Wiam Salih

Air temperature (Tair) is an important parameter to understand climate dynamics and support decision-making across various sectors. However, accurately estimating Tair at high spatial resolution remains challenging, particularly in data-sparse regions. This study addresses the research gap that arises from the scarcity of observed temperature data by developing a robust machine learning-based estimation model, TEMLI (Temperature Estimation with ML and Land Input). The approach leverages satellite-derived inputs such as Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), albedo, and wind speed. The objective is to improve the accuracy of Tair estimation by capturing the nonlinear relationships between these variables. Among the models tested within the TEMLI framework, the Multilayer Perceptron (MLP) demonstrated superior performance, achieving a Root Mean Square Error (RMSE) of less than 1.6 °C. Notably, the model excelled in estimating extreme temperatures, especially in high-altitude regions. Trained using Moroccan weather station data, the model’s generalizability across diverse climatic conditions was also validated. This study concludes that integrating satellite-derived data with machine learning offers a scalable and reliable approach for Tair estimation, particularly in regions with limited in situ observations. The findings have significant implications for advancing climate modeling, monitoring extreme events, and informing decision-making in various climate-sensitive sectors. Graphical Abstract This visual summary offers a concise and accessible overview of the study’s main objectives, methodology, and findings. The TEMLI (Temperature Estimation with Machine Learning and Land Inputs) framework aims to improve air temperature (Tair) estimation at high spatial resolution (1 km) using satellite-derived inputs. The left panel introduces the research context, emphasizing the need for accurate temperature datasets in regions with sparse observations. The central panel outlines the methodological flowchart, detailing the various inputs (LST, NDVI, NDWI, albedo, wind speed), the machine learning models tested, and the selection of the optimal model (MLP), and the evaluation metrics and process. The right panel highlights key results, demonstrating TEMLI’s improved performance over ERA5 products with RMSE < 1.6 °C and R² > 0.9, particularly in high-altitude areas. It also summarizes practical applications, such as enhanced hydrological modeling, improved agricultural climate monitoring, and better temperature-based climate adaptation strategies.

Global warming is leading to an increase in the frequency and intensity of extreme weather events. In arid and semi-arid regions such as the Tensift basin in Morocco, global warming reflected in severe water shortages. In this context, it is of crucial importance to precisely quantify the amount of precipitation and its variability with respect of time and space. In the absence of adequate coverage by weather stations, it is imperative to investigate the reliability of satellite-based precipitation products in estimating observed precipitation. The objective of this study is to assess the performance of eight satellite products, namely PERSIANN, PERSIANN CDR, IMERG, ARC2, RFE2, CHIRPS, ERA5, and MSWEP in accurately estimating extreme observed precipitation within the basin. To this end, satellite precipitation products were compared to observed data collected over fourteen weather stations between 2001-2016, at various time steps (daily, monthly, seasonal, and annual). The data were analyzed using volumetric and categorical metrics. We further evaluated the estimation of the satellite products for extreme precipitation by comparing their extreme indices with those of the weather stations. In the same vein, this study examined the behavior of satellite products during droughts by comparing their SPI with the observations to determine the most accurate satellite products in determining the onset, duration, and magnitude of droughts. Finally, in order to determine whether bias correction was necessary to enhance performance and identify the most effective method, in the context of extreme events, several bias correction methods were evaluated based on the same procedures before and after bias correction. The results showed that PERSIANN CDR, IMERG, MSWEP, and ERA5 are the most accurate, with better performance at both monthly and annual time steps relative to daily time steps. Seasonally, the worse performance of all products occurs in summer and the best in autumn. Furthermore, we found out that PERSIANN CDR is the most reliable product that water managers can use for extreme events, while MSWEP, ERA5, and PERSIANN CDR are the best products that can be used to study the impact of climate change on drought. Finally, The Cumulative Distribution Function (CDF) mapping was found to be the most effective method in correcting the bias in the Tensift basin, especially for extreme events.

The influence of growing season rainfall on agricultural production is indisputable. In Morocco, the production of crops such as barley, maize, and wheat is impacted by growing season rainfall. Due to persistent gaps in growing season rainfall and other drivers of crop yield, crops have experienced observed yields that are often below projected or potential yields. However, there are currently no studies that have quantified these gaps in yield and growing season rainfall in Morocco. To achieve this objective, time-series crop yield for all three crops and growing season rainfall data for the period 1991–2020 were collected from FAOSTAT and the World Bank climate portal, respectively. Growing season rainfall and crop yield data for the spatial variations were culled from System National de Suivi Agrometeorologique (GCMS) and the yield gaps atlas, respectively, for the same historical period. The data were subjected to bias correction to handle uncertainty. The projected/simulated crop yields and growing season rainfall were computed by regression analysis. Crop yield and growing season rainfall gaps were determined by establishing the difference between the projected and observed crop yields and rainfall data. The results show that observed and simulated wheat have a stronger relationship when compared to the other crops. Also, most years with crop yield gaps are associated with growing season rainfall gaps. Wheat records the lowest number of years with yield gaps and the highest number of years with growing season rainfall gaps during the entire data series. Therefore, even though yield gaps are strongly tied to growing season rainfall gaps, it is not the case for wheat, and therefore other drivers might be important because wheat has the lowest number of years with crop yield gaps and the highest number of years with growing season rainfall gaps. Spatially, yield and growing season rainfall gaps decline with increased latitude. The broader perspective and policy implication here is that a better understanding of yield and growing season rainfall gaps mandates an understanding of growing season rainfall and other drivers of yield. As a way forward, potential research should focus on identifying the drivers of yield gaps, sub-national experimentation at the plot level as well as on closing yield gaps through water and nutrient management.

The Tensift basin in Morocco is prominent for its ecological and hydrological diversity. This is marked by rivers flowing into areas such as Ourika. In addition to agriculture, the basin is a hub of variable land use systems. As such, it is important to gain a better understanding of the relationship between simulated and observed precipitation in this region to be able to better understand the role of precipitation in impacting the climate and water resources in the basin. This study evaluates the performance of multi-source satellite products against weather station precipitation in the basin. The satellite-product-based data were first collected for seven satellite products, namely PERSIANN, PERSIANN CDR, TRMM3B42, ARC2, RFE2, CHIRPS, and ERA5 (simulated precipitation) from the following repositories (CHRS iRain, RainSphere, NASA, EUMETSAT, NOAA, FEWS NET, ECMWF). Precipitation observation data from six weather stations, located at Tachedert (2343 m), Imskerbour (1404 m), Asni (1170 m), Grawa (550 m), Agdal (489 m), and Agafay (487 m), at different altitudes, latitudes, and temporal scales (1D, 1M, 1Y), over the period 13 May 2007 and 31 September 2019 over the Tensift basin were collected. The data were compared and analyzed through inferential statistics such as the Nash–Sutcliffe efficiency coefficient, bias, root-mean-square error (RMSE), root-mean-square deviation (RMSD), the standard deviation, the correlation coefficient (R), and the coefficient of determination (R²) and visualized through Taylor diagrams and scatterplots to visualize the closeness between the seven satellite products and the observed precipitation data. A second analysis was carried out on the monthly precipitation, resulting from the six weather stations, and based on the standardized precipitation index (SPI) to determine the onset, duration, and magnitude of the meteorological drought. The results show that PERSIANN CDR performs best and is more reliable regarding its ability to simulate precipitation over the basin. This is seen as PERSIANN CDR has significant rates for the different statistics (Bias: −0.05 (Daily Asni), RMSE: 2.86 (Daily Agdal), R: 0.83, R²:0.687 (Monthly Agdal)). The results also show that there are no major differences between the observed weather station and the satellite precipitation data. The best performance was attributed to PERSIANN CDR (for monthly and annual precipitation at all altitudes and for daily precipitation at high altitudes). However, most of the time, this product records low or negative Nash values (−6.06 (Annual Grawa)), due to the insufficient weather station data in the study area (Tensift). It was observed that TRMM overestimates precipitation during heavy precipitation and underestimates it during low precipitation. This makes it important for the latter observations to be viewed with caution due to the quality of annual comparison results and underscores the need to develop more efficient precipitation comparison approaches and datasets. Additionally, the performance of the satellite products is better at low altitudes and during wet years. Finally, it was concluded from the SPI that Tensift region has experienced 13 drought periods over the study period, with the longest event of 12 months being from Marsh 2015 to February 2016, and the most intense event with the highest drought severity (19.6) and the lowest SPI value (−2.66) being in 2019.