Caatinga biome, located in the Brazilian semi-arid region, is the most populous semi-arid region in the world, causing intensification in land degradation and loss of biodiversity over time. The main objective of this paper is to determine and analyze the changes in land cover and use, over time, on the biophysical parameters in the Caatinga biome in the semi-arid region of Brazil using remote sensing. Landsat-8 images were used, along with the Surface Energy Balance Algorithm for Land (SEBAL) in the Google Earth Engine platform, from 2013 to 2019, through spatiotemporal modeling of vegetation indices, i.e., leaf area index (LAI) and vegetation cover (V_C). Moreover, land surface temperature (LST) and actual evapotranspiration (ET_a) in Petrolina, the semi-arid region of Brazil, was used. The principal component analysis was used to select descriptive variables and multiple regression analysis to predict ET_a. The results indicated significant effects of land use and land cover changes on energy balances over time. In 2013, 70.2% of the study area was composed of Caatinga, while the lowest percentages were identified in 2015 (67.8%) and 2017 (68.7%). Rainfall records in 2013 ranged from 270 to 480 mm, with values higher than 410 mm in 46.5% of the study area, concentrated in the northern part of the municipality. On the other hand, in 2017 the lowest annual rainfall values (from 200 to 340 mm) occurred. Low vegetation cover rate was observed by LAI and V_C values, with a range of 0 to 25% vegetation cover in 52.3% of the area, which exposes the effects of the dry season on vegetation. The highest LST was mainly found in urban areas and/or exposed soil. In 2013, 40.5% of the region’s area had LST between 48.0 and 52.0 °C, raising ET_a rates (~4.7 mm day⁻¹). Our model has shown good outcomes in terms of accuracy and concordance (coefficient of determination = 0.98, root mean square error = 0.498, and Lin’s concordance correlation coefficient = 0.907). The significant increase in agricultural areas has resulted in the progressive reduction of the Caatinga biome. Therefore, mitigation and sustainable planning is vital to decrease the impacts of anthropic actions.

Artificial Intelligence (AI) has changed how processes are developed, and decisions are made in the agricultural area replacing manual and repetitive processes with automated and more efficient ones. This study presents the application of deep learning techniques to detect and segment weeds in agricultural crops by applying models with different architectures in the analysis of images captured by an Unmanned Aerial Vehicle (UAV). This study contributes to the computer vision field by comparing the performance of the You Only Look Once (YOLOv8n, YOLOv8s, YOLOv8m, and YOLOv8l), Mask R-CNN (with framework Detectron2), and U-Net models, making public the dataset with aerial images of soybeans and beans. The models were trained using a dataset consisting of 3021 images, randomly divided into test, validation, and training sets, which were annotated, resized, and increased using the Roboflow application interface. Evaluation metrics were used, which included training efficiency (mAP50 and mAP50-90), precision, accuracy, and recall in the model’s evaluation and comparison. The YOLOv8s variant achieved higher performance with an mAP50 of 97%, precision of 99.7%, and recall of 99% when compared to the other models. The data from this manuscript show that deep learning models can generate efficient results for automatic weed detection when trained with a well-labeled and large set. Furthermore, this study demonstrated the great potential of using advanced object segmentation algorithms in detecting weeds in soybean and bean crops.