1. Introduction
Precipitation is one of the most important factors affecting the global water and energy balance (Futrell et al. 2005; Kidd and Huffman 2011). Researchers can hardly conduct better simulations of the water cycle over regions without accurate precipitation inputs (Xue et al. 2013). Traditionally, there are three ways to measure precipitation, that is, rain gauges, weather radars, and satellite-based sensors (Li et al. 2013). The rain gauge is a conventional way of providing the most straightforward observations of site-based surface precipitation. However, gauge networks are sparse over most of continents and few gauges are located over the ocean (Huffman et al. 2001; Kidd and Huffman 2011; Mishra and Coulibaly 2009). The weather radar can monitor precipitation with relatively higher temporal and spatial resolutions, but it is often subject to low data quality in complex terrain, mostly due to signal blockage, attenuation by rain, and vertical variability of reflectivity (Dinku et al. 2002; Tian and Peters-Lidard 2010). Currently, the only practical way to achieve a comprehensive estimate of precipitation on a global basis comes from Earth observation satellites (Hong et al. 2012; Hou et al. 2014; Villarini and Witold 2008). Global satellite-based rainfall products are currently based on passive microwave (PMW), calibrated infrared (IR), and PMW plus IR observations. IR sensors on geostationary Earth orbit (GEO) satellites can provide precipitation estimates at high temporal resolutions, but the accuracy of IR-based estimates is generally not very good because of the indirect linkage between IR signals and precipitation. PMW sensors are more popular in precipitation estimation since its radiative signatures are more directly linked to precipitating particles. However, PMW sensors are only on board low-Earth-orbiting (LEO) satellites at present, leading to low temporal sampling. Therefore, combining GEO IR and LEO PMW sensors to improve the accuracy, coverage, and resolution of global precipitation products has been widely recognized and applied (Hong et al. 2012).
Since the launch of the Tropical Rainfall Measuring Mission (TRMM) in late 1997, rapid development of precipitation datasets based on PMW, calibrated IR, and PMW plus IR observations has provided a tremendous amount of quasi-global information for research and applications (Hong et al. 2012). To date, a number of satellite precipitation products have been released to the public with various temporal and spatial resolutions, such as TRMM Multisatellite Precipitation Analysis (TMPA; Huffman et al. 2007, 2010), Climate Prediction Center (CPC) morphing technique (CMORPH; Joyce et al. 2004), Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN; Sorooshian et al. 2000), PERSIANN Cloud Classification System (PERSIANN-CCS; Hong et al. 2004), Naval Research Laboratory (NRL)-developed blended-satellite precipitation technique (NRL-Blend; Turk and Miller 2005), and Global Satellite Mapping of Precipitation (GSMap; Kubota et al. 2007). The TMPA products used in this study are the real-time 3B42 (hereafter 3B42RT) and post-real-time 3B42, version 7 (hereafter 3B42V7), products, which have been widely studied and applied in hydrological simulations and predictions (Habib et al. 2012; Hong et al. 2007; Li et al. 2013; Long et al. 2014; Prakash et al. 2015; Zulkafli et al. 2014). Research and applications of TMPA products have produced great scientific, societal, and economic benefits, such as extreme weather event prediction, disaster forecasts, and water resources planning and management. Despite the great achievement in the TRMM era, TRMM data have some inherent limitations associated with spatiotemporal coverage and uncertainty of solid or light precipitation estimation over higher latitudes and altitudes (Hong et al. 2012; Hou et al. 2014; Huffman et al. 2007; Yong et al. 2015). In addition, the last of the instruments on TRMM was turned off on 8 April 2015, and the spacecraft reentered the Earth’s atmosphere on 15 June 2015, most of which was expected to burn up in the atmosphere (http://pmm.nasa.gov/gpm-news/trmm-spacecraft-re-enters-over-tropics).
Building on the TRMM heritage, the Global Precipitation Measurement (GPM) mission is an international network of satellites that provides the next-generation global rain and snow products at a spatial resolution of 0.1° × 0.1° with a temporal resolution of half-hourly (Hou et al. 2008, 2014). As the TRMM successor, the GPM Core Observatory was launched on 27 February 2014, marking the beginning of the GPM era. The GPM constellation consists of this Core Observatory and has approximated 10 partner satellites during the study period. Composition of the constellation will change with the launch and failure of partner satellites. The GPM Core Observatory carries a Dual-Frequency Precipitation Radar (DPR; Ku band at 13.6 GHz and Ka band at 35.5 GHz) and a multichannel GPM Microwave Imager (GMI; frequencies range between 10 and 183 GHz), both of which improve on the capabilities of measuring precipitation compared to the previous TRMM instruments (Hou et al. 2014).
As an extension and also upgrade of the highly successful TRMM mission, GPM was anticipated to provide four levels of products based on various algorithms (Hou et al. 2014). This study focuses on a Day-1 multisatellite precipitation product provided by the Integrated Multisatellite Retrievals for GPM (IMERG) algorithm, which is intended to intercalibrate, merge, and interpolate all microwave (MW) estimates of the GPM constellation, IR estimates, gauge observations, and other potential sensors’ data with a 0.1° × 0.1° spatial resolution and 30-min temporal resolution (Huffman et al. 2014). The IMERG system is run twice in near–real time to produce “early” and “late” multisatellite products. Then the system is run once after the monthly gauge analysis is received and the “final” satellite–gauge products are produced (Huffman et al. 2015), which are used in our study. The final run IMERG products were released in January 2015 with nearly a year’s time span of data since mid-March 2014. It is expected that the first retrospectively reprocessed TRMM/GPM-era IMERG datasets (from March 2000 to present) will be released in early 2017 (Huffman et al. 2015), which will supersede the TMPA level-3 3B42 products. Thus, it is necessary and meaningful to conduct a detailed comparison between IMERG and TMPA products and to examine the utility of IMERG in hydrological modeling. Please note that the IMERG product mentioned and used in the context below is the Day-1 final run product.
The objectives of this study are therefore to 1) evaluate performance and improvement of the GPM-era IMERG product compared with TRMM-era 3B42V7 and 3B42RT estimates statistically and 2) explore the continuity between IMERG, 3B42V7, and 3B42RT when used in hydrological calibration and simulation in the Ganjiang River basin, a midlatitude basin in southeast China. Our study is among the very early attempts to evaluate the newest GPM product, and we expect that the results reported here can shed light on subsequent investigations. The remaining parts of this paper are organized as follows. Section 2 introduces the study area, datasets, and methodology, including a brief description of the CREST distributed model and its newest 2.1 version. Section 3 evaluates and compares daily precipitation data of IMERG, 3B42V7, and 3B42RT. Section 4 analyzes the results of the study, and section 5 concludes and recommends future studies.
2. Study area, data, and methodology
a. Study area
The Ganjiang River basin, with a drainage area of 81 258 km2 above the Waizhou hydrological station, is the seventh-largest subcatchment of the Yangtze River, located within 24°29′–29°21′N, 113°30′–116°40′E in southeast China (Fig. 1). There are nearly 20 types of soil in this basin, dominated by krasnozem (65.86%) and paddy (22.74%), according to the China Soil Scientific Database (http://www.soil.csdb.cn). Surrounded by mountains and dominated by low hills in the central part, the topography of the study basin is complex, with elevations ranging from 11 to 1997 m MSL. The northern part of the basin is an alluvial plain, where the outlet station of the Ganjiang River basin named Waizhou is located.
Map of the Ganjiang River basin showing streams and elevation of the basin, the streamflow gauge station (Waizhou), and rain gauges that provide ground reference to evaluate satellite-based precipitation products in this study.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Map of the Ganjiang River basin showing streams and elevation of the basin, the streamflow gauge station (Waizhou), and rain gauges that provide ground reference to evaluate satellite-based precipitation products in this study.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Map of the Ganjiang River basin showing streams and elevation of the basin, the streamflow gauge station (Waizhou), and rain gauges that provide ground reference to evaluate satellite-based precipitation products in this study.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
The basin has a subtropical humid monsoon climate with mean annual precipitation of ~1500 mm. The distribution of precipitation is characterized by strong spatial and temporal variability, with precipitation decreasing from high mountains to central low hills. The rainy season spans from April through June, with about 42% of the annual rainfall. Rainfall-triggered floods often occur between May and July, but sometimes may be delayed to September. The combination of the East Asian monsoon and complex topography makes the basin a typical flood-vulnerable area of China (Hu et al. 2014). Unpredictable precipitation and the resulting frequent floods across the Ganjiang River basin have led to tremendous economic, societal, and ecological losses, making it important and urgent to obtain accurate precipitation estimates over the basin. Seven flood events were observed in 2014, which caused a loss of ~400 million U.S. dollars according to the Ministry of Civil Affairs of China (http://www.mca.gov.cn/). Hu et al. (2014) conducted multiscale evaluation of six high-resolution satellite monthly rainfall estimates over the Ganjiang River basin and concluded that TRMM 3B43V6 is the best and that 3B42RTV6 and CMORPH performed better among five other pure satellite-derived products. Therefore, it is meaningful to perform a study to quantitatively evaluate the quality of the latest GPM data and its continuity to TRMM in this area.
b. In situ and satellite precipitation datasets
1) Gauged precipitation and discharge data
A total of 310 daily precipitation gauges from 2003 through 2009 were used (Fig. 1). These rain gauges were installed and maintained by the Jiangxi Province Hydrology Bureau and are independent of the Global Precipitation Climatology Centre (GPCC) networks (Hu et al. 2014). The inverse distance weighting (IDW) method (Ahrens 2006) was used to interpolate the rain gauge data into the gridded datasets as forcing to calibrate the hydrological model as described in section 4a.
In addition to the gauge network, we acquired the China Gauge-Based Daily Precipitation Analysis (CGDPA) product developed by the National Meteorological Information Center (Shen and Xiong 2015). CGDPA was produced using data from 1955 and is updated in real time at 0.25° × 0.25° resolution on the China Meteorological Data Sharing Service System (http://cdc.nmic.cn). The product employs climatological optimal interpolation (OI) that reduces errors caused by the spatial discontinuity of precipitation to obtain the spatial distribution of precipitation over the whole country based on ~2400 national rain gauges. This product is of high quality over southeast China because of the high density of national rain gauges in east China (Shen and Xiong 2015). Therefore, CGDPA can be used as the benchmark for satellite precipitation in both statistical and hydrological assessments over the study area after 2009 when the Jiangxi gauge observations mentioned above are not available. Values of the national rain gauges used in the CGDPA production were also obtained and used in the comparison with various satellite precipitation products at the grid scale in section 3a. There are thirty-nine 0.25° × 0.25° grid boxes with at least one rain gauge in each grid box over the basin. In addition, we also obtained daily streamflow data for the Waizhou station from 2003 through 2014 for evaluating hydrological modeling performance.
2) TRMM/GPM precipitation datasets
The TRMM satellite was launched by National Aeronautics and Space Administration (NASA) and National Space Development Agency (NASDA) in 1997. The TMPA products were designed to combine precipitation estimates from different satellite systems as well as rain gauges and were intended to provide the best satellite precipitation estimates (Huffman et al. 2007). The TMPA 3B42V7 and 3B42RT products from 2008 to 2014 were obtained from NASA archive (
The GPM mission has been providing level-3 IMERG final run (research) products at 0.1° × 0.1° and half-hourly resolutions since March 2014. By the time of this study, this GPM-era dataset was available only through September 2014 (downloaded from http://pmm.nasa.gov/data-access/downloads). Thus, this study used the rainy season for the Ganjiang River basin from May through September 2014. The processing flow for IMERG is similar with that for TMPA. The merged PMW estimate is also referred to as HQ, but the resolution is improved to 0.1° × 0.1° gridded half-hourly and including more sensors [GMI, TMI, SSMIS, Advanced Microwave Scanning Radiometer 2 (AMSR2), and MHS] than TMPA in the study period. The GEO-IR satellites used in IMERG are the same as those in TMPA. In regard to precipitation gauge analysis incorporated in each product, both IMERG and 3B42V7 use the GPCC Monitoring Product (version 4) currently provided from January 2007 through near present. When the full data reanalysis is updated to a longer record (from January 1901 through January 2010 currently), the IMERG datasets are expected to be reprocessed, taking advantage of the improved data (Huffman et al. 2015).
3) Potential evapotranspiration
Potential evapotranspiration (PET) data used in this study are from the global daily database, provided by the Famine Early Warning Systems Network (http://earlywarning.usgs.gov/fews). The daily global PET was calculated using climate parameter data from Global Data Assimilation System (GDAS) analysis fields on a spatial basis using the Penman–Monteith equation (Verdin et al. 2005).
c. CREST hydrological model
The Coupled Routing and Excess Storage (CREST) model (Khan et al. 2011; Wang et al. 2011) is a grid-based distributed hydrological model developed by the University of Oklahoma (http://hydro.ou.edu) and the NASA SERVIR Project Team (www.servir.net), which aims to take advantage of remotely sensed data as well as gauge observations for local, regional, and global hydrological analysis. It has already been deployed as a core model in several operational systems including the Flooded Locations and Simulated Hydrographs Project (FLASH; http://www.nssl.noaa.gov/projects/flash/) and the Near Realtime Global Hydrological Simulation and Flood Monitoring Demonstration System (http://eos.ou.edu/). The CREST model has been implemented successfully in a variety of multiscale hydrological studies (Meng et al. 2014; Wu et al. 2012; Xue et al. 2013; Tang et al. 2015; Zhang et al. 2015). The CREST model used in this study is the latest version CREST v2.1, in which a fully distributed linear reservoir routing (LRR) method (FDLRR) is implemented to replace the previous quasi-distributed LRR (QDLRR) method, resulting in significant improvement in flow routing (Shen et al. 2015). CREST v2.1 is used in this study to evaluate the quality of GPM-era IMERG as well as its hydrological continuity to TRMM-era TMPA products.
d. Statistical metrics
Several widely used statistical metrics were used in this study to quantitatively evaluate the performance of different satellite precipitation products (Table 1). The Pearson correlation coefficient (CC) was used to assess the agreement between the “test” field and the reference field, representing degree of agreement. Five metrics were selected to demonstrate error and bias between precipitation estimates and gauge observations, including the mean error (ME), relative bias (BIAS), root-mean-square error (RMSE), centered root-mean-square error (cRMSE), and standard deviation (SD). The ME was chosen to measure the average difference between the “test” field and reference field, and BIAS is a similar index expressed as a ratio and describes the systematic bias between the two fields. RMSE was used to represent the average error magnitude.
List of the statistical metrics used in statistic and hydrologic comparison and evaluation.
SD and cRMSE were also calculated, so a Taylor diagram (Taylor 2001) can be drawn based on the cRMSE, SD, and CC. The Taylor diagram provides a way of graphically summarizing how closely a pattern (or a set of patterns) matches observations. Each point in the two-dimensional Taylor diagram can represent the three statistics simultaneously that are related by a formula similar to the law of cosines (Taylor 2001). In this study, it is used to show how closely satellite precipitation estimates match gauge observations. In addition, another popular index, the Nash–Sutcliffe coefficient efficiency (NSCE), was adopted to evaluate the performance of the hydrologic model, together with CC, BIAS, and RMSE.
3. Evaluation and comparison of precipitation
Evaluation and comparison were performed for the IMERG and TMPA products both at 39 grid boxes of 0.25° × 0.25°, with at least one rain gauge in each grid box, and over the entire basin. The study time period was from May through September 2014. To make it comparable, the 0.1° × 0.1° gridded IMERG estimates were aggregated to 0.25° × 0.25° datasets, the same as TMPA and CGDPA, using the standard bilinear interpolation method suited for gridded datasets. Please note that among all the gauges used in the production of CGDPA, 41 gauges are located in the Ganjiang River basin. So we used the 41 gauges to evaluate the quality of IMERG and TMPA products at the grid scale. The limited number of gauges could be a source of error in grid-scale comparison. In areal comparison, the CGDPA gridded product (0.25° × 0.25°) was used as ground reference. The accuracy of CGDPA over this area is high (Shen and Xiong 2015), which has also been shown by the excellent performance of CGPDA in hydrologic simulation in this study. However, lack of gauges could also bring in greater uncertainty in the evaluation.
The density-plot scatter diagrams of daily IMERG, 3B42V7, and 3B42RT for 39 grid boxes are shown in Fig. 2. Table 2 lists statistical metrics for both gridded and basin-averaged scale comparison. Comparisons between satellite precipitation products and ground-based observations at the basin scale were conducted by averaging the precipitation over the whole basin. Then we compared daily values of satellite-based basin-scale precipitation estimates and CGDPA precipitation estimates and calculated the statistics. For the comparison at the grid scale, we derived daily values of satellite products and CGDPA at all grid boxes within the study area to calculate the statistics. In general, IMERG and 3B42V7 match similarly well with gauge observations and both outperform 3B42RT (Figs. 2a,b). Both post-real-time corrections effectively reduced the biases of IMERG and 3B42V7 to single digits of underestimation, approximately −1% for grid points and −4% over the whole basin) from positive 20+% of 3B42RT, as shown in Table 2. Note that though the BIAS for IMERG and 3B42V7 is very small (close to zero) compared with 3B42RT, the RMSE for IMERG and 3B42V7 is relatively high, only a little smaller than 3B42RT. This can be attributed to the fact that the positive/negative biases are offset for IMERG and 3B42V7. As shown in Fig. 3, when comparing precipitation of 41 rain gauges and three satellite precipitation products and calculating the BIAS for each station, we found that 3B42RT overestimates precipitation for 37 rain gauges in all 41 rain gauges, consistent with the large BIAS at the grid and basin scale for 3B42RT. But for IMERG and 3B42V7, overestimation and underestimation are close in the number of gauges. As a result, cancellation of biases occurred. The post-real-time correction actually reduced the absolute biases for IMERG and 3B42V7 compared with 3B42RT at almost all stations. However, the correction effect may be exaggerated when only looking at an averaged BIAS.
Scatterplots of the grid-based daily precipitation comparison at the 39 selected 0.25° grid boxes between (a) IMERG, (b) 3B42V7, and (c) 3B42RT and gauge.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Scatterplots of the grid-based daily precipitation comparison at the 39 selected 0.25° grid boxes between (a) IMERG, (b) 3B42V7, and (c) 3B42RT and gauge.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Scatterplots of the grid-based daily precipitation comparison at the 39 selected 0.25° grid boxes between (a) IMERG, (b) 3B42V7, and (c) 3B42RT and gauge.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Statistical summary of the grid-based and basin-averaged comparison of IMERG, 3B42V7, and 3B42RT at 0.25° and daily resolutions.
BIAS (%) distribution for daily precipitation between the satellite precipitation products of (a) IMERG , (b) 3B42V7, and (c) 3B42RT and gauge from May to September 2014.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
BIAS (%) distribution for daily precipitation between the satellite precipitation products of (a) IMERG , (b) 3B42V7, and (c) 3B42RT and gauge from May to September 2014.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
BIAS (%) distribution for daily precipitation between the satellite precipitation products of (a) IMERG , (b) 3B42V7, and (c) 3B42RT and gauge from May to September 2014.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Also as shown in the Taylor diagram in Fig. 4, the post-real-time analysis significantly reduced the overestimation of real-time products, and the aggregation to basinwide comparison resulted in much improved performance for all three products in terms of their grid-scale statistics. In particular, IMERG greatly improved the correlation between satellite estimates and gauge observations, with CC increasing from 0.62 to 0.90, and the RMSE reducing from 13.09 to 4.44 mm day−1 (also see Table 2). The consistency between 3B42V7 and 3B42RT and gauge observations also increased, whereas ME and BIAS for 3B42RT decreased slightly as opposed to the trend for IMERG and 3B42V7. As shown in Fig. 4a, 3B42V7 is closer to the observation than the other two products, whereas in Fig. 4b, IMERG is closer to observation than the other two products, indicating that 3B42V7 seems slightly better than IMERG at the grid scale, but IMERG appears to be slightly better than 3B42V7 at the basin scale.
Taylor diagram for IMERG, 3B42V7, and 3B42RT at (a) grid and (b) basin scale.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Taylor diagram for IMERG, 3B42V7, and 3B42RT at (a) grid and (b) basin scale.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Taylor diagram for IMERG, 3B42V7, and 3B42RT at (a) grid and (b) basin scale.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
In general, IMERG and 3B42V7 are comparable with little difference in this study (Table 2), which may be attributed to a number of factors as follows. First, the latitude of our study area is relatively low and the study time span was the rainy season with intensive precipitation, which may contribute to the similarity between IMERG and 3B42V7. Second, IMERG is still at a very early stage (Day-1 currently), whereas 3B42V7 is the latest version of the TMPA product (version 7), partly resulting in their comparable performance. Overall, all three products demonstrated similarly acceptable (~0.63) and high (0.87) correlation at grid and basin scales, respectively, but 3B42RT showed much higher overestimation. The limited number of gauges used in this study could be a source of error in the analyses above.
4. Hydrological evaluation with two simulation scenarios
Streamflow prediction performance of the three different precipitation products was investigated using two different parameter setup scenarios for the CREST hydrological model in this section.
In scenario I (static parameters), model parameters were first calibrated using observed data of 310 rain gauges from January 2003 through December 2009, and the parameter sets were used in this scenario. Then the model was run using precipitation from CGDPA, 3B42V7, and 3B42RT in independent validation period 1 (from 1 January 2010 to 31 December 2013) with the rain gauge–calibrated model parameters. The three aforementioned precipitation products and the IMERG product were used to force the model in validation period 2 (from 1 May to 30 September 2014) with the same parameter set.
In scenario II (dynamic parameters), model parameters were dynamically recalibrated according to precipitation inputs, for example, 3B42V7 and 3B42RT, from January 2008 through December 2010. The two calibrations of the model were subsequently validated using 3B42V7 and 3B42RT in validation period 1 (from 1 January 2011 to 31 December 2013), respectively. Then we used the product-specific parameter sets to simulate streamflow based on the IMERG precipitation input to compare the performance of IMERG, 3B42V7, and 3B42RT over validation period 2 (from 1 May to 30 September 2014).
Table 3 lists the hydrological scenarios (I and II), calibration–validation setups and time periods, precipitation input, and parameter sets used in the model. Scenario I was designed to assess the hydrological utility of IMERG compared to TMPA products with reference to the static parameters calibrated using rain gauge observations, whereas scenario II was designed to investigate the applicability of the IMERG product in the TMPA product-specific calibrated model, testing continuity of the TRMM-era TMPA to the GPM-era. This product-specific calibration strategy is an alternative for ungauged basins where only satellite precipitation products are available (Xue et al. 2013). Validation period 2 was the extended rainy season of the study basin with relatively intense precipitation and streamflow and covered the time span of IMERG data.
The setup of two hydrological scenarios with their time periods, inputs, and parameters.
a. Scenario I: Static parameters
First, the CREST model was forced by the 310 rain gauge-observed data and calibrated with the daily observed streamflow from 2003 through 2009, with the first 6-month period used for spinning up the model (Liang 1994). Because of the high density of rain gauges, we considered the calibrated parameter set reliable and suitable for evaluating hydrological utility of other precipitation sources. Figure 5a shows that the simulated streamflow agrees well with the observed streamflow in the calibration period. As listed in Table 4, metrics of the calibration period show the great utility of gauge-calibrated model parameters with very high NSCE (0.90) and CC (0.95), as well as small BIAS (−6.89%) and the least RMSE (534.81 m3 s−1).
Comparison of CREST simulated streamflow with gauge-calibrated parameters and observed streamflow in both calibration (from 1 Jan 2003 to 31 Dec 2009) and validation period 1 (from 1 Jan 2010 to 31 Dec 2013). (a) Daily calibration data from 310 rain gauges; and daily validation data from (b) CGDPA, (c) 3B42V7, and (d) 3B42RT.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Comparison of CREST simulated streamflow with gauge-calibrated parameters and observed streamflow in both calibration (from 1 Jan 2003 to 31 Dec 2009) and validation period 1 (from 1 Jan 2010 to 31 Dec 2013). (a) Daily calibration data from 310 rain gauges; and daily validation data from (b) CGDPA, (c) 3B42V7, and (d) 3B42RT.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Comparison of CREST simulated streamflow with gauge-calibrated parameters and observed streamflow in both calibration (from 1 Jan 2003 to 31 Dec 2009) and validation period 1 (from 1 Jan 2010 to 31 Dec 2013). (a) Daily calibration data from 310 rain gauges; and daily validation data from (b) CGDPA, (c) 3B42V7, and (d) 3B42RT.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Comparison of daily observed and simulated streamflow under scenario I.
In validation period 1 (from 1 January 2010 to 31 December 2013), performance of the three precipitation products differed but were generally acceptable, with the NSCE greater than 0.68, small BIAS, and high CC. CGDPA matches the observed streamflow reasonably well and just underestimates a few peak flows (Fig. 5b). As shown in Table 4, some of the metrics of CGDPA even slightly improve when compared with those of the calibration period except for RMSE, which could be as a result of the climatological optimal interpolation CGDPA used. The performance of CGDPA in validation period 1 is as good as, if not better than, the calibration period using data from 310 gauges, indicating that it is reliable to use CGDPA as the reference in validation period 2. The 3B42V7 product adequately captured a majority of peak flows with NSCE up to 0.84, but still misrepresented three minor peak flows in the beginning of 2013 and forecasted some of the peak flows slightly ahead of time in 2010 (Fig. 5c). These facts could explain the degraded performance of 3B42V7, but it still came close secondarily to the reference CGDPA. The 3B42RT product overestimated most of the major peak flows because of overall precipitation overestimation, but underestimated many of the medium and small flows (Fig. 5d). The skill of 3B42RT in hydrologic simulation, though still acceptable, tended to be worse than that of 3B42V7. The coexistence of overestimates and underestimates in 3B42RT simulation resulted in a small BIAS (−1.51%), but ended with a much greater absolute error RMSE (1330.75 m3 s−1). As summarized in Table 4, the metrics of the validation period from the three different precipitation products indicate that the CREST model can reproduce daily observed streamflow in the study area effectively and robustly after the baseline calibration period 2003–09.
In validation period 2 (from 1 May to 30 September 2014), the CREST model was forced by CGDPA, 3B42V7, 3B42RT, and IMERG precipitation data (Fig. 6). Compared with validation period 1, skill scores for the first three precipitation products all fell, mainly caused by the underestimation in the second flood peak process and the last recession. Metrics in Table 4 show that CGDPA reference data have the best skill scores, as expected, closely followed by the IMERG metrics (IMERG/CGDPA: NSCE = 0.77/0.86, CC = 0.91/0.94, BIAS = −14.09/−8.76%, and RMSE = 1080.87/822.73 m3 s−1). The hydrograph of IMERG is remarkably similar to that of CGDPA (Figs. 6a,b). Among the three satellite precipitation products, IMERG performed the best and matched well with the observed streamflow, especially for the second flow peak compared with 3B42V7 and 3B42RT (Figs. 6b–d). The 3B42RT product came at the bottom as its NSCE declined to 0.46 and RMSE increased to 1637.53 m3 s−1.
Comparison of CREST simulated streamflow with gauge-calibrated parameters and observed streamflow validation period 2 (from 1 May to 30 Sep 2014). Daily data from (a) CGDPA, (b) IMERG, (c) 3B42V7, and (d) 3B42RT.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Comparison of CREST simulated streamflow with gauge-calibrated parameters and observed streamflow validation period 2 (from 1 May to 30 Sep 2014). Daily data from (a) CGDPA, (b) IMERG, (c) 3B42V7, and (d) 3B42RT.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Comparison of CREST simulated streamflow with gauge-calibrated parameters and observed streamflow validation period 2 (from 1 May to 30 Sep 2014). Daily data from (a) CGDPA, (b) IMERG, (c) 3B42V7, and (d) 3B42RT.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
We speculate that three factors led to the better performance of IMERG. First, IMERG uses the Goddard Profiling Algorithm 2014 (GPROF 2014) to retrieve rainfall and vertical structure information from satellite-based passive MW observations, whereas 3B42V7 uses GPROF 2010. The refinement in the latest MW algorithm could contribute to the improvement of IMERG upon 3B42V7 even with the same level-2 inputs. Second, the IMERG product involved more sensors than the 3B42V7 product as described in section 2b, which could improve the quality of IMERG. Third, in the statistical comparison, IMERG and 3B42V7 show comparable performance, but IMERG is slightly better at the basin scale. As such, better spatial representation of precipitation systems could most likely reduce error propagations from forcing data to hydrological prediction at the basin outlet because CREST is a distributed hydrological model that takes into account not only the intensity but also spatial variability in rainfall data. In addition, TMPA gives one number every 3 h in a 0.25° grid box, while IMERG gives a number every half hour in a 0.1° grid box. Presumably, this extra sampling gives a more accurate daily averaged precipitation.
In summary, IMERG performs very well, showing comparable hydrologic utility with gauge-based CGDPA during the extended rainy season in the Ganjiang River basin with the calibrated static parameter set during the baseline period, followed by the acceptable 3B42V7 and the less reliable 3B42RT products (because of its large uncertainty). But the difference between IMERG and 3B42V7 is modest from statistics in Tables 3 and 4.
b. Scenario II: Dynamic parameters
In scenario II, 3B42V7 and 3B42RT were separately used to recalibrate the CREST model from January 2008 through December 2010, to further investigate the applicability of the IMERG product in the TMPA product-specific calibrated model. Then the CREST model was validated for two periods with IMERG and TMPA products, based on the corresponding dynamic parameter sets.
Figure 7 compares daily time series of the simulated and observed hydrographs in both calibration and validation periods. In the calibration period, both 3B42V7 and 3B42RT capture most of the flow peaks and fit the observed streamflow reasonably well, but 3B42RT shows larger bias and higher RMSE in scenario II. In validation period 1, 3B42V7 obtained slightly worse results while the fluctuation of 3B42RT resulted in marked reductions in metrics, as were presented in Table 5, such as NSCE (0.86 for the calibration period to 0.42 for validation period 1). This indicates that the 3B42V7 product has more reliable hydrological continuity than the 3B42RT product with the precipitation product-specific parameter set.
Comparison of CREST simulated streamflow from (a) 3B42V7 and (b) 3B42RT with product-specific calibrated parameters and observed streamflow in both calibration (from 1 Jan 2008 to 31 Dec 2010) and validation (from 1 Jan 2011 to 31 Dec 2013) periods.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Comparison of CREST simulated streamflow from (a) 3B42V7 and (b) 3B42RT with product-specific calibrated parameters and observed streamflow in both calibration (from 1 Jan 2008 to 31 Dec 2010) and validation (from 1 Jan 2011 to 31 Dec 2013) periods.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Comparison of CREST simulated streamflow from (a) 3B42V7 and (b) 3B42RT with product-specific calibrated parameters and observed streamflow in both calibration (from 1 Jan 2008 to 31 Dec 2010) and validation (from 1 Jan 2011 to 31 Dec 2013) periods.
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
In validation period 2, the 3B42V7-forced simulation curve fits well with the curve of observation and captures the main flow peaks, whereas it underestimates the third main flow peak and overestimates the streamflow trend after the first flow peak (Fig. 8a). The IMERG simulation based on the 3B42V7-specific parameter set shows a similar curve with 3B42V7, with slightly higher NSCE (Fig. 8a). As for 3B42RT, deviation from observations is obvious after the first and second flow peaks, while the deviation is reduced when forcing the model using IMERG based on the 3B42RT-specific parameter set (Fig. 8b). As summarized in Table 5, all metrics of IMERG show overall very desirable hydrological continuity from TRMM-era calibrated 3B42V7-specific and 3B42RT-specific parameter sets in the previous period, except the BIAS of IMERG, which shows slight deterioration from 3B42RT.
Comparison of CREST simulated streamflow with parameters recalibrated using (a) 3B42V7 and (b) 3B42RT in validation period 2 (from 1 May to 30 Sep 2014).
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Comparison of CREST simulated streamflow with parameters recalibrated using (a) 3B42V7 and (b) 3B42RT in validation period 2 (from 1 May to 30 Sep 2014).
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
Comparison of CREST simulated streamflow with parameters recalibrated using (a) 3B42V7 and (b) 3B42RT in validation period 2 (from 1 May to 30 Sep 2014).
Citation: Journal of Hydrometeorology 17, 1; 10.1175/JHM-D-15-0059.1
The comparison results under both gauge-reference-based scenario I and forcing-specific scenario II above all showed that IMERG has high hydrologic utility and continuity (i.e., a good and robust successor) from TRMM-era standard products whether the hydrological model is calibrated using rain gauge data or TMPA specific products. If the hydrologic model is calibrated using rain gauge data, the IMERG product can be used directly in the model and the results could be comparable to that using rain gauge data. If the model is calibrated using TMPA 3B42 products, it is anticipated that consistent results can be obtained compared with 3B42V7 and 3B42RT, and some deviations from observed streamflow could even be reduced, especially for the 3B32RT product with higher uncertainty in this study. The conclusion from our basin-specific study may not apply to other basins, especially in different climates.
In terms of the uncertainty in satellite precipitation products, sampling, instrumental, algorithmic errors and geophysical features could be the major sources, and thus affect the simulation of hydrological fluxes and states (Nijssen 2004; Gebregiorgis and Hossain 2014). However, the objective of this study was to have a first look of level-3 products from the TRMM-era 3B42V7 and GPM-era Day-1 IMERG products. Remaining questions, for example, which level-2 sensor contributes how much uncertainty into satellite products, and how the sensor-specific uncertainty could affect the overall hydrologic application, warrant more thorough examinations in future studies.
c. Comparison with recent TMPA studies
Bitew et al. (2012) summarized studies that evaluated satellite precipitation products through streamflow simulation under hydrologic modeling framework. However, these studies are few in number and were published earlier than 2009. A large number of studies applying satellite precipitation products to hydrologic modeling have been published in recent years. Some recent hydrologic studies based on TMPA 3B42 V6 or V7 products are summarized in Table 6. Because of various research methods, models, and regions, different results were obtained. Among those studies, four studies divided their respective study areas into several subbasins, of which only the largest basins were demonstrated in Table 6 (Collischonn et al. 2008; Li et al. 2015; Siddique-E-Akbor et al. 2014; Zulkafli et al. 2014).Two studies adopted other satellite products apart from TMPA to perform comparisons (Li et al. 2015; Siddique-E-Akbor et al. 2014), and two studies designed several hydrologic calibration and validation scenarios, in which only representative scenarios were listed (Meng et al. 2014; Xue et al. 2013). Three metrics (NSCE, BIAS, and CC) were chosen to enable the comparison among different studies.
Summary of some hydrologic studies using TMPA 3B42 products, including 3B42RT(V6) and 3B42V6 for version 6 and 3B42RT(V7) and 3B42V7 for version 7. Those studies designed various hydrologic calibration and validation scenarios and are conducted over different basins; some studies even included more than one basin and satellite precipitation products apart from TMPA products.
Three main conclusions were drawn through the comparison: 1) hydrologic simulation performance of specific satellite precipitation products varies with hydrological models and study areas, 2) the near-real-time product usually shows worse hydrologic simulation results than post-real-time product either in TMPA V6 or V7, and 3) 3B42V7 improves over the performance of 3B42V6 in hydrologic simulation prominently for all the three metrics.
Although the recent hydrologic studies using TRMM-era level-3 products are not directly comparable because of various basins, hydrological models, and metrics, the performance of 3B42V7 and 3B42RT in our study is generally consistent with, if not better than, other studies listed in Table 6. When it comes to IMERG, it performs at least comparably, if not better, to 3B42V7 and better than 3B42RT in our study, indicating a good prospect of hydrologic utility for the IMERG product and also a smooth hydrologic continuity from TRMM-era product heritages.
However, our study site is relatively humid and located at low- to midlatitude regions; the conclusions obtained from our study may not apply to other areas. Yong et al. (2010) conducted hydrologic evaluation of TMPA precipitation products (version 6) over the Laohahe basin, a high-latitude basin in China. Results show that 3B42RT has almost no hydrologic utility, even at the monthly scale, but 3B42V6 can produce much better hydrologic predictions with reduced error propagation from input to streamflow. Yong et al. (2012) assessed TRMM-based multisatellite real-time precipitation estimation methods and their impacts on hydrologic prediction over the same basin in Yong et al. (2010) and found that the latest dataset of TMPA-RT (version 7) exhibited the best capability in capturing hydrologic response, but still with significant challenges during winter snowing events. Meng et al. (2014) studied the suitability of TRMM satellite rainfall in driving CREST in the source region of the Yellow River characterized by high latitudes and altitudes, concluding that TMPA cannot be used to drive hydrological models for daily streamflow simulation over their study area. However, TMPA rainfall performed well at the monthly time scale. Bitew et al. (2012) evaluated CMORPH, PERSIANN, TMPA 3B42RT, and 3B42V6 (version 6) through streamflow simulation using the Soil and Water Assessment Tool (SWAT) over a small mountainous watershed in Ethiopia. Their results show that 3B42RT is better than 3B42V6 in statistical comparison, whereas 3B42V6 performed better than 3B42RT in hydrologic simulation. Additional research is advocated to explore the performance of IMERG over high-latitude or altitude basins.
5. Summary and conclusions
After 17 yr of successful quasi-global precipitation measurement, TRMM was decommissioned and GPM is providing the next generation of global precipitation products. This study first focuses on the statistical evaluation of TRMM-era standard products 3B42V7/3B42RT and the GPM-era IMERG research product and then hydrologically evaluates their streamflow prediction utility using the CREST hydrologic model in the Ganjiang River basin, a midlatitude basin in China. Because of the delayed IMERG data availability and also the reliable ground gauge network, this study is limited to the extended rainy season of May–September 2014. Results from the above analyses may be specific to the Ganjiang River basin but are likely to be more generally applicable to basins situated in the subtropics.
The main conclusions from the statistical comparison of IMERG, 3B42V7, and 3B42RT, conducted at both grid and basin scales from May through September 2014, are summarized as follows:
In general, both IMERG and 3B42V7 fit similarly well with gauge observations and outperform 3B42RT. The post-real-time monthly gauge analysis-based corrections used in both datasets effectively reduce the biases of IMERG and 3B42V7 to single digits of underestimation, approximately −1% for grid points and −4% over the whole basin, compared to the positive 20+% of 3B42RT.
All three products show acceptable (~0.63) and high (0.87) correlation at grid and basin scales, respectively, indicating a similar capability of capturing trends in precipitation, but 3B42RT shows significant higher overestimation.
The Taylor diagram visually shows that the post-real-time analysis of both the TRMM-era and GPM-era has significantly reduced the overestimation of real-time products. The early Day-1 IMERG is comparable with, if not better than, TMPA and 3B42 version 7 based on both the Taylor diagram and statistics in our study.
In terms of the hydrological evaluation, two scenarios of hydrologic parameter sets are designed, with scenario I benchmarked by in situ gauges and scenario II calibrated with TMPA-based specific products. Scenario I is conventionally used over gauged basins where the hydrological model can be tuned up using ground observations. In scenario II, input-specific recalibration can be used as an alternative over regions with few or no gauges where remotely sensed data are acquired as substitutes to drive the models. The main conclusions are as follows:
In scenario I, IMERG performs well, showing comparable hydrologic utility with gauge-based CGDPA during the extended rainy season in the Ganjiang River basin with the static parameter set calibrated during the baseline period (2003–09), followed by the acceptable performance of 3B42V7 and less reliable skill of 3B42RT because of its large uncertainty.
In scenario II, the CREST model was recalibrated with two best parameter sets using 3B42V7 and 3B42RT, respectively, and then validated with the three satellite products over later periods. As anticipated, 3B42V7 shows consistent hydrologic utility over the two periods whereas the NSCE for 3B42RT reduces from 0.86 to 0.42, indicating that the recalibration for 3B42RT only improves the performance in the calibration period and its benchmarked hydrologic skill fails to carry over to later validation periods. Metrics of IMERG show overall very high hydrological continuity from TRMM-era standard precipitation-specific parameter sets calibrated in the previous period, with only slight deterioration with 3B42RT-specfic calibration.
Compared with some recent hydrologic studies using the TRMM-era products, the performance of 3B42V7 and 3B42RT in this study is generally better than the other studies summarized in Table 6, likely because of the high density of gauge networks and also the well-calibrated CREST hydrologic model.
In conclusion, IMERG performs comparably to reference data and in many cases outperforms TMPA standard products in this study, indicating a promising prospect of hydrologic utility and also a desirable hydrologic continuity from TRMM-era product heritages to the GPM-era IMERG product, even with its limited data availability to date in this well-gauged and midlatitude basin. It is reasonably anticipated that IMERG data would particularly outdistance TMPA products when it comes to high altitudes and/or high latitudes, given the capability of the GPM Core Observatory to detect light and solid precipitation (Hou et al. 2014). As more IMERG data and even retrospectively processed TRMM/GPM-era IMERG datasets are released, more studies to explore the potential of IMERG and other GPM-era products in water, weather, and climate studies are needed in the future.
Acknowledgments
This study, and the first five authors in particular, were supported by the National Natural Science Foundation of China (Project 91437214) and the Chinese Academy of Meteorological Sciences State Key Laboratory of Disaster Weather (Project 2013LASW-A09). The efforts of the TRMM and GPM research community are also highly appreciated for making the data available for international users. We are grateful to three reviewers and editors for their constructive and insightful comments. This manuscript was improved as a result of their efforts.
REFERENCES
Ahrens, B., 2006: Distance in spatial interpolation of daily rain gauge data. Hydrol. Earth Syst. Sci., 10, 197–208, doi:10.5194/hess-10-197-2006.
Bitew, M. M., Gebremichael M. , Ghebremichael L. T. , and Bayissa Y. A. , 2012: Evaluation of high-resolution satellite rainfall products through streamflow simulation in a hydrological modeling of a small mountainous watershed in Ethiopia. J. Hydrometeor., 13, 338–350, doi:10.1175/2011JHM1292.1.
Collischonn, B., Collischonn W. , and Tucci C. E. M. , 2008: Daily hydrological modeling in the Amazon basin using TRMM rainfall estimates. J. Hydrol., 360, 207–216, doi:10.1016/j.jhydrol.2008.07.032.
Dinku, T., Anagnostou E. N. , and Borga M. , 2002: Improving radar-based estimation of rainfall over complex terrain. J. Appl. Meteor., 41, 1163, doi:10.1175/1520-0450(2002)041<1163:IRBEOR>2.0.CO;2.
Futrell, J., and Coauthors, 2005: Water: Challenges at the intersection of human and natural systems. Tech. Rep. PNWD-3597, 50 pp., doi:10.2172/1046481.
Gebregiorgis, A. S., and Hossain F. , 2014: Estimation of satellite rainfall error variance using readily available geophysical features. IEEE Trans. Geosci. Remote Sens., 52, 288–304, doi:10.1109/TGRS.2013.2238636.
Habib, E., ElSaadani M. , and Haile A. T. , 2012: Climatology-focused evaluation of CMORPH and TMPA satellite rainfall products over the Nile basin. J. Appl. Meteor. Climatol., 51, 2105–2121, doi:10.1175/JAMC-D-11-0252.1.
Hong, Y., Hsu K.-L. , Sorooshian S. , and Gao X. , 2004: Precipitation estimation from remotely sensed imagery using an artificial neural network cloud classification system. J. Appl. Meteor., 43, 1834–1853, doi:10.1175/JAM2173.1.
Hong, Y., Adler R. F. , Hossain F. , Curtis S. , and Huffman G. J. , 2007: A first approach to global runoff simulation using satellite rainfall estimation. Water Resour. Res., 43, W08502, doi:10.1029/2006WR005739.
Hong, Y., Chen S. , Xue X. , and Hodges G. , 2012: Global precipitation estimation and applications. Multiscale Hydrologic Remote Sensing: Perspectives and Applications, N.-B. Chang and Y. Hong, Eds., CRC Press, 371–386.
Hou, A. Y., Skofronick-Jackson G. , Kummerow C. D. , and Shepherd J. M. , 2008: Global precipitation measurement. Precipitation: Advances in Measurement, Estimation and Prediction, S. Michaelides, Ed., Springer, 131–169.
Hou, A. Y., and Coauthors, 2014: The Global Precipitation Measurement Mission. Bull. Amer. Meteor. Soc., 95, 701–722, doi:10.1175/BAMS-D-13-00164.1.
Hu, Q., Yang D. , Li Z. , Mishra A. K. , Wang Y. , and Yang H. , 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, doi:10.1080/01431161.2013.876118.
Huffman, G. J., and Bolvin D. T. , 2015: TRMM and Other Data Precipitation Data Set Documentation, NASA Global Change Master Directory Doc., Mesoscale Atmospheric Processes Laboratory, 44 pp. [Available online at http://pmm.nasa.gov/sites/default/files/document_files/3B42_3B43_doc_V7.pdf.]
Huffman, G. J., and Coauthors, 2001: Global precipitation at one-degree daily resolution from multisatellite observations. J. Hydrometeor., 2, 36–50, doi:10.1175/1525-7541(2001)002<0036:GPAODD>2.0.CO;2.
Huffman, G. J., and Coauthors, 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 38–55, doi:10.1175/JHM560.1.
Huffman, G. J., Adler R. F. , Bolvin D. T. , and Nelkin E. J. , 2010: The TRMM Multi-Satellite Precipitation Analysis (TMPA). Satellite Rainfall Applications for Surface Hydrology, F. Hossain and M. Gebremichael, Eds., Springer, 3–22.
Huffman, G. J., Bolvin D. T. , Braithwaite D. , Hsu K. , Joyce R. , and Xie P. , 2014: NASA Global Precipitation Measurement (GPM) Integrated Multi-Satellite Retrievals for GPM (IMERG) Algorithm Theoretical Basis Doc., version 4.4. NASA GSFC, 30 pp. [Available online at http://pmm.nasa.gov/sites/default/files/document_files/IMERG_ATBD_V4.4.pdf.]
Huffman, G. J., Bolvin D. T. , and Nelkin E. J. , 2015: Integrated Multi-Satellite Retrievals for GPM (IMERG) Technical Documentation. Tech. Doc., NASA GSFC, 47 pp. [Available online at http://pmm.nasa.gov/sites/default/files/document_files/IMERG_doc.pdf.]
Joyce, R. J., Janowiak J. E. , Arkin P. A. , and Xie P. , 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, doi:10.1175/1525-7541(2004)005<0487:CAMTPG>2.0.CO;2.
Khan, S. I., and Coauthors, 2011: Hydroclimatology of Lake Victoria region using hydrologic model and satellite remote sensing data. Hydrol. Earth Syst. Sci., 15, 107–117, doi:10.5194/hess-15-107-2011.
Kidd, C., and Huffman G. , 2011: Global precipitation measurement. Meteor. Appl., 18, 334–353, doi:10.1002/met.284.
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, doi:10.1109/TGRS.2007.895337.
Li, Z., Yang D. , and Hong Y. , 2013: Multi-scale evaluation of high-resolution multi-sensor blended global precipitation products over the Yangtze River. J. Hydrol., 500, 157–169, doi:10.1016/j.jhydrol.2013.07.023.
Li, Z., Yang D. , Gao B. , Jiao Y. , Hong Y. , and Xu T. , 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, doi:10.1175/JHM-D-14-0105.1.
Liang, X., 1994: A two-layer variable infiltration capacity land surface representation for general circulation models. Ph.D. thesis, University of Washington, 208 pp.
Long, D., Shen Y. , Sun A. , Hong Y. , Longuevergne L. , Yang Y. , Li B. , and Chen L. , 2014: Drought and flood monitoring for a large karst plateau in southwest China using extended GRACE data. Remote Sens. Environ., 155, 145–160, doi:10.1016/j.rse.2014.08.006.
Meng, J., Li L. , Hao Z. , Wang J. , and Shao Q. , 2014: Suitability of TRMM satellite rainfall in driving a distributed hydrological model in the source region of Yellow River. J. Hydrol., 509, 320–332, doi:10.1016/j.jhydrol.2013.11.049.
Mishra, A. K., and Coulibaly P. , 2009: Developments in hydrometric network design: A review. Rev. Geophys., 47, RG2001, doi:10.1029/2007RG000243.
Nijssen, B., 2004: Effect of precipitation sampling error on simulated hydrological fluxes and states: Anticipating the Global Precipitation Measurement satellites. J. Geophys. Res., 109, D02103, doi:10.1029/2003JD003497.
Prakash, S., Mitra A. K. , Momin I. M. , Pai D. S. , Rajagopal E. N. , and Basu S. , 2015: Comparison of TMPA-3B42 versions 6 and 7 precipitation products with gauge-based data over India for the southwest monsoon period. J. Hydrometeor., 16, 346–362, doi:10.1175/JHM-D-14-0024.1.
Shen, X., Hong Y. , Zhang K. , and Hao Z. , 2015: Refine a distributed reservoir routing method to improve performance of the CREST model. J. Hydrol. Eng., in press.
Shen, Y., and Xiong A. , 2015: Validation and comparison of a new gauge-based precipitation analysis over mainland China. Int. J. Climatol., doi:10.1002/joc.4341, in press.
Siddique-E-Akbor, A. H. M., Hossain F. , Sikder S. , Shum C. K. , Tseng S. , Yi Y. , Turk F. J. , and Limaye A. , 2014: Satellite precipitation data–driven hydrological modeling for water resources management in the Ganges, Brahmaputra, and Meghna basins. Earth Interact., 18, doi:10.1175/EI-D-14-0017.1.
Sorooshian, S., Hsu K.-L. , Gao X. , Gupta H. V. , Imam B. , and Braithwaite D. , 2000: Evaluation of PERSIANN system satellite-based estimates of tropical rainfall. Bull. Amer. Meteor. Soc., 81, 2035–2046, doi:10.1175/1520-0477(2000)081<2035:EOPSSE>2.3.CO;2.
Tang, G., Li Z. , Xue X. , Hu Q. , and Hong Y. , 2015: A study of substitutability of TRMM remote sensing precipitation for gauge-based observation in Ganjiang River basin (in Chinese). Shuikexue Jinzhan, 26, 271–277.
Taylor, K. E., 2001: Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res., 106, 7183–7192, doi:10.1029/2000JD900719.
Tian, Y., and Peters-Lidard C. D. , 2010: A global map of uncertainties in satellite-based precipitation measurements. Geophys. Res. Lett., 37, L24407, doi:10.1029/2010GL046008.
Turk, F. J., and Miller S. D. , 2005: Toward improved characterization of remotely sensed precipitation regimes with MODIS/AMSR-E blended data techniques. IEEE Trans. Geosci. Remote Sens., 43, 1059–1069, doi:10.1109/TGRS.2004.841627.
Verdin, J., Funk C. , Senay G. , and Choularton R. , 2005: Climate science and famine early warning. Philos. Trans. Roy. Soc. London, B360, 2155–2168, doi:10.1098/rstb.2005.1754.
Villarini, G., and Witold K. , 2008: Empirically-based modeling of spatial sampling uncertainties associated with rainfall measurements by rain gauges. Adv. Water Resour., 31, 1015–1023, doi:10.1016/j.advwatres.2008.04.007.
Wang, J., and Coauthors, 2011: The coupled routing and excess storage (CREST) distributed hydrological model. Hydrol. Sci. J., 56, 84–98, doi:10.1080/02626667.2010.543087.
Wu, H., Adler R. F. , Hong Y. , Tian Y. , and Policelli F. , 2012: Evaluation of global flood detection using satellite-based rainfall and a hydrologic model. J. Hydrometeor., 13, 1268–1284, doi:10.1175/JHM-D-11-087.1.
Xue, X., Hong Y. , Limaye A. S. , Gourley J. J. , Huffman G. J. , Khan S. I. , Dorji C. , and Chen S. , 2013: Statistical and hydrological evaluation of TRMM-based Multi-Satellite Precipitation Analysis over the Wangchu basin of Bhutan: Are the latest satellite precipitation products 3B42V7 ready for use in ungauged basins? J. Hydrol., 499, 91–99, doi:10.1016/j.jhydrol.2013.06.042.
Yong, B., Ren L. L. , Hong Y. , Wang J. H. , Gourley J. J. , Jiang S. H. , and Wang W. , 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, W07542, doi:10.1029/2009WR008965.
Yong, B., Hong Y. , Ren L. L. , Gourley J. J. , Huffman G. J. , Chen X. , Wang W. , and Khan S. , 2012: Assessment of evolving TRMM-based multi-satellite real-time precipitation estimation methods and their impacts on hydrologic prediction in a high latitude basin. J. Geophys. Res., 117, D09108, doi:10.1029/2011JD017069.
Yong, B., Liu D. , Gourley J. J. , Tian Y. , Huffman G. J. , Ren L. , and Hong Y. , 2015: Global view of real-time TRMM Multisatellite Precipitation Analysis: Implication to its successor Global Precipitation Measurement mission. Bull. Amer. Meteor. Soc., 96, 283–296, doi:10.1175/BAMS-D-14-00017.1.
Zhang, Y., and Coauthors, 2015: Hydrometeorological analysis and remote sensing of extremes: Was the July 2012 Beijing flood event detectable and predictable by global satellite observing and global weather modeling systems? J. Hydrometeor., 16, 381–395, doi:10.1175/JHM-D-14-0048.1.
Zulkafli, Z., Buytaert W. , Onof C. , Manz B. , Tarnavsky E. , Lavado W. , and Guyot J.-L. , 2014: A comparative performance analysis of TRMM 3B42 (TMPA) Versions 6 and 7 for hydrological applications over Andean–Amazon River basins. J. Hydrometeor., 15, 581–592, doi:10.1175/JHM-D-13-094.1.
