Abstract

As one of the current mainstream satellite precipitation estimates, the Global Satellite Mapping of Precipitation (GSMaP) system of Japan has been developed to produce high-precision and high-resolution global rainfall products by integrating almost all of the available precipitation-related satellite sensors. To quantify the error features of GSMaP estimates and understand their hydrological potentials at short temporal scale, three widely used GSMaP products (GSMaP_NRT, GSMaP_MVK, and GSMaP_Gauge) were comprehensively investigated at 1 hourly and 0.1° × 0.1° resolutions over nine major basins of China. Assessment results show that GSMaP_NRT apparently underestimates the rainfall amounts, while GSMaP_MVK with both forward and backward propagation processes algorithm can effectively capture the most rainfall events and has the lower error and bias. GSMaP_Gauge displays the best error stability and error structure over most basins of China and also exhibits stronger rain-rate dependencies. However, its unexpected positive biases in southeastern basins, which mainly come from the overestimation at lower rain rates, still need to improve further in future developments. We expected the results documented here can both provide the retrieval developers with some valuable references and offer hydrologic users of GSMaP data a better understanding of their error features and potential utilizations for various hydrological applications.

INTRODUCTION

Accurate measurement of precipitation is crucial for modeling the hydrologic cycle processes, monitoring extreme weather events, and predicting rainfall-triggered natural hazards and disasters (e.g., floods and landslides) at local, regional, and even global scales. Over the past several decades, China has experienced many devastating hydro-meteorological extremes in the context of climate change (Piao et al. 2010). Unfortunately, with two-thirds of its total land area covered by mountains, hills, and plateaus, China suffers the common problem of uneven distribution of meteorological stations and weather radars, and these instruments are particularly sparse over the mountainous regions with complex terrains (Yan et al. 2016). Precipitation estimation from the wide variety of satellite-borne precipitation-related sensors, which can overcome the shortcomings of rain gauge and weather radar networks, has been proved to be an operational approach to capture the spatiotemporal variability of the large-scale rainfall from space particularly in remote regions and complex terrains (Peng et al. 2014).

Since the Tropical Rainfall Measuring Mission (TRMM) launched in 1997, many efforts have been conducted in producing more accurate satellite-based precipitation data at higher spatial and temporal resolutions for the satellite quantitative precipitation estimation-hydrology community. Examples of these TRMM-based quasi-global rainfall products mainly include the TRMM Multi-satellite Precipitation Analysis (TMPA) (Huffman et al. 2007), the NOAA/Climate Prediction Center (CPC) morphing technique (CMORPH) (Joyce et al. 2004), the Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN) (Sorooshian et al. 2000; Hong et al. 2004), and the GSMaP (Okamoto et al. 2005; Kubota et al. 2007). Currently, the Global Precipitation Measurement (GPM) mission led by the National Aeronautics and Space Administration (NASA) and the Japan Aerospace Exploration Agency (JAXA) (Kidd & Huffman 2011) can be regarded as the most important and attractive successor to the TRMM project. Relative to TRMM, the GPM Core Observatory, which was successfully launched on 28 February 2014, carries a dual-frequency precipitation radar and a multichannel GPM Microwave Imager instead of the corresponding precipitation radar (PR) and Microwave Imager (TMI) on the TRMM platform. It is anticipated that GPM will provide more accurate and more reliable global precipitation estimates than the previous TRMM-era products (Yong et al. 2015). However, to the best of our knowledge, the latest Integrated Multi-satellitE Retrievals for GPM system developed by NASA is not as stable as expected and still has room for improvement. Thus, comprehensively analyzing and understanding the error characteristics of other different GPM estimates (e.g., Japan's GSMaP algorithm) seems timely and meaningful at the current stage.

The GSMaP project starting in 2002 with the support of the Japan Science and Technology Agency and JAXA aims to produce high-precision and high-resolution global precipitation estimates by using almost all available satellite-borne passive microwave (PMW) and infrared (IR) sensors (Okamoto et al. 2005). The current operational GSMaP system can provide 1 hourly, 0.1° × 0.1° latitude/longitude global gridded rain rates for the latitude band 60 °N–60 °S. The GSMaP product suite mainly includes three standard hourly rainfall products, i.e., the real-time GSMaP_NRT, the post-real-time GSMaP_MVK, and the gauge-adjusted research-grade product GSMaP_Gauge, which have different input datasets and retrieval processes. Over the years, there have been many efforts to compare and validate the GSMaP and other satellite-based precipitation estimates at global, regional, or basin scale. For example, Kubota et al. (2009) compared GSMaP with five other mainstream satellite products (i.e., 3B42, 3B42RT, CMORPH, PERSSIAN, and NRL) over Japan. Validation results tend to be better for the algorithms with temporal interpolation based upon the morphed technique, such as GSMaP and CMORPH. However, all satellite estimates evidently performed worse for light rainfall during the warm season and for heavy rainfall during any season. Tian et al. (2010) studied the error characteristics of GSMaP_MVK, 3B42, CMORPH, PERSIANN, and NRL over the contiguous United States (CONUS) and they found that GSMaP_MVK significantly overestimated the strong convective events in summer and generally underestimated rainfall in winter. Qin et al. (2014) evaluated the GSMaP_MVK and three other precipitation products (3B42, 3B42RT, CMORPH) benchmarked by a high-density gauge network throughout the Chinese mainland. Their assessments illustrated that GSMaP_MVK underestimated the frequency of high-intensity rainfall events. Chen et al. (2016) intercompared the GSMaP successive versions, Version-4 and Version-5 products, in China. Comparison results indicate that the new Version-5 GSMaP generally represents substantial improvements relative to the previous Version-4 over different regions of China. Yong et al. (2016) systematically investigated four mainstream gauge-adjusted multi-satellite precipitation estimates, including GSMaP_Gauge, TMPA-V7, CMORPH-CRT, and PERSIANN-CDR, over eastern China, Japan, and Taiwan, respectively. Their evaluation results show that GSMaP_Gauge performs better than the other three products at daily scale, due to denser ground observations covering East Asia integrated in the GSMaP system. Although many prior literatures have verified that satellite-based precipitation estimates have great application potential in hydrological simulation (Meng et al. 2014; Peng et al. 2014; Tong et al. 2014; Adjei et al. 2015; Castro et al. 2015; Huang et al. 2016), their error and uncertainty over different climatic regions still need to be further quantified and understood at higher accuracy and spatiotemporal scales before these products can be operationally used in hydrological applications (Li et al. 2009; Yong et al. 2012; Kim et al. 2016; Xu et al. 2016; Yan et al. 2016). It is the high spatial and temporal resolution (0.1° × 0.1°, 1-h) that has made GSMaP very attractive to the hydrology community, especially in real-time hydrological forecasting and hazard warning. However, recently, most of the related studies were performed at relatively coarse resolutions (e.g., 0.25° × 0.25° and daily scales), while the analyses and discussions at higher resolutions (e.g., 0.1° × 0.1°, 1-h) are seldom found. Additionally, it should be noted that the previous studies primarily focused on one specific GSMaP product. In practice, there exist evident differences among different GSMaP products owing to different data inputs and the algorithm itself. Thus, the aim of this study is to quantitatively evaluate and investigate the error characteristics of GSMaP product suite against a high-density gauge-based network at 1 hourly, 0.1° × 0.1° scales over nine major basins of China. For that purpose, three types of different GSMaP products, i.e., GSMaP_NRT, GSMaP_MVK, and GSMaP_Gauge, are used here. The evaluation is performed for the 2009–2010 warm seasons (May to September). This study is expected to offer a better understanding of GSMaP's error characteristics and hydrologic applications over different basins of China.

This paper is organized as follows. The section immediately below introduces the study area and study data, followed by a section presenting the methods used in this study. Then, the results of the analysis are demonstrated, and finally, summarizing remarks and conclusions are drawn.

STUDY AREA AND DATA

Study area

China is situated in East Asia on the western shore of the Pacific Ocean and has diverse climate conditions due to its large area and varied topography, with the Tibet Plateau being the most prominent orographic feature (Zhang et al. 2011; Heike et al. 2012). Figure 1(a) shows the nine basins in China we studied and their topographic distributions. Our comparison and validation were performed over these nine main basins, including (i) Songliao River Basin, (ii) Haihe River Basin, (iii) Yellow River Basin, (iv) Inland River Basin, (v) Huaihe River Basin, (vi) Yangtze River Basin, (vii) Southeast River Basin, (viii) Pearl River Basin, and (ix) Southwest River Basin. These nine study basins are all located within the GSMaP nominal coverage (60 °N–S).

Figure 1

(a) Map of nine study basins in China and their topographic distributions. The solid black line indicates the outline of the basins: (i) Songliao River Basin, (ii) Haihe River Basin, (iii) Yellow River Basin, (iv) Inland River Basin, (v) Huaihe River Basin, (vi) Yangtze River Basin, (vii) Southeast Basin, (viii) Pearl River Basin, (ix) Southwest Basin; (b) 0.1° × 0.1° density map of gauges used in gauge observations over China.

Figure 1

(a) Map of nine study basins in China and their topographic distributions. The solid black line indicates the outline of the basins: (i) Songliao River Basin, (ii) Haihe River Basin, (iii) Yellow River Basin, (iv) Inland River Basin, (v) Huaihe River Basin, (vi) Yangtze River Basin, (vii) Southeast Basin, (viii) Pearl River Basin, (ix) Southwest Basin; (b) 0.1° × 0.1° density map of gauges used in gauge observations over China.

Study data

The GSMaP algorithm produces high spatiotemporal resolution and quasi-global (60 °N–S) satellite precipitation with various PMW and IR sensors. The incorporated PMW sensors mainly include TRMM Microwave Imager (TMI), the Special Sensor Microwave Imager, and Special Sensor Microwave Imager/Sounder on Defense Meteorological Satellite Program platforms, Aqua Advanced Microwave Scanning Radiometer-EOS, NOAA Advanced Microwave Sounding Unit-A/Unit-B (AMSU-A/-B), and Microwave Humidity Sounder. Two additional data sources were also employed in PMW retrieval algorithms: (1) 1.25° 6-hourly Japan Meteorological Agency (JMA) Global Analysis, and (2) 1° daily JMA Merged satellite and in situ data Global Daily Sea Surface Temperatures. The globally merged, full-resolution (4 km × 4 km) IR data used in GSMaP, which are merged from the 11 micron IR channels aboard the MTSAT (operated by JMA), METSOSAT-7/-8 (operated by EUMETSAT), and GOES-11/-12 (operated by NOAA), were provided by NOAA/CPC.

The GSMaP system first adopts the above PMW radiometers (i.e., imagers and sounders loaded on different satellite platforms) to retrieve the global precipitation rates within the latitude band 60 °N–60 °S (Kubota et al. 2007; Okamoto et al. 2007; Aonashi et al. 2009), and then uses the cloud moving vector derived from the IR images at the present time and 1 hour before to obtain the propagation of the precipitation distribution for filling the gaps between the microwave radiometer overpasses (Joyce et al. 2004). It is noted that the rain/no-rain classification database derived from the TRMM PR for the TRMM Microwave Imager (Seto et al. 2005) and other imagers aboard the polar-orbit satellites (Seto et al. 2008) are essential in the PMW retrieval algorithms because it can effectively help to classify radiometer observations over land as either ‘rain’ or ‘no rain’. Based on the above merging techniques, the GSMaP_NRT estimates that only consider the propagation process forward in time are developed. To further reduce the total retrieval errors, Ushio et al. (2009) employed a new Kalman filter approach to assimilate and refine the IR-based rain rates and thus generated an improved version GSMaP_MVK. The key difference between these two GSMaP products is that GSMaP_MVK contains a two-way (both forward and backward) morphing technique to propagate the rainy area from microwave radiometers. Finally, the NOAA/CPC gauge-based analysis of global daily precipitation (Xie et al. 2007) was integrated into the GSMaP_MVK estimates for realizing the satellite–gauge combination (Mega et al. 2014).

In our study, all the three GSMaP products (GSMaP_NRT, GSMaP_MVK, and GSMaP_Gauge) are assessed against one gauge-based hourly rainfall product. We chose the newly released gridded China Gauge-based Hourly Precipitation Analysis developed by the National Meteorological Information Center of the China Meteorological Administration as the ground reference data during the 2009–2010 warm seasons (May to September). The source observation data of the reference data are collected from more than 30,000 national standard automatic observation gauges distributed over the entire mainland China (Figure 1(b)). The quality of hourly gauge data was first double checked and controlled under strict guidelines (Ren et al. 2010), and then the processed data were interpolated to grid data with a spatial resolution of 0.1° using a modified climatology-based optimal interpolation algorithm (Xie et al. 2007; Shen et al. 2010a, 2010b, 2014a). Some other gauge observation datasets which used similar quality-controlled methods have been used widely over mainland China (Chen et al. 2013; Shen et al. 2014b).

Table 1 summarizes the coverage and spatiotemporal resolutions of GSMaP products and ground observation used in this study. The period analyzed consists of the 2009–2010 warm seasons (May to September), during which both observed precipitation and the GSMaP satellite datasets are all available. All of the analyses in this paper were performed just for those grid boxes with at least one gauge available to ensure statistical significance and low observed error.

Table 1

Coverage and spatiotemporal resolutions of GSMaP products and ground observation used in this study

Precipitation datasets Product Data source IR technique Corrected by gauges Temporal resolution Spatial resolution Period Coverage 
Satellite precipitation GSMaP_NRT PMW imagers, PMW sounders, GEO IR radiometers Morphing and Kalman filter, by forward process No 1 hourly 0.1° 2008.10.10 to the present 60 °N–60 °S 
GSMaP_MVK PMW imagers, PMW sounders, GEO IR radiometers Morphing and Kalman filter, by forward and backward process No 1 hourly 0.1° 2000.03.01 to 2010.11.30 60 °N–60 °S 
GSMaP_Gauge PMW imagers, PMW sounders, GEO IR radiometers, CPC global gauge data analysis Morphing and Kalman filter, by forward and backward process Yes 1 hourly 0.1° 2000.03.01 to 2010.11.30 60 °N–60 °S 
Ground observation China hourly Precipitation Analysis More than 30,000 hourly gauges 1 hourly 0.1° 2008.01.01 to the present Mainland China 
Precipitation datasets Product Data source IR technique Corrected by gauges Temporal resolution Spatial resolution Period Coverage 
Satellite precipitation GSMaP_NRT PMW imagers, PMW sounders, GEO IR radiometers Morphing and Kalman filter, by forward process No 1 hourly 0.1° 2008.10.10 to the present 60 °N–60 °S 
GSMaP_MVK PMW imagers, PMW sounders, GEO IR radiometers Morphing and Kalman filter, by forward and backward process No 1 hourly 0.1° 2000.03.01 to 2010.11.30 60 °N–60 °S 
GSMaP_Gauge PMW imagers, PMW sounders, GEO IR radiometers, CPC global gauge data analysis Morphing and Kalman filter, by forward and backward process Yes 1 hourly 0.1° 2000.03.01 to 2010.11.30 60 °N–60 °S 
Ground observation China hourly Precipitation Analysis More than 30,000 hourly gauges 1 hourly 0.1° 2008.01.01 to the present Mainland China 

METHODS

To quantify the accuracy of GSMaP estimates, we used two types of statistical indices including continuous statistics (e.g., correlation coefficient (CC), relative bias (BIAS), mean absolute error (MAE), mean error (ME), and root mean square error (RMSE)) and contingency table-based detection of rainy events (e.g., probability of detection (POD), false alarm ratio (FAR), critical success index (CSI)). A brief description of them is provided below.

The CC is used to reflect the degree of linear correlation between satellite-based precipitation and gauge observations, the BIAS measures the systematic bias of satellite precipitation estimates, the MAE represents the average magnitude of the error, the ME simply scales the average difference between the satellite precipitation estimates and ground observations, and the RMSE also measures the average error magnitude but gives greater weight to the larger errors relative to MAE. The formulas (Equations (1)–(5)) for the aforementioned indices are as follows:  
formula
(1)
 
formula
(2)
 
formula
(3)
 
formula
(4)
 
formula
(5)
where Si is satellite-based precipitation, Gi is gauged observation, n is number of the samples, is mean satellite-based precipitation, and is mean gauged observation.
On the other hand, the contingency statistics are used to examine the correspondence between the estimated and observed occurrence of rain events (Ebert et al. 2007). This is performed by POD, FAR, and CSI. They can be calculated by the following formulas (see Equations (6)–(8)):  
formula
(6)
 
formula
(7)
 
formula
(8)
where H is the number of observed rain events correctly detected, M is the number of observed rain events not detected, and F is the number of rainfall detected but not observed. The contingency table for these statistics and the detailed definition are shown in Table 2 (note that the threshold rain rate used here is 1.0 mm/h).
Table 2

Contingency table for comparing gauge observations and GSMaP estimates

  Gauge ≥ Threshold Gauge < Threshold 
Satellite ≥ Threshold H F 
Satellite < Threshold M Correct negatives 
  Gauge ≥ Threshold Gauge < Threshold 
Satellite ≥ Threshold H F 
Satellite < Threshold M Correct negatives 

RESULTS AND DISCUSSION

Average precipitation distribution in China

Figure 2(a)2(d) show the daily average precipitation for the gauge observations and three GSMaP products during the study period. The gray region in Figure 2(a) represents no gauge observations. Generally speaking, all three GSMaP products can capture the entire precipitation patterns against gauge observations but their differences are notable. One can see that the precipitation distribution of GSMaP_MVK is most similar to gauge observations in space and it performs best among the three GSMaP products in terms of average precipitation amounts. Compared to GSMaP_MVK, GSMaP_NRT shows an obvious underestimation, especially over southern humid areas (e.g., Yangtze River Basin, Southeast River Basin, and Pearl River Basin, etc.). This arises because the only-forward propagation process used in the GSMaP_NRT algorithm inevitably misses some short-lived convection storm events during the warm months (Kubota et al. 2009). Impressively, GSMaP_Gauge seems to overestimate the rainfall, especially in southeastern China. It is likely that daily CPC gauge data with the coarse resolution (0.5° × 0.5°) employed in the gauge-adjustment algorithm produced more no-raining data to raining events (Mega et al. 2014). In addition, we note that all the three GSMaP products overestimate rainfall in the upper middle reaches of the Yangtze River Basin where the terrain changes dramatically from 1,000 m to more than 3,000 m (refer to Figure 1(a)), suggesting that the current GSMaP algorithms still have room for improvement over the complex terrains. Relative to GSMaP_NRT and GSMaP_Gauge, GSMaP_MVK exhibits more rainfall amounts over Northwest China in the Inland River Basin as the current satellite-based precipitation estimates tend to have more uncertainties in high-altitude areas (Yong et al. 2015). Furthermore, we also found that both GSMaP_MVK and GSMaP_NRT underestimate the remarkable local rainfall along the Chinese southwest border in the Southwest Basin while over this region the GSMaP_Gauge can capture this feature rather well. This finding is consistent with the assessment results of Krakauer et al. (2013), in that the ungauged GSMaP estimates underestimated precipitation in mountainous regions over Nepal. This implies that GSMaP_Gauge succeeds in capturing the rainfall characteristics over this region by gauged adjustments with the CPC ground observed data.

Figure 2

Spatial distributions of the 2009–2010 mean warm seasons (May–September) precipitation (mm day−1) from gauge observations, GSMaP_NRT, GSMaP_MVK, and GSMaP_Gauge over nine basins.

Figure 2

Spatial distributions of the 2009–2010 mean warm seasons (May–September) precipitation (mm day−1) from gauge observations, GSMaP_NRT, GSMaP_MVK, and GSMaP_Gauge over nine basins.

Spatial analysis of error features

To investigate the spatial characteristics of the GSMaP products, we used four statistical indices in this section, including CC, BIAS, MAE, and RMSE. We selected the MAE because it is a more appropriate measure for average error than ME. The MAE retains the difference in magnitude as it can avoid the fact that positive and negative differences cancel each other to some degree (Willmott & Matsuura 2005; Yong et al. 2010).

Figure 3 displays the spatial distributions of CC, BIAS, MAE, and RMSE, which were computed from three GSMaP products against gauge observations on the 0.1° × 0.1° and hourly resolution. Figure 3(a)3(c) show that better CC values appear in the east and south of China with low elevation and latitude, such as the Pearl River Basin, Southeast River Basin, and Yangtze River Basin, etc. In contrast, the worse CC values appear in the northwest and southwest of China with high elevation and arid climate, such as the Inland River Basin and the Southwest River Basin. In addition, it is clear that the GSMaP_Gauge estimates have the best CC among the three products, while the CC value of GSMaP_MVK looks better than that of GSMaP_NRT. As for BIAS, it seems that the gauge adjustment algorithm in GSMaP_Gauge failed to effectively correct the systematic bias in China (Figure 3(d)3(f)), which is similar to the patterns in Figure 2. Particularly in southern China, the BIAS of GSMaP_Gauge is obviously larger than that of GSMaP_MVK. In general, the GSMaP_NRT and GSMaP_MVK products without gauged adjustments overestimated the rainfall in the northern basins (e.g., Inland River Basin, Haihe River Basin, etc.), while they underestimated the rainfall in the southern basins (e.g., Southeast River Basin, Yangtze River Basin, Pearl River Basin, etc.). Relative to GSMaP_MVK and GSMaP_Gauge, GSMaP_NRT has less overestimation in the north but the apparent underestimation occurs in the south. In terms of MAE, the three GSMaP products show rather similar spatial patterns. GSMaP_MVK performs the worst and GSMaP_Gauge performs better than GSMaP_NRT, evidently especially in the Songliao River Basin, Southeast River Basin and the Pearl River Basin. Clearly, also with very similar spatial patterns, GSMaP_NRT has the worst RMSE values and GSMaP_Gauge performs better than GSMaP_MVK but with very little difference.

Figure 3

Spatial distributions of statistical indices computed from the (1st column) GSMaP_NRT, (2nd column) GSMaP_MVK, (3rd column) GSMaP_Gauge hourly precipitation at 0.1° × 0.1° resolution over nine basins: (a)–(c) CC, (d)–(f) BIAS, (g)–(i) MAE, and (j)–(l) RMSE.

Figure 3

Spatial distributions of statistical indices computed from the (1st column) GSMaP_NRT, (2nd column) GSMaP_MVK, (3rd column) GSMaP_Gauge hourly precipitation at 0.1° × 0.1° resolution over nine basins: (a)–(c) CC, (d)–(f) BIAS, (g)–(i) MAE, and (j)–(l) RMSE.

Contingency statistics can reveal the performance of the probability of detection, the false alarm ratio, and the rate of successful hits. These statistics can help to further reveal the differences among the three GSMaP products. Figure 4 shows the spatial distributions of POD, FAR, and CSI for the three GSMaP products over nine basins.

Figure 4

The same as Figure 3, but for POD, FAR, and CSI.

Figure 4

The same as Figure 3, but for POD, FAR, and CSI.

Generally, GSMaP_Gauge and GSMaP_MVK evidently perform better than GSMaP_NRT over all basins in terms of POD, FAR, and CSI. This is likely related to the simple NRT algorithm in the GSMaP system. For the POD, the results of GSMaP_NRT indicate that the GSMaP_NRT miss some precipitation events over the nine basins. GSMaP_MVK effectively improved the performance, especially in southern and eastern China, due to both its forward and backward process in the retrieval algorithm. After the gauge was corrected, GSMaP_Gauge also improved significantly at regions where the gauge stations are dense (Southeast River Basin, the lower reaches of Yangtze River Basin and Yellow River Basin, Pearl River Basin, etc.). Interestingly, the differences among the three GSMaP products on FAR are not as obvious as POD, with their differences being very small. We can conclude that this may be related to the fact that GSMaP_Gauge and GSMaP_MVK have some excess rainfall events and produce larger F values, while GSMaP_NRT has insufficient rainfall events and produces smaller H values. Better FAR values were found in southern China while worse values occurred in the north. GSMaP_MVK performs best in the south and GSMaP_Gauge performs best in the north. Similar to the POD results, GSMaP_Gauge has the best performance on CSI among the three products, illustrating that the gauge corrected procedure can effectively improve the ability of detecting rainfall events. The worst POD, FAR, and CSI values were found over western China (Inland River Basin, Southwest River Basin) which suggests that high systematic error and FAR values may be related to altitude effect in mountain regions. In addition, compared with GSMaP_NRT, all statistics improved more over southern and wet basins than over western and arid basins. As we know, the precipitation type in southern China is almost convective during warm seasons while the precipitation type across the western and arid basins is almost stratiform. Thus, we can conclude that both the forward and backward processes may work effectively on convective rainfall events while there is almost no improvement on stratiform rainfall events.

Temporal analysis of error features

From the aforementioned results (Figure 3(d)3(f)), we found that GSMaP_NRT has the best performance in the northern regions while GSMaP_MVK performs best in the southern regions on BIAS. It seems that the gauge-adjustment tends to generate some positive bias in warm seasons.

Figures 5 and 6 show the typical time series of the daily ME and MAE over the nine basins. The daily ME values of GSMaP_MVK and GSMaP_NRT show a large fluctuation between negative and positive, while GSMaP_Gauge always shows positive values over almost all basins. Thus, we can infer that GSMaP_NRT and GSMaP_MVK have better BIAS just because the positive and negative ME differences cancel each other out. In terms of MAE (Figure 6), GSMaP_Gauge totally shows better scores than GSMaP_NRT and GSMaP_MVK over all basins.

Figure 5

Daily series of ME between various products and gauge observations over the nine study basins.

Figure 5

Daily series of ME between various products and gauge observations over the nine study basins.

Figure 6

Daily series of MAE between various products and gauge observations over the nine basins.

Figure 6

Daily series of MAE between various products and gauge observations over the nine basins.

All aforementioned results in this section suggest that GSMaP_Gauge can effectively reduce the uncertainties of GSMaP_MVK and GSMaP_NRT after the gauge-corrected procedure although there still appears to be some room for improvement due to its positive bias.

PDFs by precipitation volume and occurrence

The satellite-based precipitation estimates have been found to exhibit strong rain-rate dependency (Tian et al. 2009, 2010). To investigate the rain-rate dependency of GSMaP estimates, Figures 7 and 8 show the hourly precipitation probability distribution functions (PDFs) by volume (PDFv) and the PDFs by occurrence (PDFc), respectively, for GSMaP and gauge observations.

Figure 7

Probability distribution functions by volume (PDFv) for different basins, respectively, for gauge observations, GSMaP_NRT, GSMaP_MVK, and GSMaP_Gauge. The logarithmic scale was used to bin the precipitation rates.

Figure 7

Probability distribution functions by volume (PDFv) for different basins, respectively, for gauge observations, GSMaP_NRT, GSMaP_MVK, and GSMaP_Gauge. The logarithmic scale was used to bin the precipitation rates.

Figure 8

The same as Figure 7 but for probability distribution functions by occurrence (PDFc).

Figure 8

The same as Figure 7 but for probability distribution functions by occurrence (PDFc).

As shown in Figure 7, GSMaP_MVK and GSMaP_Gauge generally have similar distribution patterns. The main differences between them are shown at the range of 0.125–4.0 mm/h where GSMaP_Gauge obviously overestimated the rainfall amounts in all basins. GSMaP_MVK shows a very similar pattern to the gauge observations at the lower end (<0.5 mm/h) in most basins except for the Inland River Basin, the Songliao River Basin, the Huaihe River Basin, and the Haihe Basin. Meanwhile, we can note that both GSMaP_MVK and GSMaP_Gauge overestimated the rainfall at low and mid rain rate range (approximately <8 mm/h) and underestimated the rainfall at high rain rate range (>8 mm/h) in most basins. In the Southwest Basin, the underestimations of two GSMaP products were shown after the mid rain rate (4.0 mm/h), suggesting that GSMaP products have poorer performance in detecting the mid and strong precipitation in those high-altitude mountainous regions. This may be because the precipitation estimates missed the changes of precipitation caused by topographic gradients in small scale (Andermann et al. 2011). It is worth mentioning that GSMaP_Gauge showed more overestimations at low and mid rain rate range and more underestimations at high rain rate range than GSMaP_MVK. Compared with GSMaP_MVK, GSMaP_NRT almost underestimated the rainfall amounts in all the ranges of rain rate, especially in those southern basins, due to its only forward process algorithm.

The hourly precipitation PDFs by occurrence (PDFc) are shown in Figure 8. For all basins, GSMaP_Gauge seriously overestimated the rainfall events at low rain rate range (0.125–2.0 mm/h) and underestimated the heavy rainfall events (>8 mm/h). Thus, we can confirm that GSMaP_Gauge increased many excess low-level precipitation events and decreased some really heavy rainfall events after the gauge adjustments. Additionally, we find that the curves for GSMaP_Gauge began to decline and close to the curves for GSMaP_MVK and gauge observations when the precipitation rate was over 0.5 mm/h (Figure 8). However, in Figure 7, the curves for GSMaP_Gauge still rose and had no downward trend in this range. From this, we can conclude that the GSMaP_Gauge algorithm basically increased the rain rate at low rain rate range compared to GSMaP_MVK. Returning to Figure 7, we can see that the distribution curves amount for gauge observations and GSMaP_MVK are very similar at low rain rate range, while the curves for GSMaP_MVK are a little lower than the curves for gauge observations in most of the basins in Figure 8, probably as a result of GSMaP_MVK increasing the rain rate at low rain rate (<0.5 mm/h). GSMaP_NRT obviously missed many rainfall events at low rain rate range especially in the southern basins, suggesting that its underestimations mainly result from an insufficient number of rainfall events.

In summary, from all the aforementioned analysis in this section, we can draw the following three conclusions: (1) GSMaP products tend to overestimate the rainfall at low and mid rain rate range and underestimate the rainfall at high rain rate range; (2) the forward and backward propagation processes in the GSMaP_MVK algorithm can evidently reduce the underestimations of rainfall events, especially in the regions that have significant convective precipitation; (3) the gauge-based bias-correction in the GSMaP_Gauge algorithm mainly increased the rainfall at low rain rate range (approximately <2 mm/h) and decreased the rainfall at high rain rate range to a certain degree at the same time. These conclusions can help hydrology users understand the error characteristics of GSMaP products and consider what consequences will occur when they are used to drive hydrological models.

Diurnal variations

Hourly rainfall composites were created by computing grids of average rainfall rate for each hour in the dataset. Figure 9 compares the 2009–10 composite diurnal cycles of warm seasons' precipitation amount from gauge observations, and GSMaP products (at UTC hour) over nine Chinese river basins. The intention of the diurnal analysis is to determine how well each of the algorithms captures important details of the diurnal rainfall cycle and to investigate the diurnal variations' characteristics in all validation basins.

Figure 9

Time series of basin-average mean hourly rainfall over nine basins from gauge observations, GSMaP_NRT, GSMaP_MVK, and GSMaP_Gauge.

Figure 9

Time series of basin-average mean hourly rainfall over nine basins from gauge observations, GSMaP_NRT, GSMaP_MVK, and GSMaP_Gauge.

Large diurnal variations were observed for most basins except the Yellow River Basin and the Inland River Basin. For example, the gauge observations show a large peak over the Songliao Basin and Haihe Basin and a strong peak over the Southeast Basin and Pearl River Basin around 0900 UTC. Meanwhile, two diurnal peaks were seen over the Huaihe Basin, Yangtze River Basin, and Southwest Basin: for the Huaihe Basin, one in the later afternoon (0900 UTC), and another one in the early morning (2000 UTC); for the Yangtze River Basin, a strong peak occurred in the early morning (2200 UTC) and another large peak was seen in the afternoon (0700 UTC); and for the Southwest Basin, two peaks occurred in the early morning (2000 UTC) and in the afternoon (0700 UTC), respectively.

The GSMaP products generally show comparable performance of diurnal variations over the Southeast Basin. Over the Songliao Basin, Haihe Basin, Huaihe Basin, and Pearl River Basin, all three GSMaP products captured the peak (0900 UTC), but suffered from the opposite trend (obviously overestimate) from 1200 UTC to 1500 UTC. Over the Yangtze River Basin and Southwest Basin, GSMaP_Gauge and GSMaP_MVK captured two peaks while GSMaP_NRT showed no peak with continuous underestimation, and the same opposite trend from 1200 UTC to 1500 UTC were also found in these basins. Interestingly, the underestimations all appeared in the period from 2200 to 0100 UTC except in the Southwest Basin. Meanwhile, it is worth mentioning that the overestimations in daytime (2200 to 1000 UTC) are less than the overestimations at night (1000 to 2200 UTC), indicating that GSMaP products have a better performance in daytime than at night. The aforementioned results demonstrated the diurnal variations' characteristics over nine Chinese basins and the results are basically consistent with the conclusions demonstrated by Zhou et al. (2008).

Grid-based validation and comparison

In this section, we used the grid-based analysis proposed in Yong et al. (2010) to perform the comparison of GSMaP estimates against observations over all the basins. The results are summarized in Table 3. In general, the results are consistent with the aforementioned analysis. All the better CC values for three GSMaP products appear in the south of China (the Pearl River Basin, the Yangtze River Basin, the Southeast Basin, refer to Table 3). Relative to GSMaP_NRT, GSMaP_MVK and GSMaP_Gauge seem to greatly improve the correlation between satellite estimates and gauge observations in all study basins, and GSMaP_Gauge performs better than GSMaP_MVK (e.g., the CC increased from 0.165 for GSMaP_NRT to 0.341 for GSMaP_MVK while it was 0.407 for GSMaP_Gauge in the Inland River Basin at 1-hour scale). Impressively, the BIAS values for GSMaP_MVK in some basins (e.g., the Pearl River Basin, the Yangtze River Basin, and the Southeast Basin) are almost 0; this may be because the positive and the negative differences cancel each other (see Figures 5 and 6). In contrast, the BIAS values for GSMaP_Gauge are larger than GSMaP_MVK by 30% to 40% in all basins except the Inland River Basin, suggesting that the GSMaP_Gauge algorithm produced some systematic errors in rainfall estimation. GSMaP_NRT shows negative BIAS values in most southern basins and shows positive values in northern basins (e.g., the Inland River Basin, the Songliao Basin, and the Haihe Basin). When compared to GSMaP_MVK, GSMaP_Gauge has better statistical performance on RMSE over all validation basins. However, this improvement is not so notable compared to the improvement in CC. In addition, we note that all three GSMaP products present high RMSE values (>1 mm at 1-hour scale) in the coastal regions (e.g., the Southeast Basin, the Yangtze River Basin, the Huaihe Basin, and the Pearl River Basin), illustrating that GSMaP algorithms still have significant room for improving the estimation of extremely heavy rainfall in monsoon regions.

Table 3

Statistical summary of the grid-based comparison of three GSMaP estimates (the GSMaP_NRT, GSMaP_MVK, and GSMaP_Gauge) at different time scales

Basin Products CC
 
MAE (mm)
 
RMSE (mm)
 
BIAS 
1 h 3 h 6 h 12 h 24 h 1 h 3 h 6 h 12 h 24 h 1 h 3 h 6 h 12 h 24 h 1 h 
Songliao R GSMaP_NRT 0.228 0.326 0.373 0.441 0.496 0.175 0.544 1.006 1.825 3.255 0.925 2.281 3.667 5.576 8.401 9.3% 
GSMaP_MVK 0.406 0.530 0.570 0.614 0.640 0.172 0.525 0.957 1.701 2.965 0.839 1.944 3.070 4.693 7.141 20.8% 
GSMaP_Gauge 0.461 0.558 0.619 0.672 0.707 0.151 0.454 0.840 1.542 2.716 0.763 1.893 3.031 4.586 6.932 57.9% 
Haihe R GSMaP_NRT 0.238 0.323 0.366 0.413 0.455 0.180 0.545 1.017 1.897 3.511 1.051 2.623 4.207 6.395 9.548 −1.4% 
GSMaP_MVK 0.387 0.492 0.527 0.562 0.592 0.191 0.558 1.029 1.853 3.498 1.072 2.563 4.105 6.217 9.288 21.7% 
GSMaP_Gauge 0.454 0.547 0.597 0.661 0.728 0.172 0.514 0.968 1.815 3.177 0.965 2.438 3.968 5.865 8.432 59.6% 
Yellow R GSMaP_NRT 0.263 0.347 0.390 0.456 0.495 0.141 0.412 0.771 1.416 2.617 0.823 2.759 3.263 4.990 7.484 −13.2% 
GSMaP_MVK 0.424 0.525 0.548 0.588 0.614 0.160 0.466 0.857 1.524 2.635 0.780 2.009 3.065 4.755 7.151 14.2% 
GSMaP_Gauge 0.481 0.571 0.619 0.678 0.731 0.140 0.410 0.767 1.419 2.315 0.733 1.839 2.993 4.567 6.785 57.1% 
Inland R GSMaP_NRT 0.165 0.254 0.318 0.392 0.458 0.089 0.250 0.470 0.877 1.571 1.145 1.023 1.645 2.597 3.984 9.1% 
GSMaP_MVK 0.341 0.436 0.487 0.544 0.589 0.086 0.237 0.441 0.805 1.894 1.121 0.975 1.573 2.435 3.629 40.6% 
GSMaP_Gauge 0.407 0.492 0.534 0.596 0.645 0.085 0.220 0.386 0.707 1.555 1.096 0.952 1.513 2.274 3.291 49.3% 
Huaihe R GSMaP_NRT 0.299 0.390 0.447 0.507 0.551 0.209 0.592 1.103 2.029 3.777 1.173 2.531 4.063 6.146 10.121 −16.6% 
GSMaP_MVK 0.427 0.542 0.572 0.626 0.659 0.238 0.681 1.246 2.224 3.882 1.179 2.719 4.355 6.584 9.877 8.6% 
GSMaP_Gauge 0.478 0.566 0.617 0.681 0.733 0.206 0.582 1.080 1.988 3.683 1.102 2.658 4.302 6.457 9.585 46.2% 
Yangtze R GSMaP_NRT 0.281 0.371 0.442 0.494 0.536 0.241 0.721 1.338 2.452 4.426 1.188 2.922 4.775 7.360 11.044 −29.3% 
GSMaP_MVK 0.474 0.567 0.621 0.679 0.686 0.257 0.795 1.449 2.542 4.284 1.094 2.639 4.292 6.583 9.776 −7.1% 
GSMaP_Gauge 0.519 0.608 0.673 0.716 0.756 0.223 0.666 1.228 2.231 4.029 1.035 2.602 4.259 6.457 9.469 38.0% 
Southeast R GSMaP_NRT 0.252 0.352 0.396 0.468 0.533 0.339 0.966 1.782 3.205 5.653 1.498 3.492 5.650 8.622 13.212 −19.2% 
GSMaP_MVK 0.463 0.604 0.624 0.652 0.681 0.332 0.925 1.660 2.910 5.459 1.310 2.878 4.654 7.277 11.346 −5.3% 
GSMaP_Gauge 0.515 0.630 0.685 0.728 0.782 0.293 0.823 1.509 2.762 4.969 1.222 2.790 4.490 6.804 10.104 29.0% 
Pearl R GSMaP_NRT 0.206 0.292 0.354 0.410 0.467 0.342 0.972 1.798 3.261 5.720 1.664 3.848 6.214 9.469 14.167 −31.2% 
GSMaP_MVK 0.403 0.520 0.575 0.612 0.634 0.376 1.049 1.877 3.291 5.559 1.607 3.691 5.940 9.314 13.626 −3.83% 
GSMaP_Gauge 0.446 0.541 0.606 0.659 0.696 0.318 0.884 1.610 2.921 5.233 1.483 3.571 5.771 8.880 12.703 35.5% 
Southwest R GSMaP_NRT 0.211 0.299 0.355 0.405 0.449 0.276 0.800 1.482 2.699 4.725 1.124 2.593 4.147 6.382 9.336 −28.6% 
GSMaP_MVK 0.391 0.472 0.519 0.550 0.564 0.291 0.825 1.489 2.613 4.360 1.008 2.291 3.623 5.636 8.467 −26.0% 
GSMaP_Gauge 0.423 0.498 0.557 0.600 0.638 0.235 0.679 1.256 2.294 4.035 0.962 2.258 3.593 5.511 8.141 20.6% 
Basin Products CC
 
MAE (mm)
 
RMSE (mm)
 
BIAS 
1 h 3 h 6 h 12 h 24 h 1 h 3 h 6 h 12 h 24 h 1 h 3 h 6 h 12 h 24 h 1 h 
Songliao R GSMaP_NRT 0.228 0.326 0.373 0.441 0.496 0.175 0.544 1.006 1.825 3.255 0.925 2.281 3.667 5.576 8.401 9.3% 
GSMaP_MVK 0.406 0.530 0.570 0.614 0.640 0.172 0.525 0.957 1.701 2.965 0.839 1.944 3.070 4.693 7.141 20.8% 
GSMaP_Gauge 0.461 0.558 0.619 0.672 0.707 0.151 0.454 0.840 1.542 2.716 0.763 1.893 3.031 4.586 6.932 57.9% 
Haihe R GSMaP_NRT 0.238 0.323 0.366 0.413 0.455 0.180 0.545 1.017 1.897 3.511 1.051 2.623 4.207 6.395 9.548 −1.4% 
GSMaP_MVK 0.387 0.492 0.527 0.562 0.592 0.191 0.558 1.029 1.853 3.498 1.072 2.563 4.105 6.217 9.288 21.7% 
GSMaP_Gauge 0.454 0.547 0.597 0.661 0.728 0.172 0.514 0.968 1.815 3.177 0.965 2.438 3.968 5.865 8.432 59.6% 
Yellow R GSMaP_NRT 0.263 0.347 0.390 0.456 0.495 0.141 0.412 0.771 1.416 2.617 0.823 2.759 3.263 4.990 7.484 −13.2% 
GSMaP_MVK 0.424 0.525 0.548 0.588 0.614 0.160 0.466 0.857 1.524 2.635 0.780 2.009 3.065 4.755 7.151 14.2% 
GSMaP_Gauge 0.481 0.571 0.619 0.678 0.731 0.140 0.410 0.767 1.419 2.315 0.733 1.839 2.993 4.567 6.785 57.1% 
Inland R GSMaP_NRT 0.165 0.254 0.318 0.392 0.458 0.089 0.250 0.470 0.877 1.571 1.145 1.023 1.645 2.597 3.984 9.1% 
GSMaP_MVK 0.341 0.436 0.487 0.544 0.589 0.086 0.237 0.441 0.805 1.894 1.121 0.975 1.573 2.435 3.629 40.6% 
GSMaP_Gauge 0.407 0.492 0.534 0.596 0.645 0.085 0.220 0.386 0.707 1.555 1.096 0.952 1.513 2.274 3.291 49.3% 
Huaihe R GSMaP_NRT 0.299 0.390 0.447 0.507 0.551 0.209 0.592 1.103 2.029 3.777 1.173 2.531 4.063 6.146 10.121 −16.6% 
GSMaP_MVK 0.427 0.542 0.572 0.626 0.659 0.238 0.681 1.246 2.224 3.882 1.179 2.719 4.355 6.584 9.877 8.6% 
GSMaP_Gauge 0.478 0.566 0.617 0.681 0.733 0.206 0.582 1.080 1.988 3.683 1.102 2.658 4.302 6.457 9.585 46.2% 
Yangtze R GSMaP_NRT 0.281 0.371 0.442 0.494 0.536 0.241 0.721 1.338 2.452 4.426 1.188 2.922 4.775 7.360 11.044 −29.3% 
GSMaP_MVK 0.474 0.567 0.621 0.679 0.686 0.257 0.795 1.449 2.542 4.284 1.094 2.639 4.292 6.583 9.776 −7.1% 
GSMaP_Gauge 0.519 0.608 0.673 0.716 0.756 0.223 0.666 1.228 2.231 4.029 1.035 2.602 4.259 6.457 9.469 38.0% 
Southeast R GSMaP_NRT 0.252 0.352 0.396 0.468 0.533 0.339 0.966 1.782 3.205 5.653 1.498 3.492 5.650 8.622 13.212 −19.2% 
GSMaP_MVK 0.463 0.604 0.624 0.652 0.681 0.332 0.925 1.660 2.910 5.459 1.310 2.878 4.654 7.277 11.346 −5.3% 
GSMaP_Gauge 0.515 0.630 0.685 0.728 0.782 0.293 0.823 1.509 2.762 4.969 1.222 2.790 4.490 6.804 10.104 29.0% 
Pearl R GSMaP_NRT 0.206 0.292 0.354 0.410 0.467 0.342 0.972 1.798 3.261 5.720 1.664 3.848 6.214 9.469 14.167 −31.2% 
GSMaP_MVK 0.403 0.520 0.575 0.612 0.634 0.376 1.049 1.877 3.291 5.559 1.607 3.691 5.940 9.314 13.626 −3.83% 
GSMaP_Gauge 0.446 0.541 0.606 0.659 0.696 0.318 0.884 1.610 2.921 5.233 1.483 3.571 5.771 8.880 12.703 35.5% 
Southwest R GSMaP_NRT 0.211 0.299 0.355 0.405 0.449 0.276 0.800 1.482 2.699 4.725 1.124 2.593 4.147 6.382 9.336 −28.6% 
GSMaP_MVK 0.391 0.472 0.519 0.550 0.564 0.291 0.825 1.489 2.613 4.360 1.008 2.291 3.623 5.636 8.467 −26.0% 
GSMaP_Gauge 0.423 0.498 0.557 0.600 0.638 0.235 0.679 1.256 2.294 4.035 0.962 2.258 3.593 5.511 8.141 20.6% 

A comprehensive understanding of satellite estimates' errors across different temporal scales is essential for hydrological applications to use satellite-based precipitation. Meanwhile, error characteristics could be quite different when the precipitation rates are temporal accumulation (Mehran & AghaKouchak 2014). Table 3 presents the results; the CC, MAE, and the RMSE values significantly generally increase as the time interval increases. In summary, GSMaP_MVK has the best performance on BIAS, GSMaP_Gauge performs best on the CC, MAE, and RMSE. We fully believe that the results can provide much more useful information for GSMaP users.

CONCLUSIONS

The objective of the GSMaP project is to produce high-precision and high-resolution global precipitation maps using available PMW and IR data, and gauge observations. Owing to the high spatiotemporal resolution, the GSMaP products have been widely utilized in a variety of research and operational applications, especially for hydrology and meteorology. In this study, we performed a quantitative assessment for the GSMaP product suite (GSMaP_NRT, GSMaP_MVK, GSMaP_Gauge) against ground gauge observations over nine major basins of China at 1 hourly, 0.1° × 0.1° scales. Our aim is to provide data users and algorithm developers with some valuable insights into the GSMaP's error features and its hydrometeorological potential. The main findings of this study are summarized as follows:

  1. Overall, GSMaP_NRT obviously underestimates the rainfall amounts over most regions of southern China, while GSMaP_Gauge tends to overestimate precipitation, especially over southeastern China. In terms of BIAS, GSMaP_MVK seems to exhibit the best performance due to the positive and negative differences canceling each other. In practice, GSMaP_Gauge has the relatively higher CC value and lower MAE and RMSE, suggesting that it has better error stability and distribution. However, it is noted that the positive systematic biases existing in the current GSMaP_Gauge mainly come from overestimation at lower rain rates, which still need to further improve in future development.

  2. Three GSMaP products evaluated in this paper generally have similar spatial distributions of CC, BIAS, MAE, and RMSE. Better performances were observed over southern and southeastern regions (e.g., Pearl River Basin, Southeast Basin), and worse performances occurred over northwestern arid and semiarid regions (e.g., Inland River Basin). All the three GSMaP products exhibit significant overestimation at lower rain rates and underestimation at higher ones. GSMaP_MVK has the relatively better performance for the PDFc and PDFv. After the gauge adjustments, GSMaP_Gauge tends to suffer from an excessive number of lower rainy events (<1 mm/h). Accordingly, GSMaP_NRT missed many rainfall events that resulted in its relative underestimation. The difference between GSMaP_Gauge and GSMaP_MVK gradually decreases as the precipitation threshold increases, implying that the GSMaP_Gauge estimates have stronger rain-rate dependencies.

  3. Our analysis indicates that the forward and backward propagation processes in the GSMaP_MVK algorithm can effectively improve the ability of capturing rainfall events in tropical or subtropical basins relative to the forward-only GSMaP_NRT. The gauge-adjusted procedure in GSMaP_Gauge algorithm can further improve the error structures and pattern features of GSMaP estimates but the positive systematic errors were not completely and fully corrected.

In summary, our study highlights the need for special caution when using the current GSMaP products in some specific regions over the world. We expect the assessment reported here will help both algorithm developers and data users to better understand the GSMaP estimates' error features and their generation mechanisms in hydrometeorological applications.

ACKNOWLEDGEMENTS

This work was financially supported by the National Key Research and Development Program of China (2016YFA0601504) and the National Natural Science Foundation of China (91547101, 51379056, 91437214, and 91647203).

REFERENCES

REFERENCES
Adjei
,
K. A.
,
Ren
,
L. L.
,
Appiah-Adjei
,
E. K.
&
Odai
,
S. N.
2015
Application of satellite-derived rainfall for hydrological modelling in the data-scarce Black Volta trans-boundary basin
.
Hydrology Research
46
(
5
),
777
791
.
Andermann
,
C.
,
Bonnet
,
S.
&
Gloaguen
,
R.
2011
Evaluation of precipitation data sets along the Himalayan front
.
Geochemistry Geophysics Geosystems
12
(
7
),
Q07023
.
Aonashi
,
K.
,
Awaka
,
J.
,
Hirose
,
M.
,
Kozu
,
T.
,
Kubota
,
T.
,
Liu
,
G.
,
Shige
,
S.
,
Kida
,
S.
,
Seto
,
S.
,
Takahashi
,
N.
&
Takayabu
,
Y. N.
2009
GSMap passive microwave precipitation retrieval algorithm: algorithm description and validation
.
Journal of the Meteorological Society of Japan
87A
,
119
136
.
Chen
,
S.
,
Hong
,
Y.
,
Cao
,
Q.
,
Gourley
,
J. J.
,
Kirstetter
,
P. E.
,
Yong
,
B.
,
Tian
,
Y. D.
,
Zhang
,
Z. X.
,
Shen
,
Y.
,
Hu
,
J. J.
&
Hardy
,
J.
2013
Similarity and difference of the two successive V6 and V7 TRMM multisatellite precipitation analysis performance over China
.
Journal of Geophysical Research: Atmospheres
118
,
13060
13074
.
Chen
,
Z. Q.
,
Qin
,
Y. X.
,
Shen
,
Y.
&
Zhang
,
S. P.
2016
Evaluation of global satellite mapping of precipitation project daily precipitation estimates over the Chinese mainland
.
Advances in Meteorology
2016
,
9365294
,
15
.
Ebert
,
E. E.
,
Janowiak
,
J. E.
&
Kidd
,
C.
2007
Comparison of near-real-time precipitation estimates from satellite
.
Bulletin of the American Meteorological Society
88
(
1
),
47
64
.
Heike
,
H.
,
Stefan
,
B.
&
Tong
,
J.
2012
Precipitation variability in the Yangtze River subbasins
.
Water International
37
(
1
),
16
31
.
Hong
,
Y.
,
Hsu
,
K. L.
,
Sorooshian
,
S.
&
Gao
,
X. G.
2004
Precipitation estimation from remotely sensed imagery using an artificial neural network cloud classification system
.
Journal of Applied Meteorology
43
(
12
),
1834
1853
.
Huang
,
A. N.
,
Zhao
,
Y.
,
Zhou
,
Y.
,
Yang
,
B.
,
Zhang
,
L. J.
,
Dong
,
X. N.
,
Fang
,
D. X.
&
Wu
,
Y.
2016
Evaluation of multisatellite precipitation products by use of ground-based data over China
.
Journal of Geophysical Research: Atmospheres
121
,
10654
10675
.
Huffman
,
G. J.
,
Adler
,
R. F.
,
Bolvin
,
D. T.
,
Gu
,
G. J.
,
Nelkin
,
E. J.
,
Bowman
,
K. P.
,
Hong
,
Y.
,
Stocker
,
E. F.
&
Wolff
,
D. B.
2007
The TRMM Multisatellite Precipitation Analysis (TMPA): quasi-global, multiyear, combined-sensor precipitation estimates at fine scales
.
Journal of Hydrometeorology
8
(
1
),
38
55
.
Kidd
,
C.
&
Huffman
,
G. J.
2011
Global precipitation measurement
.
Meteorological Applications
18
(
3
),
334
353
.
Kim
,
J. P.
,
Jung
,
I. W.
,
Park
,
K. W.
,
Yoon
,
S. K.
&
Lee
,
D. H.
2016
Hydrological utility and uncertainty of multi-satellite precipitation products in the mountainous region of South Korea
.
Remote Sensing
8
,
608
.
doi: 10.3390/rs8070608
.
Krakauer
,
N. Y.
,
Pradhanang
,
S. M.
,
Lakhankar
,
T.
&
Jha
,
A. K.
2013
Evaluating satellite products for precipitation estimation in mountain regions: a case study for Nepal
.
Remote Sensing
5
(
8
),
4107
4123
.
Kubota
,
T.
,
Shige
,
S.
,
Hashizume
,
H.
&
Aonashi
,
K.
2007
Global precipitation map using satellite-borne microwave radiometers by the GSMaP project: production and validation
.
IEEE Transactions on Geoscience and Remote Sensing
45
(
7
),
2259
2275
.
Kubota
,
T.
,
Ushio
,
T.
,
Shige
,
S.
,
Kida
,
S.
,
Kachi
,
M.
&
Okamoto
,
K.
2009
Verification of high-resolution satellite-based rainfall estimates around Japan using gauge-calibrated ground-radar dataset
.
Journal of the Meteorological Society of Japan
87A
,
203
222
.
Li
,
L.
,
Hong
,
Y.
,
Wang
,
J. H.
,
Adler
,
R. F.
,
Policelli
,
F. S.
,
Habib
,
S.
,
Irwn
,
D.
,
Korme
,
T.
&
Okello
,
L.
2009
Evaluation of the real-time TRMM-based multi-satellite precipitation analysis for an operational flood prediction system in Nzoia Basin, Lake Victoria, Africa
.
Natural Hazards
50
(
1
),
109
123
.
Mega
,
T.
,
Ushio
,
T.
,
Kubota
,
T.
,
Kachi
,
M.
,
Aonashi
,
K.
&
Shige
,
S.
2014
Gauge Adjusted Global Satellite Mapping of Precipitation (GSMaP_Gauge)
.
General Assembly and Scientific Symposium IEEE,1–4. Available online: http://www.ursi.org/proceedings/procGA14/papers/ursi_paper2042.pdf
.
Meng
,
J.
,
Li
,
L.
,
Hao
,
Z. C.
,
Wang
,
J. H.
&
Shao
,
Q. X.
2014
Suitability of TRMM satellite rainfall in driving a distributed hydrological model in the source region of Yellow River
.
Journal of Hydrology
509
,
320
332
.
Okamoto
,
K.
,
Ushio
,
T.
,
Iguchi
,
T.
,
Takahashi
,
N.
&
Iwanami
,
K.
2005
The global satellite mapping of precipitation (GSMaP) project
. In:
25th IEEE International Geoscience and Remote Sensing Symposium
,
Seoul, South Korea
, pp.
3414
3416
.
Okamoto
,
K.
,
Iguchi
,
T.
,
Takahashi
,
N.
,
Ushio
,
T.
,
Awaka
,
J.
,
Kozu
,
T.
,
Shige
,
S.
&
Kubota
,
T.
2007
High precision and high resolution global precipitation map from satellite data
. In:
Proceedings of ISAP2007
,
Niigata, Japan
, pp.
506
509
.
Peng
,
B.
,
Shi
,
J. C.
,
Ni-Meister
,
W.
,
Zhao
,
T. J.
&
Ji
,
D. B.
2014
Evaluation of TRMM Multisatellite Precipitation Analysis (TMPA) products and their potential hydrological application at an arid and semiarid basin in China
.
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
7
(
9
),
3915
3930
.
Piao
,
S. L.
,
Ciais
,
P.
,
Huang
,
Y.
,
Shen
,
Z. H.
,
Peng
,
S. S.
,
Li
,
J. S.
,
Zhou
,
L. P.
,
Liu
,
H. Y.
,
Ma
,
Y. C.
,
Ding
,
Y. H.
,
Friedlingstein
,
P.
,
Liu
,
C. Z.
,
Tan
,
K.
,
Yu
,
Y. Q.
,
Zhang
,
T. Y.
&
Fang
,
J. Y.
2010
The impacts of climate change on water resources and agriculture in China
.
Nature
467
(
7311
),
43
51
.
Qin
,
Y. X.
,
Chen
,
Z. Q.
,
Shen
,
Y.
,
Zhang
,
S. P.
&
Shi
,
R. H.
2014
Evaluation of satellite rainfall estimates over the Chinese mainland
.
Remote Sensing
6
(
11
),
11649
11672
.
Ren
,
Z. H.
,
Zhao
,
P.
,
Zhang
,
Q.
,
Zhang
,
Z. F.
,
Cao
,
L. J.
,
Yang
,
Y. R.
,
Zou
,
F. L.
,
Zhao
,
Y. F.
,
Zhao
,
H. M.
&
Chen
,
Z.
2010
Quality control procedures for hourly precipitation data from automatic weather stations in China
.
Meteorological Monthly
36
(
7
),
123
132
(in Chinese)
.
Seto
,
S.
,
Kubota
,
T.
,
Takahashi
,
N.
,
Iguchi
,
T.
&
Oki
,
T.
2008
Advanced rain/no-rain classification methods for microwave radiometer observations over land
.
Journal of Applied Meteorology and Climatology
47
(
11
),
3016
3029
.
Shen
,
Y.
,
Feng
,
M. N.
,
Zhang
,
H. Z.
&
Gao
,
F.
2010a
Interpolation methods of China daily precipitation data
.
Journal of Applied Meteorological Science
21
(
3
),
279
286
(in Chinese)
.
Shen
,
Y.
,
Xiong
,
A. Y.
,
Wang
,
Y.
&
Xie
,
P. P.
2010b
Performance of high-resolution satellite precipitation products over China
.
Journal of Geophysical Research
115
(
D2
).
doi: 10.1029/2009JD012097
.
Shen
,
Y.
,
Zhao
,
P.
,
Pan
,
Y.
&
Yu
,
J. J.
2014a
A high spatiotemporal gauge-satellite merged precipitation analysis over China
.
Journal of Geophysical Research: Atmospheres
119
(
6
),
3063
3075
.
Shen
,
Y.
,
Xiong
,
A. Y.
,
Hong
,
Y.
,
Yu
,
J. J.
,
Pan
,
Y.
,
Chen
,
Z. Q.
&
Saharia
,
M.
2014b
Uncertainty analysis of five satellite-based precipitation products and evaluation of three optimally merged multi-algorithm products over the Tibetan Plateau
.
International Journal of Remote Sensing
35
(
19
),
6843
6858
.
Sorooshian
,
S.
,
Hsu
,
K. L.
,
Gao
,
X. G.
,
Gupta
,
H. V.
,
Imam
,
B.
&
Braithwaite
,
D.
2000
Evaluation of PERSIANN system satellite-based estimates of tropical rainfall
.
Bulletin of the American Meteorological Society
81
(
9
),
2035
2046
.
Tian
,
Y. D.
,
Peters-Lidard
,
C. D.
,
Eylander
,
J. B.
,
Joyce
,
R. J.
,
Huffman
,
G. J.
,
Adler
,
R. F.
,
Hsu
,
K. L.
,
Turk
,
F. J.
,
Garcia
,
M.
&
Zeng
,
J.
2009
Component analysis of errors in satellite-based precipitation estimates
.
Journal of Geophysical Research
114
,
D24101
.
Tian
,
Y. D.
,
Peters-Lidard
,
C. D.
,
Adler
,
R. F.
,
Kubota
,
T.
&
Ushio
,
T.
2010
Evaluation of GSMaP precipitation estimates over the contiguous United States
.
Journal of Hydrometeorology
11
(
2
),
566
574
.
Tong
,
K.
,
Su
,
F. G.
,
Yang
,
D. Q.
,
Zhang
,
L. L.
&
Hao
,
Z. C.
2014
Tibetan Plateau precipitation as depicted by gauge observations, reanalyses and satellite retrievals
.
International Journal of Climatology
34
(
2
),
265
285
.
Ushio
,
T.
,
Sasashige
,
K.
,
Kubota
,
T.
,
Shige
,
S.
,
Okamoto
,
K.
,
Aonashi
,
K.
,
Inoue
,
K.
,
Takahashi
,
N.
,
Iguchi
,
T.
,
Kachi
,
M.
,
Oki
,
R.
,
Morimoto
,
T.
&
Kawasaki
,
Z. I.
2009
A Kalman filter approach to the global satellite mapping of precipitation (GSMaP) from combined passive microwave and infrared radiometric data
.
Journal of the Meteorological Society of Japan
87A
,
137
151
.
Xie
,
P. P.
,
Yatagai
,
A.
,
Chen
,
M. Y.
,
Hayasaka
,
T.
,
Fukushima
,
Y.
,
Liu
,
C. M.
&
Yang
,
S.
2007
A gauge-based analysis of daily precipitation over East Asia
.
Journal of Hydrometeorology
8
(
3
),
607
626
.
Yan
,
D. H.
,
Liu
,
S. H.
,
Qin
,
T. L.
,
Weng
,
B. S.
,
Li
,
C. Z.
,
Lu
,
Y. J.
&
Liu
,
J. J.
2016
Evaluation of TRMM precipitation and its application into distributed hydrological model in Naqu River Basin of the Tibetan plateau
.
Hydrology Research
48
(
3
),
822
839
.
doi: 10.2166/nh.2016.090
.
Yong
,
B.
,
Ren
,
L. L.
,
Hong
,
Y.
,
Wang
,
J. H.
,
Gourley
,
J. J.
,
Jiang
,
S. H.
,
Chen
,
X.
&
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 Resources Research
46
(
7
),
759
768
.
Yong
,
B.
,
Hong
,
Y.
,
Ren
,
L. L.
,
Gourley
,
J. J.
,
Huffman
,
G. J.
,
Chen
,
X.
,
Wang
,
W.
&
Khan
,
S. I.
2015
Assessment of evolving TRMM-based multisatellite real-time precipitation estimation methods and their impacts on hydrologic prediction in a high latitude basin
.
Journal of Geophysical Research
117
.
doi: 10.1029/2011JD017069
.
Yong
,
B.
,
Liu
,
D.
,
Gourley
,
J. J.
,
Tian
,
Y. D.
,
Huffman
,
G. J.
,
Ren
,
L. L.
&
Hong
,
Y.
2015
Global view of real-time TRMM multi-satellite precipitation analysis: implications for its successor global precipitation measuring mission
.
Bulletin of the American Meteorological Society
96
(
2
),
283
296
.
Yong
,
B.
,
Wang
,
J. J.
,
Ren
,
L. L.
,
You
,
Y.
,
Xie
,
P. P.
&
Hong
,
Y.
2016
Evaluating four multi-satellite precipitation estimates over Diaoyu Islands during typhoon seasons
.
Journal of Hydrometeorology
17
(
6
),
1623
1641
.
Zhang
,
Z. X.
,
Chen
,
X.
,
Xu
,
C. Y.
,
Yuan
,
L. F.
,
Yong
,
B.
&
Yan
,
S. F.
2011
Evaluating the non-stationary relationship between precipitation and streamflow in nine major basins of China during the past 50 years
.
Journal of Hydrology
409
(
1–2
),
81
93
.