Next Article in Journal
Deep Learning-Based Daily Streamflow Prediction Model for the Hanjiang River Basin
Previous Article in Journal
Monitoring the Impacts of Human Activities on Groundwater Storage Changes Using an Integrated Approach of Remote Sensing and Google Earth Engine
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Comparative Analysis of Satellite-Based Rainfall Products for Drought Assessment in a Data-Poor Region

by
Hansini Gayanthika
1,
Dimuthu Lakshitha
1,
Manthika Chathuranga
1,
Gouri De Silva
1 and
Jeewanthi Sirisena
2,*
1
Faculty of Engineering, University of Peradeniya, Peradeniya 20400, Sri Lanka
2
Climate Service Center Germany, Helmholtz-Zentrum Hereon, 20095 Hamburg, Germany
*
Author to whom correspondence should be addressed.
Hydrology 2025, 12(7), 166; https://doi.org/10.3390/hydrology12070166
Submission received: 14 May 2025 / Revised: 15 June 2025 / Accepted: 23 June 2025 / Published: 27 June 2025

Abstract

Drought is one of the most impactful natural disasters, and it significantly impacts three main sectors of a country: the environment, society, and the economy. Therefore, drought assessment and monitoring are essential for reducing vulnerability and risk. However, insufficient and sparse long-term in situ rainfall data limit drought assessment in developing countries. Recently developed satellite-based rainfall products, available at different temporal and spatial resolutions, offer a valuable alternative in data-poor regions like Sri Lanka, where rain gauge networks are sparse and maintenance issues are prevalent. This study evaluates the accuracy of satellite-based rainfall estimates compared to in situ observations for drought assessment within the Mi Oya River Basin, Sri Lanka. We assessed the performance of various satellite-based rainfall products, including IMERG, GSMaP, CHIRPS, PERSIANN, and PERSIANN-CDR, by comparing them with ground-based observations over 20 years, from 2003 to 2022. Our methodology involved checking detection accuracy using the False Alarm Ratio (FAR), Probability of Detection (POD), and Critical Success Index (CSI), and assessing accuracy through metrics such as Root Mean Square Error (RMSE), Pearson Correlation Coefficient (CC), Percentage Bias (PBias), and Nash–Sutcliffe Efficiency (NSE). The two best-performing satellite-based rainfall products were used for meteorological and hydrological drought assessment. In the accuracy detection metrics, the results indicate that while products like IMERG and GSMaP generally provide reliable rainfall estimates, others like PERSIANN and PERSIANN-CDR tend to overestimate rainfall. For instance, IMERG shows a CSI range of 0.04–0.25 for moderate and heavy rainfall and 0.10–0.30 for light rainfall. On a monthly scale, IMERG and CHIRPS showed the highest performance, with CC (NSE) values of 0.81–0.94 (0.53–0.83) and 0.79–0.86 (0.54–0.74), respectively. However, GSMaP showed the lowest bias, with a range of −17.1–13.2%. Recorded drought periods over 1981–2022 (1998–2022) were reasonably well captured by CHIRPS (IMERG) products in the Mi Oya River Basin. Our results highlighted uncertainties and discrepancies in the capability of different rainfall products to assess drought conditions. This research provides valuable insights for optimizing the use of satellite rainfall products in hydrological modeling and disaster preparedness in the Mi Oya River Basin.

1. Introduction

Drought is one of the most destructive and economically costly natural hazards, causing significant impacts on agriculture, water resources, ecosystems, and socioeconomic development [1,2,3,4]. It is typically defined as a prolonged period of below-average precipitation that can last from months to years, and unlike permanent dryness in deserts, drought is a temporary climatic anomaly that can occur in both humid and arid regions [5]. Its effects are particularly severe in arid and semi-arid areas, where rainfall is already limited, and agriculture is predominantly rainfed [6]. Even short-term droughts in these situations have the potential to seriously affect crop productivity, jeopardize food security, and disrupt livelihoods.
Drought can be categorized into four main types, each impacting different systems and resources [4,7]. (1) Meteorological drought is characterized by a deficiency in precipitation over a specific region during a lengthy period (typically months), often assessed by comparing current precipitation levels to long-term averages or drought indices. (2) Hydrological drought occurs when surface and subsurface water resources become insufficient to meet the demands of water resource management systems. This is typically measured and assessed through streamflow data and indices incorporating streamflow, which reflect reductions in surface water availability. (3) Agricultural drought relates to inadequate soil moisture content, which can lead to crop failure. It is influenced by meteorological and hydrological drought conditions, as well as the difference between actual and potential evapotranspiration. Various drought indices incorporating precipitation, temperature, and soil moisture are employed to monitor and assess agricultural drought. (4) Socioeconomic drought arises when water shortages affect the balance between supply and demand for economic activities, particularly when the water demand exceeds the available supply due to weather-induced shortages. Each type of drought varies in intensity, duration, and spatial extent, but all have significant implications for agriculture, ecosystems, and socioeconomic systems [7].
Quantification of the severity of droughts is not a new concept, where numerous indices over 150 have been developed, many of which rely on precipitation data as a primary input [8]. Indices such as the Standardized Precipitation Index (SPI, [9,10]), the Palmer Drought Severity Index (PDSI, [11]), and the Standardized Precipitation Evapotranspiration Index (SPEI, [12]) are widely used in drought monitoring and analysis related to hydrology and agriculture. However, their effectiveness is highly dependent on the availability of accurate, high-resolution precipitation data. Many regions in Africa, Asia, and Latin America lack high-resolution, accurate observed data due to sparse or inconsistent rain gauge networks [13].
In these data-scarce regions, satellite-based rainfall products have emerged as a critical alternative. These products offer near-global coverage, varying temporal resolutions, and the capacity to monitor precipitation in remote and under-monitored regions. Datasets such as CHIRPS (Climate Hazards Group InfraRed Precipitation with Station data), TRMM (Tropical Rainfall Measuring Mission), PERSIANN (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks), and GSMaP (Global Satellite Mapping of Precipitation) have become widely used in hydrological and agricultural research [14,15,16,17,18,19,20,21,22,23,24,25]. Even though satellite products are freely available across the globe, these estimates often suffer from biases and uncertainties, particularly in regions with complex terrain or highly variable microclimates, which can limit their effectiveness in operational drought assessment [26]. To ensure the reliability of satellite data for any hydro-meteorological analysis, including drought analysis, rainfall data validation against ground-based observations is essential.
However, drought analysis and monitoring are far beyond scientific objectives; they are critical development priorities, particularly in regions where agriculture forms the backbone of the economy. It is critical to safeguard food security, sustain livelihoods, and ensure overall economic stability, particularly in developing countries, where a majority of the population relies on agriculture. Without timely and accurate information, communities remain vulnerable to adverse impacts of droughts such as crop failure, water scarcity, and subsequent socio-economic crises [27,28].
Accurate and consistent rainfall data are essential to support decision-making, including early warning systems, climate-resilient agricultural planning, and other targeted drought interventions. However, most developing countries, including Sri Lanka, continue to face significant challenges in obtaining reliable precipitation data due to limited ground-based monitoring infrastructure. Against this backdrop, this study conducts a comparative analysis of widely used satellite-based rainfall products to assess their performance in a data-poor region. The aim is to evaluate their suitability for drought monitoring in agricultural contexts, identify the most reliable datasets, and provide recommendations for enhancing drought resilience in vulnerable and under-monitored areas. Having a strong reliance on rainfed agriculture, limited hydro-meteorological infrastructure, and increasing vulnerability to climate extremes such as droughts, the Mi Oya Basin presents a good case for this type of study. Despite its agricultural and socioeconomic importance, the Mi Oya Basin has received limited attention in satellite-based drought monitoring studies, particularly in terms of validating satellite-derived precipitation against ground observations and drought identification, highlighting a critical knowledge gap. This study offers novel insights into the applicability of satellite rainfall products in a data-poor region, with their reliability and limitations. The comparative validation of multiple satellite products using available ground observations, combined with an applicability assessment of drought analysis in an agriculturally sensitive region, provides a unique contribution to both national and global research. This study not only advances scientific understanding but also supports practical applications in drought risk management and policy formulation in regions facing similar data scarcity challenges.

2. Materials and Methods

2.1. Study Area

The Mi Oya River Basin, situated in the northwestern part of Sri Lanka (as shown in Figure 1), serves as a vital hydrological system within the country. It originates from the central highlands at an elevation of approximately 600 m and discharges into the Gulf of Mannar in the Indian Ocean. The basin spans an area of approximately 1530 km2, with a total length of about 118 km, and discharges 412 million m3 of water annually [29,30]. The basin has many water reservoirs, including Inginimitiya, Thabbowa, Halmilla, and Maha Wewa (Figure 1), serving water for irrigation, drinking purposes, fisheries, and recreational activities.
The surrounding landscape of the Mi Oya River Basin predominantly consists of alluvial soil, making it highly favorable to agricultural activities. Land use in the basin is diverse, featuring a mix of crops, including paddy, tea, and rubber plantations, as well as forested areas [30]. Predominantly irrigated agriculture is fed by the main reservoirs and tanks in the basin. The region receives an average annual rainfall of approximately 1200 mm during the major rainy season (Southwestern) and approximately 840 mm during the low rainy season (Northeastern). The average temperature of the basin is above 30 °C [30].
In recent years, the Mi Oya River Basin has encountered significant challenges, including major floods (2016 and 2019) and droughts (2011, 2014, and 2017) [31,32,33]. These floods have substantial implications for water resource management and the livelihoods of the local population. Consequently, ongoing study and monitoring of this basin are essential for effective water resource management, particularly drought monitoring, and for improving the understanding of the broader hydrological dynamics in the region.

2.2. Methodology

The following sub-sections present a methodology for evaluating the suitability of satellite rainfall estimates as an alternative to ground-based data in addressing data scarcity, which are ultimately used for drought assessment over the basin. The process begins with the collection of multi-source rainfall data, followed by rigorous data pre-processing with gap filling and consistency checking. This is followed by accuracy assessment and comparative analysis to evaluate the performance of satellite data relative to ground observations. Finally, the best-performing satellite data was used for the drought assessment.

2.2.1. Data and Analysis

(a) 
Gauge Data
The daily rainfall data were collected from the Meteorological Department, Sri Lanka, for the selected rainfall gauging stations in the Mi Oya River Basin. The gauging stations selected for analysis include Puttalam, Palawi, Anamaduwa, Atharagalla, Thabbowa, Kottukachchiya, and Mediyawa (Figure 1). These stations are the best representative stations in the study region and provide a reasonable spatial distribution for rainfall analysis. The seven stations have less than 10% of missing data for the analysis period (2003–2022), which is within the recommended thresholds for climatological stations by the World Meteorological Organization. The missing data was gap-filled using the Inverse Distance Weighting (IDW) method which is a widely adopted spatial interpolation technique in hydrology and meteorology. This method is particularly suitable when a moderately dense network of stations is available, and it helps preserve local rainfall variability [34,35]. Following gap-filling, consistency checking of the rainfall data was performed using the Double Mass Curve (DMC) technique. The DMC method compares the cumulative rainfall at a station with the cumulative average of neighboring stations. This method is a well-established and most commonly used approach to detect inconsistencies in datasets (e.g., [36,37,38]). Inconsistencies in rainfall records may arise from changes in instrumentation, relocation of the gauge, observational errors, or environmental impacts such as forest fires or landslides [38]. These pre-processing procedures ensured the reliability and consistency of the rainfall data for further analysis. The summary of these stations’ data is tabulated in Table 1.
(b) 
Satellite-based Rainfall Products
In this research, five satellite-based rainfall products—IMERG, GSMaP, CHIRPS, PERSIANN, and PERSIANN-CDR—were considered for analysis. These products were selected because of their accessibility, temporal coverage, and extensive use in hydrological and climate studies. All data were obtained from their respective official websites. The characteristics of each dataset are summarized in Table 2.
The Integrated Multi-Satellite Retrievals for GPM (IMERG) product was developed by the National Aeronautics and Space Administration (NASA) as a part of the Global Precipitation Measurement (GPM) project. It offers 30 min temporal resolution and 0.1° spatial resolution precipitation estimates from 1998 to the present. The IMERG algorithm uses a multi-step processing framework that includes calibration, morphing, and bias correction stages to merge passive microwave sensor data with infrared information. According to Huffman et al. [39], IMERG (GPM IMERG V07) offers improved detection of light and moderate rainfall events through its refined multi-sensor approach. The main advantages of IMERG include its high spatial and temporal resolution and the availability of near-real-time and post-processed data. However, some studies have reported limitations in detecting light rainfall and underestimations in regions with complex terrain due to satellite viewing geometry and retrieval uncertainties [40].
The Global Satellite Mapping of Precipitation (GSMaP) is a product developed by the Japan Aerospace Exploration Agency (JAXA). GSMaP provides hourly precipitation data at a 0.1° spatial resolution from 2003 onward. According to Kubota et al. [41], GSMaP uses passive microwave radiometer observations, adjusted with infrared estimates, and applies a Kalman filter algorithm to improve temporal consistency. Further advancements in the GSMaP algorithm in the GPM era have improved its rainfall detection capabilities, especially in tropical regions [41]. Relatively high temporal resolution and high performance over tropical and subtropical areas are strengths of GSMaP. However, its drawbacks include limited accuracy in arid zones and mountainous areas, and some lag in delivering real-time data products [42].
The Climate Hazards Group InfraRed Precipitation with Stations (CHIRPS) dataset was mainly developed to support drought monitoring and food security assessments. It provides daily to monthly precipitation data at a 0.05° spatial resolution from 1981 to the present. CHIRPS combines satellite infrared data with in situ station observations using a blending algorithm that improves the reliability of rainfall estimates in data-scarce regions. Funk et al. [43] highlighted the utility of CHIRPS for long-term climate analysis and extreme weather monitoring due to its extended historical record and inclusion of ground-based data. Its high spatial resolution and long-term availability make it ideal for climate trend analysis and agricultural planning. However, CHIRPS has a lower temporal resolution compared to IMERG or GSMaP, and performance is influenced by the density and quality of ground stations.
Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIAN) is a satellite rainfall estimation product developed by the Center for Hydrometeorology and Remote Sensing (CHRS). It offers daily precipitation data at a 0.25° spatial resolution using artificial neural networks to estimate rainfall from infrared satellite observations. According to Nguyen et al. [44], PERSIANN was designed to support real-time hydrological applications, especially in data-poor regions. Sorooshian et al. [45] explained that the neural network is trained using microwave observations to improve infrared-based rainfall estimates, enhancing the system’s ability to adjust to varying meteorological conditions. The main advantages of PERSIANN include its real-time availability and flexibility in capturing diverse rainfall patterns; however, data is available from 2000.
PERSIANN-CDR (Climate Data Record) is a long-term, climate-focused version of the PERSIANN product. It provides daily precipitation data from 1983 to the present at a 0.25° spatial resolution, similar to PERSIANN. PERSIANN-CDR uses the same neural network framework as PERSIANN, but is tuned for consistency across decades, enabling its use in climate change studies. Ashouri et al. [46] noted that PERSIANN-CDR is constructed using infrared data reprocessed for homogeneity, making it a reliable source for multi-decadal precipitation trend analysis. The key strengths of PERSIANN-CDR are its long historical coverage and methodological consistency, which make it ideal for climate variability and drought analysis. However, like its parent product, it is a coarse spatial resolution dataset.

2.2.2. Accuracy Assessment of Satellite-Based Data

There are several metrics/ indices (i.e., Probability of Detection (POD), False Alarm Ratio (FAR), Critical Success Index (CSI), Nash–Sutcliffe Efficiency (NSE), Pearson Correlation Coefficient (CC), Percentage Bias (PBias), etc.) continue to be a widely accepted and methodologically rigorous approach for evaluating satellite-based precipitation products and their hydrological applicability. These metrics are widely used to quantify the statistical performance of satellite rainfall datasets, particularly in data-scarce, topographically complex, or climatically diverse regions. For example, Kumar et al. [47] demonstrated the robustness of these classical metrics in Himalayan terrain, showing their effectiveness in detecting rainfall variability across multiple satellite products. Similarly, Zhang et al. [48] employed the RMSE, NSE, CC, PBias, POD, FAR, and CSI in the upper Red River Basin in China to assess the performance of satellite rainfall datasets in hydrologically sensitive environments, revealing meaningful differences in bias, correlation, and event detection accuracy. In drought-related studies, Zhu et al. [49] applied these same metrics to evaluate precipitation inputs for the Standardized Precipitation Index (SPI) in the Xiang River Basin, China, concluding that they were essential for validating satellite-derived drought estimates. Wang et al. [50] conducted a detailed intercomparison of the Global Precipitation Measurement Mission (GPM) and Tropical Rainfall Measuring Mission (TRMM) rainfall products using POD, FAR, bias, and correlation metrics against ground-based radar observations over the continental United States, further affirming the operational value of these classical statistics. The use of these metrics in arid and complex terrains has also been validated by Wang et al. [51], who applied RMSE, CC, PBias, POD, FAR, and CSI to assess the accuracy of TRMM estimates in northwestern China.
The accuracy of satellite data was evaluated using categorical indices such as the Critical Success Index (CSI), Probability of Detection (POD), and False Alarm Ratio (FAR). These metrics were applied to daily rainfall data to assess the performance of the satellite products in capturing precipitation events and are explained in Equations (1)–(3). Considering the availability of data, 20 years (2003–2022) were selected for this analysis.
P r o b a b i l i t y   o f   D e t e c t i o n   P O D = H H + M
F a l s e   A l a r m   R a t i o   F A R = F H + F
C r i t i c a l   S u c c e s s   I n d e x   C S I = F H + M + F
where H represents the accurately detected rainfall (correct hits), M represents the undetected rainfall (missed data), and F represents the falsely detected precipitation (false alarms), as shown in Table 3. The intensity class of rainfall was determined based on the following criteria: “Light rainfall” ( 0 P < 25 mm/day), “Moderate rainfall” ( 25     P < 100 mm/day), and “Heavy rainfall” ( 100     P mm/day). If the rainfall value fell within the specified range, it was classified as “Rain”; otherwise, it was classified as “No Rain.”
After the detection check, further accuracy assessment was conducted using daily, monthly, seasonal, and annual data. For the seasonal analysis, rainfall gauge data and satellite rainfall estimates were compared based on basin average values. Four distinct seasons were considered in this analysis: First Inter-Monsoon, Northeast Monsoon, Second Inter-Monsoon, and Southwest Monsoon. The evaluation involved comparing observed rainfall with satellite estimates using specific accuracy metrics, such as NRMSE, CC, NSE, and PBias. Equations (4)–(7) represent the accuracy metrics.
N o r m a l i z e d   R o o t   M e a n   S q u a r e   E r r o r   N R M S E = i = 1 n S i G i 2 n G m a x G m i n  
P e a r s o n   C o r r e l a t i o n   C o e f f i c i e n t   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
N a s h S u t c l i f f i e   E f f i c i e n c y   N S E = 1 i = 1 n S i     G i 2 i = 1 n S i G ¯ 2
P B i a s = i = 1 n S i G i i = 1 n G i
S i and G i represent the satellite rainfall estimation and the Gauged rainfall, respectively, for the ith season rainfall data of the relevant season, S ¯ and   G ¯   represent the mean value of satellite and gauge data products, respectively, and n represents the total number of relevant seasons. G m a x and G m i n are the maximum and minimum gauge rainfall, respectively.
A lower NRMSE indicates better agreement between satellite rainfall estimates and gauge observations. The correlation between both datasets can be assessed according to Table A1. An NSE value approaching 1 signifies a stronger correlation between satellite estimates and gauge observations, indicating high model performance. Similarly, a lower PBias value reflects a closer agreement between satellite estimates and ground-based observations, with minimal deviation from the actual measured rainfall. Following the accuracy assessment, the satellite data with the highest correlation coefficient and the lowest root mean square value was selected for further analysis. This selection criterion ensured that the chosen satellite estimates demonstrated the strongest agreement with ground-based observations and the lowest error in rainfall estimation.

2.2.3. Drought Analysis

The Standardized Precipitation Index (SPI, [9,10]) is one of the most widely used indices for drought monitoring and defining drought severity (Alahacoon et al. [52]) and is recognized as a meteorological drought index. One of the main advantages of using the SPI is its simple calculation across various timescales (e.g., 1, 3, 6, or 12 or more months) [7], making it both practical and effective. Numerous studies have evaluated the performance of the SPI for drought monitoring in comparison to other indices across different climatic regions, with results showing that the SPI is highly successful in detecting meteorological drought events [53]. Additionally, the SPI has been applied in various drought-related studies, including drought forecasting, frequency analysis, and spatiotemporal drought analysis, as well as in investigations into drought duration, severity, and the impacts of climate change [52,54,55,56,57]. For these reasons, SPI was adopted as the primary drought monitoring index in this study.
In general, the SPI fits the rainfall data to a probability distribution for calculating a reduced variate. In this study, the most commonly used Gamma distribution [10,58,59] was applied to fit the rainfall data in the SPI calculation process. According to the different SPIs based on the number of months considered (e.g., SPI-1, SPI-3, SPI-6, SPI-12, and SPI-24), short-term and long-term drought can be analyzed. We calculated SPI-1, SPI-3, SPI-6, and SPI-12 by using the best-performing global precipitation product(s) over the Mi-Oya Basin. The drought severity classification was adopted from McKee et al. [10] in Table A1. This analysis was carried out over four sub-basins (Figure 1), taking the average rainfall over each basin from the satellite data. The arithmetic average of gridded rainfall data was calculated for each sub-basin. The resulting monthly time series data were then used in SPI calculations. The selection of the sub-basin was purely dependent on the agricultural and topography features of the Mi Oya Basin. We assessed the SPI indices with recorded drought conditions in the basin over the last 3 decades to verify the applicability of these global products for drought assessment. The overall methodology is presented in Figure 2.

3. Results

3.1. Performance of Satellite Data

3.1.1. Detection of Daily Rainfall

The detection accuracy of each satellite was checked using categorical metrics such as the Critical Success Index (CSI), Probability of Detection (POD), and False Alarm Ratio (FAR). Daily rainfall gauge data and satellite-derived rainfall data from 2003 to 2022 were utilized for this analysis. The evaluation was performed for each of the seven-gauge data stations. The daily rainfall events were classified into three categories based on their intensities: light rainfall (less than 25 mm), moderate rainfall (between 25 mm and 100 mm), and heavy rainfall (greater than 100 mm).
An analysis of the CSI for detecting light rainfall across five satellite products (IMERG, GSMaP, CHIRPS, PERSIANN, and PERSIANN-CDR) at seven locations indicates that GSMaP consistently performs the best with the highest accuracy (Table 4). For example, in Puttalam, GSMaP achieved a CSI of 0.39 and an FAR of 0.46, indicating higher accuracy and fewer false detections. While PERSIANN-CDR exhibited the highest POD, GSMaP’s overall better performance in both CSI and FAR made it the most reliable satellite product for light rainfall detection. Overall, IMERG showed the lowest accuracy among the selected satellite products. CHIRPS and PERSIANN offered moderate performance, while PERSIANN-CDR performed slightly better than IMERG. These results suggest that GSMaP is the most reliable satellite product for detecting light rainfall in the study area.
IMERG consistently demonstrates the highest CSI values and the highest POD, making it the most effective and reliable satellite product for moderate rainfall detection, while PERSIANN-CDR showed the lowest CSI and POD values, indicating lower reliability in capturing such events (Table 4). Additionally, IMERG reported the lowest FAR, while PERSIANN generally showed the highest. Overall, IMERG is the most effective and reliable satellite product for moderate rainfall detection, whereas PERSIANN performs the weakest across these key metrics.
Even for heavy rainfall detection, IMERG outperformed other satellite products across key performance metrics. In Thabbowa, it recorded a CSI of 0.19 and a POD of 0.38, while in Anamaduwa, it showed relatively low FAR values, such as 0.71 (Table 4). IMERG showed a more consistent performance across locations, particularly with lower FAR values compared to GSMaP, indicating fewer false detections of heavy rainfall events. Notably, FAR values were not reported for CHIRPS and PERSIANN-CDR, probably due to their failure to detect any heavy rainfall events, which in turn reflects the limited number of such events in the dataset (Table 4). Therefore, while IMERG appears most reliable, the small sample size of heavy rainfall events demands careful interpretation of these results. Overall, GSMaP proved to be the most effective satellite product for detecting light rainfall, while IMERG excelled in the detection of moderate and heavy rainfall, followed by CHIRPS.

3.1.2. Overall Accuracy

Accuracy assessment was also carried out for daily, monthly, seasonal, and annual scales to identify the most suitable satellite data product(s) using statistical metrics such as Normalized Root Mean Square Error (NRMSE), Pearson Correlation Coefficient (CC), Nash Sutcliffe Efficiency (NSE), and Percentage Bias (PBias). However, results are presented for the monthly, seasonal, and annual scales as these are more relevant and informative for drought assessment.
Monthly Scale
Among the five satellite-based rainfall products evaluated, IMERG and CHIRPS consistently provided the highest accuracy across all gauge stations, with lower NRMSE values and higher CC (Figure 3 and Table A3). IMERG in particular showed the highest correlation with the observed rainfall data, with correlation coefficients ranging from 0.81 to 0.94 and the highest NSE values ranging from 0.53 to 0.83. In contrast, GSMaP and PERSIANN-CDR exhibited moderate accuracy, while PERSIANN showed the weakest performance among the five satellite-based products considered.
In terms of total bias, PERSIANN and PERSIANN-CDR datasets exhibit the highest positive biases, consistently overestimating rainfall, particularly at Ataragalla station. IMERG shows moderate positive biases across all stations, indicating relatively consistent overestimation, while GSMaP and CHIRPS show more variable performance, including instances of negative bias at locations such as Madiyawa, where rainfall is underestimated. Overall, the analysis indicates that while some satellite products, such as IMERG and CHIRPS, provide relatively stable and balanced estimates, others, like PERSIANN and PERSIANN-CDR, tend to significantly overestimate rainfall, with biases varying notably by location (Table A3).
Seasonal Scale
At the seasonal scale, the performance of basin average rainfall estimates was evaluated across four main monsoon seasons: First Inter-Monsoon (FIM), Northeast Monsoon (NEM), Second Inter-Monsoon (SIM), and Southwest Monsoon (SWM). NRMSE results for satellite rainfall products reveal that IMERG and CHIRPS generally have the lowest error ranges, from 0.13 to 0.26 and 0.08 to 0.37, respectively, indicating better performance, especially during the Northeast Monsoon (Figure 4a). In contrast, PERSIANN shows the highest error levels, particularly in the First Inter-Monsoon and Southwest Monsoon seasons. GSMaP and PERSIANN-CDR perform moderately, with varying error levels across seasons. The CC values further support the performance assessment. IMERG consistently achieves high CC values, ranging from 0.75 to 0.97, indicating strong performance (Figure 4b). CHIRPS performs exceptionally well during the Northeast Monsoon, with a CC of 0.97, and maintains good performance across other seasons. GSMaP and PERSIANN-CDR show moderate performance, with relatively lower CC values. PERSIANN shows moderate to good correlation but generally has the lowest CC value among satellite products. Similarly, the NSE results show that PERSIANN performs poorly, particularly during the First Inter-Monsoon and Southwest Monsoon seasons (Figure 4c). CHIRPS and GSMaP demonstrate better performance, especially during the Northeast and Southwest Monsoon seasons. IMERG has moderate to good performance across all seasons, with the highest NSE during the Second Inter-Monsoon. The performance of PERSIANN-CDR is more variable, ranging from moderate to poor, depending on the season. Percentage Bias (PBIAS) values further highlight the accuracy of each satellite rainfall product in estimating seasonal rainfall (Figure 4d). IMERG and CHIRPS generally show lower PBIAS values, indicating better performance. IMERG has PBIAS values ranging from 7.1% to 52.2%, while CHIRPS performs slightly better, ranging from 6.2% to 27.7%. GSMaP shows moderate performance with values between −21.1% and 17.4%. PERSIANN and PERSIANN-CDR exhibit higher biases, with PERSIANN particularly overestimating rainfall during the Southwest Monsoon, reaching a PBIAS of 112.6%. Overall, IMERG and CHIRPS are more reliable rainfall products consistently delivering better accuracy and correlation across seasons, while PERSIANN tends to have significant performance limitations, particularly in terms of bias and efficiency.
Annual Scale
The accuracy assessment of annual basin-averaged rainfall data from satellite products reveals notable differences in performance (Table 5). GSMaP demonstrates the highest overall accuracy, with the lowest NRMSE of 0.12, the highest CC of 0.89, a positive NSE of 0.74, and minimal bias at −0.04, indicating strong agreement with gauge data. CHIRPS also performs reasonably, with moderate NRMSE and NSE values, a CC of 0.78, and a slight positive bias of 7.28%, suggesting reliable rainfall estimates. IMERG, while showing a strong correlation (CC of 0.92), recorded a higher NRMSE of 0.20 and moderate NSE of 0.22, along with a positive bias of 17.70%, indicating some overestimation. PERSIANN and PERSIANN CDR perform poorly, with PERSIANN showing a high NRMSE of 0.49, a low CC of 0.58, a negative NSE of −3.27, and a substantial positive bias of 44.18%. PERSIANN CDR fares slightly better, but still exhibits a negative NSE of −0.22 and a bias of 19.13%, indicating limited reliability. Overall, GSMaP provides the most accurate and consistent rainfall estimates, while PERSIANN shows significant deviations from the observed data, making it the least reliable among the evaluated satellite products.

3.2. Drought Assessment over the Mi-Oya Basin

Drought patterns were analyzed in four sub-basin areas in the Mi Oya River Basin: (a) Upper Basin, (b) Mid Basin 1, (c) Mid Basin 2, and (d) Downstream Basin (refer to Figure 1). Here, we calculated the SPI indices for 1, 3, 6, 12, and 24 months (Figure 5 and Figure 6, Figure A1, Figure A2 and Figure A3), representing short-term droughts (meteorological and hydrological by SPI-1, 3, and 6) and long-term droughts (agricultural by SPI-12 and 24).

3.2.1. Short-Term Droughts

The SPI-1, 3, and 6, which represent cumulative rainfall anomalies over one-, three-, and six-month periods, respectively, were calculated using IMERG and CHIRPS satellite-based rainfall products to assess the temporal and spatial variability in drought conditions over the Mi Oya River Basin. This provides a robust indication of short-term drought conditions relevant to hydrological impacts.
Figure 5 shows the variation in SPI-6 for both datasets across the four sub-basins. The results indicate that both CHIRPS and IMERG follow similar drought patterns after the year 2000, capturing the same periods of rainfall deficits and recovery. This consistency reinforces confidence in the reliability of satellite-derived drought monitoring in recent decades. However, CHIRPS data extends further back in time, allowing for the identification of historical drought events prior to 2000. SPI-6 time series derived by CHIRPS highlights significant drought events in the years 1983, 1994, 1999, 2006, 2011, 2014, and 2017. All of these SPI values are below 1.5, which indicates severe drought conditions. All four basins were affected by these droughts, suggesting that such events were spatially extensive across the entire Mi Oya River Basin. SPI-6 derived from CHIRPS data shows that the most extreme drought occurred in 1983, whereas IMERG shows 1999 in all the sub-basins of the Mi Oya Basin. In 1999, CHIRPS data represented a severe drought in the basin. Therefore, drought severity derived from the two datasets shows different magnitudes, resulting in distinct drought classes. Furthermore, according to the SPI-1 and SPI-3 analyses based on the CHIRPS dataset (Figure A1 and Figure A2), the years 1982 and 2012, respectively, were derived as the most extreme droughts over the four sub-basins. In contrast, SPI-1 and SPI-3 from IMERG show the most extreme drought occurrences in 2001 and 2020, respectively. These findings confirm that meteorological droughts can prevail for one to six months, but they do not always translate into short- to mid-term droughts affecting the hydrology in the basin.

3.2.2. Long-Term Droughts

The SPI-12 and SPI-24, which reflect precipitation anomalies over 12- and 24-month periods, respectively, provide valuable insight into the long-term drought conditions affecting the Mi Oya River Basin (Figure 6 and Figure A3). This extended timescale provides insights into assessing how prolonged rainfall deficits impact overall water availability. SPI-12 values derived from both CHIRPS and IMERG datasets reveal consistent long-term drought patterns, with notable events occurring in 1983, 1987, 2000, 2004, 2014, and 2017. During these years, SPI-12 values remained below −1.0, indicating moderate to extreme drought conditions throughout the basin. SPI-12 from CHIRPS data confirms that 1983 was the most extreme drought that occurred in all the sub-basins. Similarly, IMERG derived the most extreme drought in 2004 in the most downstream basin and in 2017 in the other three basins. Based on the SPI-24 analysis by CHIRPS data, all four basins experienced the most extreme drought in 1982–1983, and in Middle-Stream Basin 2, the most extreme drought occurred in 1987–1988. With the IMERG dataset, in the Upstream Basin, the most extreme drought occurred in 1998–1999, and the Middle- and Downstream Basins experienced the most extreme drought in 1999–2000 (Figure A3). These temporal and spatial differences in drought occurrence and severity can greatly affect water and food security in the Mi Oya River Basin.
Overall, the results clearly demonstrate that long-term and medium-term rainfall shortages, as indicated by SPI-6 and SPI-12, are closely linked to agricultural impacts, particularly in regions reliant on surface water for irrigation. The close agreement between the two datasets confirms their reliability for drought monitoring, while the longer historical record of CHIRPS enhances its utility for detecting older drought events not captured by IMERG due to its limited temporal coverage. The ability of CHIRPS to capture these long-term drought events highlights its suitability for historical drought analysis and water resource planning, particularly in data-scarce regions. Meanwhile, the good agreement between CHIRPS and IMERG after 2000 supports the use of both rainfall products for drought identification, monitoring, and early warning applications.

4. Discussion

4.1. Performance of the Satellite-Based Rainfall

The categorical analysis revealed that IMERG is the most reliable satellite rainfall product compared to other similar products, particularly in detecting precipitation events. However, similar to Perera et al. [24], the study also noted certain limitations, including over-estimations in specific regions and during particular seasons, underscoring the need for localized calibration to improve accuracy. Moreover, an evaluation of satellite rainfall products over the Mahaweli River Basin in Sri Lanka confirmed that IMERG exhibited the strongest correlation with ground-based rainfall observations compared to other selected products [24]. This finding aligns with the results of the present study and further supports the reliability of IMERG for accurate rainfall estimation in the South and Southeast Asian regions [60,61]. Furthermore, Alahacoon et al. [52] used CHIRPS data and other remote-sensing-based data for meteorological and agricultural drought monitoring over Sri Lanka, capturing the 2012 and 2016–2017 droughts. Overall, IMERG and CHIRPS are the most reliable products for multiscale rainfall analysis, while GSMaP performs best on an annual scale, and PERSIANN products require careful consideration due to their tendency to overestimate rainfall.

4.2. Suitability of Satellite-Based Rainfall Products in Drought Assessment

To evaluate the impact of these droughts on agriculture, annual paddy harvest data from the Kurunegala and Puttalam districts were examined. We obtained data from the Department of Census and Statistics, Sri Lanka, particularly on paddy statistics over the last four decades (1979–2024). These areas, which depend heavily on water from the Mi Oya River, experienced significant declines in harvested areas during major drought years such as 1987, 1996, 2004, and 2017. For example, in 1987, the harvested area dropped to 25% of the peak values in Kurunegala and 19% in Puttalam, illustrating the severity of crop losses. Similar reductions were observed during other drought years, highlighting the sensitivity of agriculture to meteorological droughts in this region.
The significant drought years identified using SPI-6 analysis—1983, 1994, 1998, 2006, and 2011—are supported by independent institutional and scientific evidence. The Disaster Management Centre [62] confirms that both 1983 and 2006 were major drought years, with widespread agricultural and hydrological impacts across multiple districts in Sri Lanka. Somasundaram et al. [63] identified 1994 and 1998 as significant drought years linked to strong El Niño events, which are known to weaken South Asian monsoons and trigger regional rainfall deficits, including in Sri Lanka. Additionally, the International Federation of Red Cross and Red Crescent Societies [33] reports that the 2011 drought in Sri Lanka affected over 1.8 million people, particularly in the Northern, Eastern, and North Central Provinces, highlighting its severity and spatial extent. The FAO [64] highlighted that the 1996 drought resulted in a 22% reduction in rice production, particularly in dry zone areas like Kurunegala (where the upstream of the Mi Oya Basin is located) and Anuradhapura. However, none of the datasets could capture this severe drought event (Figure 6). The UNCCD [65] report noted that the 2000 drought led to severe land degradation and affected agriculture. Furthermore, the Disaster Management Centre [31] reported that the 2014 drought caused major water shortages and crop failures in the North and North Central Provinces, while the FAO [32] confirmed that the 2017 drought significantly reduced rice production, impacting over 900,000 people. These sources provide consistent and independent confirmation of SPI-derived drought years and support the reliability of satellite-based drought monitoring.
The validated satellite rainfall products (IMERG and CHIRPS) and derived SPI indices could be integrated into national and local-level decision support systems aimed at drought preparedness and agricultural planning, up to a certain extent knowing the limitations of the global products. For instance, the SPI derived from satellite rainfall can be routinely monitored to provide early warnings about developing drought conditions. This information could be shared with agricultural officers and local farmers to guide short-term decisions such as crop selection, irrigation timing, and fertilizer application. Over longer timescales, SPI trends can inform seasonal planning and preparedness activities, including the allocation of water resources across competing sectors and reservoir operations during anticipated dry periods. Furthermore, the spatially continuous nature of satellite rainfall data enables monitoring in remote or ungauged areas, ensuring more equitable coverage across the regions. This supports more informed decision-making at both the administrative and farm levels, enabling proactive rather than reactive responses to drought conditions. However, before further use of these results, they should undergo in-depth analysis using bias-corrected satellite-based products.

5. Conclusions

This study focused on the accuracy assessment of freely available satellite-based global rainfall products for drought analysis in the Mi Oya Basin, Sri Lanka. Here, five satellite-based rainfall products were assessed, and IMERG was identified as the most reliable for rainfall estimation in the Mi Oya River Basin, followed by CHIRPS. SPI-6 and SPI-12 indices derived from CHIRPS and IMERG satellite rainfall data products were used to analyze drought patterns across four sub-basins in the Mi Oya River Basin. CHIRPS, with its extended historical coverage, effectively captured major drought events from 1980 onward (e.g., 1983, 1994, 1998, 2006, and 2011), while IMERG reliably detected droughts occurring after 2000 (e.g., 2006 and 2011). These findings highlight the suitability of both rainfall products for long-term drought analysis and water resource planning, particularly in data-scarce regions. Overall, CHIRPS and IMERG are reliable datasets for drought detection, early warning, and impact assessment, offering a practical framework for improving climate resilience in vulnerable regions.
As the Mi Oya River Basin is heavily managed by irrigation schemes, assessing future drought risk for managing reservoir operations and crop rotations and selection is crucial. Therefore, this study could be extended to a more comprehensive analysis of agriculture and socio-economic droughts under a changing climate with available in situ data and freely available global datasets.

Author Contributions

Conceptualization, H.G., G.D.S. and J.S.; methodology, H.G., G.D.S. and J.S.; validation, H.G. and D.L.; formal analysis, H.G. and D.L.; investigation, H.G.; data curation, H.G. and D.L.; writing—original draft preparation, H.G., D.L. and M.C.; writing—review and editing, H.G., D.L., M.C., G.D.S. and J.S.; visualization, H.G. and J.S.; supervision, G.D.S. and J.S. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

The data used in this study will be available upon request.

Conflicts of Interest

The authors declare no conflicts of interest.

Appendix A

Table A1. Interpretation of Correlation Coefficient.
Table A1. Interpretation of Correlation Coefficient.
Correlation Coefficient (CC)Interpretation
+1−1Perfect
+0.9 to +0.7−0.9 to −0.7Very strong
+0.6 to +0.4−0.6 to −0.4Strong
+0.3−0.3Moderate
+0.2−0.2Weak
+0.1−0.1Negligible
Table A2. Drought severity classification.
Table A2. Drought severity classification.
Drought CategorySPI
Extremely wet>2
Very wet1.5 to 2
Moderately wet1 to 1.5
Near Normal−1.0 to 1.0
Moderately dry−1 to −1.5
Severely dry−1.5 to −2.0
Extremely dry≤−2
Table A3. Performance of satellite rainfall data on a monthly scale. The best performances are highlighted in bold.
Table A3. Performance of satellite rainfall data on a monthly scale. The best performances are highlighted in bold.
StationIMERGGSMaPCHIRPSPERSIANNPERSIANN -CDR
NRMSE
Anamaduwa0.090.130.110.140.12
Ataragalla0.130.140.130.200.14
Kottukachchiya0.090.110.100.140.11
Madiyawa0.090.110.090.140.10
Palawi0.090.090.080.130.09
Puttalam0.080.120.100.170.11
Thabbowa0.100.110.110.150.11
CC
Anamaduwa0.860.730.800.760.76
Ataragalla0.810.760.790.700.77
Kottukachchiya0.870.790.810.770.79
Madiyawa0.860.810.860.760.82
Palawi0.930.780.840.720.80
Puttalam0.940.830.860.770.84
Thabbowa0.910.830.850.780.83
NSE
Anamaduwa0.710.430.620.290.54
Ataragalla0.530.480.54−0.070.42
Kottukachchiya0.710.560.650.340.58
Madiyawa0.740.630.720.390.67
Palawi0.740.550.690.070.56
Puttalam0.830.620.740.240.65
Thabbowa0.780.640.670.360.65
PBias
Anamaduwa18.1−2.38.037.218.9
Ataragalla43.813.230.370.549.3
Kottukachchiya26.66.910.842.423.0
Madiyawa9.2−17.1−10.028.210.9
Palawi40.97.513.855.934.3
Puttalam24.92.24.641.523.1
Thabbowa24.73.619.237.419.1
Figure A1. SPI-1 for (a) Upstream Basin, (b) Middle-Stream Basin 1, (c) Middle-Stream Basin 2, and (d) Downstream Basin of the Mi Oya Basin. SPIs are derived from IMERG and CHIRPS data starting from 1998 and 1981, respectively.
Figure A1. SPI-1 for (a) Upstream Basin, (b) Middle-Stream Basin 1, (c) Middle-Stream Basin 2, and (d) Downstream Basin of the Mi Oya Basin. SPIs are derived from IMERG and CHIRPS data starting from 1998 and 1981, respectively.
Hydrology 12 00166 g0a1
Figure A2. SPI-3 for (a) Upstream Basin, (b) Middle-Stream Basin 1, (c) Middle-Stream Basin 2, and (d) Downstream Basin of the Mi Oya Basin. SPIs are derived from IMERG and CHIRPS data starting from 1998 and 1981, respectively.
Figure A2. SPI-3 for (a) Upstream Basin, (b) Middle-Stream Basin 1, (c) Middle-Stream Basin 2, and (d) Downstream Basin of the Mi Oya Basin. SPIs are derived from IMERG and CHIRPS data starting from 1998 and 1981, respectively.
Hydrology 12 00166 g0a2
Figure A3. SPI-24 for (a) Upstream Basin, (b) Middle-Stream Basin 1, (c) Middle-Stream Basin 2, and (d) Downstream Basin of the Mi Oya Basin. SPIs are derived from IMERG and CHIRPS data starting from 1998 and 1981, respectively.
Figure A3. SPI-24 for (a) Upstream Basin, (b) Middle-Stream Basin 1, (c) Middle-Stream Basin 2, and (d) Downstream Basin of the Mi Oya Basin. SPIs are derived from IMERG and CHIRPS data starting from 1998 and 1981, respectively.
Hydrology 12 00166 g0a3

References

  1. Dutta, R. Drought Monitoring in the Dry Zone of Myanmar Using MODIS Derived NDVI and Satellite Derived CHIRPS Precipitation Data. Sustain. Agric. Res. 2018, 7, 46. [Google Scholar] [CrossRef]
  2. Shah, D.; Mishra, V. Drought Onset and Termination in India. J. Geophys. Res. Atmos. 2020, 125, e2020JD032871. [Google Scholar] [CrossRef]
  3. Bhardwaj, K.; Mishra, V. Drought Detection and Declaration in India. Water Secur. 2021, 14, 100104. [Google Scholar] [CrossRef]
  4. Wilhite, D.A. Drought as a Natural Hazards Concept and Definition in Wilhite. In Drought: A Global Assessment; Routledge: London, UK, 2000. [Google Scholar]
  5. Wilhite, D.A.; Glantz, M.H. Understanding: The Drought Phenomenon: The Role of Definitions. Water Int. 1985, 10, 111–120. [Google Scholar] [CrossRef]
  6. Sheffield, J.; Wood, E.F. Drought: Past Problems and Future Scenarios; Earthscan: London, UK, 2011. [Google Scholar]
  7. Mishra, A.K.; Singh, V.P. A Review of Drought Concepts. J. Hydrol. 2010, 391, 202–216. [Google Scholar] [CrossRef]
  8. Zargar, A.; Sadiq, R.; Naser, B.; Khan, F.I. A Review of Drought Indices. Environ. Rev. 2011, 19, 333–349. [Google Scholar] [CrossRef]
  9. McKee, T.B.; Doesken, N.J.; Kleist, J. Drought Monitoring with Multiple Time Scales. In Proceedings of the 9th Conference on Applied Climatology, Dallas, TX, USA, 15–20 January 1995; pp. 233–236. [Google Scholar]
  10. McKee, T.B.; Dosken, N.J.; Kleist, J. The Relationship of Drought Frequency and Duration to Time Scales. In Proceedings of the Eighth Conference on Applied Climatology, Anaheim, CA, USA, 17–22 January 1993; pp. 1–6. [Google Scholar]
  11. Palmer, W.C. Meteorological Drought; US Department of Commerce, Weather Bureau: Silver Spring, MD, USA, 1965; Volume 30. [Google Scholar]
  12. Vicente-Serrano, S.M.; Beguería, S.; López-Moreno, J.I. A Multiscalar Drought Index Sensitive to Global Warming: The Standardized Precipitation Evapotranspiration Index. J. Clim. 2010, 23, 1696–1718. [Google Scholar] [CrossRef]
  13. Tuttle, S.E.; Roof, S.R.; Retelle, M.J.; Werner, A.; Gunn, G.E.; Bunting, E.L. Evaluation of Satellite-Derived Estimates of Lake Ice Cover Timing on Linnévatnet, Kapp Linné, Svalbard Using In-Situ Data. Remote Sens. 2022, 14, 1311. [Google Scholar] [CrossRef]
  14. Sirisena, T.A.J.G. Process Based Modelling of Future Variations in River Flows and Fluvial Sediment Supply to Coasts Due to Climate Change and Human Activities: Data Poor Regions. Ph.D. Thesis, University of Twente, Enschede, The Netherlands, 2020. [Google Scholar]
  15. Sirisena, T.A.J.G.; Maskey, S.; Ranasinghe, R.; Babel, S. Effects of Different Precipitation Inputs on Streamflow Simulation in the Irrawaddy River Basin, Myanmar. J. Hydrol. Reg. Stud. 2018, 19, 265–278. [Google Scholar] [CrossRef]
  16. De Zoysa, S.; Sirisena, J.; Perera, H.; Fernando, S.; Gunathilake, M.; Rathnayake, U. Development of Intensity-Duration-Frequency Curves for Sri Lanka Using Satellite-Based Precipitation Products—Understanding Environmental Conditions and Concerns. Case Stud. Chem. Environ. Eng. 2024, 9, 100713. [Google Scholar] [CrossRef]
  17. Lv, X.; Guo, H.; Tian, Y.; Meng, X.; Bao, A.; De Maeyer, P. Evaluation of GSMaP Version 8 Precipitation Products on an Hourly Timescale over Mainland China. Remote Sens. 2024, 16, 210. [Google Scholar] [CrossRef]
  18. Perera, H.; Senaratne, N.; Gunathilake, M.B.; Mutill, N.; Rathnayake, U. Appraisal of Satellite Rainfall Products for Malwathu, Deduru, and Kalu River Basins, Sri Lanka. Climate 2022, 10, 156. [Google Scholar] [CrossRef]
  19. Cavalcante, R.B.L.; da Silva Ferreira, D.B.; Pontes, P.R.M.; Tedeschi, R.G.; da Costa, C.P.W.; de Souza, E.B. Evaluation of Extreme Rainfall Indices from CHIRPS Precipitation Estimates over the Brazilian Amazonia. Atmos. Res. 2020, 238, 104879. [Google Scholar] [CrossRef]
  20. Ghorbanian, A.; Mohammadzadeh, A.; Jamali, S.; Duan, Z. Performance Evaluation of Six Gridded Precipitation Products throughout Iran Using Ground Observations over the Last Two Decades (2000–2020). Remote Sens. 2022, 14, 3783. [Google Scholar] [CrossRef]
  21. Gado, T.A.; Zamzam, D.H.; Guo, Y.; Zeidan, B.A. Evaluation of Satellite-Based Rainfall Estimates in the Upper Blue Nile Basin. J. Earth Syst. Sci. 2024, 133, 27. [Google Scholar] [CrossRef]
  22. Gulakhmadov, M.; Chen, X.; Gulakhmadov, A.; Umar Nadeem, M.; Gulahmadov, N.; Liu, T. Multiscale Evaluation of Gridded Precipitation Datasets across Varied Elevation Zones in Central Asia’s Hilly Region. Remote Sens. 2023, 15, 4990. [Google Scholar] [CrossRef]
  23. Gharnouki, I.; Aouissi, J.; Benabdallah, S.; Tramblay, Y. Assessing the Variability of Satellite and Reanalysis Rainfall Products over a Semiarid Catchment in Tunisia. Acta Geophys. 2024, 72, 1257–1273. [Google Scholar] [CrossRef]
  24. Perera, H.; Fernando, S.; Gunathilake, M.B.; Sirisena, T.A.J.G.; Rathnayake, U. Evaluation of Satellite Rainfall Products over the Mahaweli River Basin in Sri Lanka. Adv. Meteorol. 2022, 2022, 1926854. [Google Scholar] [CrossRef]
  25. Nicholson, S.E.; Klotter, D.A. Assessing the Reliability of Satellite and Reanalysis Estimates of Rainfall in Equatorial Africa. Remote Sens. 2021, 13, 3609. [Google Scholar] [CrossRef]
  26. Foster, T.; Mieno, T.; Brozović, N. Satellite-Based Monitoring of Irrigation Water Use: Assessing Measurement Errors and Their Implications for Agricultural Water Management Policy. Water Resour. Res. 2020, 56, e2020WR028378. [Google Scholar] [CrossRef]
  27. Shiferaw, B.; Tesfaye, K.; Kassie, M.; Abate, T.; Prasanna, B.M.; Menkir, A. Managing Vulnerability to Drought and Enhancing Livelihood Resilience in Sub-Saharan Africa: Technological, Institutional and Policy Options. Weather. Clim. Extrem. 2014, 3, 67–79. [Google Scholar] [CrossRef]
  28. Rahman, G.; Jung, M.-K.; Kim, T.-W.; Kwon, H.-H. Drought Impact, Vulnerability, Risk Assessment, Management and Mitigation under Climate Change: A Comprehensive Review. KSCE J. Civil. Eng. 2025, 29, 100120. [Google Scholar] [CrossRef]
  29. Wanasinghe, W.M.A.Y.; Gamage, K.H.; Neluwala, N.G.P.B.; Gimhan, P.G.S. Performance of Different Parameterization Configurations of WRF-ARW Model during Heavy Rainfall over Mi Oya River Basin. Eng. J. Inst. Eng. Sri Lanka 2023, 56, 31–41. [Google Scholar] [CrossRef]
  30. Madumali, G.V.H.M.; Manamperi, M.M.S.B. Impact of Water Scarcity on Agriculture in Mi Oya River Basin. TRIVALENT ත්‍රිසංයුජ J. Archaeol. Tour. Anthropol. 2020, 1, 80–89. [Google Scholar] [CrossRef]
  31. Disaster Management Centre. Annual Report 2014; Disaster Management Centre: Colombo, Sri Lanka, 2014. [Google Scholar]
  32. Food and Agriculture Organization of the United Nations. Sri Lanka’s Food Production Hit by Extreme Drought Followed by Floods. Available online: https://www.fao.org/newsroom/detail/Sri-Lanka-s-food-production-hit-by-extreme-drought-followed-by-floods/en (accessed on 20 June 2025).
  33. International Federation of Red Cross and Red Crescent. Revised Emergency Appeal Sri Lanka: Drought; International Federation of Red Cross and Red Crescent: Geneva, Switzerland, 2013. [Google Scholar]
  34. Lu, G.Y.; Wong, D.W. An Adaptive Inverse-Distance Weighting Spatial Interpolation Technique. Comput. Geosci. 2008, 34, 1044–1055. [Google Scholar] [CrossRef]
  35. Chen, F.-W.; Liu, C.-W. Estimation of the Spatial Rainfall Distribution Using Inverse Distance Weighting (IDW) in the Middle of Taiwan. Paddy Water Environ. 2012, 10, 209–222. [Google Scholar] [CrossRef]
  36. Sriwongsitanon, N.; Jandang, W.; Williams, J.; Suwawong, T.; Maekan, E.; Savenije, H.H.G. Using Normalised Difference Infrared Index Patterns to Constrain Semi-Distributed Rainfall–Runoff Models in Tropical Nested Catchments. Hydrol. Earth Syst. Sci. 2023, 27, 2149–2171. [Google Scholar] [CrossRef]
  37. Ahmed, K.; Nawaz, N.; Khan, N.; Rasheed, B.; Baloch, A. Inhomogeneity Detection in the Precipitation Series: Case of Arid Province of Pakistan. Environ. Dev. Sustain. 2021, 23, 7176–7192. [Google Scholar] [CrossRef]
  38. Khalil, A. Inhomogeneity Detection in the Rainfall Series for the Mae Klong River Basin, Thailand. Appl. Water Sci. 2021, 11, 147. [Google Scholar] [CrossRef]
  39. Huffman, G.J.; Stocker, E.F.; Bolvin, D.T.; Nelkin, E.J.; Tan, J. GPM IMERG Final Precipitation L3 1 Day 0.1 Degree x 0.1 Degree V07; Goddard Earth Sciences Data and Information Services Center (GES DISC): Greenbelt, MD, USA, 2023. [Google Scholar]
  40. Li, X.; O, S.; Wang, N.; Liu, L.; Huang, Y. Evaluation of the GPM IMERG V06 Products for Light Rain over Mainland China. Atmos. Res. 2021, 253, 105510. [Google Scholar] [CrossRef]
  41. Kubota, T.; Aonashi, K.; Ushio, T.; Shige, S.; Takayabu, Y.N.; Kachi, M.; Arai, Y.; Tashima, T.; Masaki, T.; Kawamoto, N.; et al. Global Satellite Mapping of Precipitation (GSMaP) Products in the GPM Era. In Satellite Precipitation Measurement. Advances in Global Change Research; Springer: Cham, Switzerland, 2020; pp. 355–373. [Google Scholar] [CrossRef]
  42. Nepal, B.; Shrestha, D.; Sharma, S.; Shrestha, M.S.; Aryal, D.; Shrestha, N. Assessment of GPM-Era Satellite Products’ (IMERG and GSMaP) Ability to Detect Precipitation Extremes over Mountainous Country Nepal. Atmosphere 2021, 12, 254. [Google Scholar] [CrossRef]
  43. Funk, C.; Peterson, P.; Landsfeld, M.; Pedreros, D.; Verdin, J.; Shukla, S.; Husak, G.; Rowland, J.; Harrison, L.; Hoell, A.; et al. The Climate Hazards Infrared Precipitation with Stations—A New Environmental Record for Monitoring Extremes. Sci. Data 2015, 2, 150066. [Google Scholar] [CrossRef]
  44. Nguyen, P.; Shearer, E.J.; Tran, H.; Ombadi, M.; Hayatbini, N.; Palacios, T.; Huynh, P.; Braithwaite, D.; Updegraff, G.; Hsu, K.; et al. The CHRS Data Portal, an Easily Accessible Public Repository for PERSIANN Global Satellite Precipitation Data. Sci. Data 2019, 6, 180296. [Google Scholar] [CrossRef] [PubMed]
  45. Sorooshian, S.; Nguyen, P.; Sellars, S.; Braithwaite, D.; AghaKouchak, A.; Hsu, K. Satellite-Based Remote Sensing Estimation of Precipitation for Early Warning Systems. In Extreme Natural Hazards, Disaster Risks and Societal Implications; Cambridge University Press: Cambridge, UK, 2014; pp. 99–112. [Google Scholar]
  46. Ashouri, H.; Hsu, K.L.; Sorooshian, S.; Braithwaite, D.K.; Knapp, K.R.; Cecil, L.D.; Nelson, B.R.; Prat, O.P. PERSIANN-CDR: Daily Precipitation Climate Data Record from Multisatellite Observations for Hydrological and Climate Studies. Bull. Am. Meteorol. Soc. 2015, 96, 69–83. [Google Scholar] [CrossRef]
  47. Kumar, S.; Amarnath, G.; Ghosh, S.; Park, E.; Baghel, T.; Wang, J.; Pramanik, M.; Belbase, D. Assessing the Performance of the Satellite-Based Precipitation Products (SPP) in the Data-Sparse Himalayan Terrain. Remote Sens. 2022, 14, 4810. [Google Scholar] [CrossRef]
  48. Zhang, Y.; Li, Y.; Ji, X.; Luo, X.; Li, X. Evaluation and Hydrologic Validation of Three Satellite-Based Precipitation Products in the Upper Catchment of the Red River Basin, China. Remote Sens. 2018, 10, 1881. [Google Scholar] [CrossRef]
  49. Zhu, Q.; Luo, Y.; Zhou, D.; Xu, Y.-P.; Wang, G.; Gao, H. Drought Monitoring Utility Using Satellite-Based Precipitation Products over the Xiang River Basin in China. Remote Sens. 2019, 11, 1483. [Google Scholar] [CrossRef]
  50. Wang, J.; Petersen, W.A.; Wolff, D.B. Validation of Satellite-Based Precipitation Products from TRMM to GPM. Remote Sens. 2021, 13, 1745. [Google Scholar] [CrossRef]
  51. Wang, X.; Ding, Y.; Zhao, C.; Wang, J. Validation of TRMM 3B42V7 Rainfall Product under Complex Topographic and Climatic Conditions over Hexi Region in the Northwest Arid Region of China. Water 2018, 10, 1006. [Google Scholar] [CrossRef]
  52. Alahacoon, N.; Edirisinghe, M.; Ranagalage, M. Satellite-Based Meteorological and Agricultural Drought Monitoring for Agricultural Sustainability in Sri Lanka. Sustainability 2021, 13, 3427. [Google Scholar] [CrossRef]
  53. Jafari, S.M.; Nikoo, M.R.; Dehghani, M.; Alijanian, M. Evaluation of Two Satellite-Based Products against Ground-Based Observation for Drought Analysis in the Southern Part of Iran. Nat. Hazards 2020, 102, 1249–1267. [Google Scholar] [CrossRef]
  54. Sirisena, J.; Augustijn, D.; Nazeer, A.; Bamunawala, J. Use of Remote-Sensing-Based Global Products for Agricultural Drought Assessment in the Narmada Basin, India. Sustainability 2022, 14, 13050. [Google Scholar] [CrossRef]
  55. Senatilleke, U.; Sirisena, J.; Gunathilake, M.B.; Muttil, N.; Rathnayake, U. Monitoring the Meteorological and Hydrological Droughts in the Largest River Basin (Mahaweli River) in Sri Lanka. Climate 2023, 11, 57. [Google Scholar] [CrossRef]
  56. Guttman, N.B. Comparing the Palmer Drought Index and the Standardized Precipitation Index. JAWRA J. Am. Water Resour. Assoc. 1998, 34, 113–121. [Google Scholar] [CrossRef]
  57. Abeysingha, N.S.; Rajapaksha, U.R.L.N. SPI-Based Spatiotemporal Drought over Sri Lanka. Adv. Meteorol. 2020, 2020, 9753279. [Google Scholar] [CrossRef]
  58. Yuan, X.; Wood, E.F. Multimodel Seasonal Forecasting of Global Drought Onset. Geophys. Res. Lett. 2013, 40, 4900–4905. [Google Scholar] [CrossRef]
  59. Ma, F.; Yuan, X.; Ye, A. Seasonal Drought Predictability and Forecast Skill over China. J. Geophys. Res. Atmos. 2015, 120, 8264–8275. [Google Scholar] [CrossRef]
  60. Arshad, M.; Ma, X.; Yin, J.; Ullah, W.; Ali, G.; Ullah, S.; Liu, M.; Shahzaman, M.; Ullah, I. Evaluation of GPM-IMERG and TRMM-3B42 Precipitation Products over Pakistan. Atmos. Res. 2021, 249, 105341. [Google Scholar] [CrossRef]
  61. Ghimire, U.; Akhtar, T.; Shrestha, N.K.; Paul, P.K.; Schürz, C.; Srinivasan, R.; Daggupati, P. A Long-Term Global Comparison of IMERG and CFSR with Surface Precipitation Stations. Water Resour. Manag. 2022, 36, 5695–5709. [Google Scholar] [CrossRef]
  62. Disaster Management Centre. Historical Disaster Information System in Sri Lanka: Preliminary Analysis; Disaster Management Centre: Colombo, Sri Lanka, 2007. [Google Scholar]
  63. Somasundaram, D.; Zhu, J.; Zhang, Y.; Nie, Y.; Zhang, Z.; Yu, L. Drought Characteristics and Drought-Induced Effects on Vegetation in Sri Lanka. Climate 2024, 12, 172. [Google Scholar] [CrossRef]
  64. Food and Agriculture Organization. GIEWS—Global Information and Early Warning System on Food and Agriculture: Special Alert No. 269—Sri Lanka—25 July 1996; Food and Agriculture Organization of United Nations: Rome, Italy, 1996. [Google Scholar]
  65. UNCCD. National Report on Desertification/Land Degradation in Sri Lanka; UNCCD: Bonn, Germany, 2000. [Google Scholar]
Figure 1. Study area: Mi Oya Basin, Sri Lanka. Note: US_Basin: Upstream Basin; Mid_Basin1: Middle-Stream Basin 1; Mid_Basin2: Middle-Stream Basin 2; and DS_Basin: Downstream Basin.
Figure 1. Study area: Mi Oya Basin, Sri Lanka. Note: US_Basin: Upstream Basin; Mid_Basin1: Middle-Stream Basin 1; Mid_Basin2: Middle-Stream Basin 2; and DS_Basin: Downstream Basin.
Hydrology 12 00166 g001
Figure 2. Overall methodology flowchart.
Figure 2. Overall methodology flowchart.
Hydrology 12 00166 g002
Figure 3. Normalized Root Mean Square Error (NRMSE, (a)), Pearson Correlation Coefficient (CC, (b)) of monthly satellite rainfall data for the period 2003–2022.
Figure 3. Normalized Root Mean Square Error (NRMSE, (a)), Pearson Correlation Coefficient (CC, (b)) of monthly satellite rainfall data for the period 2003–2022.
Hydrology 12 00166 g003
Figure 4. Performance of seasonal rainfall over the Mi Oya Basin: (a) Normalized Root-Mean Square Error, (b) Correlation Coefficient, (c) Nash–Sutcliffe Efficiency, and (d) Percentage of Bias for the analysis period (2003–2022). Note: FIM—First Inter-Monsoon, NEM—Northeast Monsoon, SIM—Second Inter-Monsoon, and SWM—Southwest Monsoon.
Figure 4. Performance of seasonal rainfall over the Mi Oya Basin: (a) Normalized Root-Mean Square Error, (b) Correlation Coefficient, (c) Nash–Sutcliffe Efficiency, and (d) Percentage of Bias for the analysis period (2003–2022). Note: FIM—First Inter-Monsoon, NEM—Northeast Monsoon, SIM—Second Inter-Monsoon, and SWM—Southwest Monsoon.
Hydrology 12 00166 g004
Figure 5. SPI-6 for (a) Upstream Basin, (b) Middle-Stream Basin 1, (c) Middle-Stream Basin 2, and (d) Downstream Basin of the Mi Oya Basin. SPIs are derived from IMERG and CHIRPS data starting from 1998 and 1981, respectively.
Figure 5. SPI-6 for (a) Upstream Basin, (b) Middle-Stream Basin 1, (c) Middle-Stream Basin 2, and (d) Downstream Basin of the Mi Oya Basin. SPIs are derived from IMERG and CHIRPS data starting from 1998 and 1981, respectively.
Hydrology 12 00166 g005
Figure 6. SPI-12 for (a) Upstream Basin, (b) Middle-Stream Basin 1, (c) Middle-Stream Basin 2, and (d) Downstream Basin of the Mi Oya Basin. SPIs are derived from IMERG and CHIRPS data starting from 1998 and 1981, respectively.
Figure 6. SPI-12 for (a) Upstream Basin, (b) Middle-Stream Basin 1, (c) Middle-Stream Basin 2, and (d) Downstream Basin of the Mi Oya Basin. SPIs are derived from IMERG and CHIRPS data starting from 1998 and 1981, respectively.
Hydrology 12 00166 g006
Table 1. Summary of gauge data (2003–2022).
Table 1. Summary of gauge data (2003–2022).
Rain Gauge
Station
LocationAnnual
Rainfall (mm)
Maximum Daily Rainfall (mm)Daily Mean ± Std (mm)
LatitudeLongitude
Anamaduwa7.87880.01115733544 ± 15
Ataragalla7.92780.28713032194 ± 13
Kottukachchiya7.93979.94814262744 ± 14
Mediyawa7.88380.28516082444 ± 14
Puttalam8.03179.83113198314 ± 15
Palawi7.97979.84512792984 ± 12
Thabbowa8.08679.92813382854 ± 13
Table 2. Overview of satellite-based global rainfall products used in this study.
Table 2. Overview of satellite-based global rainfall products used in this study.
ProductSpatial ResolutionSpatial CoverageTemporal ResolutionTemporal Coverage
IMERG0.1°90° N–90° S30 min/ daily1998—NRT
GSMaP0.1°60° N–60° Shourly2003—NRT
CHIRPS0.05°50° N–50° SDaily1981—NRT
PERSIANN0.25°60° N–60° SDaily *2000—NRT
PERSIANN-CDR0.25°60° N–60° SDaily *1983—NRT
Note: NRT: Near-real-time and *—finer resolution data is available in the new versions.
Table 3. Contingency table for categorical indices.
Table 3. Contingency table for categorical indices.
Satellite DataGauge Observation
RainNo-Rain
RainH- HitF = False Detection
No rainM- MissCorrect No Rain
Table 4. Detection accuracy of light, moderate, and heavy rainfall at seven stations for the 2003–2022 period. Note: The best performances are highlighted in bold.
Table 4. Detection accuracy of light, moderate, and heavy rainfall at seven stations for the 2003–2022 period. Note: The best performances are highlighted in bold.
StationIMERGGSMaPCHIRPSPERSIANNPERSIANN CDR
PODFARCSIPODFARCSIPODFARCSIPODFARCSIPODFARCSI
For Light Rainfall
Anamaduwa0.830.780.210.560.660.270.570.690.250.600.730.230.850.770.22
Ataragalla0.750.900.100.670.800.180.610.830.160.650.850.140.830.890.11
Kottukachchiya0.780.850.150.600.740.220.570.770.200.580.810.170.850.830.16
Madiyawa0.810.790.200.640.660.290.570.710.240.600.750.220.850.790.21
Palawi0.810.740.240.570.550.340.560.610.300.590.670.270.860.720.27
Puttalam0.830.680.300.590.460.390.550.530.340.580.590.310.840.660.32
Thabbowa0.830.780.210.620.620.310.570.670.260.600.720.240.850.770.22
For Moderate Rainfall
Anamaduwa0.390.620.240.290.690.180.200.690.140.350.760.170.150.700.11
Ataragalla0.380.700.200.300.680.180.190.760.120.400.760.180.150.780.10
Kottukachchiya0.380.660.220.330.670.200.170.740.120.370.750.180.150.680.11
Madiyawa0.360.640.220.290.580.210.160.670.120.340.740.170.160.670.12
Palawi0.430.680.220.270.730.150.200.700.140.370.770.160.150.720.11
Puttalam0.410.620.250.310.670.190.230.620.170.370.750.170.140.700.11
Thabbowa0.410.600.250.320.660.200.260.660.170.350.750.170.140.690.11
For Heavy Rainfall
Anamaduwa0.200.710.130.090.910.050N/A00.090.800.070N/A0
Ataragalla0.060.750.050.001.000.000N/A00.060.800.050N/A0
Kottukachchiya0.190.750.120.060.920.030N/A00.120.330.110N/A0
Madiyawa0.050.880.040.100.780.070N/A00.001.000.000N/A0
Palawi0.170.880.080.130.920.050N/A00.130.670.100N/A0
Puttalam0.200.830.100.001.000.000N/A00.180.500.150N/A0
Thabbowa0.380.730.190.200.850.100N/A00.200.500.170N/A0
Table 5. Performance of satellite-based rainfall products over the Mi Oya Basin at the annual scale.
Table 5. Performance of satellite-based rainfall products over the Mi Oya Basin at the annual scale.
Rainfall ProductNRMSECCNSEPBIAS (%)
IMERG0.200.920.2217.70
GSMaP0.120.890.74−0.04
CHIRPS 0.150.780.487.28
PERSIANN0.490.58−3.2744.18
PERSIANN-CDR0.240.77−0.2219.13
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

Gayanthika, H.; Lakshitha, D.; Chathuranga, M.; De Silva, G.; Sirisena, J. Comparative Analysis of Satellite-Based Rainfall Products for Drought Assessment in a Data-Poor Region. Hydrology 2025, 12, 166. https://doi.org/10.3390/hydrology12070166

AMA Style

Gayanthika H, Lakshitha D, Chathuranga M, De Silva G, Sirisena J. Comparative Analysis of Satellite-Based Rainfall Products for Drought Assessment in a Data-Poor Region. Hydrology. 2025; 12(7):166. https://doi.org/10.3390/hydrology12070166

Chicago/Turabian Style

Gayanthika, Hansini, Dimuthu Lakshitha, Manthika Chathuranga, Gouri De Silva, and Jeewanthi Sirisena. 2025. "Comparative Analysis of Satellite-Based Rainfall Products for Drought Assessment in a Data-Poor Region" Hydrology 12, no. 7: 166. https://doi.org/10.3390/hydrology12070166

APA Style

Gayanthika, H., Lakshitha, D., Chathuranga, M., De Silva, G., & Sirisena, J. (2025). Comparative Analysis of Satellite-Based Rainfall Products for Drought Assessment in a Data-Poor Region. Hydrology, 12(7), 166. https://doi.org/10.3390/hydrology12070166

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