<p>Dryland mountain forests are increasingly exposed to interacting climatic and ecological stressors whose spatially heterogeneous impacts remain insufficiently understood. Here, we present an integrative, spatially explicit framework to assess two decades (2002–2022) of forest biomass dynamics in the Helen Forest Protected Area (HFPA), a climate-sensitive forest ecosystem in the Zagros Mountains of southwestern Iran. The analysis combines field-based tree biomass inventories, Landsat-derived vegetation indices, machine learning–based pest risk modeling, and multi-hazard spatial assessments encompassing drought intensity, dust deposition, thermal conditions, evapotranspiration, wind exposure, and biotic disturbance. Results indicate a pervasive and spatially clustered decline in tree total biomass (TB), with a mean loss of − 3.92 t ha⁻¹ and localized reductions exceeding − 9.6 t ha⁻¹ in highly exposed forest compartments. Spatial autocorrelation analyses reveal extensive High–High clusters of the TB loss, covering 34.6% of the protected area, highlighting structurally vulnerable zones under compounded stress. Multicollinearity diagnostics show that climatic variables constitute a largely global and spatially smooth stress regime, whereas dust deposition and predicted pest presence exhibit pronounced local variability. Consistent with these spatial characteristics, geographically weighted regression (GWR) substantially outperforms global models (adjusted R² = 0.819), capturing strong spatial non-stationarity and identifying dust and pest pressure as key proximate drivers of localized TB decline. By explicitly separating large-scale climatic forcing from spatially heterogeneous ecological stress pathways, this study demonstrates that forest degradation in dryland mountains is shaped by locally amplified processes operating within broader climate constraints. The proposed framework informs precision adaptation climate-risk policy.</p>