The study of land use/land cover (LULC) has become an increasingly important stage in the development of forest ecosystems strategies. Hence, the main goal of this study was to describe the vegetation change of Azrou Forest in the Middle Atlas, Morocco, between 1987 and 2017. To achieve this, a set of Landsat images, including one Multispectral Scanner (MSS) scene from 1987; one Enhanced Thematic Mapper Plus (ETM+) scene from 2000; two Thematic Mapper (TM) scenes from 1995 and 2011; and one Landsat 8 Operational Land Imager (OLI) scene from 2017; were acquired and processed. Ground-based survey data and the normalized difference vegetation index (NDVI) were used to identify and to improve the discrimination between LULC categories. Then, the maximum likelihood (ML) classification method was applied was applied, in order to produce land cover maps for each year. Three classes were considered by the classification of NDVI value: low-density vegetation; moderate-density vegetation, and high-density vegetation. Our study achieved classification accuracies of 66.8% (1987), 99.9% (1995), 99.8% (2000), 99.9% (2011), and 99.9% (2017). The results from the Landsat-based image analysis show that the area of low-density vegetation was decreased from 27.4% to 2.1% over the past 30 years. While, in 2017, the class of high-density vegetation was increased to 64.6% of the total area of study area. The results of this study show that the total forest cover remained stable. The present study highlights the importance of the image classification algorithms combined with NDVI index for better understanding the changes that have occurred in this forest. Therefore, the findings of this study could assist planners and decision-makers to guide, in a good manner, the sustainable land development of areas with similar backgrounds.
Machine learning (ML) models are commonly used in solar modeling due to their high predictive accuracy. However, the predictions of these models are difficult to explain and trust. This paper aims to demonstrate the utility of two interpretation techniques to explain and improve the predictions of ML models. We compared first the predictive performance of Light Gradient Boosting (LightGBM) with three benchmark models, including multilayer perceptron (MLP), multiple linear regression (MLR), and support-vector regression (SVR), for estimating the global solar radiation (H) in the city of Fez, Morocco. Then, the predictions of the most accurate model were explained by two model-agnostic explanation techniques: permutation feature importance (PFI) and Shapley additive explanations (SHAP). The results indicated that LightGBM (R2 = 0.9377, RMSE = 0.4827 kWh/m2, MAE = 0.3614 kWh/m2) provides similar predictive accuracy as SVR, and outperformed MLP and MLR in the testing stage. Both PFI and SHAP methods showed that extraterrestrial solar radiation (H0) and sunshine duration fraction (SF) are the two most important parameters that affect H estimation. Moreover, the SHAP method established how each feature influences the LightGBM estimations. The predictive accuracy of the LightGBM model was further improved slightly after re-examination of features, where the model combining H0, SF, and RH was better than the model with all features.
Groundwater is a most important resource in arid and semi-arid regions and is required for drinking, irrigation and industrialization. Assessing the potential zone of groundwater recharge is extremely crucial for the protection of water quality and the management of groundwater systems. To identify the groundwater potential zone in the study area, thematic layers of lithology, slope, karst degrees, land cover, lineament and drainage density were generated using topographic maps, thematic maps, field data and satellite image, and were prepared, classified, weighted and integrated in a geographic information system (GIS) environment by the means of fuzzy logic. The fuzzy membership values have been assigned to different thematic layers according to their classification on respect for their contribution and their occurrence in groundwater. Based on the generated groundwater potential map, it was found that about 8% of the investigation area was categorized as very high potential for groundwater recharge, 31% as high, 28% as moderate, 17% as low and 16% as very low potential for groundwater recharge. Finally, the results were verified using well-yield data. The highest recharge potential area is located towards the downstream regions related to more fractured and karstified limestone.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.