satellite precipitation estimates were relatively consistent with the rain gauge observations, among which CHIRPS had the closest values. CHIRPS also reproduced the mean values and interannual variations of CDD and CWD, whereas TMPA and PERSIANN-CDR did not capture the interannual variations of CDD and CWD.
Interannual variations of four extreme rainfall indices of rain gauges, TMPA, CHIRPS, and PERSIANN-CDR from 2000 to 2015.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Interannual variations of four extreme rainfall indices of rain gauges, TMPA, CHIRPS, and PERSIANN-CDR from 2000 to 2015.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Interannual variations of four extreme rainfall indices of rain gauges, TMPA, CHIRPS, and PERSIANN-CDR from 2000 to 2015.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
d. Influence of geographical factors on precipitation spatial heterogeneity
Statistic regression analysis results between the mean daily satellite precipitation products (TMPA, CHIRPS, and PERSIANN-CDR) and the geographical factors (DEM, NDVI, latitude, and longitude) are shown in Table 3. It could be found that DEM and latitude had significant (P < 0.01) positive correlations with satellite precipitation estimates, indicating that the precipitation increased with the elevation and latitude in the Kai-Kong River basin. The large CC between NDVI and mean daily precipitation indicated that rainfall has a large impact on the growth of vegetation. The CC between PERSIANN-CDR and DEM (0.539) was obviously smaller than those of TMPA and CHIRPS (0.704 and 0.747, respectively), while CC values between PERSIANN-CDR and NDVI as well as latitude were the largest. Negative correlation between longitude and precipitation were observed, but it was not significant. Therefore, only three geographical factors (DEM, NDVI, and latitude) with large CC values were selected to calculate the residual errors. Figure 9 shows the residual errors between satellite precipitation and geographical factors. The DEM could represent the spatial distribution of precipitation with small residual errors, except for the water surface anomaly in the TMPA precipitation. The residual errors of NDVI were largely positive in water and glacier covered area, but negative in Yanqi plain, where the land-cover type is dominated by arable land. For the regions covered by mountains and water, latitude was poorly related to precipitation.
Statistic regression between the mean daily satellite precipitation estimates (TMPA, CHIRPS, and PERSIANN-CDR) and the geographical factors (DEM, NDVI, latitude, and longitude).
Residual error maps of satellite precipitation estimates against DEM, NDVI, and latitude.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Residual error maps of satellite precipitation estimates against DEM, NDVI, and latitude.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
Residual error maps of satellite precipitation estimates against DEM, NDVI, and latitude.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0167.1
4. Discussion
Previous studies have shown that the precipitation detection capability of satellite estimates at daily or even smaller scale is poor when compared to the monthly scale (Duan et al. 2016; Chen et al. 2018). Table 2 showed that the CC values of each of the satellite precipitation estimates at the daily scale were only about 0.25. Poor agreement between satellite precipitation products and rain gauge data at daily scale may be caused by errors in satellite sampling, errors in precipitation retrieval algorithms, and errors in bias correction algorithms (Shen et al. 2010; Duan et al. 2016). With increasing time scale, the accuracy of satellite precipitation increases (Chen et al. 2018). A large number of studies have shown that the accuracy (CC) of satellite precipitation at the monthly scale over homogeneous regions can reach about 0.80 and therefore meets the requirements for hydrological modeling (Yong et al. 2010; Zhang et al. 2019). This is because monthly gauge measurements data (such as the monthly GPCC data) are used to correct monthly satellite precipitation data (Ashouri et al. 2015; Chen et al. 2018), which significantly improves the satellite precipitation estimation accuracy at the monthly scale. As expected, the monthly CC of three satellite precipitation products reached about 0.70 in this study. CHIRPS had the best performance at monthly, seasonal and annual scales, compared with TMPA and PERSIANN-CDR (Table 2). It is mainly attributed to the blending of multiple source data in the precipitation retrieval algorithms, including the CHPclim, CCD information, TMPA 3B42 v7 data, version 2 atmospheric model rainfall field from the CFS, and the rain gauge stations data from multiple sources (Funk et al. 2015; Duan et al. 2016; Prakash 2019), which makes CHIRPS outperform others and shows the greatest consistency with rain gauge data. PERSIANN-CDR had the worst performance with the smallest CC and the largest RRMSE. Similar findings have also been reported in many other regions (Hu et al. 2014; Liu et al. 2015; Tan et al. 2018). This is mainly because the PERSIANN-CDR product uses the 2.5° monthly GPCP data in bias correction, while the other satellite precipitation products use the 1° monthly GPCC product (Tan et al. 2018).
In terms of the seasonal precipitation, due to the water vapor transported from the Atlantic Ocean and the Arctic Ocean (Zhang 2013; Jin et al. 2016), precipitation across different seasons decreased from northwest to southeast over the Kai-Kong River basin, with the summer precipitation being the largest (Fig. 2). PERSIANN-CDR failed to capture spatial differences and temporal trends in precipitation due to the fusion of the coarse 2.5° monthly GPCP data. Compared to CHIRPS and PERSIANN-CDR, TMPA significantly overestimated precipitation over water bodies and glaciers (Fig. 3) and had a significant difference from gauge observations in temporal trends of mean annual precipitation (Fig. 4). This phenomenon has also been found by Tang et al. (2016b). It is reported that the presence of ice and snow cover produces strong erroneous results in microwave-based precipitation retrievals of TMPA (Ebert et al. 2007; Tian et al. 2014), suggesting that TMPA may not be a good alternative for rain gauges in the arid mountainous area. Due to the limited extent of annual rainfall data in the time series of satellite precipitation estimates, it is difficult to completely capture the annual rainfall trend. However, the annual precipitation was well captured by CHIRPS based on the 16 years of data, showing a similar mean values and interannual variations to the rain gauge data.
The performance of satellite precipitation estimates especially for the extreme rainfall with different rain/no-rain thresholds showed that the difference of CDF between CHIRPS and rain gauge data was small, while CHIRPS had high POD and CSI values and low FAR values (Fig. 6), indicating a better ability of CHIRPS for precipitation detection at different rainfall thresholds, especially when the rainfall rates were greater than 1.0 mm day−1. With the increase in rainfall intensity, the POD and CSI decreased and the FAR increased significantly (Fig. 6), suggesting the limited capability of satellite sensors for detecting heavy rainfall events (Sun et al. 2016). However, it was found that the number of data used for the calculation of three metrics decreased with the increase in rainfall intensity, which may introduce noise into the results, especially when the daily rainfall was greater than 25 mm day−1. Results of the four extreme rainfall indices in Fig. 7 showed an obvious underestimation of CDD and an overestimation of CWD by TMPA and PERSIANN-CDR, demonstrating a limited ability for extreme rainfall detection. Satellite precipitation estimates are usually greater than zero when the gauges show no rainfall (Xu et al. 2017), which results in poor statistical results and interannual variations of CWD and CDD of satellite precipitation products. Furthermore, satellite sensors cannot adequately sample all rainfall events because of the coarse temporal resolutions (3 hourly or daily), leading to an obvious underestimation of Rx1day and Rx5day by TMPA and PERSIANN-CDR. Overall, the evaluation results showed that CHIRPS was more suitable for extreme rainfall detection over the Kai-Kong River basin.
Precipitation in the arid mountainous regions usually shows large spatial heterogeneity, even over the same underlying land surface (Chen et al. 2015). This study attempted to explain the spatial variations between satellite precipitation and geographical factors through the residual error analysis. It was found (Table 3) that the elevation had a significant impact on precipitation. Satellite precipitation products showed heavy precipitation in high elevation areas of the studied area, which may be caused by the orographic enhancement (Tang et al. 2018). Vegetation cover has a significant positive correlation with precipitation distribution over the arid mountainous areas due to the fact that the water required for vegetation growth is mainly supplied by precipitation over the arid area (Immerzeel et al. 2009; Jia et al. 2011). It should be noted that mean daily precipitation is equivalent to mean annual precipitation/365, and mean NDVI during 2000–15 is equivalent to mean annual NDVI. Therefore, the time scale of this analysis is equivalent to the annual time scale. The time lag between NDVI and precipitation can then be neglected. The positive CC values between mean daily satellite precipitation and NDVI could be found in Table 3. However, there was a large residual error between NDVI and precipitation over the Yanqi plain region in Fig. 9. Qu and Ma (1995) have noted that northwest China is a typical arid area with little precipitation, and the growth of green vegetation depends mainly on irrigation and groundwater. It was found that the negative residual error in Yanqi plain was mainly distributed in arable land (Figs. 1 and 9) where the vegetation growth is mainly fed by irrigation or groundwater, and therefore, rainfall cannot be explained by NDVI in such areas.
It should be highlighted that all of the evaluations were based on the assumption that the true value at the ground within each 0.25° pixel can be represented by the corresponding rain gauges. However, satellite precipitation is a homogenization of data at the 0.25° × 0.25° (~625 km2) grid scale, whereas the rain gauge data are the point-based observation. There is a typical scale mismatch issue between (point based) rain gauge data and (grid based) satellite precipitation estimates (Ebrahimi et al. 2017). Because the local precipitation variation typically occurs at scales of 2 km and larger (Orlanski 1975; Immerzeel et al. 2009), it is difficult for the sparse rain gauges to reflect the rainfall characteristics of satellite estimates at 0.25° grid scale, especially at the daily scale. To assess the spatial mismatch between (point based) rain gauge data and (grid based) satellite precipitation estimates, this study compared two grid-to-point extraction techniques (NN and BWI). Results showed that the BWI method performed better with relatively higher CC and lower RRMSE (Table 2) than those of the NN method especially at monthly, seasonal, and annual scales. This demonstrated that the BWI method was effective for exacting precipitation in arid mountainous areas with high accuracy. Since the main analysis in this study was based on daily precipitation and the BWI method calculated the average of nearby satellite pixel values (Ebrahimi et al. 2017), which may not be suitable for extreme rainfall extraction, all the analysis was conducted using the NN method.
The limited number of gauges could be a source of error in evaluation of satellite precipitation products (Tang et al. 2016a; Shen et al. 2014). Prakash et al. (2019) quantitatively assessed the uncertainty of monthly mean monsoon rainfall incurred by changes in rain gauge density over India, using a theoretical model and Monte Carlo simulations. It is shown that the uncertainty in spatially averaged monthly rainfall increases with the omitted fraction of rain gauges. The ungauged regions are exactly the places that could benefit the most from the satellite precipitation products; however, these regions lack of ground-based observations for validation. At present, the mainstream satellite precipitation products combine high quality PMW observations from low Earth orbit satellites and high spatiotemporal coverage VIRS/IR data from geostationary Earth orbiting satellites to derive precipitation (Ashouri et al. 2015). Since the rain gauge observations are generally considered as true values for validation (Tang et al. 2018), they are widely used for bias correction that is further applied to the PMW-IR blended products. For example, all of the three products (TMPA, CHIRPS, and PERSIANN-CDR) used in this study are satellite–gauge blended precipitation estimates. Limited by the harsh natural environment and the complex terrain, the available rain gauges used for evaluation are sparse in this study. However, the gauge-based evaluation results may be reasonable to certain extent and can be extended into the entire region (even for those places without gauges), because the satellite-based sensors are spatially continuous measurements. Future work should focus on the impact of rain gauge density on precipitation estimates; especially dense rain gauge stations should be obtained to make a more objective assessment for satellite precipitation products.
Overall, it was found in this study that, when compared with TMPA and PERSIANN-CDR, CHIRPS could capture the spatiotemporal distributions, interannual variations and extreme rainfalls of the gauged precipitation over the Kai-Kong River basin, which is dominated by a typical arid mountainous system. However, the CHIRPS satellite precipitation product at the daily scale still contains relatively large errors (Duan et al. 2016). More in-depth investigations should be taken to improve the applicability of CHIRPS to capture extreme rainfall using dense rain gauges, particularly in remote mountainous areas such as the Tibetan Plateau.
5. Conclusions
A comparative evaluation of multiple satellite precipitation products (TMPA, CHIRPS, and PERSIANN-CDR) was carried out over the Kai-Kong River basin for 2000–15. Satellite precipitation estimates had poor performance at the daily scale, but reasonably estimated rainfall at the monthly, seasonal, and annual scales. The daily precipitation detection ability of satellite estimates decreased gradually as rainfall rates increase. CHIRPS tended to overestimate the gauge precipitation across all time scales, whereas TMPA and PERSIANN-CDR tended to underestimate. The three satellite estimates could overall capture the interannual variations of intensity indices of extreme rainfall (Rx1day and Rx5day), but TMPA and PERSIANN-CDR failed to reproduce the temporal trends of continuity indices (CDD and CWD). Compared with TMPA and PERSIANN-CDR, CHIRPS had the closest interannual variation in annual precipitation and extreme rainfall indices to rain gauge data, and the largest CSI as well as POD when rainfall rates were greater than 1.0 mm day−1. Based on all the evaluation results, it can be concluded that CHIRPS was the best satellite precipitation product for application to the Kai-Kong River basin. Since precipitation in the arid mountainous regions is typically spatially heterogeneous, it is necessary to develop algorithms for fusing multisource data to enhance the spatiotemporal accuracy of satellite precipitation in future research.
Acknowledgments
This work is supported by the by the National Key Research and Development Program of China (2017YFC0404501), the National Natural Science Foundation of China (51709179, 51609230, and 51809175), and the fund of Nanjing Hydraulic Research Institute (Y918011).
REFERENCES
Almazroui, M., 2011: Calibration of TRMM rainfall climatology over Saudi Arabia during 1998–2009. Atmos. Res., 99, 400–414, https://doi.org/10.1016/j.atmosres.2010.11.006.
An, Y., W. Gao, Z. Gao, C. Liu, and R. Shi, 2015: Trend analysis for evaluating the consistency of Terra MODIS and SPOT VGT NDVI time series products in China. Front. Earth Sci., 9, 125–136, https://doi.org/10.1007/s11707-014-0428-9.
Ashouri, H., K. Hsu, S. Sorooshian, D. K. Braithwaite, K. R. Knapp, L. D. Cecil, B. R. Nelson, and O. P. Prat, 2015: PERSIANN-CDR: Daily precipitation climate data record from multisatellite observations for hydrological and climate studies. Bull. Amer. Meteor. Soc., 96, 69–83, https://doi.org/10.1175/BAMS-D-13-00068.1.
Chen, C., and S. Zhao, 2016: Drought monitoring and analysis of Huanghuai Hai Plain based on TRMM precipitation data. Remote Sens. Land Resour., 28, 122–129.
Chen, C., S. Zhao, Z. Duan, and Z. Qin, 2015: An improved spatial downscaling procedure for TRMM 3B43 precipitation product using geographically weighted regression. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 8, 4592–4604, https://doi.org/10.1109/JSTARS.2015.2441734.
Chen, C., and Coauthors, 2018: Multiscale comparative evaluation of the GPM IMERG v5 and TRMM 3B42 v7 precipitation products from 2015 to 2017 over a climate transition area of China. Remote Sens., 10, 944, https://doi.org/10.3390/rs10060944.
Chen, F., and X. Li, 2016: Evaluation of IMERG and TRMM 3B43 monthly precipitation products over mainland China. Remote Sens., 8, 472, https://doi.org/10.3390/rs8060472.
Duan, Z., J. Liu, Y. Tuo, G. Chiogna, and M. Disse, 2016: Evaluation of eight high spatial resolution gridded precipitation products in Adige basin (Italy) at multiple temporal and spatial scales. Sci. Total Environ., 573, 1536–1553, https://doi.org/10.1016/j.scitotenv.2016.08.213.
Ebert, E. E., J. E. Janowiak, and C. Kidd, 2007: Comparison of near-real-time precipitation estimates from satellite observations and numerical models. Bull. Amer. Meteor. Soc., 88, 47–64, https://doi.org/10.1175/BAMS-88-1-47.
Ebrahimi, S., C. Chen, Q. Chen, Y. Zhang, N. Ma, and Q. Zaman, 2017: Effects of temporal scales and space mismatches on the TRMM 3B42 v7 precipitation product in a remote mountainous area. Hydrol. Processes, 31, 4315–4327, https://doi.org/10.1002/hyp.11357.
Funk, C., P. Peterson, M. Landsfeld, D. Pedreros, J. Verdin, S. Shukla, and J. Michaelsen, 2015: The Climate Hazards Infrared Precipitation with Stations—A new environmental record for monitoring extremes. Sci. Data, 2, 150066, https://doi.org/10.1038/sdata.2015.66.
Hamid, E., Z. Kawasaki, and R. Mardiana, 2001: Impact of the 1997–98 El Niño event on lightning activity over Indonesia. Geophys. Res. Lett., 28, 147–150, https://doi.org/10.1029/2000GL011374.
Hou, A. Y., R. K. Kakar, S. Neeck, A. A. Azarbarzin, C. D. Kummerow, M. Kojima, and T. Iguchi, 2014: The Global Precipitation Measurement Mission. Bull. Amer. Meteor. Soc., 95, 701–722, https://doi.org/10.1175/BAMS-D-13-00164.1.
Hu, Q., D. Yang, Z. Li, A. K. Mishra, Y. Wang, and H. Yang, 2014: Multi-scale evaluation of six high-resolution satellite monthly rainfall estimates over a humid region in China with dense rain gauges. Int. J. Remote Sens., 35, 1272–1294, https://doi.org/10.1080/01431161.2013.876118.
Huang, A., Y. Zhao, Y. Zhou, B. Yang, L. Zhang, X. Dong, D. Fang, and Y. Wu, 2016: Evaluation of multisatellite precipitation products by use of ground-based data over China. J. Geophys. Res. Atmos., 121, 10 654–10 675, https://doi.org/10.1002/2016JD025456.
Huffman, G. J., D. T. Bolvin, E. J. Nelkin, D. B. Wolff, R. F. Adler, G. Gu, and E. F. Stocker, 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 38–55, https://doi.org/10.1175/JHM560.1.
Immerzeel, W. W., M. M. Rutten, and P. Droogers, 2009: Spatial downscaling of TRMM precipitation using vegetative response on the Iberian Peninsula. Remote Sens. Environ., 113, 362–370, https://doi.org/10.1016/j.rse.2008.10.004.
Ji, X., and Y. Chen, 2012: Characterizing spatial patterns of precipitation based on corrected TRMM 3B43 data over the mid Tianshan Mountains of China. J. Mt. Sci., 9, 628–645, https://doi.org/10.1007/s11629-012-2283-z.
Jia, S., W. Zhu, A. Lű, and T. Yan, 2011: A statistical spatial downscaling algorithm of TRMM precipitation based on NDVI and DEM in the Qaidam basin of China. Remote Sens. Environ., 115, 3069–3079, https://doi.org/10.1016/j.rse.2011.06.009.
Jiang, S., Z. Zhang, Y. Huang, X. Chen, and S. Chen, 2017: Evaluating the TRMM Multisatellite Precipitation Analysis for extreme precipitation and streamflow in Ganjiang River basin, China. Adv. Meteor., 2017, 2902493, https://doi.org/10.1155/2017/2902493.
Jin, X., H. Shao, C. Zhang, and Y. Yan, 2016: The applicability evaluation of three satellite products in Tianshan mountains. Ziran Ziyuan Xuebao, 31, 106–117.
Joyce, R. J., J. E. Janowiak, P. A. Arkin, and P. Xie, 2004: CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeor., 5, 487–503, https://doi.org/10.1175/1525-7541(2004)005<0487:CAMTPG>2.0.CO;2.
Katiraie-Boroujerdy, P., H. Ashouri, K. Hsu, and S. Sorooshian, 2017: Trends of precipitation extreme indices over a subtropical semi-arid area using PERSIANN-CDR. Theor. Appl. Climatol., 130, 249–260, https://doi.org/10.1007/s00704-016-1884-9.
Kuang, D., Y. Shen, Z. Niu, and L. Wang, 2012: Analysis on uncertainty of satellite retrieved precipitation products at various temporal & spatial resolutions and rain rate levels. Remote Sens. Inf., 2012, 75–81.
Kubota, T., and Coauthors, 2007: Global precipitation map using satellite-borne microwave radiometers by the GSMaP project: Production and validation. IEEE Trans. Geosci. Remote Sens., 45, 2259–2275, https://doi.org/10.1109/TGRS.2007.895337.
Li, D., G. Christakos, X. Ding, and J. Wu, 2018: Adequacy of TRMM satellite rainfall data in driving the SWAT modeling of Tiaoxi catchment (Taihu Lake basin, China). J. Hydrol., 556, 1139–1152, https://doi.org/10.1016/j.jhydrol.2017.01.006.
Li, Z., D. Yang, and Y. Hong, 2013: Multi-scale evaluation of high-resolution multi-sensor blended global precipitation products over the Yangtze River. J. Hydrol., 500, 157–169, https://doi.org/10.1016/j.jhydrol.2013.07.023.
Li, Z., D. Yang, B. Gao, Y. Jiao, Y. Hong, and T. Xu, 2015: Multiscale hydrologic applications of the latest satellite precipitation products in the Yangtze River basin using a distributed hydrologic model. J. Hydrometeor., 16, 407–426, https://doi.org/10.1175/JHM-D-14-0105.1.
Liu, J., Z. Duan, J. Jiang, and A. Zhu, 2015: Evaluation of three satellite precipitation products TRMM 3B42, CMORPH, and PERSIANN over a subtropical watershed in China. Adv. Meteor., 2015, 151239, https://doi.org/10.1155/2015/151239.
Liu, Y., Y. Wu, Z. Feng, Z. Huang, and D. Wang, 2017: Evaluation of a variety of satellite retrieved precipitation products based on extreme rainfall in China. Trop. Geogr., 37, 417–433.
Lu, X., M. Wei, G. Tang, and Y. Zhang, 2018: Evaluation and correction of the TRMM 3B43V7 and GPM 3IMERGM satellite precipitation products by use of ground-based data over Xinjiang, China. Environ. Earth Sci., 77, 209, https://doi.org/10.1007/s12665-018-7378-6.
Ma, Y., G. Tang, D. Long, B. Yong, L. Zhong, W. Wan, and Y. Hong, 2016: Similarity and error intercomparison of the GPM and its predecessor-TRMM Multisatellite Precipitation Analysis using the best available hourly gauge network over the Tibetan Plateau. Remote Sens., 8, 569, https://doi.org/10.3390/rs8070569.
Maisongrande, P., B. Duchemin, and G. Dedieu, 2004: VEGETATION/SPOT: An operational mission for the Earth monitoring; presentation of new standard products. Int. J. Remote Sens., 25, 9–14, https://doi.org/10.1080/0143116031000115265.
Miao, C., H. Ashouri, K. L. Hsu, S. Sorooshian, and Q. Duan, 2015: Evaluation of the PERSIANN-CDR daily rainfall estimates in capturing the behavior of extreme precipitation events over China. J. Hydrometeor., 16, 1387–1396, https://doi.org/10.1175/JHM-D-14-0174.1.
Orlanski, I., 1975: A rational subdivision of scales for atmospheric processes. Bull. Amer. Meteor. Soc., 56, 527–530, https://doi.org/10.1175/1520-0477-56.5.527.
Prakash, S., 2019: Performance assessment of CHIRPS, MSWEP, SM2RAIN-CCI, and TMPA precipitation products across India. J. Hydrol., 571, 50–59, https://doi.org/10.1016/j.jhydrol.2019.01.036.
Prakash, S., A. K. Mitra, A. AghaKouchak, Z. Liu, H. Norouzi, and D. S. Pai, 2018: A preliminary assessment of GPM-based multi-satellite precipitation estimates over a monsoon dominated region. J. Hydrol., 556, 865–876, https://doi.org/10.1016/j.jhydrol.2016.01.029.
Prakash, S., A. Seshadri, J. Srinivasan, and D. S. Pai, 2019: A new parameter to assess impact of rain gauge density on uncertainty in the estimate of monthly rainfall over India. J. Hydrometeor., 20, 821–832, https://doi.org/10.1175/JHM-D-18-0161.1.
Qu, Y., and S. Ma, 1995: Water and oasis in Hexi Corridor area of Gansu province, China. Ganhanqu Ziyuan Yu Huanjing, 9, 93–99.
Rabus, B., M. Eineder, A. Roth, and R. Bamler, 2003: The Shuttle Radar Topography Mission—A new class of digital elevation models acquired by spaceborne radar. ISPRS J. Photogramm. Remote Sens., 57, 241–262, https://doi.org/10.1016/S0924-2716(02)00124-7.
Shen, Y., A. Xiong, Y. Wang, and P. Xie, 2010: Performance of high-resolution satellite precipitation products over China. J. Geophys. Res., 115, D02114, https://doi.org/10.1029/2009JD012097.
Shen, Y., P. Zhao, Y. Pan, and J. Yu, 2014: A high spatiotemporal gauge-satellite merged precipitation analysis over China. J. Geophys. Res. Atmos., 119, 3063–3075, https://doi.org/10.1002%2F2013JD020686.
Sorooshian, S., K.-L. Hsu, X. Gao, H. V. Gupta, and D. Braithwaite, 2000: Evaluation of PERSIANN system satellite-based estimates of tropical rainfall. Bull. Amer. Meteor. Soc., 81, 2035–2046, https://doi.org/10.1175/1520-0477(2000)081<2035:EOPSSE>2.3.CO;2.
Sorooshian, S., and Coauthors, 2011: Advanced concepts on remote sensing of precipitation at multiple scales. Bull. Amer. Meteor. Soc., 92, 1353–1357, https://doi.org/10.1175/2011BAMS3158.1.
Sorooshian, S., and Coauthors, 2014: NOAA Climate Data Record (CDR) of Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Climate Data Record (PERSIANN-CDR), version 1. NOAA National Centers for Environmental Information, accessed 12 October 2018, https://doi.org/10.7289/V51V5BWQ.
Sun, R., H. Yuan, X. Liu, and X. Jiang, 2016: Evaluation of the latest satellite–gauge precipitation products and their hydrologic applications over the Huaihe River basin. J. Hydrol., 536, 302–319, https://doi.org/10.1016/j.jhydrol.2016.02.054.
Tan, J., W. A. Petersen, G. Kirchengast, D. C. Goodrich, and D. B. Wolff, 2018: Evaluation of global precipitation measurement rainfall estimates against three dense gauge networks. J. Hydrometeor., 19, 517–532, https://doi.org/10.1175/JHM-D-17-0174.1.
Tan, M. L., and Z. Duan, 2017: Assessment of GPM and TRMM precipitation products over Singapore. Remote Sens., 9, 720, https://doi.org/10.3390/rs9070720.
Tan, M. L., and H. Santo, 2018: Comparison of GPM IMERG, TMPA 3B42 and PERSIANN-CDR satellite precipitation products over Malaysia. Atmos. Res., 202, 63–76, https://doi.org/10.1016/j.atmosres.2017.11.006.
Tang, G., Y. Ma, D. Long, L. Zhong, and Y. Hong, 2016a: Evaluation of GPM day-1 IMERG and TMPA version-7 legacy products over mainland China at multiple spatiotemporal scales. J. Hydrol., 533, 152–167, https://doi.org/10.1016/j.jhydrol.2015.12.008.
Tang, G., D. Long, and Y. Hong, 2016b: Systematic anomalies over inland water bodies of high mountain Asia in TRMM precipitation estimates: No longer a problem for the GPM era? IEEE Geosci. Remote Sens. Lett., 13, 1762–1766, https://doi.org/10.1109/LGRS.2016.2606769.
Tang, G., A. Behrangi, D. Long, C. Li, and Y. Hong, 2018: Accounting for spatiotemporal errors of gauges: A critical step to evaluate gridded precipitation products. J. Hydrol., 559, 294–306, https://doi.org/10.1016/j.jhydrol.2018.02.057.
Tian, Y., Y. Liu, K. R. Arsenault, and A. Behrangi, 2014: A new approach to satellite-based estimation of precipitation over snow cover. Int. J. Remote Sens., 35, 4940–4951, https://doi.org/10.1080/01431161.2014.930208.
Wake, B., 2013: Rain from space. Nat. Climate Change, 3, 950, https://doi.org/10.1038/nclimate2042.
Xia, J., D. She, Y. Zhang, and H. Du, 2012: Spatio-temporal trend and statistical distribution of extreme precipitation events in Huaihe River basin during 1960–2009. J. Geogr. Sci., 22, 195–208, https://doi.org/10.1007/s11442-012-0921-6.
Xu, R., F. Tian, L. Yang, H. Hu, H. Lu, and A. Hou, 2017: Ground validation of GPM IMERG and TRMM 3B42V7 rainfall products over southern Tibetan Plateau based on a high-density rain gauge network. J. Geophys. Res. Atmos., 122, 910–924, https://doi.org/10.1002/2016JD025418.
Yong, B., L. L. Ren, Y. Hong, J. H. Wang, and W. Wang, 2010: Hydrologic evaluation of multisatellite precipitation analysis standard precipitation products in basins beyond its inclined latitude band: A case study in Laohahe basin, China. Water Resour. Res., 46, 759–768, https://doi.org/10.1029/2009WR008965.
Yong, B., D. Liu, J. J. Gourley, Y. Tian, G. J. Huffman, L. Ren, and Y. Hong, 2015: Global view of real-time TRMM Multisatellite Precipitation Analysis: Implications for its successor Global Precipitation Measurement Mission. Bull. Amer. Meteor. Soc., 96, 283–296, https://doi.org/10.1175/BAMS-D-14-00017.1.
Zhang, Q., V. P. Singh, J. Li, and X. Chen, 2011: Analysis of the periods of maximum consecutive wet days in China. J. Geophys. Res., 116, D23106, https://doi.org/10.1029/2011JD016088.
Zhang, X., 2013: Study on the downscaling algorithm of remote sensing precipitation and analysis of temporal-spatial characteristic in the middle section of Mount TianShan. M.D. dissertation, Northwest Normal University, 42 pp.
Zhang, Z., J. Tian, Y. Huang, X. Chen, S. Chen, and Z. Duan, 2019: Hydrologic evaluation of TRMM and GPM IMERG satellite-based precipitation in a humid basin of China. Remote Sens., 11, 431, https://doi.org/10.3390/rs11040431.