Author(s): Maria Belen Pedoja Balparda; Juan Antonio Rodriguez Diaz; Maria Pilar Montesinos Barrios; Rafael Gonzalez Perea

Linked Author(s):

Keywords: Machine learning; Precision agriculture; Irrigation; Crop modelling; Remote sensors

Abstract: This study develops a tool to create Digital Twins for precision irrigation agriculture, integrating the hydraulic model of the irrigation system with the soil-crop-atmosphere system. As a structural component, an automated satellite image download module was integrated into the system, establishing a direct API connection with the data distribution platforms corresponding to the PlanetScope and Sentinel-2 constellations. Using the images obtained through this process, various machine learning algorithms for canopy cover estimation were comparatively evaluated. This stage employed a data fusion strategy, correlating the high spatial resolution of UAV images with the medium resolution of PlanetScope images. The selected model (multilayer perceptron neural network) exhibited the best statistical performance, reporting an RMSE of 0.0855 and an R² of 0.944. Finally, the incorporation of the AquaCrop water productivity model into the Digital Twin environment is described, validating the system by estimating de canopy cover of a broccoli crop from the time series images with the optimal model. The integral digital twin can generate irrigation management scenarios to facilitate strategic decision-making focused on water efficiency and sustainable production.

DOI:

Year: 2026