Abstract
Satellite-based precipitation products, with simultaneously high spatial and temporal resolutions, are mostly needed to assess climate change repercussions. Previous research used datasets neglecting either good temporal or good spatial resolution, PERSIANN-CCSCDR, ERA5, and SM2RAIN-ASCAT are some of the projects aiming to remedy these limitations. This study's goal is to evaluate the accuracy of the PERSIANN-CCS-CDR, ERA5, and SM2RAIN-ASCAT at a monthly scale and their suitability for drought assessment in a Moroccan semiarid watershed. Several statistical indices were computed, the drought SPI was calculated using PERSIANN-CCS-CDR estimates, ERA5 products, and observed records as an input in the SPI formula using Gamma distribution to simulate drought from 1983 to 2017. The preliminary comparison and evaluation results of PERSIANN-CCS-CDR estimates and ERA5 datasets showed good CC on a basin scale for monthly precipitation, with a slight overestimation of the observed precipitation shown by the PBIAS. The NSE scored 0.41 for PERSIANN-CCS-CDR and 0.72 for ERA5. The results for SM2RAIN-ASCAT showed an overestimation of the observed precipitation data. At the basin scale, the SPI3 correlation coefficients between the PERSIANN-CCS-CDR monthly estimates and observed gauge rainfall data were greater than 0.67, and the RMSE was closer to 0, outperforming ERA5 in the SPI3 evaluation.
HIGHLIGHTS
The use of remotely sensed precipitation data for climatological and hydrological studies.
Evaluation of one of the rarest evaluated products (PERSIANN-CCS-CDR).
Recommendation for alternative precipitation datasets for areas lacking precipitation data.
INTRODUCTION
Water supply is one of the most important variables impacting socioeconomic development in arid and semi-arid regions, particularly in areas that rely heavily on agricultural production (Esper et al. 2007; Schilling et al. 2012; Jarlan et al. 2015). Moreover, drought has become increasingly common and severe in Morocco (Brahim et al. 2017) and especially the Tensift basin region as yearly rainfall has declined in recent decades (Zbiri et al. 2019; Zkhiri et al. 2019; Habitou et al. 2020). In the Tensift region, droughts have reduced water inflows to the Takerkoust dam, hence the amount of water available for agricultural irrigation. Drought years resulted in a reduction of more than 50% of the water volume intended for irrigation. Farmers restrict or skip irrigation of annual crops during periods of water constraint to save water for irrigation of perennial crops. Water scarcity for irrigation has resulted in a drop of up to 100% of the surface allocated in some circumstances (Bennani et al. 2016; Meliho et al. 2019). The primary component of the hydrological cycle is precipitation. The spatial and temporal distribution of precipitation is an important characteristic of water resource management and natural disaster avoidance. Precipitation data from rain gauges perform poorly in determining the spatial distribution of precipitation (Yan et al. 2014). Nonetheless, in underdeveloped countries, the poor coverage of rainfall systems can significantly constrain the quantification of these inputs in the watershed (Ouatiki et al. 2017). Global precipitation estimates are becoming more vital for climatological and hydrological studies. Satellite-based precipitation products (PPs), with simultaneously, high spatial and temporal resolutions are mostly needed to assess climate change repercussions (Qureshi et al. 2022).
Satellite precipitation recovery algorithms, such as Tropical Rainfall Measuring Mission (TRMM), the Climate Prediction Center morphing technique (CMORPH), Global Satellite Mapping of Precipitation and Precipitation Estimation From Remotely Sensed Information Using Artificial Neural Networks (PERSIANN) (Hsu et al. 1997; Saouabe et al. 2020), have been developed over the past few decades by combining infrared (IR) and passive microwave estimates from various sensors (Hsu et al. 1997). Satellite PPs also behave differently in different regions (Gao et al. 2018). In the study by Ouatiki et al. (2017), the authors have demonstrated that TRMM data, at monthly and annual time intervals, can be a helpful source of rainfall data for water resource monitoring and management in semi-arid ungauged basins.
PERSIANN-CCS-CDR (Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks-Cloud Classification System-Climate Data Record) is a near-global precipitation dataset with high spatial and temporal resolutions that spans more than 37 years. Developed by the Center for Hydrometeorology and Remote Sensing (CHRS) at the University of California, Irvine (UCI), PERSIANN-CCS-CDR provides precipitation estimates at 0.04° spatial and 3-hourly temporal resolutions from 1983 to the present throughout the 60°S to 60°N geographical domain (Sadeghi et al. 2019, 2021). There has been no assessment of the accuracy and applicability of this new set of satellite precipitation data in the Tensift basin. This region would benefit from accurate precipitation data in order to tackle the drought and water resource issues. As a result, it is critical to analyze the use of this new satellite precipitation dataset and its potential for drought monitoring.
The SM2RAIN-ASCAT (Soil Moisture to Rain) rainfall data record covers the period 2007–2021, with a geographical and temporal sample rate of almost 12.5 km/day. The new SM2RAIN-ASCAT data record was created by applying the SM2RAIN algorithm (Brocca et al. 2014, 2019) to ASCAT soil moisture data records H113 and H114 provided by the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) Satellite Application Facility on Support to Operational Hydrology and Water Management (H SAF). It is the first SM2RAIN-ASCAT data record available at the same spatial resolution as the ASCAT soil moisture product (Brocca et al. 2014).
ERA5 is based on the operationalized Integrated Forecasting System (IFS) Cy41r2 in 2016. Thus, ERA5 takes use of a decade's worth of improvements in model physics, core dynamics, and data assimilation. In addition to a greatly improved horizontal resolution of 31 km compared to 80 km for ERA-Interim, ERA5 offers hourly output across the board and an ensemble estimate of uncertain accuracy (3 h at half the horizontal resolution) (Hersbach et al. 2020).
The standard precipitation index (SPI), initially established by McKee et al. (McKee et al. 1993; Edwards & McKee 1997), is a recommended meteorological drought index by the World Meteorological Organization. The SPI is commonly employed for monitoring and predicting droughts. Only long-term precipitation records are considered in the SPI computation (Hayes et al. 2011). In a study conducted in the Tensift basin (Salih et al. 2022), the authors determined that the Tensift region underwent 13 drought periods throughout the study period from 2007 to 2019, with the longest occurrence lasting 12 months from March 2015 to February 2016, and the most intense event with the highest drought severity and the lowest SPI value occurring in 2019.
The accuracy of PERSIANN-CCS-CDR, ERA5, and SM2RAIN-ASCAT data over the Tensift basin in Morocco was examined in this work by comparison with ground observations. Based on long-term satellite/reanalysis products and the SPI3 drought index, this study evaluates the three products using statistical indices and applies the datasets to the assessment of meteorological droughts in the Tensift basin. This is the first study to evaluate and implement PERSIANN-CCS-CDR for drought monitoring in the Tensift basin from January 1983 to August 2017. The study area and data are illustrated initially. It is important to note that this semi-arid area is highly varied, with mountains and snow at high altitudes. The lack of sufficient meteorological stations and observed data, in addition to the number of droughts and extreme meteorological events that the Tensift basin is experiencing (Sinan & Belhouji 2016; Hajhouji et al. 2018), led us to choose this region as a study area. Furthermore, the technique and SPI are presented. The evaluations and comparisons of satellite/reanalysis products and estimations with gauge observations are then presented, followed by the application of PERSIANN-CCS-CDR and ERA5 datasets for drought monitoring. The conclusion and suggestions are therefore provided.
STUDY AREA AND DATA
Study area and rainfall observed data
Stations . | Longitude . | Latitude . | Altitude . | Period . |
---|---|---|---|---|
ABADLA | − 8.56559 | 31.72092 | 246 | 1983–2017 |
AGHBALOU | − 7.74648 | 31.31408 | 1005 | 1983–2017 |
CHICHAOUA | − 8.75297 | 31.54842 | 337 | 1983–2017 |
ILOUDJANE | − 8.79945 | 31.1786 | 752 | 1989–2017 |
Imn. HAMMAM | − 8.11067 | 31.21183 | 744 | 1983–2017 |
MARRAKECH | − 8.03154 | 31.55526 | 460 | 1983–2017 |
N'KOURIS | − 8.138 | 31.05518 | 1060 | 1983–2017 |
Sd. BOATMAN | − 8.44734 | 31.22391 | 816 | 1989–2017 |
SIDI HSAIN | − 8.23885 | 31.19329 | 1021 | 1998–2017 |
SIDI RAHAL | − 7.4736 | 31.63578 | 687 | 1983–2017 |
TAFERIAT | − 7.59854 | 31.54454 | 761 | 1983–2017 |
TAHANAOUT | − 7.96305 | 31.29197 | 1043 | 1983–2017 |
TAKERKOUST | − 8.13656 | 31.36954 | 676 | 1983–2017 |
TALMEST | − 9.26986 | 31.86334 | 35 | 1985–2017 |
Stations . | Longitude . | Latitude . | Altitude . | Period . |
---|---|---|---|---|
ABADLA | − 8.56559 | 31.72092 | 246 | 1983–2017 |
AGHBALOU | − 7.74648 | 31.31408 | 1005 | 1983–2017 |
CHICHAOUA | − 8.75297 | 31.54842 | 337 | 1983–2017 |
ILOUDJANE | − 8.79945 | 31.1786 | 752 | 1989–2017 |
Imn. HAMMAM | − 8.11067 | 31.21183 | 744 | 1983–2017 |
MARRAKECH | − 8.03154 | 31.55526 | 460 | 1983–2017 |
N'KOURIS | − 8.138 | 31.05518 | 1060 | 1983–2017 |
Sd. BOATMAN | − 8.44734 | 31.22391 | 816 | 1989–2017 |
SIDI HSAIN | − 8.23885 | 31.19329 | 1021 | 1998–2017 |
SIDI RAHAL | − 7.4736 | 31.63578 | 687 | 1983–2017 |
TAFERIAT | − 7.59854 | 31.54454 | 761 | 1983–2017 |
TAHANAOUT | − 7.96305 | 31.29197 | 1043 | 1983–2017 |
TAKERKOUST | − 8.13656 | 31.36954 | 676 | 1983–2017 |
TALMEST | − 9.26986 | 31.86334 | 35 | 1985–2017 |
Satellite and reanalysis data
PERSIANN-CCS-CDR
PERSIANN-CCS-CDR (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Cloud Classification System-Climate Data Record) is developed by the CHRS at the UCI. PERSIANN-CCS-CDR provides precipitation dataset with both high spatial and temporal resolutions at 0.04° spatial and 3-hourly temporal resolutions from 1983 to present over the global domain of 60°S to 60°N. It combines the algorithms used in the development of PERSIANN-CCS and PERSIANN-CDR and uses data from GEO satellites as input data, the PERSIANN-CCS method is applied to gridded satellite data (GridSat-B1) and to CPC-4 km merged IR datasets. The bias evaluations are then corrected by applying the GPCP product to the entire period of the dataset (Sadeghi et al. 2019, 2021; Eini et al. 2022). Monthly data was obtained from http://chrsdata.eng.uci.edu/PERSIANN-CCS-CDR (accessed September 2020). For this study, the timeframe of PERSIANN-CCS-CDR precipitation data used starts from January 1983 to 2017. The choice of using this product is based on the lack of evaluation studies of PERSIANN-CCS-CDR, and more precisely in a semi-arid context.
ERA5 monthly averaged
The fifth generation of atmospheric global climate reanalysis ERA5 is produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). This product provides a dataset of several variables including surface temperature and multiple types of precipitation estimates, by combining observational and reanalysis data, from 1979 to present, ERA5 has a maximum temporal resolution of 1 h and a 31 km spatial resolution (Hersbach et al. 2020). Previous studies conducted in North America (Tarek et al. 2020) and in Morocco over the Tensift watershed (El Bouazzaoui et al. 2022; Salih et al. 2022) showed that the performance of reanalysis products is good in lowland or low altitude areas and can be considered as one of the best estimates of PPs. The monthly averaged precipitation data which contains the accumulated monthly precipitation data used in this study is freely accessible from https://cds.climate.copernicus.eu/cdsapp#!/home (accessed October 2022). The ERA5 reanalysis rainfall product is evaluated in this study to give more visibility about the difference in the evaluation results of different satellite-based rainfall sources. A period of 34 years (1983–2017) were used in this study.
SM2RAIN-ASCAT
One of the newest global PPs is SM2RAIN-ASCAT, which was created using the SM2RAIN algorithm, the advanced scatterometer (ASCAT) satellite moisture data (Brocca et al. 2014, 2019). The dataset provides precipitation estimates measured using the soil as a natural rain gauge with a spatial resolution starting from 1 km downscaled from the initial dataset of a 10 km spatial resolution, the dataset period starts from 2007 to present. SM2Rain-ASCAT data can be downloaded freely from https://zenodo.org/record/6459152#.Y317z0mZND- (accessed October 2022). In this study, the initial 10 km spatial resolution dataset was used. As for PERSIANN-CCS-CDR, evaluating the SM2RAIN-ASCAT product in this study is based on the lack of evaluation studies conducted in a semi-arid basin in north Africa generally (Tramblay et al. 2023). The analysis was conducted over the period of 10 years (2007–2017).
METHODS
Statistical evaluation and metrics
The evaluation between the satellite/reanalysis products and ground-based observation precipitations data has been conducted using the data from 14 stations and the different PPs. The precipitation estimates datasets for each station were obtained by extracting the values of the corresponding pixels in the raster images of all products. To evaluate and comprehend the performance of the different PPs, several widely used statistical metrics were computed (Table 2). Pearson's correlation coefficient (CC) is a measure which quantifies the strength of a linear relationship between two variables; it describes the correlation between the observed data and the satellite/reanalysis time series, it has no unit and its values range from −1 to 1 which both are described as perfect correlations (Salih et al. 2022). The coefficient of determination (R2) is the square of the CC when the relationship between the two variables is linear. It assesses the quality of the linear regression. The values of R2 range from 0 to 1 and measure the fit between the estimated and observed datasets. When R2 = 1, it means that both datasets are perfectly aligned (Salih et al. 2022). The root-mean-square error (RMSE) indicates the average magnitude of the error between satellite/reanalysis data and the observed precipitation data, when it equals 0 it indicates a perfect match between the datasets, the more the value increases the more the two datasets series have a poor match (Golmohammadi et al. 2014). The Nash–Sutcliffe efficiency coefficient (NSE) varies from negative infinity to 1. When NSE = 1, it signifies a perfect fit between observed and projected values. Values between 0 and 1 are deemed acceptable levels of effectiveness, however, values less than 0 imply that the mean observed value is higher than the simulated value, indicating unacceptable performance (Golmohammadi et al. 2014). The percentage of bias (PBIAS) reflects the average propensity of the simulated data to be greater or smaller than their measured counterparts. PBIAS has an optimum value of 0, with low amplitude values suggesting an accurate simulation model. Positive numbers imply underestimation bias, whereas negative values suggest overestimation bias (Vijai et al. 1999). To determine the significance of the correlations, we calculated the P-value of the correlations and we determined that alpha is 5% (0.05). When the P-value of a correlation is less than alpha, the correlation is considered statistically significant (Wilcox 2016).
Statistical metrics . | Formula . | Optimal value . |
---|---|---|
CC | − 1 or 1 | |
R2 | 1 | |
RMSE | 0 | |
NSE | 1 | |
PBIAS | 0 |
Statistical metrics . | Formula . | Optimal value . |
---|---|---|
CC | − 1 or 1 | |
R2 | 1 | |
RMSE | 0 | |
NSE | 1 | |
PBIAS | 0 |
Where n is the sample size; Gi is the satellite precipitation estimate; Si is the gauge observation; is the mean satellite precipitation estimate; and is the mean gauge observation.
SPI
McKee et al. (1993) created the SPI as an index for defining and tracking drought occurrences. It is a basic indicator that uses simply precipitation data to examine both wet and dry periods/cycles. The SPI compares precipitation over time (typically 1–24 months) to the average long-term precipitation at the same location. After modifying a probability density, the SPI is calculated by normalizing precipitation for a specific station. The gamma distribution is the finest representation of the development of precipitation data (Edwards & McKee 1997). Table 3 describes a categorization method for defining the severity of drought occurrences based on the SPI index value at any time frame (Habitou et al. 2020).
SPI Classes . | SPI values . |
---|---|
Extreme drought | − 2 or less |
Severe drought | − 2 to −1.5 |
Moderate drought | − 1.5 to −1 |
Mild drought | − 1 to 0 |
Mildly wet | 0–1 |
Moderately wet | 1–1.5 |
Severely wet | 1.5–2 |
Extremely wet | 2 or more |
SPI Classes . | SPI values . |
---|---|
Extreme drought | − 2 or less |
Severe drought | − 2 to −1.5 |
Moderate drought | − 1.5 to −1 |
Mild drought | − 1 to 0 |
Mildly wet | 0–1 |
Moderately wet | 1–1.5 |
Severely wet | 1.5–2 |
Extremely wet | 2 or more |
The constants are c0 = 2.515517, c1 = 0.802853, c2 = 0.010328, d1 = 1.432788, d2 = 0.189269, and d3 = 0.001308.
RESULTS AND DISCUSSION
Evaluation of satellite and reanalysis estimates
PERSIANN-CCS-CDR evaluation
Five statistical indices (CC, RMSE, PBIAS, R2, and NSE) are used to compare the monthly PERSIANN-CCS-CDR product with the monthly ground-based observation data from January 1983 to August 2017 at both the basin scale and the individual grid-cells corresponding to rain gauges. Table 4 displays the statistical evaluation indices for the PERSIANN-CCS-CDR product at both the basin and grid-cell scales.
Stations . | Altitudes . | CC . | NSE . | PBIAS . | RMSE . | R2 . | P-values . |
---|---|---|---|---|---|---|---|
TALMEST | 35 | 0.71 | 0.15 | − 19.00 | 35.23 | 0.51 | 2.89 × 10 10−60 |
ABADLA | 246 | 0.67 | − 0.35 | − 45.90 | 24.78 | 0.45 | 7.45 × 10−56 |
CHICHAOUA | 337 | 0.70 | 0.07 | − 26.90 | 21.58 | 0.49 | 4.87 × 10−63 |
MARRAKECH | 460 | 0.43 | − 1.72 | − 48.10 | 46.59 | 0.18 | 3.69 × 10−20 |
TAKERKOUST | 676 | 0.50 | − 2.26 | − 43.10 | 44.84 | 0.25 | 2.53 × 10−27 |
SIDI RAHAL | 687 | 0.48 | − 0.90 | − 13.90 | 44.85 | 0.23 | 6.06 × 10−25 |
Imn. HAMMAM | 744 | 0.48 | − 0.53 | 1.30 | 43.63 | 0.23 | 5.26 × 10−25 |
ILOUDJANE | 752 | 0.55 | − 0.06 | 2.50 | 33.96 | 0.30 | 6.58 × 10−28 |
TAFERIAT | 761 | 0.42 | − 0.73 | − 2.10 | 48.36 | 0.17 | 1.78 × 10−18 |
Sd. BOATMAN | 816 | 0.48 | − 0.15 | 14.30 | 37.68 | 0.23 | 4.57 × 10−22 |
AGHBALOU | 1005 | 0.44 | − 0.35 | 16.70 | 52.38 | 0.20 | 1.41 × 10−21 |
SIDI HSAIN | 1021 | 0.51 | 0.10 | 25.50 | 37.48 | 0.26 | 2.62 × 10−16 |
TAHANAOUT | 1043 | 0.51 | − 0.35 | 6.10 | 41.95 | 0.26 | 1.9 × 10−28 |
N'KOURIS | 1060 | 0.63 | − 0.40 | − 29.70 | 32.34 | 0.40 | 1.02 × 10−47 |
Basin scale (Tensift) | 0.77 | 0.41 | − 3.88 | 20.91 | 0.59 | 1.32 × 10−82 |
Stations . | Altitudes . | CC . | NSE . | PBIAS . | RMSE . | R2 . | P-values . |
---|---|---|---|---|---|---|---|
TALMEST | 35 | 0.71 | 0.15 | − 19.00 | 35.23 | 0.51 | 2.89 × 10 10−60 |
ABADLA | 246 | 0.67 | − 0.35 | − 45.90 | 24.78 | 0.45 | 7.45 × 10−56 |
CHICHAOUA | 337 | 0.70 | 0.07 | − 26.90 | 21.58 | 0.49 | 4.87 × 10−63 |
MARRAKECH | 460 | 0.43 | − 1.72 | − 48.10 | 46.59 | 0.18 | 3.69 × 10−20 |
TAKERKOUST | 676 | 0.50 | − 2.26 | − 43.10 | 44.84 | 0.25 | 2.53 × 10−27 |
SIDI RAHAL | 687 | 0.48 | − 0.90 | − 13.90 | 44.85 | 0.23 | 6.06 × 10−25 |
Imn. HAMMAM | 744 | 0.48 | − 0.53 | 1.30 | 43.63 | 0.23 | 5.26 × 10−25 |
ILOUDJANE | 752 | 0.55 | − 0.06 | 2.50 | 33.96 | 0.30 | 6.58 × 10−28 |
TAFERIAT | 761 | 0.42 | − 0.73 | − 2.10 | 48.36 | 0.17 | 1.78 × 10−18 |
Sd. BOATMAN | 816 | 0.48 | − 0.15 | 14.30 | 37.68 | 0.23 | 4.57 × 10−22 |
AGHBALOU | 1005 | 0.44 | − 0.35 | 16.70 | 52.38 | 0.20 | 1.41 × 10−21 |
SIDI HSAIN | 1021 | 0.51 | 0.10 | 25.50 | 37.48 | 0.26 | 2.62 × 10−16 |
TAHANAOUT | 1043 | 0.51 | − 0.35 | 6.10 | 41.95 | 0.26 | 1.9 × 10−28 |
N'KOURIS | 1060 | 0.63 | − 0.40 | − 29.70 | 32.34 | 0.40 | 1.02 × 10−47 |
Basin scale (Tensift) | 0.77 | 0.41 | − 3.88 | 20.91 | 0.59 | 1.32 × 10−82 |
ERA5 reanalysis evaluation
CC, RMSE, PBIAS, R2, and NSE are the five statistical indicators used to compare the monthly ERA5 reanalysis result with the monthly ground-based observation data from January 1983 to August 2017 at both the basin scale and individual grid cells corresponding to rain gauges. The results of statistical evaluation indices for the ERA5 reanalysis product at both the basin and grid-cell scales are displayed in Table 5.
Stations . | Altitudes . | CC . | NSE . | PBIAS . | RMSE . | R2 . | P-values . |
---|---|---|---|---|---|---|---|
TALMEST | 35 | 0.92 | 0.84 | 5.92 | 14.99 | 0.84 | 5.4 × 10−155 |
ABADLA | 246 | 0.83 | 0.65 | − 20.35 | 12.60 | 0.69 | 1.6 × 10−106 |
CHICHAOUA | 337 | 0.87 | 0.67 | − 34.27 | 12.81 | 0.76 | 1.8 × 10−131 |
MARRAKECH | 460 | 0.75 | 0.42 | − 40.27 | 21.46 | 0.57 | 3.09 × 10−77 |
TAKERKOUST | 676 | 0.74 | − 1.00 | − 109.03 | 35.13 | 0.55 | 1.68 × 10−74 |
SIDI RAHAL | 687 | 0.82 | 0.33 | − 54.27 | 26.59 | 0.67 | 4.2 × 10−102 |
Imn. HAMMAM | 744 | 0.76 | 0.44 | − 30.88 | 26.43 | 0.58 | 6.09 × 10−81 |
ILOUDJANE | 752 | 0.75 | 0.56 | − 21.54 | 20.98 | 0.57 | 1.31 × 10−62 |
TAFERIAT | 761 | 0.65 | 0.09 | − 59.86 | 34.88 | 0.42 | 1.65 × 10−49 |
Sd. BOATMAN | 816 | 0.79 | 0.48 | − 32.23 | 24.63 | 0.57 | 1.32 × 10−62 |
AGHBALOU | 1005 | 0.56 | 0.07 | − 35.98 | 43.49 | 0.31 | 1.81 × 10−35 |
SIDI HSAIN | 1021 | 0.84 | 0.68 | − 6.13 | 21.15 | 0.51 | 6.09 × 10−37 |
TAHANAOUT | 1043 | 0.69 | 0.03 | − 47.76 | 35.60 | 0.48 | 4.74 × 10−61 |
N'KOURIS | 1060 | 0.73 | − 0.04 | − 89.91 | 27.81 | 0.54 | 7.51 × 10−71 |
Basin scale (Tensift) | 0.91 | 0.72 | − 31.88 | 13.80 | 0.82 | 9.9 × 10−157 |
Stations . | Altitudes . | CC . | NSE . | PBIAS . | RMSE . | R2 . | P-values . |
---|---|---|---|---|---|---|---|
TALMEST | 35 | 0.92 | 0.84 | 5.92 | 14.99 | 0.84 | 5.4 × 10−155 |
ABADLA | 246 | 0.83 | 0.65 | − 20.35 | 12.60 | 0.69 | 1.6 × 10−106 |
CHICHAOUA | 337 | 0.87 | 0.67 | − 34.27 | 12.81 | 0.76 | 1.8 × 10−131 |
MARRAKECH | 460 | 0.75 | 0.42 | − 40.27 | 21.46 | 0.57 | 3.09 × 10−77 |
TAKERKOUST | 676 | 0.74 | − 1.00 | − 109.03 | 35.13 | 0.55 | 1.68 × 10−74 |
SIDI RAHAL | 687 | 0.82 | 0.33 | − 54.27 | 26.59 | 0.67 | 4.2 × 10−102 |
Imn. HAMMAM | 744 | 0.76 | 0.44 | − 30.88 | 26.43 | 0.58 | 6.09 × 10−81 |
ILOUDJANE | 752 | 0.75 | 0.56 | − 21.54 | 20.98 | 0.57 | 1.31 × 10−62 |
TAFERIAT | 761 | 0.65 | 0.09 | − 59.86 | 34.88 | 0.42 | 1.65 × 10−49 |
Sd. BOATMAN | 816 | 0.79 | 0.48 | − 32.23 | 24.63 | 0.57 | 1.32 × 10−62 |
AGHBALOU | 1005 | 0.56 | 0.07 | − 35.98 | 43.49 | 0.31 | 1.81 × 10−35 |
SIDI HSAIN | 1021 | 0.84 | 0.68 | − 6.13 | 21.15 | 0.51 | 6.09 × 10−37 |
TAHANAOUT | 1043 | 0.69 | 0.03 | − 47.76 | 35.60 | 0.48 | 4.74 × 10−61 |
N'KOURIS | 1060 | 0.73 | − 0.04 | − 89.91 | 27.81 | 0.54 | 7.51 × 10−71 |
Basin scale (Tensift) | 0.91 | 0.72 | − 31.88 | 13.80 | 0.82 | 9.9 × 10−157 |
SM2RAIN-ASCAT evaluation
The same five statistical indices (CC, RMSE, PBIAS, R2, and NSE) were utilized to compare the monthly SM2RAIN-ASCAT product with the monthly observed precipitation data for the period from January 2007 to August 2017 at both the basin scale and the individual grid-cells corresponding to rain gauges. The results of the statistical evaluation indices for the SM2RAIN-ASCAT product at both the basin and grid-cell scales are summarized in Table 6.
Stations . | Altitudes . | CC . | NSE . | PBIAS . | RMSE . | R2 . | P-values . |
---|---|---|---|---|---|---|---|
TALMEST | 35 | 0.88 | 0.87 | 0.36 | 8.94 | 0.77 | 2.69 × 10-42 |
ABADLA | 246 | 0.81 | 0.83 | −24.88 | 6.33 | 0.66 | 3.55 × 10-31 |
CHICHAOUA | 337 | 0.78 | 0.79 | −19.17 | 7.83 | 0.61 | 1.92 × 10-27 |
MARRAKECH | 460 | 0.67 | 0.69 | −26.20 | 14.62 | 0.45 | 3.14 × 10-18 |
TAKERKOUST | 676 | 0.82 | 0.77 | −57.21 | 9.44 | 0.67 | 4.43 × 10-32 |
SIDI RAHAL | 687 | 0.72 | 0.74 | −51.03 | 14.50 | 0.52 | 1 × 10-21 |
Imn. HAMMAM | 744 | 0.79 | 0.86 | 4.77 | 12.82 | 0.63 | 6.86 × 10-29 |
ILOUDJANE | 752 | 0.64 | 0.75 | −22.60 | 12.32 | 0.41 | 3.66 × 10-16 |
TAFERIAT | 761 | 0.58 | 0.63 | −53.77 | 16.74 | 0.33 | 1.24 × 10-12 |
Sd. BOATMAN | 816 | 0.65 | 0.78 | 0.59 | 15.46 | 0.42 | 1.83 × 10-16 |
AGHBALOU | 1005 | 0.62 | 0.79 | 6.22 | 20.30 | 0.39 | 5.62 × 10-15 |
SIDI HSAIN | 1021 | 0.70 | 0.78 | 16.66 | 18.10 | 0.48 | 7.45 × 10-20 |
TAHANAOUT | 1043 | 0.75 | 0.84 | −1.40 | 13.61 | 0.56 | 3.09 × 10-24 |
N'KOURIS | 1060 | 0.64 | 0.75 | −23.18 | 12.04 | 0.41 | 5.76 × 10-16 |
Basin scale (Tensift) | 0.88 | 0.90 | −4.24 | 7.83 | 0.77 | 1.29 × 10-42 |
Stations . | Altitudes . | CC . | NSE . | PBIAS . | RMSE . | R2 . | P-values . |
---|---|---|---|---|---|---|---|
TALMEST | 35 | 0.88 | 0.87 | 0.36 | 8.94 | 0.77 | 2.69 × 10-42 |
ABADLA | 246 | 0.81 | 0.83 | −24.88 | 6.33 | 0.66 | 3.55 × 10-31 |
CHICHAOUA | 337 | 0.78 | 0.79 | −19.17 | 7.83 | 0.61 | 1.92 × 10-27 |
MARRAKECH | 460 | 0.67 | 0.69 | −26.20 | 14.62 | 0.45 | 3.14 × 10-18 |
TAKERKOUST | 676 | 0.82 | 0.77 | −57.21 | 9.44 | 0.67 | 4.43 × 10-32 |
SIDI RAHAL | 687 | 0.72 | 0.74 | −51.03 | 14.50 | 0.52 | 1 × 10-21 |
Imn. HAMMAM | 744 | 0.79 | 0.86 | 4.77 | 12.82 | 0.63 | 6.86 × 10-29 |
ILOUDJANE | 752 | 0.64 | 0.75 | −22.60 | 12.32 | 0.41 | 3.66 × 10-16 |
TAFERIAT | 761 | 0.58 | 0.63 | −53.77 | 16.74 | 0.33 | 1.24 × 10-12 |
Sd. BOATMAN | 816 | 0.65 | 0.78 | 0.59 | 15.46 | 0.42 | 1.83 × 10-16 |
AGHBALOU | 1005 | 0.62 | 0.79 | 6.22 | 20.30 | 0.39 | 5.62 × 10-15 |
SIDI HSAIN | 1021 | 0.70 | 0.78 | 16.66 | 18.10 | 0.48 | 7.45 × 10-20 |
TAHANAOUT | 1043 | 0.75 | 0.84 | −1.40 | 13.61 | 0.56 | 3.09 × 10-24 |
N'KOURIS | 1060 | 0.64 | 0.75 | −23.18 | 12.04 | 0.41 | 5.76 × 10-16 |
Basin scale (Tensift) | 0.88 | 0.90 | −4.24 | 7.83 | 0.77 | 1.29 × 10-42 |
The evaluation of SM2RAIN-ASCAT concludes with this part, as the results showed an overestimation of monthly precipitation at the pixel and basin scale compared with the observed precipitation data, despite the fact that the CC and R2 values were promising, the SM2RAIN-ASCAT product is too short to be considered in this study for SPI analysis (10 years, from 2007 to 2017) and in terms of homogeneity with the other products, it will not be considered for the SPI analysis.
SPI3 analysis
This section has been used as a substitute to evaluate the efficiency of the two remaining PPs products for drought monitoring; however, it does not reflect the drought conditions in the Tensift basin. In this study, only SPI 3 months (SPI3) was calculated for observed rain gauge data, PERSIANN-CCS-CDR and ERA5 over the entire basin for the period from January 1983 to August 2017. It is useful for detecting drought conditions in a relatively short time period and can help to predict water resources in the near future (McKee et al. 1993), it is necessary to note that long time scales drought indices are associated with changes in groundwater storage, whereas short time scales indices, such as SPI3, are mostly related to differences in soil water content and river flow in headwater region (Vicente-Serrano et al. 2010). The choice of SPI3 will allow the operational application of precipitation satellite products in our region. For the observed rain data, the inverse distance weighed (IDW) interpolation method is used to interpolate the precipitation dataset over the Tensift basin. This method has its limits, especially over regions with complex terrain and with inhomogeneous ground station spatial distribution over the whole region, which may lead to large uncertainty. The IDW method was largely used in our area of interest in previous studies (Marchane et al. 2017; Ruelland et al. 2015; Zkhiri et al. 2017; El Khalki et al. 2018, 2020).
. | NSE . | PBIAS . | RMSE . | CC . | R2 . | P-value . |
---|---|---|---|---|---|---|
ERA5 | − 0.08 | 827.62 | 0.98 | 0.46 | 0.22 | 1.72 × 10−23 |
PERSIANN-CCS-CDR | 0.41 | 7086.30 | 0.72 | 0.67 | 0.45 | 2.34 × 10−55 |
. | NSE . | PBIAS . | RMSE . | CC . | R2 . | P-value . |
---|---|---|---|---|---|---|
ERA5 | − 0.08 | 827.62 | 0.98 | 0.46 | 0.22 | 1.72 × 10−23 |
PERSIANN-CCS-CDR | 0.41 | 7086.30 | 0.72 | 0.67 | 0.45 | 2.34 × 10−55 |
CONCLUSION
In this work, the performances of three satellite/reanalysis PPs and estimates (PERSIANN-CCS-CDR, ERA5, and SM2RAIN-ASCAT) were assessed on different spatial scales (grid-cell and basin sizes) using readings from 14 meteorological stations in Morocco's Tensift basin. From 1983 to 2017, two satellite/reanalysis PPs (PERSIANN-CCS-CDR and ERA5) and from 2007 to 2017 SM2RAIN-ASCAT were evaluated.
- (1)
The PERSIANN-CCS-CDR and ERA5 products showed promising results compared to the SM2RAIN-ASCAT product in terms of their ability to describe the geographical distribution of the observed precipitation over the Tensift basin.
- (2)
The analysis demonstrates that basin-averaged PERSIANN-CCS-CDR and ERA5 precipitation are more consistent with interpolated precipitation over the catchment from observed data than pixel-to-station precipitation.
- (3)
SM2RAIN-ASCAT overestimate observed precipitation data despite promising CC and R2 values; we should take into consideration the small period used for the evaluation of SM2RAIN-ASCAT which may influence the results compared to the long periods used to evaluate the two other rainfall products (PERSIANN-CCS-CDR and ERA5).
- (4)
PERSIANN-CCS-CDR outperforms ERA5 in the evaluation of the products using SPI3.
Our results revealed that the PERSIANN-CCS-CDR and ERA5 products outperform the SM2RAIN-ASCAT product in terms of their ability to describe the spatial distribution of observed precipitation over the Tensift basin. Due to the fact that both of these products (PERSIANN-CCS-CDR and ERA5) were able to accurately show the spatiotemporal distribution of the observed rainfall over the basin, we advocate using them to comprehend the spatial and temporal variations of rainfall in semi-arid basins in Morocco. The SPI3 correlation coefficients between the PERSIANN-CCS-CDR monthly estimates and observed gauge rainfall data were greater than 0.67, and the RMSE was closer to 0. Therefore, we recommend that the monthly PERSIANN-CCS-CDR products be employed as a supplement to the drought assessment in Morocco's semi-arid regions.
DATA AVAILABILITY STATEMENT
All relevant data are available from an online repository or repositories: PERSIANN: http://chrsdata.eng.uci.edu/PERSIANN-CCS-CDR; ERA5: https://cds.climate.copernicus.eu/cdsapp#!/home; SM2RAIN: https://zenodo.org/record/6459152#.Y317z0mZND-.
CONFLICT OF INTEREST
The authors declare there is no conflict.