Complementarity of Camera Traps and Ranger Patrols Shapes Rhino Detection Patterns in Matobo National Park, Zimbabwe

Abstract

Poaching remains a critical threat to African rhinoceros' populations, necessitating monitoring systems that are both ecologically informative and operationally effective. However, detection data are shaped not only by animal distribution but also by how monitoring design interacts with the environment. We evaluated how camera traps and ranger patrols differ in their environmental detection signatures and spatial coverage for rhino monitoring in Matobo National Park, Zimbabwe. We modelled rhino detections using 19 bioclimatic variables, Normalised Difference Vegetation Index (NDVI), and distances to roads and water sources. We built separate Random Forest models for patrol-only, camera trap-only, and combined datasets. We found that camera trap detections were most strongly associated with NDVI and proximity to water, whereas patrol detections were primarily structured by proximity to roads. Precipitation-related variables contributed more strongly than temperature variables across models, indicating that rainfall gradients play an important role in structuring detection patterns. Model performance improved substantially when datasets were combined (AUC = 0.92) relative to patrol-only (AUC = 0.84) or camera-only (AUC = 0.76) models, and the integrated model predicted the largest spatial detection extent. Our findings demonstrate that monitoring outputs reflect both ecological processes and sampling design. Camera traps and patrols capture complementary components of rhino space use, and integrating both methods reduces method-specific bias. Effective rhino conservation therefore requires explicitly aligning monitoring strategies with ecological objectives while maintaining broad spatial security coverage.