Capturing early signs of drought stress in tree saplings with a low-cost thermal camera to support irrigation monitoring in drought-prone reforestation sites

As climate change intensifies drought events worldwide, monitoring plant water status has become a key challenge for sustainable ecosystem management. Traditional methods for assessing drought stress in trees are often time-consuming, expensive, or destructive. This study evaluates the potential of low-cost thermal cameras to detect early drought stress in tree saplings and support irrigation scheduling in reforestation contexts. In a greenhouse experiment, thermal images were acquired along with ecophysiological and environmental variables for three commonly planted tree species in Central Europe. We used these thermal images to assess drought stress with the crop water stress index (CWSI), which has been used for a variety of tree species. For CWSI calculation, saplings were isolated from the background using Segment Anything 2. Across species, CWSI showed moderate yet significant (p ' 0.001) correlations with drought stress indicators such as the daily minimum tree water deficit (TWD_min, ρ = 0.42), predawn water potential (ρ = −0.42), or soil water content (SWC, ρ = −0.45). Early drought stress, defined by TWD_min recovery, was predicted using generalized additive models. With CWSI as the main predictor, a balanced accuracy (BA) of 70.7% (Cohen's Kappa κ = 0.44) was achieved. Including additional predictors (SWC, vapour pressure deficit) improved model performance (BA 92.4%, κ = 0.85), at the cost of higher measurement efforts compared to using CWSI alone. For sites with skeletal soils, where measuring soil moisture is challenging, the CWSI-based approach offers a practical tool to enhance drought stress monitoring and irrigation scheduling. Finally, thresholds indicating early drought stress were derived, providing a basis for operational irrigation scheduling.