https://orcid.org/0000-0002-8998-8160
Figures
Abstract
Climate change is expected to change precipitation and temperature patterns, which will impact the hydrological regime in Asia. Most river systems in the region originate from the Hindu Kush-Himalayas, and the altered precipitation patterns pose a threat to their sustainability, making it a major concern for planners and stakeholders. Obtaining accurate data on precipitation distribution is crucial for water accounting, which poses challenge. To address this, gridded precipitation products developed from satellite imagery and modeling techniques have become a viable alternative or addition to observed rainfall. However, the accuracy of these products in the region is uncertain. In this study, we aim to evaluate and compare the seven most commonly used precipitation products for the regions to address this gap. The study evaluated seven rainfall products, namely APHRODITE, TRMM, CHIRPS, PERSIANN-CDR, CMORPH, WFDEI, and GPCC by comparing daily, dekadal, and monthly rainfall data to 168 stations data in six countries and 11 river basins in the HKH region. The analysis used four continuous statistical indicators (Pearson correlation coefficient, Bias, Root Mean Square Error, and Nash–Sutcliffe Efficiency coefficient) and two categorical indicators (Probability of Detection and False Alarm Ratio). APHRODITE consistently performed well in several basins with high r values and low RMSE values, but had positive or negative bias values in different basins. CMORPH had the lowest positive bias value in the Ganga_Brahmaputra basin, while GPCC showed the largest r value and lowest RMSE value in the Sindha basin. CHIRPS performed well in Afghanistan, but had positive bias values. GPCC performed well in Myanmar and Pakistan, but had negative or positive bias values. APHRODITE performed consistently well in Nepal, but had negative bias values. Overall, the performance of different gridded precipitation products varies depending on the country and type of evaluation.
Citation: Mishra B, Panthi S, Ghimire BR, Poudel S, Maharjan B, Mishra Y (2023) Gridded precipitation products on the Hindu Kush-Himalaya: Performance and accuracy of seven precipitation products. PLOS Water 2(8): e0000145. https://doi.org/10.1371/journal.pwat.0000145
Editor: Sher Muhammad, ICIMOD: International Centre for Integrated Mountain Development, NEPAL
Received: January 5, 2023; Accepted: June 11, 2023; Published: August 8, 2023
Copyright: © 2023 Mishra et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data were acquired from the public repositories and incorporated all sources in the manuscripts.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
The Himalayas have been called the ‘water tower’ of Asia as more than 3 billion people are estimated to depend on the combined flows of the major rivers originating in the region [1]. The pressure on the area’s water resources is immense [2] and is expected to increase significantly in the future due to a combination of climate change and socioeconomic developments [1, 3]. The combination of steep topography, rapidly draining soils and a strongly seasonal climate delivering 80% of the rainfall during the summer monsoon in the greater part of the Himalayas, renders the natural ecosystems of the region particularly fragile and thus vulnerable to disturbance [4]. Studies in the Lesser Himalayas (3,700 to 4,500 m) by [4] suggested a significant increase in rainfall quantity and intensity during the monsoon and decreased in dry season rainfall. Such changes in precipitation patterns are a big threat to the sustainability of the river systems in the region. Owing to the socioeconomic and ecological importance of these river systems, the effects of climate change are of extreme concern among the planners and stakeholders in the region.
Accurate and detailed information on spatial variation in precipitation is critical in understanding the effects of climate change on Himalayan River systems. A robust climate change analysis and adaptation planning will not be possible without long-term hydro-meteorological monitoring. However, a lack of reliable and consistent data on precipitations severely limits our scientific understanding of the climate change impacts on these rivers [5]. Moreover, long-term observed data are rare in the higher altitude area of the Himalayan region, largely due to the remoteness. On top of that, systematic bias is also present through the urbanization effect on meteorological observations, and wind effect observed precipitations [6]. Therefore, a better understanding of the spatial and temporal extents of Himalayan precipitation is instrumental in water resources planning in the region.
Gridded precipitation products developed from satellite imagery sources have well served these gaps and stood as a strong alternative to the in-situ stations. These products come with varying spatial and temporal resolution along with individual strengths and weaknesses [7]. The use of gridded precipitation data has been common for a long time for various applications such as hydrological, disaster management, and agricultural studies. With the increasing importance of these data for applications such as hydrological modeling, assessing climate change impacts, extreme climate induced hazards, and vegetation studies, it is important assess their quality as different gridded datasets may perform differently. Li et al. (2018) found that the distribution of simulated daily discharge values agreed well with observations while using in the WRF-Hydro (v3.5.1) modeling system in the Beas Basin [8]. In another study, Immerzeel et. al. (2009) found that the discharge obtained from the TRMM 3B43 matched with the observed discharge in the Himalayan river basins [9]. However it is noted that the snowmelt constitutes up to 50% of the total annual discharge in the Indus catchments, (far western) and ~25% in the Tsangpo (far eastern) catchment [10], therefore the performance depends on various factors such as other input data sources like temperature, soil moisture, land cover etc. as well as the model structures and parameters. The accumulated precipitation was found in overestimation in Southern China [11] while GPM-IERG products perform satisfactorily to catch the flash flood in Yunnan China even though some products over-estimation the precipitation [12]. Therefore, such products perform differently in a different region of the Himalayas and therefore separate evaluation is important.
Those products are validated for the different locations of the world however, their general consistencies are not ensured all over the world as the accuracy varies from place to place, and product by product. Same product does not perform equally all over the world [13–17].
A number of studies have already been conducted to compare the gridded precipitation estimations with the ground measurement (Dembélé and Zwart, 2016; Serrat-Capdevila et al., 2016; Bhardwaj et al., 2017; Ghulami, Hussain et al., 2017; Nawaz et al., 2017; Shukla et al., 2019) [18]. Results from such studies have depicted huge variances in performances, in different location topography, climate season, etc. [8, 17, 19–24]. The pattern is determined by a number of factors such as Global geography topography distance from the ocean [25]. A large part of the HKH region still remains unstudied.
Therefore, in this study we aim to evaluate the most common precipitation products in the HKH regions. This study considered the seven most commonly used precipitation products based on the literature review. The product includes Asian Precipitation Highly Resolved Observational Data Integration Towards Evaluation of Water Resources (APHRODITE) 1801_R1 [26, 27], A Climate Hazard Group InfraRed Precipitation with Station Data (CHIRPS) [27, 28], The climate prediction center morphing method (CMORPH) [29], The Precipitation Estimation from Remotely Sensed Information using the Artificial Neural Networks Climate Data Record (PERSIANN-CDR) [27, 30]. The WFDEI meteorological forcing data set has been produced using the WATCH Forcing Data (WFD) [31, 32], Tropical Rainfall Measuring Mission (TRMM 3B42) [29, 33, 34] and Global Precipitation Climatology Centre (GPCC) [30, 35, 36]. These products were identified as the most commonly used products in the region. We compared the identified gridded precipitation with the ground measurement obtained from the 168 stations scattered throughout the HKH region. Six statistical measures, with four continuous and two categorical statics, were considered to compare with three different time accumulation, daily, 10-day, and 30-day.
This study distinguishes itself through its notable features, including a broader spatial coverage, a comprehensive evaluation of multiple precipitation products, an analysis considering three-time accumulation durations, and localized assessments at the national and basin levels. These contributions play a significant role in advancing our understanding of gridded precipitation patterns and their reliability within the HKH region.
Study area
HKH is the largest continuous mountainous range extending over 3500km from Afghanistan in the west to Myanmar in the east and hosts all 14 mountains above 8000m heigh (Fig 1). Ten major river systems of Asia–Amu Darya, Tarim, Indus, Ganges, Brahmaputra, Irrawaddy, Yangtze, Yellow, Mekong, and Salween are originated from the higher mountainous region of HKH. The total population of HKH is more than 240 million covering more than 4.2 million km2 [37].
Source base layer: [42].
Monsoon precipitation contributes higher in the Siwalik and Pir-Panjal range of lower Himalayas and continuously decreases towards the northwards into the High Himalaya. The western part of the Karakoram Himalaya is likely to have heavy snowfall during winter due to the western disturbances that emanate from upper tropospheric westerlies [38, 39]. Monsoon contributes above 80% of annual precipitation in the central and eastern Himalayas. Precipitation is highly variable across the Himalayans river basins with annual mean precipitation is estimated 435, 1094 and 2143mm in the Indus, the Ganges, and the Brahmaputra river basins respectively [40, 41].
Materials and methods
Precipitation data sets
Ground observations.
The Global Historical Climatology Network (GHCN) is an integrated database of daily climate summaries from land surface stations across the globe [43]. A total of 317 stations are available in the Hindu Kush Himalayas region, however, only 168 stations have data for more than a year which are considered for analysis in this study. As in many other places, their distribution is not even and is mostly concentrated in low-altitude areas of the Indian Himalayas. The highest elevation of the measurement station is 4880m. The data are maintained and provided by the National Climatic Data Centre (NCDC) of the National Oceanic and Atmospheric Administration (NOAA). Although data are available for a long duration (less than one year to more than 175 years), we have selected the period from 2001 to 2016 in this study, as the gridded precipitation of all products and the observed precipitation overlap during this period.
Gridded precipitation products.
APHRODITE is a 0.25° spatial resolution, daily, gridded precipitation of monsoon Asia (MA) (60.0°E– 150.0°E, 15.0°S– 55.0°N). The dataset is available from 1998 to 2015 and generated through the interpolation of the ground observation through the modified version of the distance weighting interpolation method [44]. It includes the rainfall information from the Global Telecommunication System (GTS) network, and precompiled datasets from meteorological organizations of Asian countries [23]. The series of monthly and daily rain-gauge data having a period of 5 years or more were considered while constructing this product. We obtained the dataset from the Research Institute for Humanity and Nature (RIHN) and the Meteorological Research Institute of Japan Meteorological Agency (MRI/JMA).
A CHIRPS is a near-global (50°S to 50°N) precipitation product of 0.05° spatial resolution available from 1981 to present [45]. It was developed to support the United States Agency for International Development Famine Early Warning Systems Network (FEWS NET) and in general agricultural drought monitoring. The major input data includes (i) expected annual rainfall sequence, computed through CHPClim (ii) thermal infrared images from the Climate Prediction Center (CPC) in every 30 minutes at 4km spatial resolution and B1 IR, a 3-hourly dataset of 8km spatial resolution from National Climatic Data Center (NCDC); (iii) 3B42 dataset from NASA obtained from the Tropical Rainfall Measuring Mission (TRMM 3B42), (iv) atmospheric model rainfall fields from climate forecast system version 2, NOAA, and (v) ground measurement from multiple sources. The data product is generated into two steps first includes the precipitation estimation from infrared images and in the second the estimated precipitation is blended with stations to produce the final CHIRPS product [46].
The CMORPH is a 30-minute global precipitation data that is derived from passive microwave images obtained from geostationary satellites [47]. CMORPH uses motion vectors for precipitation estimation. Time-weighted linear precipitation is used to modify precipitation features such as shape and intensity.
The PERSIANN-CDR is available from 1983 to the near future from 60° S to 60° N with the spatial resolution of 0.25° on the daily basis. It uses an artificial neural network that was trained using National Centers of Environmental Prediction (NCEP) stage IV hourly precipitation data [22]. The input dataset includes the B1 infrared window (IRWIN) Channel of the Gridded Satellite Data (GridSat-B), Global Precipitation Climatology Project (GPCP) V2.2. The accuracy is improved by adaptive adjustment of the network parameters using rainfall estimates from the TMI.
The WFDEI meteorological forcing data set has been produced using the WATCH Forcing Data (WFD) by using of ERA-Interim reanalysis data [48]. The ERA-Interim reanalysis data improves in precipitation and wind speed data while changes in the aerosol corrections improve downward shortwave fluxes. All precipitation corrections are conducted based on the CRU option.
GPCC is the product based on quality-controlled 67000 stations around the Globe that feature a record duration of 10 years or longer [49]. Source from national meteorological and hydrological services, global and regional data collections as well as WMO GIS data. This is a daily product of 0.5°× 0.5° is an hourly dataset. The rainfall rate is based on GPCC bias-correction, under catch corrected measures.
TRMM 3B42 V7, a peri-global precipitation product, is a 0.25° spatial resolution available in 50°S to 50°N from 1998 to the present [34]. The estimated precipitation was generated through the TRMM Multi-satellite Precipitation Analysis (TMPA) system that utilizes multi-satellite gauge data for bias correction. The daily precipitation was computed from a 3-hourly dataset. The major input includes TRMM Precipitation Radar (PR), TRMM Microwave Imager (TMI) the advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) on Aqua SSM/I and space sensor microwave imager (SSMIS), the AMSU-B and Microwave humidity Sounder, the IR data from an international constellation of Geosynchronous Earth Orbit (GEO) satellite and GPCP precipitation gauge analysis data.
We have summarized the details of all data products in Table 1 and visually represented the selected gridded precipitation products in Fig 2. The figure shows that the precipitation distribution and intensity vary among the different products. We have highlighted some of the identified precipitation disparities between different products using red and purple circles in the figure.
Data analysis
The gridded products were compared to the ground station data on a point-to-pixel basis [50]. The timing and intensity of rainfall event and the number of rainy days is highly spatially variables in the topographically heterogeneous region of HKH Mountain. Thus, we overlaid the stations over the raster images and extracted the timeseries dataset for the respective pixels on which intersected stations lie. We wrote a python script to get the timeseries dataset of the intersected pixel in the gridded product. The time series dataset in the selected pixels were subjected to analysis for three different temporal scales i.e., daily, dekadal (10-day), and monthly (30-day). The 10-day and 30-day data were considered for analysis when at least 90% of daily observations were available. This resulted in 13,511 dekadal totals, and 3458 monthly totals over the 168 synoptic stations for the satellite products during the 16 years of study.
Evaluation statistics.
A pair-wise continuous statistics were used to evaluate the performance of the gridded products in estimating the rainfall quantity and categorical statistics were used to assess the rainfall detection capabilities. We considered 6 statistical indicators with 4 continuous and two categorical indicators for the analysis. The statistical measures considered in this study for the pairwise comparison are summarized in Table 2, it includes (i) the Pearson correlation coefficient (r-squared), this gives the degree of matching on estimating the gridded products against the observed values; its range is -1 to +1 with the +1 is the perfect results, (ii) bias, the degree of estimated values over or under-estimation was measured through bias, its range is -∞ to ∞, with 0 is the perfect value, (iii) RMSE, the commonly used measures of differences between two variables are measured through the root mean square error (RMSE)–it measures the average magnitude of the estimated errors; lower the RMSE values, greater the central tendencies and generally smaller extreme errors; (iv) Nash–Sutcliffe Efficiency coefficient (E) measures the pattern of predicted observed time series. It varies from -∞ to 1, with 1 being the perfect value. (v) The probability of detection (POD) is the proportion of the correctly detected rainfall event to the total number of forecasted events. It gives the capacity of the correctly detected capacity of rainfall estimation; the range is 0 to 1. The perfect score is 1, (vi) False alarm ratio, a statistic for the categorical forecast, is the ratio of false alarm to the total number of forecasted events. It measures the reliability of the forecast. The perfect score is 0. The summary of the statistics is depicted in Table 2.
Results and discussion
Ground observation availability and coverage
The availability of ground observation stations is very low in the Hindu Kush Himalayan regions in comparison to their counterparts in Europe and America. Table 3 shows the statistics of available ground observation stations globally based on GHCN [43]. It is very high in North America, followed by Australia, Europe, South America, Asia, and Africa. There are 317 stations in the Hindu Kush Himalayas; however, 47% of the total stations were maintained for less than a year. Only the remaining 168 stations depicted in Fig 3 were subjected to analysis. Most of the stations in the HKH were concentrated in the Southeast (Bangladesh) and the southern part of the Indian Himalayas.
Tibetan plateau, the upstream part of the Yangtze River, has sparse stations but they are uniform. The central part of the Tibetan Plateau has very few or no stations. Most of the stations are located at an altitude less than 3000m whereas 57 stations are located at an altitude less than 1000m, 54 stations are in between 1000 and 2000m, 29 in between 2000 and 3000m while only 20 in between 3000 and 4000m and only eight stations lie above 4000m, where 4880m is the highest station. The details of the stations are presented in Fig 3.
Statistical analysis was performed for seven daily gridded precipitation products with observed precipitation (GHCN) collected from the ground stations at a point to pixel basis. The statistical analysis of seven gridded precipitation products overall six HKH countries and all eleven river basins are presented in the following subsections.
Results based on the country.
This section provides an overview of the findings from a study that compared the performance of seven different gridded precipitation products in six countries in HKH regions. Fig 4 illustrates the statistical matrices for seven gridded precipitation products over six countries. Daily, dekadal and monthly dispersion are presented in the first, second and third columns respectively. The countrywide best-performing products are summarized in Table 4.
In the daily dataset, CHIRPS showed the highest r value (0.04) with less RMSE (13.62 mm) and 1.36 mm positive bias value among all seven gridded precipitation products in Afghanistan. In China, APHRODITE product exhibited the largest r value (0.51) and smallest RMSE value (4.87 mm) with 0.63 mm of positive bias value. Likewise in India, again APHRODITE product exhibited the largest r value (0.05) and less RMSE value (13.67 mm) with 0.05 mm of positive bias value. In Myanmar, GPCC displayed r value (0.11) and smallest RMSE value (12.05 mm) with 4.04 mm of negative bias value. In Nepal, APHRODITE product showed the highest r value (0.39) with less RMSE (11.05 mm) and 1.47 mm of negative bias value, whereas in Pakistan, GPCC displayed the largest r value (0.16) and less RMSE value (11.43 mm) with 1.46 mm of positive bias value.
The dekadal evaluation based on country data, CHIRPS showed the highest r value (0.16) with less RMSE (59.77 mm) and 15.72 mm of positive bias value among all seven gridded precipitation products in Afghanistan. In China, APHRODITE product exhibited the largest r value (0.76) and smallest RMSE value (18.87 mm) with 6.29 of positive bias value. Likewise in India, again APHRODITE product exhibited r value (0.15) and less RMSE value (73.62 mm) with the lowest 3.55 mm of negative bias value. In Nepal, APHRODITE product showed the highest r value (0.79) with less RMSE (48.52 mm) and 16.12 mm negative bias value, whereas in Pakistan, GPCC displayed the largest r value (0.46) and less RMSE value (31.14 mm) with 6.42 mm positive bias value.
The monthly evaluation based on country data, CHIRPS showed the highest r value (0.29) with the lowest RMSE (121.5 mm) and 41.41 mm of positive bias value among all seven gridded precipitation products in Afghanistan. In China, APHRODITE product exhibited the largest r value (0.8) and less RMSE value (44.3 mm) with 18.49 mm positive bias value. Likewise in India, again APHRODITE product exhibited r value (0.22) and less RMSE value (165.04 mm) with the lowest 18.23 mm of negative bias value. In Nepal, APHRODITE product showed the highest r value (0.84) with less RMSE (140.61 mm) and 74.22 mm negative bias value, whereas in Pakistan, GPCC displayed the largest r value (0.54) and the lowest RMSE value (65.36 mm) with 17.75 mm positive bias value.
In overall, the performance of different gridded precipitation products in estimation of precipitation in different countries varies. CHIRPS performed well in Afghanistan with high r values and low RMSE, but had positive bias values. APHRODITE performed well in China and India with high r values and less RMSE, but had positive or negative bias values in different evaluations. GPCC performed well in Myanmar and Pakistan with high r values and low RMSE, but had negative or positive bias values in different evaluations. APHRODITE performed consistently well in Nepal with high r values and less RMSE, but had negative bias values in some evaluations. Overall, different products had different strengths and weaknesses in predicting precipitation in different countries and types of evaluations.
Results based on the basins.
The section presents a comparison of seven gridded precipitation products in eleven river basins across the HKH region. Fig 5 shows the statistical matrices for the different products, and Table 5 summarizes the best-performing products in each basin.
The daily evaluation based on basin data, APHRODITE showed the highest r value (0.13) with less RMSE (16.85 mm) and 0.35 mm of positive bias value among all seven gridded precipitation products in the Brahmaputra basin. In the Ganga basin, APHRODITE product exhibited the largest r value (0.09) and less RMSE value (14.99 mm) with 0.87 mm of positive bias value. Likewise in Ganga_Bhramaputra basin, CMORPH product exhibited the largest r value (0.12) and lowest RMSE value (12.4 mm) with 3.66 mm of positive bias value. In the Huang basin, APHRODITE displayed the highest r value (0.42) and smallest RMSE value (5.62 mm) with 0.98 mm positive bias value. In the Irrawaddy basin, again APHRODITE product showed the highest r value (0.19) with less RMSE (13.82 mm) and 0.18 mm of positive bias value. In the same way, in the Mekong basin, APHRODITE displayed the largest r value (0.73) and lowest RMSE value (3.71 mm) with no bias value. In Salween basin, APHRODITE product exhibited the largest r value (0.76) and lowest RMSE value (3.27 mm) with 0.7 mm of positive bias value. Likewise, in Sindha basin, APHRODITE product exhibited largest r value (0.05) and less RMSE value (11.6 mm) with 0.85 mm of positive bias value. In Tarim basin, APHRODITE displayed highest r value (0.4) and smallest RMSE value (2.75 mm) with 0.3 mm of positive bias value. In Tibet_interior basin, again APHRODITE product showed the highest r value (0.46) with lowest RMSE (5.62 mm) and 1.31 mm of positive bias value. In the same way, in Yangtze basin, again APHRODITE displayed largest r value (0.5) and lowest RMSE value (5.8 mm) with 0.67 mm of positive bias value.
The dekadal evaluation based on basin data, APHRODITE showed the highest r value (0.22) with less RMSE (89.65 mm) and 4.6 mm of negative bias value among all seven gridded precipitation products in Bhramaputra basin. In Ganga basin, APHRODITE product exhibited the largest r value (0.26) and less RMSE value (82.37 mm) with 8.32 mm of positive bias value, whereas, in Ganga_Bhramaputra basin, CMORPH product exhibited the smallest r value (0.29) but lowest RMSE value (58.75 mm) with 26.45 mm (low) of positive bias value. In the Huang basin, APHRODITE displayed the highest r value (0.62) and smallest RMSE value (23.3 mm) with 9.79 mm positive bias value. In the Irrawaddy basin, again APHRODITE product showed the high r value (0.45) with the lowest RMSE (68.7 mm) and 2.8 mm of positive bias value. In the same way, in the Mekong basin, APHRODITE displayed the largest r value (0.85) and lowest RMSE value (16.29 mm) with 0.35 mm of positive bias value. In Salween basin, APHRODITE product exhibited the largest r value (0.86) and lowest RMSE value (17.23 mm) with 7.69 mm of positive bias value. In Sindha basin, the GPCC product exhibited the largest r value (0.39) and lowest RMSE value (39.08 mm) with 8.54 mm positive bias value. In the Tarim basin, APHRODITE displayed the highest r value (0.69) and smallest RMSE value (9.35 mm) with 2.7 mm of positive bias value. In Tibet_interior basin, again APHRODITE product showed the highest r value (0.63) with the lowest RMSE (24.92 mm) and 12.6 mm of positive bias value. In the same way, in the Yangtze basin, again APHRODITE displayed the largest r value (0.81) and lowest RMSE value (19.79 mm) with 6.38 mm of positive bias value.
The monthly evaluation based on basin data, APHRODITE showed the high r value (0.29) with RMSE (201.51) and 27.62 (low) of negative bias value among all seven gridded precipitation products in the Brahmaputra basin. In Ganga basin, TRMM product exhibited the largest r value (0.31) and RMSE value (179.93) with 5.14 (low) positive bias value. Likewise in Ganga_Bhramaputra basin, CMORPH product exhibited r value (0.45) and the lowest RMSE value (141.32) with 80.47 (low) positive bias value. In Huang basin, APHRODITE displayed the highest r value (0.66) and smallest RMSE value (53.29) with 29.08 of positive bias value. In Irrawaddy basin, CHIRPS product showed the highest r value (0.61) with the lowest RMSE (155.93) and 7.33 positive bias value. In Mekong basin, APHRODITE displayed largest r value (0.86) and RMSE value (41.33) with 2.13 of bias value. In Salween basin, APHRODITE product exhibited largest r value (0.88) and RMSE value (45.36) with 22.92 positive bias values. Likewise, in Sindha basin, CDR product exhibited the largest r value (0.22) and lowest RMSE value (98.5) with 6.63 of positive bias value. In Tarim basin, APHRODITE displayed the highest r value (0.74) and smallest RMSE value (22.78) with 8.03 of positive bias value. In Tibet_interior basin, again APHRODITE product showed the highest r value (0.68) with RMSE (58.32) and 37.25 of positive bias value. In the same way, in Yangtze basin, again APHRODITE displayed the largest r value (0.86) and RMSE value (44.74) with 18.64 of positive bias value.
In summary, APHRODITE consistently showed the highest r values and lowest RMSE values across several basins, including the Brahmaputra, Ganga, Huang, Irrawaddy, Mekong, Salween, Tarim, Tibet_interior, and Yangtze basins. The product exhibited positive bias values in most basins, except for the dekadal evaluation in the Bhramaputra basin where it showed negative bias. CMORPH had the lowest positive bias value in the Ganga_Brahmaputra basin, and GPCC showed the largest r value and lowest RMSE value in the Sindha basin.
Discussion
The gridded precipitation datasets used in this study are typically based on observations from rain gauges, radar, and satellite remote sensing, which are interpolated onto a regular grid to create a spatially continuous dataset. APHRODITE has consistently demonstrated superior performance in most of the river basins and countries within the HKH region, when compared to the seven other products considered. Similar results were obtained in a comparative study with CHIRPS and PERSIANN-CDR in the Tibetan Plateau and its surroundings [26]. In another study, APHRODITE outperformed the other five precipitation products (CPC-RFE, GSMaP, TRMM-3B42, TRMM-2B31) in the five small river basins in the Himalayas [51]. However, all products were found to be quite reliable when evaluated on a monthly basis, as they demonstrated high correlation coefficients and low RMSE. A similar observation was made in an analysis of the Northwest Himalayan Region [36]. Overall, APHRODITE consistently outperforms other products by better tracking above-threshold precipitation events, in which other products tend to underestimate the precipitation quantity [51].
The source of error in different products, obtained from various methodology, might be different, including the spatial and temporal resolution of the dataset, the interpolation method used, and the quality and density of the input observations. In the higher altitude regions of the Himalayas, the quality and density of input observations are reported to be relatively low and mostly very sparse [35, 40], making it challenging to accurately represent precipitation patterns and amounts. Additionally, the high-altitude regions in the Himalayas are characterized by steep slopes, deep valleys, and complex atmospheric processes [52], which further complicates precipitation measurement and estimation. Furthermore, the rain shadow regions in the Himalayas, such as the regions to the north of the Great Himalayas, receive less precipitation due to the orographic effect [51], which can be difficult to accurately capture in gridded precipitation products.
The higher temporal/spatial resolution datasets (hourly or few km scale) may have lower accuracy compared to datasets with lower temporal/spatial resolution (longer time of several km scales) due to the higher variability and uncertainty in precipitation measurements at this resolution. In the Himalayas, where precipitation is highly variable in space and time, accurately representing precipitation patterns and amounts can be particularly challenging. However, if gridded precipitation datasets are carefully validated and calibrated against observed streamflow data, they can still provide useful information for hydrological modeling in the Himalayas.
By considering a wider geographic area, evaluating multiple products, analyzing different time scales, and providing localized insights, our study contributes greatly to enhancing knowledge in this field. Ultimately, these findings strengthen our understanding of gridded precipitation in the HKH region. However, a large number of precipitation products, including ERA5 and ERA5-Land, Multi-Source Weighted-Ensemble Precipitation (MSWEP) v2.2, Modern-Era Retrospective Analysis for Research and Applications 2 (MERRA-2), and Integrated Multi-Satellite Retrievals (IMERG) for Global Precipitation Mission (GPM), near-real-time precipitation estimations based on machine learning algorithms, such as FY4QPE-MSA and PECA-FY4A etc. among others, are available for use in this region and have gained popularity in various applications. Therefore, we encourage future studies to assess the quality of these products. Furthermore, it would be worthwhile to study how different precipitation products perform in various topographical settings, as we have not taken this into account in this study.
Conclusion and recommendation
The performance of gridded precipitation products varies across countries and basins. After considering statistical parameters, it was found that CHIRPS is the better product for Afghanistan, APHRODITE for China, India, and Nepal, and GPCC for Pakistan, for all daily, dekadal, and monthly analyses. However, the performance of the products is not consistent across different basins. APHRODITE performs better than other products in most basins, except for the Ganga_Bhramaputra basin where CMORPH had favorable results in daily analysis. In dekadal analysis, APHRODITE outperformed other products in all basins except Ganga_Bhramaputra (CMORPH) and Sindha (GPCC). When it comes to monthly analysis, APHRODITE performed better in most basins except for Ganga (TRMM), Ganga_Bhramaputra (CMORPH), Irrawaddy (CHIRPS), and Sindha (CDR). In general, APHRODITE and TRMM performed best in all the indicators for all stations in the region, while CHIRPS performed the worst.
Due to the highly heterogeneous topography, gridded precipitation may not capture the timing and spatial variation within a short distance. As the gridded precipitation is an aggregation of a spatial extent of a single-pixel, it may moderate the results, and the timing could also have a significant impact on it. Given the highly heterogeneous topography and potential impact of timing and spatial variation on the accuracy of gridded precipitation products, it is recommended that daily estimates be used with caution in the HKH region. Nonetheless, some products’ dekadal and monthly estimates can be used with confidence. It is advisable for researchers and professionals to evaluate the products for their specific study area and objectives before incorporating these gridded products into their analysis.
Supporting information
S1 Table. Precipitation stations considered in this study.
https://doi.org/10.1371/journal.pwat.0000145.s001
(XLSX)
Acknowledgments
Thanks also to Global Historical Climatology Network (GHCN) for the collection of an integrated database of daily climate summaries from land surface stations across the globe.
References
- 1. Bandyopadhyay J. Securing the Himalayas as the Water Tower of Asia: An Environmental Perspective. Asia Policy. 2013;16: 45–50.
- 2. Merz B, Kreibich H, Thieken A, Schmidtke R, Hydrology SE. Estimation uncertainty of direct monetary flood damage to buildings_Merz et al_2004.pdf. Nat Hazards Earth Syst Sci. 2004;4: 153–163. Available: https://www.nat-hazards-earth-syst-sci.net/4/153/2004/nhess-4-153-2004.pdf
- 3.
Wester P, Mishra A, Mukherji A, Shrestha AB, Change C. The Hindu Kush Himalaya Assessment. Bajracharya SR, Shrestha B, editors. ICIMOD; 2019.
- 4. Tsering K, Sharma E, Chettri N, Shrestha A. Climate Change and its socio-cultural impact in the Himalayan region of Nepal–A Visual Documentation. ICIMOD. 2013.
- 5. Mishra Y, Babel MS, Nakamura T, Mishra B. Impacts of climate change on irrigation water management in the babai river basin, Nepal. Hydrology. 2021;8: 1–20.
- 6.
Krishnan R, Shrestha AB, Ren G, Rajbhandari R, Saeed S, Sanjay J, et al. Unravelling Climate Change in the Hindu Kush Himalaya: Rapid Warming in the Mountains and Increasing Extremes. Wester P., Mishra A., Mukherji A. SA, editor. The Hindu Kush Himalaya Assessment. Springer, Cham; 2019. https://doi.org/10.1007/978-3-319-92288-1_3
- 7. Bhardwaj A, Ziegler AD, Wasson RJ, Chow W. Accuracy of rainfall estimates at high altitude in the Garhwal Himalaya (India): a comparison of secondary precipitation products and station rainfall measurements. Atmos Res. 2017.
- 8. Li H, Haugen JE, Xu C. Precipitation pattern in the Western Himalayas revealed by four datasets. Hydrol Earth Syst Sci. 2018;22: 5097–5110.
- 9. Immerzeel WW, Droogers P, de Jong SM, Bierkens MFP. Large-scale monitoring of snow cover and runoff simulation in Himalayan river basins using remote sensing. Remote Sens Environ. 2009;113: 40–49.
- 10. Bookhagen B, Burbank DW. Toward a complete Himalayan hydrological budget: Spatiotemporal distribution of snowmelt and rainfall and their impact on river discharge. 2010;115.
- 11. Gao Z, Huang B, Ma Z, Chen X, Qiu J. Comprehensive Comparisons of State-of-the-Art Gridded Precipitation Estimates for Hydrological Applications over Southern China. Remote Sens. 2020;12.
- 12. Ma M, Wang H, Jia P, Tang G, Wang D. Application of the GPM-IMERG Products in Flash Flood Warning: A Case Study in Yunnan, China. Remote Sens. 2020;12: 1–15.
- 13. Dembélé M, Zwart SJ. Evaluation and comparison of satellite-based rainfall products in Burkina Faso, West Africa. Int J Remote Sens. 2016;37: 3995–4014.
- 14. Sun W, Ma J, Yang G. Statistical and Hydrological Evaluations of Multi-Satellite Precipitation Products over Fujiang River Basin in Humid Southeast China. 2018.
- 15. Peralta J, Narisma G, Cruz F. Validation of High-Resolution Gridded Rainfall Datasets for Climate Applications in the Philippines. 2020; 1571–1587.
- 16. Dilinuer T, Yao J, Chen J, Zhao Y, Mao W, Li J. Systematical Evaluation of Three Gridded Daily Precipitation Products Against Rain Gauge Observations Over Central Asia. 2021;9: 1–17.
- 17. Shukla AK, Shekhar C, Ojha P, Singh RP, Pal L, Fu D. Evaluation of TRMM Precipitation Dataset over Himalayan Catchment: The Upper Ganga Basin, India. Water. 2019.
- 18. Serrat-Capdevila A, Merino M, Valdes JB, Durcik M. Evaluation of the performance of three satellite precipitation products over Africa. Remote Sens. 2016;8.
- 19. Hussain S, Song X, Ren G, Hussain I, Zaman DH& H. Evaluation of gridded precipitation data in the Hindu-Kush-Karakoram- Himalaya mountainous area Evaluation of gridded precipitation data in the Hindu Kush–Karakoram–Himalaya mountainous area. Hydrol Sci J. 2017;00: 1–13.
- 20. Ghulami M, Babel MS, Shrestha MS. Evaluation of gridded precipitation datasets for the Kabul Basin, Afghanistan. Int J Remote Sens. 2017;38: 3317–3332.
- 21. Nawaz Z, Ijaz MW, Sajid M, Farid HU. Evaluation and Comparison of TRMM Multi-Satellite Precipitation Products With Reference to Rain Gauge Observations in Hunza River Basin, Karakoram. 2017.
- 22. Ashouri H, Hsu K-L, Sorooshian S, BraitHwaite DK, Knapp KR, dewayne CeCiL L, et al. Daily Precipitation Climate Data Record from Multisatellite Observations for Hydrological and Climate Studies. Am Meteorol Soc. 2015; 69–84.
- 23. Ak Yatagai, Kamiguchi K, AraKawa O, Hamada A, Yasutomi N, Kitoh Ak. Constructing a Long-Term Daily Gridded Precipitation Dataset for Asia Based on a Dense Network of Rain Gauges. Am Meteorol Soc. 2012; 1401–1416.
- 24. De Roo A, Zambrano-Bigiarini M, Rojas R, Thiemig V, Levizzani V. Validation of Satellite-Based Precipitation Products over Sparsely Gauged African River Basins. J Hydrometeorol. 2012;13: 1760–1783.
- 25. Meher JK, Das L, Benestad E, Mezghani A. Analysis of winter rainfall change statistics over the Western Himalaya: the influence of internal variability and topography. Int J Climatol. 2017.
- 26. Tan X, Ma Z, He K, Han X, Ji Q, He Y. Evaluations on gridded precipitation products spanning more than half a century over the Tibetan Plateau and its surroundings. J Hydrol. 2020;582.
- 27. Hasson Sul, Böhner J, Chishtie F. Low fidelity of CORDEX and their driving experiments indicates future climatic uncertainty over Himalayan watersheds of Indus basin. Clim Dyn. 2019;52: 777–798.
- 28. Shrestha NK, Qamer FM, Pedreros D, Murthy MSR, Wahid SM, Shrestha M. Evaluating the accuracy of Climate Hazard Group (CHG) satellite rainfall estimates for precipitation based drought monitoring in Koshi basin, Nepal. J Hydrol Reg Stud. 2017;13: 138–151.
- 29. Bajracharya SR, Palash W, Shrestha MS, Khadgi VR, Duo C, Das PJ, et al. Systematic evaluation of satellite-based rainfall products over the brahmaputra basin for hydrological applications. Adv Meteorol. 2015;2015.
- 30. Palazzi E, Hardenberg J Von, Provenzale A. Precipitation in the Hindu-Kush Karakoram Himalaya: Observations and future scenarios. 2013;118: 85–100.
- 31. Dahri ZH, Ludwig F, Moors E, Ahmad B, Khan A, Kabat P. An appraisal of precipitation distribution in the high-altitude catchments of the Indus basin. Sci Total Environ. 2016;548–549: 289–306. pmid:26802357
- 32. Wijngaard RR, Biemans H, Friedrich Lutz A, Bhakta Shrestha A, Wester P, Willem Immerzeel W. Climate change vs. socio-economic development: Understanding the future South Asian water gap. Hydrol Earth Syst Sci. 2018;22: 6297–6321.
- 33. Palazzi E, Von Hardenberg J, Provenzale A. Precipitation in the hindu-kush karakoram himalaya: Observations and future scenarios. J Geophys Res Atmos. 2013;118: 85–100.
- 34. Bharti V, Singh C. Evaluation of error in TRMM 3B42V7 precipitation estimates over the Himalayan region. J Geophys Res Atmos Res. 2015.
- 35. Zandler H, Haag I, Samimi C. Evaluation needs and temporal performance differences of gridded precipitation products in peripheral mountain regions. Sci Rep. 2019;9: 1–15. pmid:31641198
- 36. Kanda N, Negi HS, Rishi MS, Kumar A. Performance of various gridded temperature and precipitation datasets over northwest himalayan region. Environ Res Commun. 2020;2.
- 37.
Bajracharya SR, Shrestha B. The Status of Glaciers in the Hindu Kush-Himalayan Region. Bajracharya SR, Shrestha B, editors. ICIMOD. 2011.
- 38. Madhura RK, Krishnan R, Revadekar J V., Mujumdar M, Goswami BN. Changes in western disturbances over the Western Himalayas in a warming environment. Clim Dyn. 2014.
- 39. Dimri AP, Niyogi D, Barros AP, Ridley J, Mohanty UC, Yasunari T, et al. Reviews of Geophysics Western Disturbances: A review. Rev Geophys. 2015; 225–246.
- 40. Mishra B, Babel MS, Tripathi NK. Analysis of climatic variability and snow cover in the Kaligandaki River Basin, Himalaya, Nepal. Theor Appl Clim. 2014;116: 681–694.
- 41. Nepal S, Shrestha AB. Impact of climate change on the hydrological regime of the Indus, Ganges and Brahmaputra river basins: a review of the literature. Int J Water Resour Dev ISSN. 2015;0627.
- 42. ICIMOD. Outline Boundary of Hindu Kush Himalayan (HKH) Region [dataset]. In: ICIMOD [Internet]. 2008 [cited 13 Jul 2023]. Available: https://doi.org/10.26066/RDS.3924
- 43. Menne MJ, Durre I, Vose RS, Gleaason BE, Houston TG. An Overview of the Global Historical Climatology Network-Daily Database. J Atmos Ocean Technol. 2012; 897–910.
- 44. Shepard D. Two- dimensional interpolation function for irregularly- spaced data. Proc 23rd Nat Conf. 1968; 517–524.
- 45. Funk C, Peterson P, Landsfeld M, Pedreros D, Verdin J, Shukla S, et al. The climate hazards infrared precipitation with stations—A new environmental record for monitoring extremes. Sci Data. 2015;2: 1–21. pmid:26646728
- 46. Funk CC, Peterson PJ, Landsfeld MF, Pedreros DH, Verdin JP, Rowland JD, et al. A Quasi-Global Precipitation Time Series for Drought Monitoring Data Series 832. U.S. Geological Survey Data Series 832. 2014.
- 47. Joyce RJ, Janowiak JE, Arkin PA, Xie P. CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J Hydrometeorol. 2004;5: 487–503.
- 48. Weedon GP, Balsamo G, Bellouin N, Gomes S, Best MJ, Viterbo P. The WFDEI meteorological forcing data set: WATCH Forcing data methodology applied to ERA-Interim reanalysis data. Water Resour Res. 2014;50: 7505–7514.
- 49. Schneider U, Becker A, Finger P, Meyer-Christoffer A, Ziese M, Rudolf B. GPCC’s new land surface precipitation climatology based on quality-controlled in situ data and its role in quantifying the global water cycle. Theor Appl Climatol. 2014;115: 15–40.
- 50. Dembélé M, Schaefli B, van de Giesen N, Mariéthoz G. Suitability of 17 rainfall and temperature gridded datasets for largescale hydrological modelling in West Africa. Hydrol Earth Syst Sci Discuss. 2020; 1–39.
- 51. Freiberg D. Evaluation of precipitation data sets along the Himalayan front. Geochemistry Geophys Geosystems. 2011;12.
- 52. Poudel S, Mishra B, Ghimire S, Luintel N, Thapa P, Sapkota R. Climate change and agroecosystem: impacts, adaption, and mitigation in South Asia. Climate Change, Community Response, and Resilience: Insight for Socio-Ecological Sustainability. 2023. pp. 329–344.