Water scarcity, exacerbated by climate change and increasing agricultural water demands, highlights the necessity for efficient irrigation management. This study focused on estimating actual evapotranspiration (ETa) in watermelons under semi-arid Mediterranean conditions by integrating high-resolution satellite imagery and agro-meteorological data. Field experiments were conducted in Rutigliano, southern Italy, over a 2.80 ha area. ETa was measured with the eddy covariance (EC) technique and predicted using machine learning models. Multispectral reflectance data from Planet SuperDove satellites and local meteorological records were used as predictors. Partial least squares, the generalized linear model and three machine learning algorithms (Random Forest, Elastic Net, and Support Vector Machine) were evaluated. Random Forest yielded the highest predictive accuracy with an average R² of 0.74, RMSE of 0.577 mm, and MBE of 0.03 mm. Model interpretability was performed through permutation importance and SHAP, identifying the near-infrared and red spectral bands, average daily temperature, and relative humidity as key predictors. This integrated approach could provide a scalable, precise method for watermelon ETa estimation, supporting data-driven irrigation management and improving water use efficiency in Mediterranean horticultural systems.

This study investigated the application of high-resolution satellite imagery from SuperDove satellites combined with machine learning algorithms to estimate the spatiotemporal variability of some winter wheat parameters, including the relative leaf chlorophyll content (RCC), relative water content (RWC), and aboveground dry matter (DM). The research was carried out within an experimental field in Southern Italy during the 2024 growing season. Different machine learning (ML) algorithms were trained and compared using spectral band data and calculated vegetation indices (VIs) as predictors. Model performance was assessed using R² and RMSE. The ML models tested were random forest (RF), support vector regressor (SVR), and extreme gradient boosting (XGB). RF outperformed the other ML algorithms in the prediction of RCC when using VIs as predictors (R² = 0.81) and in the prediction of the RWC and DM when using spectral bands data as predictors (R² = 0.71 and 0.87, respectively). Model explainability was assessed with the SHAP method. A SHAP analysis highlighted that GNDVI, Cl1, and NDRE were the most important VIs for predicting RCC, while yellow and red bands were the most important for DM prediction, and yellow and nir bands for RWC prediction. The best model found for each target was used to model its seasonal trend and produce a variability map. This approach highlights the potential of integrating ML and high-resolution satellite imagery for the remote monitoring of wheat, which can support sustainable farming practices.

Chlorophyll fluorescence is a useful indicator of a plant’s physiological status, particularly under stress conditions. Remote sensing is an increasingly adopted technology in modern agriculture, allowing the acquisition of crop information (e.g., chlorophyll fluorescence) without direct contact, reducing fieldwork. The objective of this study is to improve the monitoring of olive tree fluorescence (Fv′/Fm′) via remote sensing in a Mediterranean environment, where the frequency of stress factors, such as drought, is increasing. An advanced approach combining explainable artificial intelligence and multispectral Sentinel-2 satellite data was developed to predict olive tree fluorescence. Field measurements were conducted in southeastern Italy on two olive groves: one irrigated and the other one under rainfed conditions. Sentinel-2 reflectance bands and vegetation indices were used as predictors and different machine learning algorithms were tested and compared. Random Forest showed the highest predictive accuracy, particularly when Sentinel-2 reflectance bands were used as predictors. Using spectral bands preserves more information per observation, enabling models to detect variations that VIs might miss. Additionally, raw reflectance data minimizes potential bias that could arise from selecting specific indices. SHapley Additive exPlanations (SHAP) analysis was performed to explain the model. Random Forest showed the highest predictive accuracy, particularly when using Sentinel-2 reflectance bands as predictors. Key spectral regions associated with Fv′/Fm′, such as red-edge and NIR, were identified. The results highlight the potential of integrating remote sensing and machine learning to improve olive grove management, providing a useful tool for early stress detection and targeted interventions.