Data: 374 Scandinavian grid cells (Norway + Sweden) on a ~50 km lattice. Svalbard and edge-case cells with incomplete prediction coverage have been excluded. Model version v2.0. Native browser build using Leaflet, Plotly.js, and Arquero; no Python runtime required; no data leaves your device.

Disclaimer. Interface-testing release. The deer variables D_roe, D_moose, pred_roe_density and pred_moose_density are relative density indices on [0, 1] (min–max-normalised hunter-harvest rates per Eq. 1a; Random-Forest projections per Eq. 1c), not absolute densities in animals km⁻². H is the wolf habitat suitability (HurdleGBM-predicted relative wolf density per Eq. 2), not a prey quantity. Moose refers to moose calves (Alces alces calves). Annex 6.6 of the underlying paper reports roe-deer validation R² = 0.855 and moose-calf R² = 0.446, with moose RMSE rising from 0.072 (2016) to 0.176 (2024); the vegetation–deer stationarity assumption partially fails for moose after 2020. These outputs are under active review by Dr. Barbara Zimmermann (INN) and must not be cited, redistributed, or used for management decisions until a validated v3 release is announced.

GBIF-calibrated overlay. When available in the variable menu, columns suffixed _cal show a capped-weight per-grid bias correction against independent GBIF occurrence records (Canis lupus, Capreolus capreolus, Alces alces; Norway and Sweden, 1999–2024) aggregated to the same 50 km grid and min–max-normalised on log(1 + n) per Isaac et al. (2014, Methods Ecol. Evol.). Maximum GBIF influence is capped at 50 % of the per-grid Spearman ρ over the 1999–2015 training window; grids with ρ ≤ 0 or fewer than 5 overlap years are passed through unchanged. GBIF cannot distinguish moose calves from adults, so D_moose_cal corrects spatial pattern only, not magnitude. gbif_agreement reports the per-grid average ρ across all three species.