Next Article in Journal
Optimization-Based Downscaling of Satellite-Derived Isotropic Broadband Albedo to High Resolution
Previous Article in Journal
Soil Organic Carbon Prediction and Mapping in Morocco Using PRISMA Hyperspectral Imagery and Meta-Learner Model
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Assessment of Integrated Multi-Satellite Retrievals for Global Precipitation Measurement (IMERG) Precipitation Products in Northwest China

1
College of Atmospheric Sciences, Lanzhou University, Lanzhou 730000, China
2
Lanzhou Central Meteorological Observatory, Lanzhou 730020, China
3
Tianjin Climate Center, Tianjin 300074, China
*
Author to whom correspondence should be addressed.
Remote Sens. 2025, 17(8), 1364; https://doi.org/10.3390/rs17081364
Submission received: 11 February 2025 / Revised: 23 March 2025 / Accepted: 24 March 2025 / Published: 11 April 2025

Abstract

:
This study evaluates the applicability of the IMERG satellite precipitation product in Northwest China using data from more than 6000 ground-level meteorological stations during the warm season (April–September) from 2016 to 2023. The evaluation spans climatological, annual, monthly, and daily time scales with different precipitation intensities. IMERG precipitation can well capture the spatial and temporal precipitation climatology, with precipitation decreasing from southeast to Northwest China, and peaking in August. The correlation coefficient (CC) between IMERG precipitation and ground-observed precipitation is 0.69. However, IMERG precipitation systematically overestimates precipitation at climatological, annual, and monthly scales, especially in areas with relatively low precipitation climatology. At the daily time scale, IMERG precipitation data can represent precipitation events very well, especially in the southeastern part of Northwest China. IMERG precipitation overestimates light rainfall while underestimating precipitation of other intensities. While IMERG precipitation performs well in detecting light rain events, its accuracy diminishes for heavier rainfall, highlighting limitations for monitoring extreme precipitation. The Probability of Detection (POD) for light rainfall events is consistently above 0.9, while for Torrential Rainfall events, the POD is below 0.7. These findings provide insights into the effective application of IMERG data in precipitation monitoring and forecasting in Northwest China.

Graphical Abstract

1. Introduction

Precipitation plays a significant role in climate change, weather forecasting, and the interactions between the global hydrosphere, atmosphere, and biosphere [1]. Obtaining high-resolution spatiotemporal precipitation data and improving monitoring capabilities for precipitation distribution are essential for disaster monitoring, weather forecasting, and climate prediction [2]. Northwest China spans the northern Tibetan Plateau and the Loess Plateau, serving as a transitional zone between monsoon and non-monsoon regions, as well as between arid and semi-arid zones [3]. This region, characterized by fragile ecosystems and complex geological structures, is one of the most sensitive areas to global climate change [4]. On one hand, precipitation in Northwest China exhibits significant spatial variability in both quantity and intensity [5]. The southeastern region, influenced by monsoon and westerly circulation, experiences relatively abundant precipitation, while the northwestern part suffers from water vapor scarcity, resulting in much lower precipitation levels. Consequently, the climate elements in Northwest China are extremely unevenly distributed [6]. On the other hand, soil moisture resources are scarce, and groundwater levels are deep, making precipitation distribution and variation vital for the region’s ecological stability and climate evolution [7,8]. Therefore, enhancing precipitation monitoring, forecasting, and early warning capabilities in Northwest China is of critical importance.
Rain gauges and radar observations are traditional ground-based methods for direct precipitation monitoring. However, rain gauges provide highly accurate precipitation data, but their spatial density is often insufficient, and their distribution is uneven [9]. In arid and semi-arid regions with complex underlying surfaces, such as Northwest China, data from conventional meteorological stations struggle to achieve high spatiotemporal resolution. Ground-based radar measurements are limited by their spatial coverage and are susceptible to interference from environmental factors, resulting in significant uncertainties in complex terrain areas [10,11]. Therefore, conventional ground-based precipitation observations cannot accurately capture precipitation spatial distribution and intensity variations. Satellite remote sensing, on the other hand, offers wide spatial coverage and high spatiotemporal resolution, effectively compensating for the limitations of ground-based observations [12,13,14]. While satellite remote sensing has made significant strides in accurately measuring precipitation spatiotemporal distribution, systematic biases still exist due to the limitations of remote sensing instruments and retrieval algorithms [15,16]. Correcting these biases remains key to improving the quality of satellite precipitation products. Extensive research has been conducted both domestically and internationally to evaluate the applicability of satellite-based precipitation products in China, including their performance across different regions, complex terrains, and extreme weather events. These studies have demonstrated that satellite precipitation data have a certain capability for dynamic monitoring of regional precipitation processes, but their accuracy exhibits significant regional variability [17,18,19].
In recent years, satellite observation has significantly improved in both spatial and temporal resolution [20]. The Global Precipitation Measurement (GPM) mission, the successor to the Tropical Rainfall Measuring Mission (TRMM), offers high-resolution, high-timeliness, and full-coverage precipitation data [21,22]. The GPM-IMERG precipitation product demonstrates strong monitoring capabilities for extreme precipitation events across mainland China, effectively capturing their spatial distribution trends [23,24]. However, retrieval accuracy of IMERG precipitation is significantly influenced by precipitation types and terrain, with better detection for light precipitation than for torrential rain [25], and superior estimation of liquid precipitation over solid precipitation [26]. Additionally, IMERG tends to overestimate precipitation frequency but underestimate precipitation intensity, as small raindrops may evaporate before reaching the ground but can still be detected by remote sensors [27]. The applicability of IMERG precipitation has been extensively studied at global and intercontinental scales. IMERG’s ability to distinguish rain/snow phases globally has been evaluated, yet the research focused on global-scale precipitation phase classification without detailed regional assessments [28]. IMERG’s performance over the continental United States has been assessed using Stage-IV gauge-radar fusion data, but the analysis emphasized continental-scale evaluation while neglecting errors in arid regions, complex terrain, and extreme precipitation events [29]. IMERG has also been validated in tropical and coastal zones of Brazil using local rain gauge data, yet the study concentrated on humid climates [30]. Discrepancies among IMERG versions (Early/Late/Final Run) for extreme precipitation in the U.S. have been analyzed, but the large-scale approach lacked localized terrain-specific insight [31]. Additionally, the performance of IMERG precipitation products in monsoon-dominated humid regions has been verified [32]. IMERG’s accuracy over the North China Plain has been evaluated using national station data, yet the focus remained on humid monsoon zones rather than arid or plateau regions [33].
In China, studies have shown that satellite precipitation accuracy is generally lower in the mountainous and plateau regions of western China compared to the eastern plains. IMERG tends to overestimate precipitation in arid western regions while underestimating it in wetter southeastern regions [34]. In addition, different scholars have evaluated the applicability of satellite precipitation data in complex terrain regions with sparse meteorological stations. Xu and Zhang evaluated GPM-IMERG and TRMM products in the southern Tibetan Plateau, finding that IMERG better captured the regional precipitation distribution but still overestimated maximum precipitation intensity [35,36]. IMERG performance on the Mongolian Plateau has been examined using ERA5 reanalysis and national station data [37]. However, as this study primarily relies on reanalysis data, the accuracy of precipitation estimation may be affected by model parameterization [38]. IMERG has also been assessed over the Tibetan Plateau using national station data, but transitional zones of the plateau slope in Northwest China (e.g., Gansu, Ningxia, eastern Qinghai), which differ climatically and topographically from the plateau core, were excluded [39]. Moreover, the spatial distribution of national stations is sparse, especially in the arid region of Northwest China, making it difficult to capture localized precipitation variations [40].
In summary, satellite-based precipitation estimates offer high spatiotemporal continuity and can complement ground-based observations, particularly in complex terrains such as Northwest China. However, both global-scale and regional-scale studies indicate that the applicability of IMERG varies across different regions. Given the substantial annual variation in precipitation and the region’s complex geological conditions in Northwest China, the predictability of precipitation is relatively low, and the applicability of IMERG products in this region remains to be further verified [37]. Previous assessments relied on coarse-resolution national stations or model data, lacking high-density automatic weather station (AWS) validation. To address these gaps, our study focuses on Northwest China, leveraging high spatiotemporal resolution AWS data to systematically evaluate IMERG errors across diverse terrains and precipitation intensities.
This study systematically evaluates the accuracy and applicability of IMERG precipitation based on hourly ground-based observations from 6625 automatic meteorological stations in Northwest China during the warm season (April to September) from 2016 to 2023. Based on hourly ground-based observations from 6625 regional AWSs in Northwest China during the warm season (April to September) from 2016 to 2023., a multi-scale method was used to carry out high spatio-temporal resolution assessment, revealing IMERG inversion characteristics of precipitation of different magnitudes. The research results aim to provide a scientific basis for the selection of satellite precipitation products in Northwest China and provide valuable data for precipitation monitoring, forecasting, and early warning systems in Northwest China.

2. Data and Methods

2.1. Study Area

Northwest China spans longitudes 89°E to 112°E and latitudes 31°N to 43°N, including Shaanxi, Ningxia, Gansu, and Qinghai provinces (Figure 1a). It is an arid and semi-arid region with precipitation gradually decreasing from southeast to northwest [41,42]. The diverse and complex terrain here results in highly uneven distributions of meteorological elements. Additionally, ground-level meteorological stations in Northwest China are unevenly distributed, denser in the southeastern lower-altitude areas and sparser in the northwestern higher-altitude regions (Figure 1b).

2.2. Precipitation Data

The GPM mission, jointly launched by the National Aeronautics and Space Administration (NASA) and the Japan Aerospace Exploration Agency (JAXA) in February 2014, represents a next-generation global precipitation observation program. The mission comprises the GPM Core Observatory (GPMCO) and a constellation of partner satellites, operating at an altitude of 407 km with an orbital inclination of 65°. The GPMCO is equipped with the world’s first spaceborne Dual-frequency Precipitation Radar (DPR) and the GPM Microwave Imager (GMI). These advanced instruments integrate state-of-the-art microwave sensing technologies and data correction algorithms, providing broader coverage and higher spatiotemporal resolution. The GPM mission offers global precipitation data at intervals of less than 3 h using microwave-based observations and 30-min intervals using the IMERG, which integrates both microwave and infrared data [43].
IMERG, as a Level-3 satellite precipitation product, is one of the mainstream remote sensing precipitation products, covering 90°N to 90°S with a spatiotemporal resolution of 0.1°/30 min. It combines Passive Microwave (PMW) data, geostationary Infrared (IR) data, and monthly ground rain gauge observations, effectively leveraging the strengths of multi-source sensors. This integration significantly enhances capabilities in real-time typhoon monitoring, cumulative precipitation estimation, and the detection of light precipitation events (<0.5 mm/h), solid precipitation, and microphysical precipitation processes [43]. IMERG provides three types of satellite precipitation data products: Early Run, Late Run, and Final Run. The Final Run product incorporates bias correction using monthly-scale data from the Global Precipitation Climatology Centre (GPCC). Additionally, the IMERG_final precipitation undergoes bias correction based on ground rain gauge station data, offering the highest observational accuracy. The IMERG precipitation used in this study was obtained from NASA’s data portal (https://disc.gsfc.nasa.gov/ (accessed on 10 March 2024)).
In this study, hourly precipitation observational data from 6625 automatic weather stations across Northwest China during the warm season (April–September) from 2016 to 2023 were provided by the China Meteorological Administration (CMA). April to September is the warm season in Northwest China. The precipitation in this period accounts for the main part of the annual precipitation. Additionally, ground-based observation data in this period are more complete than that in the cold season, which only has precipitation data from national stations with lower resolution. To ensure consistency in spatiotemporal scales between IMERG precipitation and ground-based observational data, the 30-min precipitation data from IMERG at half-hour and hourly intervals were averaged. Additionally, the nearest-neighbor interpolation method was applied to map the IMERG gridded data onto the nearest ground-based observation. The method is computationally simple, highly efficient, and widely used in validation studies of satellite precipitation products [44]. However, it may introduce certain errors, especially in areas with high altitudes and large terrain fluctuation [45]. The method can lead to spatial discontinuities, thereby increasing data bias.

2.3. Evaluation Methods

This study uses three statistical indicators, namely the bias, Correlation Coefficient (CC), Root Mean Square Error (RMSE), and Relative Root Mean Square Error (rRMSE) to quantitatively evaluate the accuracy of IMERG precipitation in Northwest China. Bias is the difference between IMERG and observed precipitation. CC measures the linear correlation between satellite precipitation and ground-based observations, ranging from [−1, 1], with higher values indicating better performance. RMSE reflects the degree of dispersion between satellite precipitation and ground-based observations, with smaller values indicating higher estimation accuracy. rRMSE reflects the relative error between satellite precipitation and ground-based observed precipitation. The formulas are as follows:
C C = i = 1 n G i G ¯ S i S ¯ i = 1 n G i G ¯ 2 × i = 1 n S i S ¯ 2
R M S E = 1 n i = 1 n S i G i 2
r R M S E = R M S E G ¯ × 100 %
where n is the sample size, Si and Gi represent satellite and ground-based precipitation values, and S ¯ and G ¯ represent their respective means.
Additionally, classification indices such as Probability of Detection (POD) and False Alarm Ratio (FAR) are used to assess the ability of IMERG to detect daily precipitation events across four precipitation categories: light rain (0.1 mm ≤ R < 10 mm), moderate rain (10 mm ≤ R < 25 mm), heavy rain (25 mm ≤ R < 50 mm), and torrential rain (R ≥ 50 mm). POD reflects the probability of correctly detecting precipitation events, while FAR reflects the probability of false alarms. Their formulas are as follows:
P O D = H H + M
F A R = F H + F
where H is the number of correctly detected precipitation days (or stations), M is the number of missed events, and F is the number of false alarms.

3. Results

3.1. Climatological Precipitation

Figure 2 presents the spatial distribution of precipitation amount from April to September (2016–2023) based on IMERG precipitation and ground-based observations. The results indicate that precipitation in Northwest China exhibits a spatial pattern of lower values in the northwest and higher values in the southeast, with maximum precipitation amount exceeding 1400 mm in southern Shaanxi, southeastern Gansu, and highland slopes. IMERG precipitation accurately captures this spatial distribution pattern but generally overestimates precipitation, especially in regions with significant terrain fluctuations, including southeastern Qinghai, southeastern Gansu, and southern Shaanxi. Specifically, the maximum bias in eastern Qinghai reached 639 mm, and the bias in southeastern Gansu, southern Shaanxi, and the Guanzhong Plain all exceeded 1000 mm. On the slopes of plateaus and the windward sides of mountains, the uplifting effect of the terrain may lead to enhanced precipitation. Additionally, the maximum cumulative precipitation of the IMERG precipitation products is less than 1200 mm, indicating limited capability in retrieving localized heavy precipitation. This is mainly due to IMERG’s limited ability to capture localized precipitation enhancement effects and its lower retrieval accuracy in regions with complex terrain [46]. Additionally, the IMERG precipitation products are lower than the ground-based observations in the regions with large, accumulated precipitation. Figure 2d shows the probability density distribution of IMERG and ground-based precipitation. The colored area denotes the number of data points within a radius of 2.5 m. The probability density distribution further reveals a high correlation between IMERG precipitation and ground-based precipitation data, but most data points lie below the diagonal, indicating a systematic overestimation in IMERG precipitation estimates.
The time series of regionally averaged precipitation amounts reveals a consistent annual variation between the IMERG precipitation and ground-based observations (Figure 3a). This consistency indicates that IMERG precipitation can reasonably capture the interannual variation of precipitation across Northwest China. However, the accuracy of IMERG precipitation varies across different years, with a systematic overestimation observed in all cases. This overestimation aligns with the findings from the spatial distribution analysis, and it is particularly evident in 2020, where IMERG precipitation exceeded ground-based observations by 45 mm. The monthly mean precipitation peaks during July and August, both exceeding 90 mm (Figure 3b). This is followed by September, while April records the lowest monthly precipitation, with values below 40 mm. The IMERG satellite precipitation product effectively captures this monthly distribution pattern. However, the accuracy of IMERG varies across different months. The largest bias occurs in July and August, where IMERG precipitation values exceed ground-based observations by over 20 mm. In contrast, April exhibits the smallest discrepancy, with IMERG precipitation closest to ground-based observations.

3.2. Annual Precipitation

To quantitatively evaluate the annual-scale monitoring capability of IMERG precipitation in Northwest China, Figure 4 presents the probability density distribution between IMERG precipitation and ground-based observations for April to September across multiple years. It is evident that IMERG precipitation shows a clear positive correlation with observed precipitation, with the highest CC value reaching 0.9 in 2021. Additionally, the probability density distributions for most years show higher density zones below the diagonal line, further confirming IMERG precipitation tends to overestimate ground-based precipitation measurements. When ground-based observed precipitation approaches zero, the scatters reveal high-density zones near the X-axis, indicating that IMERG precipitation tends to overestimate light precipitation events. Notably, when the cumulative precipitation amount exceeds 1000 mm, IMERG precipitation tends to underestimate precipitation compared to ground-based observations. Furthermore, as precipitation intensity increases, the scatter plot shows greater dispersion, highlighting the reduced stability of IMERG precipitation’s performance in capturing extreme precipitation events.
To further assess the accuracy of the IMERG precipitation across Northwest China, data from 2909 stations with continuous observations over eight years were selected and CC and rRMSE between annual IMERG precipitation and ground-based observations from 2016 to 2023 were analyzed (Figure 5). IMERG precipitation exhibits a positive correlation with ground-based observations across most parts of Northwest China. The highest CC values (mostly exceeding 0.7) are concentrated in the regions with more climatological precipitation, such as Hexi Corridor, highland slopes, south–central Gansu, southern Gansu, and southern Shaanxi, indicating a high level of consistency between IMERG precipitation and ground-based observations in these areas. However, localized areas of low correlation are found in central–northern areas, including north–central Gansu, southwest Gansu, eastern Gansu, and northern Shaanxi. Particularly, the CC values over northern Ningxia are negative, indicating limited correlation applicability of IMERG precipitation. This is primarily because northern Ningxia has significant terrain fluctuations and relatively low annual precipitation. The precipitation base in northern Ningxia is relatively small, which makes the error in northern Ningxia more pronounced. As a result, IMERG fails to accurately capture the interannual variability of precipitation. The spatial distribution of rRMSE exhibits substantial regional variability. Areas with high rRMSE values (exceeding 200%) are primarily located in northwest Gansu and northern Ningxia, where the annual precipitation is small. The results show that the deviation between IMERG precipitation products and ground-based observations in these areas is large, and the estimation accuracy is low. It indicates a lower bias of IMERG precipitation in eastern and southern Qinghai, central Gansu, the plateau slope, and southern Shaanxi. Based on the comprehensive evaluation of CC and rRMSE, IMERG precipitation demonstrates superior performance in the central Gansu, the plateau slope, and southern Shaanxi.

3.3. Monthly Precipitation

Monthly comparison of regionally averaged precipitation between the IMERG precipitation and ground-based observations across Northwest China (Figure 6) reveals a consistent trend: both datasets show the lowest precipitation in April and the highest in July and August. However, the accuracy of IMERG precipitation varies across years and months. Among them, 2017 exhibited the closest agreement between IMERG precipitation and ground-based observations, followed by 2016. In contrast, 2020 showed significant overestimation, particularly from June to August, with peak overestimates exceeding 70 mm.
The performance of IMERG precipitation varies notably across months (Figure 7). Generally, July displays the largest bias, followed by August, while April demonstrates the closest agreement between IMERG precipitation and ground-based observations. Areas with the largest bias are primarily concentrated in southeastern Gansu, southern Shaanxi, and eastern Qinghai, which are consistent with the annual results. Notably, southwest Gansu exhibits unusually high bias in April, reaching up to 40 mm, while June to August shows significant bias in the central–eastern Hexi Corridor, and September reveals relatively smaller bias in southern Shaanxi. The Probability density distribution can present the comparison of two sets of precipitation data more clearly. Overall, IMERG precipitation from April to September systematically overestimates the observed precipitation, consistent with the results of the climatology and annual average. In April, the scatter points are evenly distributed around the diagonal, indicating strong agreement between IMERG and observations. While the scatter becomes increasingly dispersed with higher precipitation, suggesting reduced estimation stability for intense rainfall events. This feature is most obvious in September.
Figure 7 also illustrates the spatial distributions of CC and RMSE across different months. Overall, IMERG’s agreement with ground-based observations is weakest in April, with a higher correlation observed in August and September. Across months, southwest Gansu consistently shows lower CC values, indicating limited applicability of IMERG precipitation in this region. Conversely, eastern Qinghai, western Hexi Corridor, central Gansu, and northern Shaanxi show higher CC values, reflecting relatively stable performance. In August and September, CC values exceed 0.8 across most regions, except for southwest Gansu and central–eastern Hexi Corridor. In terms of RMSE, lower values are observed in April and May, indicating less variability and higher consistency in IMERG estimates. In contrast, RMSE peaks in July and August, particularly in southeastern Gansu and southern Shaanxi, highlighting greater quantitative bias. Combining the spatial patterns of CC and RMSE, IMERG precipitation performs best in the western Hexi Corridor, central Gansu, and eastern Qinghai, where CC values are high and RMSE values are low, reflecting higher estimation accuracy. However, it is worth noting that in July, both southern Shaanxi and southern Gansu exhibit low CC values, with some stations even showing negative correlations and high RMSE values, indicating poor estimation accuracy in these areas.

3.4. Daily Precipitation

IMERG precipitation has limitations in detecting light precipitation and exhibits significant uncertainties in estimating torrential rain. Annual and monthly precipitation characteristics are often insufficient for capturing extreme weather events such as storms and floods, making it essential to investigate daily-scale IMERG precipitation bias. To quantify the accuracy of IMERG precipitation estimates on a daily scale in Northwest China, Figure 8 presents the probability density distribution and bias between daily IMERG precipitation and ground-based observations. Overall, IMERG precipitation shows a relatively high correlation with ground-based observations at the daily scale. The scatter points are distributed along both sides of the diagonal line. For lower daily mean precipitation amounts, the probability density’s peak tends to fall below the diagonal, indicating a significant overestimation of light precipitation by IMERG. As the daily mean precipitation increases, the probability density peak aligns more closely with the diagonal. However, when the daily mean precipitation exceeds 3 mm, the scatter points tend to fall above the diagonal, suggesting that IMERG underestimates ground-based observations for moderate to heavy precipitation. The error distribution highlights the bias of satellite-derived precipitation from actual rainfall values. Figure 8b shows noticeable underestimations in most parts of Qinghai, southeastern Gansu, mountainous regions along the plateau margins, southern Ningxia, and mountainous areas of southern Shaanxi, including the Daba Mountains and Qinling Mountains, with maximum underestimations exceeding 3 mm. Conversely, significant overestimations occur in the Hexi Corridor, central and southern Gansu, northern Ningxia, and western Shaanxi, with the southern part of Gansu experiencing overestimations up to 3.8 mm. To evaluate IMERG’s ability to detect daily precipitation events, key statistical indices such as POD and FAR are analyzed. A higher POD and a lower FAR indicate better retrieval performance. Figure 8c,d demonstrate that IMERG precipitation exhibits a generally high POD (>0.7) across Northwest China. However, northern Qinghai, northern Hexi Corridor, and northern Ningxia show higher FAR values (>0.7). In contrast, eastern and southern Qinghai, central Hexi Corridor, southern Gansu, plateau margins, and southern Shaanxi demonstrate higher POD values (>0.9) and relatively low FAR values (0.3–0.45), indicating more reliable precipitation detection in these regions. The above results show that IMERG precipitation can capture precipitation events well on a daily scale.
The temporal variability of IMERG precipitation performance across different years is shown in Figure 9a. The POD remains consistently high (~0.8) over the years, with minimal interannual variation in FAR. The lowest FAR (0.52) was observed in 2021, while the highest (0.58) occurred in 2016. In terms of Mean Absolute Error (MAE), 2020 exhibited the highest daily mean precipitation error (2.7 mm), while 2022 showed the closest alignment with observations (MAE < 1.6 mm). On a monthly scale (Figure 9b), April recorded the lowest POD (0.72) and the highest FAR (0.7), indicating poorer detection performance during this month. In contrast, June achieved the highest POD (0.87), reflecting the most accurate precipitation detection, while August recorded the lowest FAR (0.47), indicating fewer false precipitation detections.

3.5. Precipitation at Different Intensities

The evaluation results at annual, monthly, and daily scales consistently indicate that IMERG precipitation product bias is strongly related to precipitation intensity and that product accuracy varies significantly across regions. To further analyze the performance of IMERG under different precipitation thresholds and quantify the relationship between accuracy and precipitation intensity, the estimation capabilities of IMERG precipitation were assessed for four precipitation categories: light rain, moderate rain, heavy rain, and torrential rain. From the spatial distribution of bias across different precipitation intensities (Figure 10), it is evident that IMERG precipitation significantly overestimates light rainfall while underestimating moderate, heavy, and torrential rainfall. Additionally, regions such as northern Ningxia, the southwestern part of Gansu, western Shaanxi, and mountainous areas in southern Shaanxi experience the most significant overestimation for light rain and underestimation for other precipitation intensities.
Using POD and FAR, the performance of IMERG precipitation across the four precipitation intensity categories was evaluated (Figure 11). For light rain, IMERG precipitation achieves the highest POD scores (>0.7), with values exceeding 0.9 in the Hexi Corridor and eastern, central, and southern parts of Gansu. As precipitation intensity increases, POD decreases, particularly for torrential rain. In southeastern Gansu, south–central Gansu, southern Ningxia, northern Shaanxi, and western Shaanxi, POD exceeds 0.6 for torrential rain, while in other regions, POD values are close to zero. Spatial analysis of FAR across precipitation intensities reveals that for light rain, FAR is relatively low in southern Qinghai and southwest Gansu (<0.5). However, in northern Qinghai, the Hexi Corridor, northern Ningxia, and western Shaanxi, FAR approaches or exceeds 0.75. Moderate rain exhibits higher FAR values than light rain, with elevated FAR levels across most of Northwest China, except for eastern Qinghai. For heavy and torrential rain, FAR improves slightly compared to light and moderate rain, especially in southeastern Gansu and mountainous regions of southern Shaanxi and northern Shaanxi. However, regions such as eastern Qinghai, the Hexi Corridor, southwest Gansu, plateau margins, northern Ningxia, the Qinling Mountains, and western Shaanxi show FAR values close to 1, indicating frequent false signals. In summary, IMERG precipitation performs poorly in estimating high-intensity precipitation, particularly in complex topographic regions such as eastern Qinghai, plateau margins, and the Qinling Mountains. For light and moderate rain, bias primarily stems from false signals, further verifying that inaccuracies for light rain largely arise from overestimations of both rainfall intensity and frequency.
A comprehensive evaluation of the temporal variability of IMERG precipitation product error characteristics (Figure 12) reveals that the MAE for light and moderate rain shows minimal variation over time, while for heavy and torrential rain, MAE exhibits significant inter-monthly and inter-annual fluctuations. Specifically, the MAE for heavy and torrential rain is highest in April and lowest in July and August, with 2017 showing the highest MAE for heavy and torrential rain, while 2021 had the lowest. A more detailed analysis of the POD and FAR results shows that POD for light rain is highest and FAR is lowest in April. However, for other precipitation intensities, POD is significantly lower and FAR is higher in April compared to other months, indicating that bias in April is primarily driven by moderate or higher-intensity rainfall. In July and August, POD for light rain is markedly higher and FAR is lower compared to other months, whereas the performance for other precipitation categories shows the opposite trend. This suggests that IMERG precipitation performs better for moderate to torrential rain in July and August, but has lower detection rates for light rain during these months. Additionally, the analysis reveals that POD decreases significantly with increasing rainfall intensity, with POD values exceeding 0.9 for light rain and dropping below 0.7 for torrential rain. FAR is lowest for light rain but increases significantly for moderate and higher intensity rainfall, reaching near 1 for torrential rain. This implies that IMERG precipitation has the highest detection accuracy for light rain, while the detection of torrential rain events exhibits considerable uncertainty.

4. Discussion

This study evaluates the bias characteristics of IMERG precipitation in Northwest China across multiple scales, spanning climatological, annual, monthly, and daily time scales with different precipitation intensities. The results show that IMERG precipitation can well capture the spatial and temporal precipitation climatology but exhibits significant bias differences in different regions and precipitation intensities. IMERG precipitation demonstrates certain reliability at the regional scale, well capturing the spatial distribution characteristics of cumulative precipitation in Northwest China, which is less in the northwest and more in the southeast. It also reasonably represents the temporal variation trend of precipitation in Northwest China. However, IMERG precipitation exhibits notably increased estimation errors in areas with higher precipitation, such as eastern Qinghai, southeastern Gansu, southern Shaanxi, and highland slopes, which is closely related to the complex topography of these regions. An evaluation of IMERG and ERA5 precipitation products over the Mongolian Plateau highlights that IMERG precipitation exhibits reduced detection capability in the forested regions of the Mongolian Plateau with complex terrain, primarily due to insufficient sensitivity of its algorithm to topographic effects [37]. Further analysis reveals that topographic undulation (such as the windward slope/leeward slope effect) is an important factor influencing the accuracy of IMERG precipitation [46]. Additionally, research has demonstrated that IMERG precipitation has limitations in precipitation retrieval in complex terrain areas (such as mountains and plateaus), as it may be difficult to accurately distinguish between clouds and precipitation in some of these regions [47].
Furthermore, the evaluation reveals that IMERG precipitation tends to overestimate compared to ground-based observations, with this bias being more pronounced for light precipitation, and underestimates heavy precipitation with considerable instability in detecting high-intensity rainfall events. Research has shown that this pattern is consistent with results from studies conducted over the Mongolian Plateau, where it was pointed out that IMERG precipitation generally overestimates light precipitation and underestimates heavy precipitation [37]. Further analysis has demonstrated that IMERG precipitation’s overestimation of extreme precipitation over the Tibetan Plateau is primarily due to frequency rather than intensity. From the perspective of time scale, the bias of IMERG precipitation is greatest in July and August, while it is smallest in April [39]. In addition, during the extreme precipitation event in the North China region in the summer of 2023, it was found that IMERG precipitation exhibited a significant decline in detection capability when the rainfall rate was ≥30 mm/h (the POD was approaching zero, and the FAR was nearly 100%), indicating that the satellite algorithm was insufficient in responding to the suddenness and small-scale characteristics of short-term heavy convective precipitation [48].
From the daily scale perspective, the areas where IMERG precipitation underestimates precipitation are mainly concentrated in regions with complex terrain and high altitudes, such as eastern and southern Qinghai, southeastern Gansu, and along the mountain slopes of the plateau, southern Ningxia, and along the mountainous areas of the Dabashan and Qinling Mountains in southern Shaanxi. Research has shown that there is persistent significant bias in IMERG precipitation estimates over northeastern India, a region heavily influenced by complex terrain [32]. In such complex terrain areas, the precipitation scale range is relatively small due to the local topographic uplift effect, making it difficult for satellite algorithms to accurately capture the precipitation. Moreover, the overestimation by IMERG precipitation predominantly occurs in piedmont plains, river terraces, and corridor landscapes—such as the Hexi Corridor (Gansu), central–southern Gansu, northern Ningxia, and the Guanzhong Plain (Shaanxi), which are plateau margins and mountain-plain transitional belts.
The results of the evaluation of IMERG precipitation at different magnitudes show that the IMERG precipitation products significantly overestimated the precipitation at light rain but underestimated the precipitation at other magnitudes, with a pronounced overestimation of the occurrence probability of weak precipitation. They also seriously overestimated the possibility of weak precipitation occurrence, that is, the contribution of the bias to the light rain magnitude mainly originated from the significant overestimation of the precipitation intensity of weak precipitation. This bias contribution for light rain primarily stems from a substantial overestimation of precipitation intensity in weak-intensity rainfall regimes. Additionally, IMERG precipitation demonstrates relatively stable detection performance for light precipitation, as precipitation intensity increases, POD shows a marked decline, accompanied by a sharp rise in FAR. Notably, the FAR approaches unity for torrential rain, further highlighting elevated uncertainties in IMERG precipitation’s detection of heavy rain and torrential rain events.
In conclusion, the estimation accuracy of IMERG precipitation in complex topographic regions still requires further optimization, as the distribution characteristics of heavy precipitation under such terrain conditions remain a significant challenge for satellite-based precipitation estimation. This study systematically validates and evaluates the bias characteristics of IMERG precipitation in Northwest China, elucidating their performance variations across different geographic regions and precipitation intensity thresholds. However, the bias mechanisms of IMERG precipitation in complex terrain zones (e.g., the Tibetan Plateau escarpment and Qinling Mountains) and their relationships with precipitation types (e.g., convective precipitation and orographic cloud precipitation) warrant further investigation. Future research could employ multi-source data fusion techniques (e.g., integrating radar observations, ground-based station data, and reanalysis datasets) to refine retrieval algorithms, thereby mitigating the observational limitations of IMERG precipitation in high-altitude and steep-slope terrains.

5. Conclusions

This study analyzes the error characteristics of the IMERG precipitation in Northwest China, using ground-based automatic station observations from April to September 2016–2023 as the reference. The analysis was conducted at multiple time scales, including climate state, annual, monthly, daily, and different precipitation intensities. The following key conclusions were drawn:
(1)
Precipitation in the Northwest region shows a spatial distribution pattern characterized by low precipitation in the northwest and high precipitation in the southeast. IMERG precipitation effectively captures this spatial distribution trend across different scales and can also generally reflect the climatological characteristic of precipitation in different years and months.
(2)
The evaluation results at the climatological, annual, and monthly scales show a high correlation between IMERG precipitation and observed ground-based precipitation. However, there is a consistent overestimation of precipitation, with the largest bias occurring in July and August, and April being closest to the observed data. The areas with the largest bias are concentrated in southeastern regions. Furthermore, IMERG precipitation tends to overestimate light precipitation more significantly and underestimate heavy precipitation, with considerable instability in detecting high-intensity rainfall events. Given the significant errors in IMERG during July and August, empirical bias correction is necessary for operational applications. For example, regional correction models based on ground observations can be developed to enhance their monitoring capability for summer heavy precipitation. In areas with complex terrain, such as southeastern Gansu, southern Shaanxi, and eastern Qinghai, multi-source data fusion combining radar, automatic weather station data, and numerical model forecasts should be implemented to improve the reliability of heavy precipitation estimation.
(3)
At the daily scale, IMERG precipitation significantly underestimates precipitation in most of Qinghai, southeastern Gansu, the plateau edge, southern Ningxia, and southern Shaanxi, while overestimating precipitation in western Gansu, central and southern Gansu, northern Ningxia, and the western Shaanxi. IMERG precipitation performs well in detecting daily precipitation events, showing high POD (>0.9) and low FAR (0.3–0.45) in eastern and southern Qinghai, central Gansu, southern Gansu, plateau edge regions, and southern Shaanxi. In April, the POD is the lowest (0.72), and FAR is the highest (0.7), while in June, the POD is the highest (0.87), and in August, the FAR is the lowest (0.47).
(4)
The evaluation of the performance of IMERG precipitation at different precipitation intensities shows a clear overestimation of light precipitation and an underestimation of higher-intensity rainfalls. IMERG precipitation detected more light rain events and significantly overestimated the occurrence of weak precipitation while underestimating the number of heavy and torrential rain days. The MAE for light and moderate rain shows minimal variation over time, while for heavy and torrential rain, the MAE is highest in April and lowest in July and August. The POD and FAR for light rain are both high, and their biases are primarily due to a significant overestimation of weak precipitation intensity and frequency. As precipitation intensity increases, the POD decreases sharply, and the FAR increases significantly. IMERG precipitation shows the highest detection accuracy for light rain but exhibits considerable uncertainty in detecting torrential rain events. For different precipitation events, it is recommended to introduce machine learning methods to optimize the false alarm rate for light rain. For heavy rain and torrential rain, quantitative precipitation correction should be performed based on meteorological station data to enhance the accuracy of precipitation estimation.

Author Contributions

Writing—original draft, D.W., W.D., S.C., H.X., L.X. and Z.J.; Methodology, W.T. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by the National Natural Science Foundation of China (42205083), the Natural Science Foundation of Gansu (23JRRA1329, 22JR5RA749), the CMA Innovation and Development Project (CXFZ2025Q012), and the Gansu Meteorological Research Project (2425rczx-B-QNBJRC-07, GSQXCXTD-2024-01, ZcQn2024-B-29).

Data Availability Statement

The original data can be made available upon request by contacting the corresponding author.

Acknowledgments

We thank NASA and the China Meteorological Administration for providing the original data. The authors are deeply appreciative of the meticulous attention of the editors and reviewers.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Allen, M.R.; Ingram, W.J. Constraints on Future Changes in Climate and the Hydrologic Cycle. Nature 2002, 419, 224–232. [Google Scholar] [CrossRef] [PubMed]
  2. Chen, S.; Hong, Y.; Cao, Q.; Gourley, J.J.; Kirstetter, P.-E.; Yong, B.; Tian, Y.; Zhang, Z.; Shen, Y.; Hu, J.; et al. Similarity and Difference of the Two Successive V6 and V7 TRMM Multisatellite Precipitation Analysis Performance over China. J. Geophys. Res. Atmos. 2013, 118, 13060–13074. [Google Scholar] [CrossRef]
  3. Sui, X.; Xu, Q.; Tao, H.; Zhu, B.; Li, G.; Zhang, Z. Vegetation Dynamics and Recovery Potential in Arid and Semi-Arid Northwest China. Plants 2024, 13, 3412. [Google Scholar] [CrossRef]
  4. Ren, X.; Hou, P.; Ma, Y.; Ma, R.; Wang, J.; Xie, L. Analysis of Vegetation Changes and Driving Factors on the Qinghai-Tibet Plateau from 2000 to 2022. Forests 2024, 15, 2188. [Google Scholar] [CrossRef]
  5. Wang, Y.; Zhang, W.; Zhang, P.; Tang, L.; Huang, Y.; Feng, N. Spatial and Temporal Variations of Precipitation in Northwest China during 1973-2019. Theor. Appl. Clim. 2024, 155, 4347–4356. [Google Scholar] [CrossRef]
  6. Ge, Q.; Zheng, J.; Hao, Z.; Liu, H. General Characteristics of Climate Changes during the Past 2000 Years in China. Sci. China Earth Sci. 2013, 56, 321–329. [Google Scholar] [CrossRef]
  7. Mehta, P. The Impact of Climate Change on the Environment, Water Resources, and Agriculture: A Comprehensive Review. In Climate, Environment and Agricultural Development: A Sustainable Approach Towards Society; Springer: Berlin/Heidelberg, Germany, 2024; pp. 189–201. [Google Scholar]
  8. Wang, Z.; Li, Z.; Zhan, H.; Yang, S. Effect of Long-Term Saline Mulched Drip Irrigation on Soil-Groundwater Environment in Arid Northwest China. Sci. Total Environ. 2022, 820, 153222. [Google Scholar] [CrossRef]
  9. Lee, J.; Kim, S.; Jun, H. A Study of the Influence of the Spatial Distribution of Rain Gauge Networks on Areal Average Rainfall Calculation. Water 2018, 10, 1635. [Google Scholar] [CrossRef]
  10. Kumari, M.; Singh, C.K.; Bakimchandra, O.; Basistha, A. Geographically Weighted Regression Based Quantification of Rainfall–Topography Relationship and Rainfall Gradient in Central Himalayas. Int. J. Climatol. 2017, 37, 1299–1309. [Google Scholar] [CrossRef]
  11. Ranhao, S.; Baiping, Z.; Jing, T. A Multivariate Regression Model for Predicting Precipitation in the Daqing Mountains. MRED 2008, 28, 318–325. [Google Scholar] [CrossRef]
  12. Kidd, C.; Takayabu, Y.N.; Skofronick-Jackson, G.M.; Huffman, G.J.; Braun, S.A.; Kubota, T.; Turk, F.J. The Global Precipitation Measurement (GPM) Mission. In Satellite Precipitation Measurement: Volume 1; Levizzani, V., Kidd, C., Kirschbaum, D.B., Kummerow, C.D., Nakamura, K., Turk, F.J., Eds.; Springer International Publishing: Cham, Switzerland, 2020; pp. 3–23. ISBN 978-3-030-24568-9. [Google Scholar]
  13. Ouatiki, H.; Boudhar, A.; Chehbouni, A. Accuracy Assessment and Bias Correction of Remote Sensing–Based Rainfall Products over Semiarid Watersheds. Theor. Appl. Clim. 2023, 154, 763–780. [Google Scholar] [CrossRef]
  14. Pan, X.; Wu, H.; Chen, S.; Nanding, N.; Huang, Z.; Chen, W.; Li, C.; Li, X. Evaluation and Applicability Analysis of GPM Satellite Precipitation over Mainland China. Remote Sens. 2023, 15, 2866. [Google Scholar] [CrossRef]
  15. Dubovik, O.; Schuster, G.L.; Xu, F.; Hu, Y.; Bösch, H.; Landgraf, J.; Li, Z. Grand Challenges in Satellite Remote Sensing. Front. Remote Sens. 2021, 2, 619818. [Google Scholar] [CrossRef]
  16. Kolster, M.E.; Wigh, M.D.; Lima Simões da Silva, E.; Bjerg Vilhelmsen, T.; Døssing, A. High-Speed Magnetic Surveying for Unexploded Ordnance Using UAV Systems. Remote Sens. 2022, 14, 1134. [Google Scholar] [CrossRef]
  17. Lasmono, F.; Risyanto; Nauval, F.; Saufina, E.; Trismidianto; Harjana, T. Satellite Rainfall Estimation from Himawari-8 Multi Channels Observation Based on AWS Data Trained Machine Learning Methods. In Proceedings of the International Conference on Radioscience, Equatorial Atmospheric Science and Environment and Humanosphere Science, 2021; Yulihastin, E., Abadi, P., Sitompul, P., Harjupa, W., Eds.; Springer Nature: Singapore, 2022; pp. 495–506. [Google Scholar]
  18. Lu, H.; Ding, L.; Ma, Z.; Li, H.; Lu, T.; Su, M.; Xu, J. Spatiotemporal Assessments on the Satellite-Based Precipitation Products From Fengyun and GPM Over the Yunnan-Kweichow Plateau, China. Earth Space Sci. 2020, 7, e2019EA000857. [Google Scholar] [CrossRef]
  19. Wu, H.; Yong, B.; Shen, Z. Spatial Reconstruction of Quantitative Precipitation Estimates Derived from Fengyun-2G Geostationary Satellite in Northeast China. Remote Sens. 2023, 15, 5251. [Google Scholar] [CrossRef]
  20. Prigent, C. Precipitation Retrieval from Space: An Overview. Comptes Rendus Geosci. 2010, 342, 380–389. [Google Scholar] [CrossRef]
  21. Liu, Z. Comparison of Integrated Multisatellite Retrievals for GPM (IMERG) and TRMM Multisatellite Precipitation Analysis (TMPA) Monthly Precipitation Products: Initial Results. J. Hydrometeorol. 2016, 17, 777–790. [Google Scholar] [CrossRef]
  22. Smith, E.A.; Asrar, G.; Furuhama, Y.; Ginati, A.; Mugnai, A.; Nakamura, K.; Adler, R.F.; Chou, M.-D.; Desbois, M.; Durning, J.F.; et al. International Global Precipitation Measurement (GPM) Program and Mission: An Overview. In Measuring Precipitation From Space: EURAINSAT and the Future; Levizzani, V., Bauer, P., Turk, F.J., Eds.; Springer: Dordrecht, The Netherlands, 2007; pp. 611–653. ISBN 978-1-4020-5835-6. [Google Scholar]
  23. Tang, G.; Clark, M.P.; Papalexiou, S.M.; Ma, Z.; Hong, Y. Have Satellite Precipitation Products Improved over Last Two Decades? A Comprehensive Comparison of GPM IMERG with Nine Satellite and Reanalysis Datasets. Remote Sens. Environ. 2020, 240, 111697. [Google Scholar] [CrossRef]
  24. Jiang, Q.; Li, W.; Fan, Z.; He, X.; Sun, W.; Chen, S.; Wen, J.; Gao, J.; Wang, J. Evaluation of the ERA5 reanalysis precipitation dataset over Chinese Mainland. J. Hydrol. 2021, 595, 125660. [Google Scholar] [CrossRef]
  25. Mahmoud, M.T.; Mohammed, S.A.; Hamouda, M.A.; Dal Maso, M.; Mohamed, M.M. Performance of the IMERG Precipitation Products over High-Latitudes Region of Finland. Remote Sens. 2021, 13, 2073. [Google Scholar] [CrossRef]
  26. Meng, C.; Mo, X.; Liu, S.; Hu, S. Extensive Evaluation of IMERG Precipitation for Both Liquid and Solid in Yellow River Source Region. Atmos. Res. 2021, 256, 105570. [Google Scholar] [CrossRef]
  27. Lv, P.; Wu, G. The Performance of GPM IMERG Product Validated on Hourly Observations over Land Areas of Northern Hemisphere. Remote Sens. 2024, 16, 4334. [Google Scholar] [CrossRef]
  28. Xiong, W.; Tang, G.; Wang, T.; Ma, Z.; Wan, W. Evaluation of IMERG and ERA5 Precipitation-Phase Partitioning on the Global Scale. Water 2022, 14, 1122. [Google Scholar] [CrossRef]
  29. Beck, H.E.; Pan, M.; Roy, T.; Weedon, G.P.; Pappenberger, F.; van Dijk, A.I.J.M.; Huffman, G.J.; Adler, R.F.; Wood, E.F. Daily Evaluation of 26 Precipitation Datasets Using Stage-IV Gauge-Radar Data for the CONUS. Hydrol. Earth Syst. Sci. 2019, 23, 207–224. [Google Scholar] [CrossRef]
  30. Rozante, J.R.; Rozante, G. IMERG V07B and V06B: A Comparative Study of Precipitation Estimates Across South America with a Detailed Evaluation of Brazilian Rainfall Patterns. Remote Sens. 2024, 16, 4722. [Google Scholar] [CrossRef]
  31. Li, Z.; Tang, G.; Kirstetter, P.; Gao, S.; Li, J.-L.F.; Wen, Y.; Hong, Y. Evaluation of GPM IMERG and Its Constellations in Extreme Events over the Conterminous United States. J. Hydrol. 2022, 606, 127357. [Google Scholar] [CrossRef]
  32. Prakash, S.; Mitra, A.K.; AghaKouchak, A.; Liu, Z.; Norouzi, H.; Pai, D.S. A Preliminary Assessment of GPM-Based Multi-Satellite Precipitation Estimates over a Monsoon Dominated Region. J. Hydrol. 2018, 556, 865–876. [Google Scholar] [CrossRef]
  33. Zhang, D.; Yang, M.; Ma, M.; Tang, G.; Wang, T.; Zhao, X.; Ma, S.; Wu, J.; Wang, W. Can GPM IMERG Capture Extreme Precipitation in North China Plain? Remote Sens. 2022, 14, 928. [Google Scholar] [CrossRef]
  34. Pradhan, R.K.; Markonis, Y.; Vargas Godoy, M.R.; Villalba-Pradas, A.; Andreadis, K.M.; Nikolopoulos, E.I.; Papalexiou, S.M.; Rahim, A.; Tapiador, F.J.; Hanel, M. Review of GPM IMERG Performance: A Global Perspective. Remote Sens. Environ. 2022, 268, 112754. [Google Scholar] [CrossRef]
  35. Xu, R.; Tian, F.; Yang, L.; Hu, H.; Lu, H.; Hou, A. Ground Validation of GPM IMERG and TRMM 3B42V7 Rainfall Products over Southern Tibetan Plateau Based on a High-Density Rain Gauge Network. J. Geophys. Res. Atmos. 2017, 122, 910–924. [Google Scholar] [CrossRef]
  36. Zhang, S.; Wang, D.; Qin, Z.; Zheng, Y.; Guo, J. Assessment of the GPM and TRMM Precipitation Products Using the Rain Gauge Network over the Tibetan Plateau. J. Meteorol. Res. 2018, 32, 324–336. [Google Scholar] [CrossRef]
  37. Xin, Y.; Yang, Y.; Chen, X.; Yue, X.; Liu, Y.; Yin, C. Evaluation of IMERG and ERA5 Precipitation Products over the Mongolian Plateau. Sci. Rep. 2022, 12, 21776. [Google Scholar] [CrossRef]
  38. Reichle, R.H.; Liu, Q.; Koster, R.D.; Draper, C.S.; Mahanama, S.P.P.; Partyka, G.S. Land Surface Precipitation in MERRA-2. J. Clim. 2017, 30, 1643–1664. [Google Scholar] [CrossRef]
  39. Zhang, Y.; Huang, Y.; Xu, H.; Wang, S.; Long, T.; Zhao, Q. Evaluation of Precipitation Frequency and Intensity as Estimated by the GPM IMERG Precipitation Product at Daily and Hourly Scales over the Tibetan Plateau. Atmosphere 2023, 14, 1653. [Google Scholar] [CrossRef]
  40. Li, L.; Chen, X.; Ma, Y.; Zhao, W.; Zuo, H.; Liu, Y.; Cao, D.; Xu, X. Implications for Validation of IMERG Satellite Precipitation in a Complex Mountainous Region. Remote Sens. 2023, 15, 4380. [Google Scholar] [CrossRef]
  41. Huang, J.; Yu, H.; Guan, X.; Wang, G.; Guo, R. Accelerated Dryland Expansion under Climate Change. Nat. Clim. Chang. 2016, 6, 166–171. [Google Scholar] [CrossRef]
  42. Huang, J.; Guan, X.; Ji, F. Enhanced Cold-Season Warming in Semi-Arid Regions. Atmos. Chem. Phys. 2012, 12, 5391–5398. [Google Scholar] [CrossRef]
  43. Huffman, G.; Bolvin, D.; Nelkin, E.; Kidd, C.; Park, C.; Park, C. Improving User Access to the Integrated Multi-Satellite Retrievals for GPM (IMERG) Products. In Proceedings of the EGU General Assembly, Vienna, Austria, 17–22 April 2016. [Google Scholar]
  44. Derin, Y.; Yilmaz, K.K. Evaluation of Multiple Satellite-Based Precipitation Products over Complex Topography. J. Hydrometeorol. 2014, 15, 1498–1516. [Google Scholar] [CrossRef]
  45. Bharti, V.; Singh, C. Evaluation of Error in TRMM 3B42V7 Precipitation Estimates over the Himalayan Region. J. Geophys. Res. Atmos. 2015, 120, 12458–12473. [Google Scholar] [CrossRef]
  46. Acosta-Reyes, M.-Á.; Hernández-Anguiano, J.-H.; Carrión-Viramontes, F.-J.; Li, Y.; Quintana-Rodríguez, J.-A.; Rojo-Hernández, J.-D.; Navarro-Céspedes, J.-M.; Piña-González, V.; Zha, X.; Liao, X. Analysis of IMERG Precipitation Product in Mexican States to Evaluate the Spatial and Temporal Variation in Precipitation. Sustain. Water Resour. Manag. 2024, 10, 190. [Google Scholar] [CrossRef]
  47. Skofronick-Jackson, G.; Kirschbaum, D.; Petersen, W.; Huffman, G.; Kidd, C.; Stocker, E.; Kakar, R. The Global Precipitation Measurement (GPM) Mission’s Scientific Achievements and Societal Contributions: Reviewing Four Years of Advanced Rain and Snow Observations. QJR Meteorol. Soc. 2018, 144, 27–48. [Google Scholar] [CrossRef]
  48. Li, Z.; Liang, H.; Chen, S.; Li, X.; Li, Y.; Wei, C. Performance Assessment of Satellite-Based Precipitation Products in the 2023 Summer Extreme Precipitation Events over North China. Atmosphere 2024, 15, 1315. [Google Scholar] [CrossRef]
Figure 1. Spatial distribution of the study area and ground-level meteorological stations. (a) Study area; (b) Ground-level meteorological station.
Figure 1. Spatial distribution of the study area and ground-level meteorological stations. (a) Study area; (b) Ground-level meteorological station.
Remotesensing 17 01364 g001
Figure 2. Average precipitation amounts in the warm season (April to September) in Northwest China from 2016 to 2023 based on (a) observation, (b) IMERG, (c) bias, and (d) probability density distribution between the two precipitation datasets (the colored area denotes the number of data points within a radius of 2.5 mm).
Figure 2. Average precipitation amounts in the warm season (April to September) in Northwest China from 2016 to 2023 based on (a) observation, (b) IMERG, (c) bias, and (d) probability density distribution between the two precipitation datasets (the colored area denotes the number of data points within a radius of 2.5 mm).
Remotesensing 17 01364 g002
Figure 3. Mean precipitation amount over Northwest China for (a) annual and (b) monthly scale.
Figure 3. Mean precipitation amount over Northwest China for (a) annual and (b) monthly scale.
Remotesensing 17 01364 g003
Figure 4. Probability density distribution between IMERG precipitation and ground-based observations in different years (the colored area denotes the number of data points within a radius of 2.5 mm). (a) 2016; (b) 2017; (c) 2018; (d) 2019; (e) 2020; (f) 2021; (g) 2022; (h) 2023.
Figure 4. Probability density distribution between IMERG precipitation and ground-based observations in different years (the colored area denotes the number of data points within a radius of 2.5 mm). (a) 2016; (b) 2017; (c) 2018; (d) 2019; (e) 2020; (f) 2021; (g) 2022; (h) 2023.
Remotesensing 17 01364 g004
Figure 5. Spatial distribution of error statistical indices between IMERG precipitation and ground-based observations: (a) CC and (b) rRMSE.
Figure 5. Spatial distribution of error statistical indices between IMERG precipitation and ground-based observations: (a) CC and (b) rRMSE.
Remotesensing 17 01364 g005
Figure 6. Time series of regional average precipitation from April to September during 2016–2023. The dotted line represents IMERG precipitation, the solid line represents the ground-based observations, and the bars represent the difference between IMERG and observations.
Figure 6. Time series of regional average precipitation from April to September during 2016–2023. The dotted line represents IMERG precipitation, the solid line represents the ground-based observations, and the bars represent the difference between IMERG and observations.
Remotesensing 17 01364 g006
Figure 7. Bias, probability density distribution, CC, and RMSE between monthly IMERG precipitation and ground-based observations from April to September. (a) Bias in April; (b) probability density distribution in April; (c) CC in April; (d) RMSE in April; (e) Bias in May; (f) probability density distribution in May; (g) CC in May; (h) RMSE in May; (i) Bias in June; (j) probability density distribution in June; (k) CC in June; (l) RMSE in June; (m) Bias in July; (n) probability density distribution in July; (o) CC in July; (p) RMSE in July; (q) Bias in August; (r) probability density distribution in August; (s) CC in August; (t) RMSE in August; (u) Bias in September; (v) probability density distribution in September; (w) CC in September; (x) RMSE in September.
Figure 7. Bias, probability density distribution, CC, and RMSE between monthly IMERG precipitation and ground-based observations from April to September. (a) Bias in April; (b) probability density distribution in April; (c) CC in April; (d) RMSE in April; (e) Bias in May; (f) probability density distribution in May; (g) CC in May; (h) RMSE in May; (i) Bias in June; (j) probability density distribution in June; (k) CC in June; (l) RMSE in June; (m) Bias in July; (n) probability density distribution in July; (o) CC in July; (p) RMSE in July; (q) Bias in August; (r) probability density distribution in August; (s) CC in August; (t) RMSE in August; (u) Bias in September; (v) probability density distribution in September; (w) CC in September; (x) RMSE in September.
Remotesensing 17 01364 g007
Figure 8. Spatial distribution of IMERG daily average precipitation error statistical indices: (a) Probability density distribution, (b) Bias between daily IMERG precipitation and ground-based observations, (c) PDO, and (d) FAR.
Figure 8. Spatial distribution of IMERG daily average precipitation error statistical indices: (a) Probability density distribution, (b) Bias between daily IMERG precipitation and ground-based observations, (c) PDO, and (d) FAR.
Remotesensing 17 01364 g008
Figure 9. MAE, POD, and FAR of daily average precipitation of (a) annual and (b) monthly series.
Figure 9. MAE, POD, and FAR of daily average precipitation of (a) annual and (b) monthly series.
Remotesensing 17 01364 g009
Figure 10. Average precipitation amounts in Northwest China at different magnitudes: (ac) light rain, (df) moderate rain, (gi) heavy rain, and (jl) torrential rain.
Figure 10. Average precipitation amounts in Northwest China at different magnitudes: (ac) light rain, (df) moderate rain, (gi) heavy rain, and (jl) torrential rain.
Remotesensing 17 01364 g010
Figure 11. Spatial distribution of PDO and FAR for precipitation at different magnitudes: (a,b) light rain, (c,d) moderate rain, (e,f) heavy rain, and (g,h) torrential rain.
Figure 11. Spatial distribution of PDO and FAR for precipitation at different magnitudes: (a,b) light rain, (c,d) moderate rain, (e,f) heavy rain, and (g,h) torrential rain.
Remotesensing 17 01364 g011
Figure 12. Time series of MAE (a,b), POD (c,d), and FAR (e,f) for precipitation at different magnitudes.
Figure 12. Time series of MAE (a,b), POD (c,d), and FAR (e,f) for precipitation at different magnitudes.
Remotesensing 17 01364 g012
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Wei, D.; Di, W.; Tian, W.; Cheng, S.; Xie, H.; Xie, L.; Jing, Z. Assessment of Integrated Multi-Satellite Retrievals for Global Precipitation Measurement (IMERG) Precipitation Products in Northwest China. Remote Sens. 2025, 17, 1364. https://doi.org/10.3390/rs17081364

AMA Style

Wei D, Di W, Tian W, Cheng S, Xie H, Xie L, Jing Z. Assessment of Integrated Multi-Satellite Retrievals for Global Precipitation Measurement (IMERG) Precipitation Products in Northwest China. Remote Sensing. 2025; 17(8):1364. https://doi.org/10.3390/rs17081364

Chicago/Turabian Style

Wei, Dong, Wenjing Di, Wenshou Tian, Shanjun Cheng, Hongfei Xie, Lijun Xie, and Zhikun Jing. 2025. "Assessment of Integrated Multi-Satellite Retrievals for Global Precipitation Measurement (IMERG) Precipitation Products in Northwest China" Remote Sensing 17, no. 8: 1364. https://doi.org/10.3390/rs17081364

APA Style

Wei, D., Di, W., Tian, W., Cheng, S., Xie, H., Xie, L., & Jing, Z. (2025). Assessment of Integrated Multi-Satellite Retrievals for Global Precipitation Measurement (IMERG) Precipitation Products in Northwest China. Remote Sensing, 17(8), 1364. https://doi.org/10.3390/rs17081364

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

Article Metrics

Back to TopTop