Sensor choice matters: How multi-source remote sensing reshapes ecosystem service hotspot footprints

Spatial planning of ecosystem services (ES) requires maps that indicate where interventions may be most effective, yet the sensitivity of ES hotspot locations to alternative remote sensing (RS) data configurations is rarely examined. Using a 41-ha forest area in southern Finland as a case study, we assessed how stable ES hotspot footprints are across RS predictor tiers of increasing information content: Sentinel-2 (S2) imagery, S2 combined with aerial imagery (S2+AER), and S2 combined with aerial imagery and airborne laser scanning (S2+AER+ALS). We modeled six ESs (bilberry and cowberry potential, visual amenity, biodiversity conservation, carbon storage, and timber production) using Random Forest and derived tier-specific priority maps with Zonation. Statistically significant ES hotspots were identified with Local Moran’s I , and cross-tier agreement was quantified using complementary overlap and stability metrics with spatial block-bootstrap uncertainty. Across ESs, adding aerial imagery to S2 largely preserved hotspot footprints, with high spatial overlap and directional stability. In contrast, adding ALS improved predictive accuracy but frequently reorganized hotspot locations, with the strongest changes observed for carbon storage, biodiversity conservation, and visual amenity. Bilberry and cowberry remained stable across predictor tiers. Hotspots common to all tiers formed compact core areas but covered only a limited fraction of the landscape. These case-study results demonstrate that predictive accuracy and hotspot stability represent distinct properties of ES maps. Additional RS information may alter prioritization outcomes even when predictive performance improves. The workflow can complement conventional accuracy assessment, but the observed sensor-specific effects require validation across other ecological and management contexts.