Environmental monitoring agencies often operate under strict intervention budgets, which makes the practical value of a forecast depend less on overall cell-wise accuracy and more on whether the model correctly prioritizes the small subset of locations that matter most. In this thesis, the deforestation forecasting problem is reformulated as a spatial ranking problem, and a differentiable adaptation of probabilistic Learning-to-Rank to dense two-dimensional prediction is presented. The proposed framework combines a Top-K-oriented objective with a tournament- based sampling strategy that includes ground-truth injection to expose severely underestimated targets during training. Experiments on DETER-B data from the Brazilian Amazon, using historical observations from 2018 to 2024 for training and a fully unseen 2025 period for testing, show that the proposed approach improves target prioritization and rank quality compared to conventional regression-based losses, while also distributing predictions more widely across distinct deforestation hotspots.