Featured Application UAV-based monitoring combined with machine learning can estimate soil respiration, moisture, and temperature behavior to optimize environmental monitoring and aid decision-making.Abstract This study aimed to characterize the spatiotemporal dynamics of soil respiration (CO2 efflux), soil moisture, and soil temperature across different land-use systems in a semi-arid environment through in situ monthly monitoring and to evaluate the potential of UAV-based imagery combined with Random Forest modeling to spatialize these variables within the agroforestry system. The variables were monitored monthly using an Infrared Gas Analyzer (IRGA) over 9 months, and UAV imagery was acquired at two distinct time points. The 11-month experimental campaign enabled evaluation of seasonal and spatial variability and of soil physical and hydraulic properties. Soil CO2 efflux ranged from 1.0 to 6.7 mu mol m-2 s-1, with higher values observed during the rainy period, closely following soil moisture dynamics. Soil moisture and temperature exhibited clear seasonal patterns driven by rainfall variability. The pasture system showed higher CO2 efflux in most months, while AFS2 presented more stable fluxes over time. In contrast, AFS1 exhibited lower CO2 efflux, likely associated with its soil characteristics. Despite these patterns, no significant differences were observed among land-use systems for most soil physical properties. UAV-derived data combined with machine learning techniques proved effective for modeling soil CO2 efflux, soil temperature, and soil moisture, demonstrating their potential for monitoring soil processes in semi-arid environments. Overall, agroforestry systems did not significantly differ from other land uses in terms of CO2 efflux, likely due to their early stage of development. These findings indicate that the effects of agroforestry systems on soil processes occur gradually and highlight the importance of long-term monitoring to fully capture system dynamics.