Remote sensing of bird communities in the Peruvian Amazon: evaluating Landsat predictors for biodiversity monitoring

Over one-third of the Amazon rainforest has been lost or degraded through human activities. While Earth observation satellites effectively monitor forest cover, assessing the impacts of degradation on biodiversity at large scales remains challenging. We analysed 3,129 bird surveys conducted over 16 years in the Tambopata Forest, south-eastern Peru, to evaluate whether remote sensing (RS) variables derived from Landsat imagery can predict variation in bird community composition. We compared the performance of RS-based predictors with traditional habitat descriptors in modelling the occurrence probabilities of 135 commonly recorded bird species. Models using Landsat reflectance outperformed those based on habitat data for predicting species occupancy (mean AUC = 0.68 vs 0.58) and achieved high predictive accuracy (AUC > 0.7) for more species (49 vs 20). However, low detection rates limited the ability of all models to estimate true community composition and to detect temporal change. To address this, we recommend survey designs that prioritise greater replication at fewer sites, thereby improving detection rates and the power to monitor biodiversity trends. Our findings highlight the potential for integrating satellite-based environmental variables with improved survey designs to enhance biodiversity monitoring in tropical forests.