Post-fire regeneration in Mediterranean pine forests is increasingly constrained by recurrent high-severity fires and changing climate conditions, while post-disturbance interventions such as salvage logging may further hinder natural recovery at the local scale. We assessed the drivers of early natural regeneration in a coastal Pinus halepensis Mill. forest in Spotorno (NW Italy) affected by two subsequent high-severity wildfires (September 2006 and July 2015) followed by salvage logging. Pine seedlings were mapped in spring 2025 with an RTK GNSS antenna, while high-resolution UAV images and LiDAR products were used to derive terrain- and forest structurebased predictors. Topographically mediated constraints on regeneration were quantified using the Topographic Wetness Index (TWI) and the Heat Load Index (HLI), which capture spatial variation in soil moisture accumulation and heat exposure. Seed availability was represented by the distance from each seedling to the nearest remnant adult pine, identified from the canopy height model using a local-maximum filtering approach. Spatial point pattern analysis was used to test whether empirically evident regeneration clusters reflected plant-plant interactions, or environmentally-driven density variation. Drivers of regeneration were modelled using GLMs, GAMs and Random Forests (RF), and two pseudo-absence strategies in the RF were explicitly compared by training models with (i) ecologically informed, spatially homogeneous pseudo-absences and (ii) randomly sampled pseudo-absences. The informed pseudo-absence Random Forest achieved substantially higher discrimination (AUC = 0.895; ACC = 0.821; SEN = 0.785; SPE = 0.864) than the random-absence model (AUC = 0.653; ACC = 0.607; SEN = 0.648; SPE = 0.571).). The best model was applied to generate a 5 & times; 5 m ecological suitability map identifying regeneration "hotspots", i.e., near seed sources, in warm, well-drained microsite conditions, and persistent "coldspots" in convergent terrain and seed-limited areas. This workflow provides an operational, transferable basis for precision-oriented post-fire restoration planning in Mediterranean landscapes where passive recovery is uncertain.
Fine-scale topographic filtering and seed limitation shape post-fire regeneration patterns in Mediterranean pine forests / Atzeni, F., Taccaliti, F., Marangon, D., Lingua, E.. - In: FOREST ECOLOGY AND MANAGEMENT. - ISSN 0378-1127. - 619:(2026). [10.1016/j.foreco.2026.124069]
Fine-scale topographic filtering and seed limitation shape post-fire regeneration patterns in Mediterranean pine forests
Atzeni F.
Primo
Methodology
;Taccaliti F.Investigation
;Lingua E.Ultimo
Validation
2026-01-01
Abstract
Post-fire regeneration in Mediterranean pine forests is increasingly constrained by recurrent high-severity fires and changing climate conditions, while post-disturbance interventions such as salvage logging may further hinder natural recovery at the local scale. We assessed the drivers of early natural regeneration in a coastal Pinus halepensis Mill. forest in Spotorno (NW Italy) affected by two subsequent high-severity wildfires (September 2006 and July 2015) followed by salvage logging. Pine seedlings were mapped in spring 2025 with an RTK GNSS antenna, while high-resolution UAV images and LiDAR products were used to derive terrain- and forest structurebased predictors. Topographically mediated constraints on regeneration were quantified using the Topographic Wetness Index (TWI) and the Heat Load Index (HLI), which capture spatial variation in soil moisture accumulation and heat exposure. Seed availability was represented by the distance from each seedling to the nearest remnant adult pine, identified from the canopy height model using a local-maximum filtering approach. Spatial point pattern analysis was used to test whether empirically evident regeneration clusters reflected plant-plant interactions, or environmentally-driven density variation. Drivers of regeneration were modelled using GLMs, GAMs and Random Forests (RF), and two pseudo-absence strategies in the RF were explicitly compared by training models with (i) ecologically informed, spatially homogeneous pseudo-absences and (ii) randomly sampled pseudo-absences. The informed pseudo-absence Random Forest achieved substantially higher discrimination (AUC = 0.895; ACC = 0.821; SEN = 0.785; SPE = 0.864) than the random-absence model (AUC = 0.653; ACC = 0.607; SEN = 0.648; SPE = 0.571).). The best model was applied to generate a 5 & times; 5 m ecological suitability map identifying regeneration "hotspots", i.e., near seed sources, in warm, well-drained microsite conditions, and persistent "coldspots" in convergent terrain and seed-limited areas. This workflow provides an operational, transferable basis for precision-oriented post-fire restoration planning in Mediterranean landscapes where passive recovery is uncertain.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
