“…and the forests will echo with laughter!”: Remote sensing and bioacoustic integration for assessing Picus viridis distribution in structurally heterogeneous riparian forests

Riparian forests are biodiversity hotspots, yet their structural complexity has long hindered systematic habitat monitoring. Until recently, the lack of integrated technologies limited the identification of priority conservation areas at fine spatial scales. We investigated the habitat suitability of P. viridis , a bioindicator of mature and structurally complex woodlands, in three riparian biotopes of South Tyrol (Italy). Acoustic occurrence data were obtained from AudioMoth recorders and processed with BirdNET to derive daily presence/absence observations. Environmental predictors included LiDAR-derived structural metrics, Sentinel-2 NDVI descriptors, and spectral heterogeneity metrics aggregated at 90 m resolution. We modelled daily acoustic occurrence using Partial Least Squares Regression (PLSR) and generated relative acoustic habitat suitability maps, assessing model robustness through leave-one-logger-out and leave-one-area-out validation frameworks. Local predictive performance varied among study areas, and suitability maps revealed spatial associations between predicted relative acoustic suitability and local forest-structure gradients. However, transferability among biotopes was limited, with marked declines in predictive performance when models were transferred across study areas. Predictor importance also varied substantially among riparian systems, suggesting strong local ecological context dependence. Logger level validation showed only limited agreement between observed and predicted occurrence rates, while between-area transferability was poor. This study shows that integrating passive acoustics and remote sensing can help characterize local patterns of relative acoustic habitat suitability, while also highlighting important limitations in transferring species distribution models across heterogeneous landscapes. These findings emphasize the need for caution when extrapolating acoustic SDMs beyond local calibration contexts.