Vegetation recovery drivers at short-term after fire are plant community-dependent in mediterranean burned landscapes

Abstract

Large, severe wildfires can cause unpredictable changes in Mediterranean plant communities that undermine ecosystem resilience. Understanding the geophysical drivers shaping vegetation dynamics after fire is vital to provide integrated knowledge into plant community recovery feedbacks in the context of global change. We investigated the role of fire severity and a comprehensive set of geophysical drivers in the post-fire vegetation recovery of Mediterranean landscapes. We used physical-based remote sensing techniques to estimate vegetation recovery and examined whether this recovery is plant-community dependent in the short-term (one year) after the fire. The fractional vegetation cover (FCOVER) recovery was selected as a resilience indicator for Pinus sylvestris L. (Scots pine) forests, Quercus ilex L. (holm oak) woodlands, shrublands dominated by Cytisus oromediterraneus Rivas Mart. et al. (black broom) and grasslands. To obtain the pre- and post-fire FCOVER, we used the Gaussian processes regression algorithm trained with the PROSAIL-D radiative transfer model. The algorithm was parameterized to account for the variability of biophysical conditions in the plant communities considered and was applied to the Sentinel-2 multispectral imagery. Validation of remotely-sensed FCOVER with field data in burned and control plots was successful (R2 = 0.84 and RMSE = 10.83%), without significant under or overestimation. We selected 31 variables pertaining to fire severity, climate, post-fire soil moisture, pre-fire soil condition, and landscape configuration as putative predictors of FCOVER recovery percentage (FCOVER%rec), calculated as the ratio of post-fire to pre-fire FCOVER. Random Forest regression (RFR) was used to disentangle the influences of fire severity and geophysical drivers on community-specific FCOVER%rec. Fire severity and post-fire surface soil moisture variables were relevant predictors in RFR models in all plant communities. However, their importance was largely community-dependent. In grasslands, surface soil moisture in the early summer season one year after fire was the major variable explaining variability in FCOVER%rec. In holm oak woodlands, the contribution of fire severity and surface soil moisture was balanced. Conversely, fire severity largely outweighed other variables in driving FCOVER%rec in black broom shrublands and Scots pine forests. Pre-fire climate, soil condition and landscape configuration variables were not meaningful predictors in general. Our findings offer novel insights into the processes that underlie resilience in fire-prone plant communities located in the western Mediterranean Basin. This establishes a baseline to help prioritize adaptive management strategies for the most vulnerable areas in the region.