Abstract
The evaluation of open-source precipitation data is crucial to enable the selection of the most appropriate product for a specific research. This study aims to evaluate the capability of the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) with a spatial resolution of 0.05° for estimating monthly and annual precipitation in the Wala basin, Jordan, from 1987 to 2017 using a point-to-pixel comparison approach. Eleven precipitation extreme indices, recommended by the Expert Team on Climate Change Detection and Indices (ETCCDI), were used in this study. The findings show that CHIRPS correlated moderately with stations in monthly precipitation estimation, with the Pearson correlation coefficient values ranging from 0.50 to 0.73. However, CHIRPS had low correlations with stations in most of the extreme indices, except PRCPTOT, R10mm, and R20mm. The CHIRPS, particularly in the extreme years, overestimated low precipitation amounts and underestimated high ones. Moreover, CHIRPS underestimated the calculation of consecutive dry days, consecutive wet days, R10mm, R20mm, and R30mm, while an overestimation was found for the R95p, R99p, and Rx1day. The trend analysis and Wilcox text showed a lack of resemblance between the CHIRPS and gauges, showing a bias correction is needed before applying an extreme analysis in this region.
HIGHLIGHTS
CHIRPS captured monthly precipitation well over the Wala Basin, Jordan.
CHIRPS underestimated high precipitation amounts and overestimated low ones.
CHIRPS is not suitable for calculating some extreme indices.
A bias correction is needed before applying CHIRPS for extreme analysis.
INTRODUCTION
Climate change is a global phenomenon of paramount importance due to its interrelationship and impact on the various components of ecosystems. Many studies have indicated tangible changes in temperatures. Cold events where the temperature is less than −15 °C have increased in the northwest of Iran (Aalijahan et al. 2019; Aalijahan et al. 2022). In contrast, increased temperatures have been observed over Bajestan in the north-eastern part of Iran (Khosravichenar et al. 2023). Increases in temperatures and extreme heat trends in the Middle East and North Africa regions showed that extreme heat events were attributed to global warming (Rezaei et al. 2023). A decreasing trend in the discharge of the Aras River was related to carbon dioxide and methane emissions (Aalijahan et al. 2021). In fact, previous research indicated that carbon emissions from crops, livestock, and industrial sectors are one of the reasons for climate change and extreme weather events (Abbas et al. 2022b, 2022c; Elahi et al. 2022, 2024). It is worth noting that global warming and climate change have a significant impact on precipitation. For example, precipitation trend analysis for 15 rainfall stations in the Marmara region (Türkiye) showed that precipitation did not follow a specific trend. Downward trends were observed in three stations, and there was only a significant decline in precipitation during the early 1990s (Aalijahan et al. 2023). A projection study of future precipitation trends in Pakistan using the Coupled Model Intercomparison Project Phase 6 (CIMP6) showed that the annual precipitation will increase under the Shared Socio-Economic Scenarios SSP3-7.0 and SSP5-8.5 (Abbas et al. 2022a). Also, SSP5-8.5 pointed to the probability of increasing precipitation extremes during the second half of the 21st century (Abbas et al. 2023).
Precipitation is a crucial component for studying agriculture, hydrology, climate, water resource management, hydraulic structure design, disaster, and ecological modelling (Cavalcante et al. 2020). In arid and semi-arid regions, producing continuous time series precipitation data might be difficult due to a lack of observation gauges and maintenance (Alsilibe et al. 2023). Furthermore, places with sparse rain gauge networks often perform poorly in capturing precipitation variability over a large area. Therefore, satellite technology plays an important role in providing long-term precipitation data over any region in the world.
Open-source gridded precipitation products have become available in recent years. These products are designed to address different objectives and include different data sources, geographical and temporal resolutions, spatial coverage, and latencies (Beck et al. 2017). The Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS), a well-known blended gauge-satellite precipitation product, spans the 50°S to 50°N latitude area from 1981 to the present. CHIRPS has low latency, temporal resolutions ranging from daily to yearly, and spatial resolutions of 0.25° and 0.05° (Funk et al. 2015). A critical step in selecting the optimum precipitation product for a particular research or operational application is to evaluate these open-source precipitation datasets. Many research studies evaluated the advantages and disadvantages of the available datasets. Even some precipitation products employ the same data sources, and there are noticeable differences.
Studies on the validation of open-source gridded precipitation products have been conducted in different regions of the globe, including Brazil (Cavalcante et al. 2020; Mu et al. 2021), Colombia (Ocampo-Marulanda et al. 2022), Mali (Fofana et al. 2022), Ghana (Larbi et al. 2018), India (Prakash 2019; Shen et al. 2020), West Africa (Sacré Regis et al. 2020), China (Zhang et al. 2022), Malaysia (Tan et al. 2023), and Argentina (Rivera et al. 2018). These studies typically compare open-source gridded precipitation products with ground observations at different temporal scales while applying various further applications such as streamflow simulation, extreme precipitation detection, and drought monitoring. Some studies have reported that CHIRPS can work reasonably and outperform other datasets (Katsanos et al. 2016).
In Iran, CHIRPS performed well in the highlands but underestimated precipitation during the wet season (Saeidizand et al. 2018), making it less reliable for estimating extreme precipitation indices (Keikhosravi-Kiany et al. 2023). In comparison to the dry months, CHIRPS performs better during the rainy season in Syria (Alsilibe et al. 2023). In Cyprus, there is a significant correlation between CHIRPS and observed precipitation (Katsanos et al. 2016). CHIRPS performed well when estimating precipitation less than 1 mm/day in Egypt, which represented 30% of wet days throughout the study period (Nashwan et al. 2020). To evaluate the suitability of five sets of satellite data as an alternative to rain gauges, including CHIRPS, in the arid region, Saudi Arabia was selected as a representative of the arid regions (Helmi & Abdelhamed 2023). CHIRPS was regarded as one of the best alternatives to run hydrological modelling in this region. However, the performance of CHIRPS varies across the geographical and climate conditions.
The Mujib Basin in Jordan is a crucial groundwater basin for supplying fresh water to local populations. CHIRPS has been proven to be useful for estimating monthly precipitation, with a good precipitation detection capability (Alsalal et al. 2023). However, the effectiveness of CHIRPS in estimating extreme precipitation over the Wala Basin within the Mujib Basin remains unclear. Therefore, this study aims to evaluate the performance of CHIRPS over the Wala basin from 1987 to 2017 using nine observation stations from three aspects: (1) annual and monthly scales, (2) extreme indices, and (3) trend analysis. In addition, as part of the quality control, this study also determines the homogeneity status of gauge observations. It advances knowledge of the climate extremes changes at the local level. Moreover, the findings contribute to decision-making in the usage of CHIRPS data for climate change and water resources management in arid regions. This work also contributes to our understanding of localized climatic extremes.
STUDY AREA
Most of the precipitation days occur in January and February with high spatial variabilities from north to south and from west to east. Therefore, the Wala basin has more than one climate zone that stretches over a relief. The highest elevation is 968 m above mean sea level and the lowest is 440 m above and below mean sea level, respectively. Precipitation in the Wala basin decreases from over 500 mm/year to less than 100 mm/year as it moves from the northwest to the southeast.
DATA AND METHODOLOGIES
Observed precipitation and quality control
The Jordanian Ministry of Water and Irrigation (MWI) provided observed daily precipitation data from 10 stations distributed over the Wala basin from 1987 to 2017. The month of January had the most monthly precipitation, whereas June through September had no precipitation. In contrast to the central and eastern parts of the basin, where monthly precipitation averages less than 40 mm for all months, the majority of the stations in the northern and western part of the basin experiences monthly rainfall between 60 and 100 mm during the winter months of December, January, and February.
A preliminary data quality check was performed to eliminate stations with lower reliability in terms of missing values and homogeneity status. An observational climate data time series mean may vary due to station relocation, land use changes, equipment adjustments, and observational issues. These influences may lead to homogeneity issues and first-order autoregressive errors. Four homogeneity tests including the Pettitt, Buishand, SNHT, and von Neumann ratio test were selected for the homogeneous analysis. The four tests are included in the XLSTAT time series analysis module, which is a Microsoft Excel add-in that can be used to identify the homogenous trend of the daily precipitation data for each station (Tan et al. 2019).
CHIRPS
The CHIRPS data from 1987 to 2017 was used in this study. The Climate Hazards Group of the University of California developed the CHIRPS version 2 rainfall data using infrared precipitation with station data. CHIRPS is a quasi-global rainfall dataset that spans from 1981 to almost the present and covers 50° S to 50° N. It combines 0.05° × 0.05° resolution satellite imagery with in situ station data to construct gridded rainfall time series suited for trend analysis and seasonal drought monitoring (Funk et al. 2015).
Extreme indices
The precipitation extreme indices were selected from the set established by the Expert Team on Climate Change Detection and Indices (ETCCDI) as listed in Table 1. The PRCPTOT, R95p, R99p, RX1d, RX5d, and SDII indices are indicated by mm depth. The consecutive dry days (CDDs), consecutive wet days (CWDs), R10mm, R20mm, and R30mm are indicated by the number of days. These precipitation extremes have been widely applied to the assessment of climate change and the study of water-related disasters such as floods and drought (Tan et al. 2019). The user-defined daily violent precipitation threshold was set to 30 mm/day based on the arid condition (Larbi et al. 2018). The RClimDex tool was utilized to measure precipitation extreme indices trends for each station (precipitation station data and CHIRPS). The RClimDex tool incorporates Mann–Kendall and Sen's slope tests for the trend evaluations. A p-value <0.05 was considered to reject the null hypothesis of no trend and accept the alternative hypothesis of a significant trend.
Index . | Name . | Detail . | Unit . |
---|---|---|---|
SDII | Simple daily intensity index | Annual total precipitation divided by the number of wet days in the year | mm/days |
R10mm | Number of heavy precipitation days | Annual count of days when precipitation ≥10 mm | Days |
R20mm | Number of very heavy precipitation days | Annual count of days when precipitation ≥20 m | Days |
R30mm | Number of violent precipitation days | Annual count of days when precipitation ≥30 mm | Days |
CDD | Consecutive dry days | Maximum number of consecutive days with rainfall < 1 mm | Days |
CWD | Consecutive wet days | Maximum number of consecutive days with rainfall ≥ 1 mm | Days |
R95p | Very wet days | Annual total precipitation when rainfall > 95th percentile | mm |
R99p | Extremely wet days | Annual total precipitation when rainfall > 99th percentile | mm |
PRCPTOT | Annual total wet-day precipitation | Annual total precipitation on wet days (rainfall > 1 mm) | mm |
Rx1day | Max 1-day precipitation amount | Annual or Monthly maximum 1-day precipitation | mm |
Rx5day | Max 5-day precipitation amount | Annual or monthly maximum 5-day precipitation | mm |
Index . | Name . | Detail . | Unit . |
---|---|---|---|
SDII | Simple daily intensity index | Annual total precipitation divided by the number of wet days in the year | mm/days |
R10mm | Number of heavy precipitation days | Annual count of days when precipitation ≥10 mm | Days |
R20mm | Number of very heavy precipitation days | Annual count of days when precipitation ≥20 m | Days |
R30mm | Number of violent precipitation days | Annual count of days when precipitation ≥30 mm | Days |
CDD | Consecutive dry days | Maximum number of consecutive days with rainfall < 1 mm | Days |
CWD | Consecutive wet days | Maximum number of consecutive days with rainfall ≥ 1 mm | Days |
R95p | Very wet days | Annual total precipitation when rainfall > 95th percentile | mm |
R99p | Extremely wet days | Annual total precipitation when rainfall > 99th percentile | mm |
PRCPTOT | Annual total wet-day precipitation | Annual total precipitation on wet days (rainfall > 1 mm) | mm |
Rx1day | Max 1-day precipitation amount | Annual or Monthly maximum 1-day precipitation | mm |
Rx5day | Max 5-day precipitation amount | Annual or monthly maximum 5-day precipitation | mm |
Data comparison
A point-to-pixel technique was employed to compare the observed precipitation in the Wala basin with CHIRPS to avoid errors caused by the spatial interpolation of poorly spaced and unevenly distributed climate stations. This approach has been widely utilized to evaluate the reliability of open-source gridded precipitation products around the world (Baez-Villanueva et al. 2018). Daily rainfall was extracted over the pixel in which each rainfall station is located to construct the CHIRPS rainfall data series. For the two data series (stations and CHIRPS), the monthly and annual scales were considered. The percentage bias (PBIAS), Pearson correlation coefficient (CC), the root-mean-square error (RMSE), and the Nash–Sutcliffe efficiency coefficient (NSE) were used to assess the effectiveness of the CHIRPS product in estimating monthly rainfall and rainfall indices. PBIAS determines how likely it is on average for the simulated values to differ from the observed values. PBIAS should be ideally at 0.0, with low magnification values denoting precise model simulation. CC measures the strength of the relationship between two variables. The CC values range is −1 to −1. CC value of 1 indicates perfect positive correlation, and CC value of −1 indicates perfect negative correlation. RMSE measures the average difference between predicted and observed values. The RMSE of zero indicates a perfect fit of the predicted to the observed data. The NSE is one minus the ratio between the variance of the predicted to that of the observed. The NSE value of 1 indicates a perfect fit, and the NSE value of zero indicates that the model performance is the same as the mean of the observed data, while NSE values less than zero to minus infinity indicate poor model performance. In addition, the Wilcox test was also applied to compare monthly data from rainfall stations and CHIRPS. A statistical significance level of 0.05 was taken into account for all tests.
RESULTS AND DISCUSSION
Homogenous analysis
All stations passed the four homogeneity tests, except Dhab'a Nursery station (CD0015), which passed only the von Neumann ratio test (Table 2). As the computed p-value is lower than the significance level of 0.05, we should reject the null hypothesis (data are homogeneous) and accept the alternative hypothesis (there is a change in the data). The station was ultimately eliminated due to both the high number of missing values and the inhomogeneity of the data. Nine stations were selected for this study after passing the four homogeneity criteria. This study made use of historical rainfall data spanning from 1987 to 2017. Since the summer months of June, July, August, and September are entirely dry, the CHIRPS evaluation was focused on the remaining months.
Test . | AN0003 . | CC0001 . | CC0004 . | CD0003 . | CD0005 . | CD0006 . | CD0007 . | CD0017 . | CD0028 . | CD0015 . |
---|---|---|---|---|---|---|---|---|---|---|
Pettitt | 0.501 | 0.556 | 0.440 | 0.680 | 0.798 | 0.689 | 0.255 | 0.404 | 0.713 | 0.021* |
SNHT test | 0.532 | 0.687 | 0.667 | 0.813 | 0.124 | 0.963 | 0.425 | 0.758 | 0.504 | 0.005* |
Buishand | 0.463 | 0.339 | 0.277 | 0.676 | 0.799 | 0.528 | 0.186 | 0.355 | 0.774 | 0.003* |
von Neumann | 0.191 | 0.664 | 0.719 | 0.694 | 0.126 | 0.510 | 0.610 | 0.514 | 0.204 | 0.211 |
Test . | AN0003 . | CC0001 . | CC0004 . | CD0003 . | CD0005 . | CD0006 . | CD0007 . | CD0017 . | CD0028 . | CD0015 . |
---|---|---|---|---|---|---|---|---|---|---|
Pettitt | 0.501 | 0.556 | 0.440 | 0.680 | 0.798 | 0.689 | 0.255 | 0.404 | 0.713 | 0.021* |
SNHT test | 0.532 | 0.687 | 0.667 | 0.813 | 0.124 | 0.963 | 0.425 | 0.758 | 0.504 | 0.005* |
Buishand | 0.463 | 0.339 | 0.277 | 0.676 | 0.799 | 0.528 | 0.186 | 0.355 | 0.774 | 0.003* |
von Neumann | 0.191 | 0.664 | 0.719 | 0.694 | 0.126 | 0.510 | 0.610 | 0.514 | 0.204 | 0.211 |
*Represents significance level at 0.05, when the p-value suppose less than 0.05.
Monthly precipitation assessment
The spatial distribution of CC for monthly precipitation between observed and CHIRPS estimates is shown in Figure 3. The correlation values of April and November of 0.8 were higher than January (CC = 0.72) and February (CC = 0.65). The CC values of March, October, and December were 0.57, 0.52, and 0.57 respectively, while May has the minimum CC value of 0.37. These results were consistent with previous findings that showed that the correlations between observed and CHIRPS were stronger for the months of spring and autumn months of low rainfall than for winter months characterized by intense rainfall (Saeidizand et al. 2018).
It should be noted that CC indicated association and not necessarily resemblance or predictability. The NSE, which better tracks the data overestimation or underestimation, indicated a poor performance of the CHIRPS monthly; hence, all NSE values were below 0.5. Even two negative NSE values were calculated, which indicates that the mean values represent a better prediction of the station data than CHIRPS data. The shortcomings of CHIRPS were highlighted by previous studies. López-Bermeo et al. (2022) indicated that CHIRPS for tropical mountainous regions in South America has limited use because of the low accuracy in daily values.
The non-parametric Wilcox test was performed to test the resemblance between the station and CHIRPS data (Table 3). The CD0017 was the only null hypothesis that was rejected during all months, indicating a lack of resemblance at this station. The p-value >0.05 indicated a similarity between CHIRPS and station data; however, significant differences were observed during all months at every station. For May, p-values <0.05 were observed for all stations except at CD0028, while during October, p-values >0.05 were dominant except at CC0001 and CD0017.
. | Jan . | Feb . | Mar . | Apr . | May . | Oct . | Nov . | Dec . |
---|---|---|---|---|---|---|---|---|
AN0003 | 0.124* | 0.021 | 0.803* | 0.791* | 0.023 | 0.622* | 0.934* | 0.861* |
CC0001 | 0.011 | 0.039 | 0.043 | 0.713* | 0.009 | 0.048 | 0.058* | 0.931* |
CC0004 | 0.041 | 0.638* | 0.057* | 0.379* | 0.010 | 0.871* | 0.224* | 0.636* |
CD0003 | 1.000* | 0.144* | 0.003 | 0.023 | 0.006 | 0.896* | 0.053* | 0.041 |
CD0005 | 0.000 | 0.001 | 0.000 | 0.124* | 0.010 | 0.444* | 0.004 | 0.000 |
CD0006 | 0.017 | 0.184* | 0.012 | 0.000 | 0.020 | 0.636* | 0.232* | 0.256* |
CD0007 | 0.692* | 0.692* | 0.071* | 0.003 | 0.012 | 0.276* | 0.021 | 0.141* |
CD0017 | 0.007 | 0.018 | 0.000 | 0.003 | 0.017 | 0.029 | 0.000 | 0.000 |
CD0028 | 0.379* | 0.818* | 0.001 | 0.065* | 0.075* | 0.888* | 0.029 | 0.014 |
. | Jan . | Feb . | Mar . | Apr . | May . | Oct . | Nov . | Dec . |
---|---|---|---|---|---|---|---|---|
AN0003 | 0.124* | 0.021 | 0.803* | 0.791* | 0.023 | 0.622* | 0.934* | 0.861* |
CC0001 | 0.011 | 0.039 | 0.043 | 0.713* | 0.009 | 0.048 | 0.058* | 0.931* |
CC0004 | 0.041 | 0.638* | 0.057* | 0.379* | 0.010 | 0.871* | 0.224* | 0.636* |
CD0003 | 1.000* | 0.144* | 0.003 | 0.023 | 0.006 | 0.896* | 0.053* | 0.041 |
CD0005 | 0.000 | 0.001 | 0.000 | 0.124* | 0.010 | 0.444* | 0.004 | 0.000 |
CD0006 | 0.017 | 0.184* | 0.012 | 0.000 | 0.020 | 0.636* | 0.232* | 0.256* |
CD0007 | 0.692* | 0.692* | 0.071* | 0.003 | 0.012 | 0.276* | 0.021 | 0.141* |
CD0017 | 0.007 | 0.018 | 0.000 | 0.003 | 0.017 | 0.029 | 0.000 | 0.000 |
CD0028 | 0.379* | 0.818* | 0.001 | 0.065* | 0.075* | 0.888* | 0.029 | 0.014 |
Note: A p-value > 0.05 indicates a non-significant difference and the null hypothesis cannot be rejected, while a p-value < 0.05 indicates significant differences and the rejection of the null hypothesis.
*Represents significance level at 0.05, when the p-value suppose less than 0.05.
The usefulness of CHIRPS depends on its ability to correctly estimate monthly precipitation. Previous research produced mixed results concerning the suitability of CHIRPS for extreme analysis. CHIRPS performed reasonably well in Pakistan and southeastern Iran (Nawaz et al. 2021; Mianabadi et al. 2022). However, discrepancies between CHIRPS and observed data sets were reported in other studies. López-Bermeo et al. (2022) cautioned against the use of the monthly CHIRPS in evaluating the risk of drought and floods because they found significant overestimation and underestimation of magnitudes in the month, even though the monthly Pearson's correlation coefficients between CHIRPS and the observed data were in the range of 0.52–0.86. Evaluating the CHIRPS dataset in a coastal region showed that it yielded incorrect estimates, which were attributed to local factors such as relief, vegetation, and the atmospheric precipitation system (Arregocés et al. 2023). The CHIRPS data did not perform well in the semi-arid region in Brazil (Paredes-Trejo et al. 2017). In an arid region characterized by low rainfall of 100–200 mm, the comparison between CHIRPS and Satellite precipitation datasets in general yielded high uncertainties (Helmi & Abdelhamed 2023). As a result, the Wilcox test comparison of the CHIRPS and station data for the Wala station revealed that the null hypothesis cannot be accepted for all months, suggesting that CHIRPS differs for some wet months like December, January, and February.
Annual precipitation assessment
Statistical . | AN0003 . | CC0001 . | CC0004 . | CD0003 . | CD0005 . | CD0006 . | CD0007 . | CD0017 . | CD0028 . |
---|---|---|---|---|---|---|---|---|---|
Monthly scale | |||||||||
PBIAS | −15% | −17% | −5% | 18% | 45% | 4.4% | −8% | 54% | 7% |
Pearson correlation coefficient (CC) | 0.64 | 0.70 | 0.73 | 0.58 | 0.63 | 0.55 | 0.67 | 0.50 | 0.61 |
RMSE | 7.52 | 8.67 | 5.45 | 4.03 | 7.52 | 6.59 | 4.37 | 9.41 | 5.39 |
NSE | 0.28 | 0.32 | 0.48 | 0.04 | −2.49 | 0.22 | 0.37 | −2.03 | 0.18 |
Annual scale | |||||||||
PBIAS | −7% | −11% | −3% | 21% | 51% | 23% | 7% | 53% | 19% |
Pearson correlation coefficient (CC) | 0.64 | 0.73 | 0.74 | 0.53 | 0.35 | 0.53 | 0.63 | 0.17 | 0.48 |
Mean annual precipitation (mm) station | 360.9 | 302.4 | 342.1 | 147.9 | 126.9 | 212.3 | 240.0 | 120.5 | 209.4 |
Mean annual precipitation (mm) CHIRPS | 311.1 | 258.4 | 318.2 | 173.2 | 205.2 | 235.5 | 238.2 | 194.4 | 231.6 |
Standard deviation station | 150.27 | 110.65 | 116.36 | 55.05 | 52.55 | 107.10 | 99.82 | 57.11 | 98.49 |
Standard deviation CHIRPS | 67.47 | 53.10 | 67.70 | 39.01 | 47.93 | 45.91 | 46.05 | 42.83 | 47.32 |
Statistical . | AN0003 . | CC0001 . | CC0004 . | CD0003 . | CD0005 . | CD0006 . | CD0007 . | CD0017 . | CD0028 . |
---|---|---|---|---|---|---|---|---|---|
Monthly scale | |||||||||
PBIAS | −15% | −17% | −5% | 18% | 45% | 4.4% | −8% | 54% | 7% |
Pearson correlation coefficient (CC) | 0.64 | 0.70 | 0.73 | 0.58 | 0.63 | 0.55 | 0.67 | 0.50 | 0.61 |
RMSE | 7.52 | 8.67 | 5.45 | 4.03 | 7.52 | 6.59 | 4.37 | 9.41 | 5.39 |
NSE | 0.28 | 0.32 | 0.48 | 0.04 | −2.49 | 0.22 | 0.37 | −2.03 | 0.18 |
Annual scale | |||||||||
PBIAS | −7% | −11% | −3% | 21% | 51% | 23% | 7% | 53% | 19% |
Pearson correlation coefficient (CC) | 0.64 | 0.73 | 0.74 | 0.53 | 0.35 | 0.53 | 0.63 | 0.17 | 0.48 |
Mean annual precipitation (mm) station | 360.9 | 302.4 | 342.1 | 147.9 | 126.9 | 212.3 | 240.0 | 120.5 | 209.4 |
Mean annual precipitation (mm) CHIRPS | 311.1 | 258.4 | 318.2 | 173.2 | 205.2 | 235.5 | 238.2 | 194.4 | 231.6 |
Standard deviation station | 150.27 | 110.65 | 116.36 | 55.05 | 52.55 | 107.10 | 99.82 | 57.11 | 98.49 |
Standard deviation CHIRPS | 67.47 | 53.10 | 67.70 | 39.01 | 47.93 | 45.91 | 46.05 | 42.83 | 47.32 |
Statistical analysis
The statistical metrics of CHIRPS at each station on the monthly and annual scales are shown in Table 4. The average Pearson CC values for the monthly and annual precipitation were 0.62 and 0.53, respectively. The lower values on the annual scale are similar to the results obtained. The CC0001 and CD0003 stations showed that the annual correlations were higher at the annual scale than the monthly scale. In addition, the annual and monthly scale correlations for the AN0003 were equal, while for all other stations, the monthly correlations were higher than the annual correlations, which are similar to the results obtained by Katsanos et al. (2016). Also, by definition, the monthly dataset contains more data points than the annual dataset, which may have suppressed some of the noise and improved the monthly scale correlation coefficients.
Extreme precipitation
The modest performance of CHIRPS in capturing the extreme precipitation indices may be explained, at least partially, by the fast disappearing clouds in arid regions, which provides a small window of time for the infrared sensors to detect temperature of the top of the clouds (Mianabadi et al. 2022), which in turn yields inaccuracies in the daily measurements and creates discrepancies in the extreme precipitation indices.
Precipitation is the primary input for hydrological models, and the reliability of the hydrological simulations is inherent to the quality of rainfall data (McMillan et al. 2011; Stephens et al. 2022; Wang et al. 2023). Therefore, an alternative source to the observed rainfall should conform to several requirements related to the quantity and distribution of rainfall. In this regard, an increase in extreme precipitation indices such as R10mm, R20mm, R30mm, RX1day, RX5day, or SDII may increase the risk of flooding (Abbas et al. 2023). Furthermore, an investigation in southern Brazil indicated that RX1day and RX5day were significantly and positively correlated with flash floods (Ávila et al. 2016), and a hydrological study on the impacts of precipitation extremes in the Huaihe River Basin (China) found that extreme precipitation events increased the streamflow extremes (Yang et al. 2016). In Indian River basins, multiday precipitation was the main flood driver in large watersheds, while extreme precipitation was more important in smaller basins (Nanditha & Mishra 2022). In the Yellow River basin (China), the PRCPTOT, R95p, and R99p were the main indices affecting runoff (Ren et al. 2023). Pearson's coefficients indicate poor to medium correlation between CHIRPS and station data for R10mm, R20mm, and R30mm indices, using CHIRPS in hydrological analysis for the Wala basin, may reduce the performance of the model due to data uncertainty. However, it should be noted the relationship between precipitation indices and runoff may be interdependent with several other factors such as land use, crop cover, soil, and terrain (Ren et al. 2023). Also, runoff simulation depends on the hydrological model dynamics and specific input (Worqlul et al. 2018). The calibration and validation of soil water assessment in the Zarqa River basin, which is located to the north of Wala Basin, showed the direct influence of rainfall on annual runoff (Rahbeh et al. 2019), and it is anticipated that PRCPTOT may be similar in the Wala basin.
Trends in precipitation extreme
The station data did not show significant trends for the PRCPTOT; however, a significant PRCPTOT trend was observed using CHIRPS data for the CD0005 and CD0017 stations, which are in the middle and south parts of the Wala basin. Furthermore, significant trends for CWD were observed from the CHIRPS data for most stations. The majority of the rainfall indices could not be detected at the same stations using CHIRPS data as compared to the usage of rain station data. A similar result was obtained by Cavalcante et al. (2020).
The trend analysis for the station data showed the absence of a significant trend for the PRTPTOT for all stations, indicating no change in the mean annual rainfall. However, the RX1day index showed a significant trend in the increasing direction for six of the nine stations. Also, an increase in SDII trend was found in three stations. Hence, this suggests a change in rainfall distribution towards the occurrence of more wet or extreme, which was also supported by the R99p significant trend analysis reported by Villafuerte & Matsumoto (2015) who found an increasing RX1day trend in Indochina and east-central Philippines and a decreasing trend in the maritime Continent, which they inferred to the rising global temperature. Global assessment of precipitation projections showed that global warming (Ju et al. 2021) as well as urbanization (Lu et al. 2019) will exacerbate extreme precipitation indices such as RX1day. However, CHIRPS data did not capture the RX1day trend correctly and instead, it pointed to different trends for the CWD and PRCPTOT indices, which were not observed from the station data. CHIRPS data showed a tendency to underestimate the values of the wettest months.
CONCLUSIONS
In this study, the performance of CHIRPS with a spatial resolution of 0.05o was evaluated using nine rainfall stations located within the Wala Basin, Jordan, for the period of 1987–2017. Most of the rainfall stations are located on the western side of the Wala Basin, while the eastern side is not well represented due to the lack of rainfall stations with continuous and uninterrupted time series. Despite this limitation, the study provided important insight regarding the suitability of CHIRPS data and climate change trends. The performance assessment was conducted using the point-to-pixel comparison approach from the aspects of monthly scale, annual scale, extreme indices calculation, and trend analysis.
In general, CHIRPS correlated moderately with precipitation of the stations in estimating monthly and annual precipitation as well as the R10mm and R20mm indices, while low correlation values were found for other precipitation extreme indices. The mean monthly rainfall calculated from CHIRPS showed the tendency to underestimate the values for the wettest months. Most of the extreme precipitation indices were not reflected properly by the CHIRPS product. In addition, the trend analysis of precipitation extreme indices determined for observed rainfall data showed the onset of climate in the Wala basin, more specifically towards increasing rainfall magnitude and intensities, with fewer rainfall events.
The findings of this study demonstrate that there is still potential for algorithm development, particularly in arid areas, despite ongoing improvements in the resilience of satellite-based precipitation products. It is recommended that more studies should be done at the local and regional level to evaluate CHIRPS data more accurately and also to conduct local bias correction of CHIRPS before assessing climate change in the region and for other hydrological studies.
The article would benefit from suggestions for future research directions, such as potential improvements to CHIRPS data or alternative methods for precipitation estimation in the region. This could guide researchers and contribute to the advancement of climate data quality.
ACKNOWLEDGEMENT
This work was supported by the Universiti Sains Malaysia, Research University Team (RUTeam) Grant Scheme (Grant Number: 1001/PHUMANITI/8580014).
DATA AVAILABILITY STATEMENT
All relevant data are available from an online repository or repositories. CHIRPS is available at: https://data.chc.ucsb.edu/products/CHIRPS-2.0/.
CONFLICT OF INTEREST
The authors declare there is no conflict.