Indicator-based assessment of post-fire recovery dynamics using satellite NDVI time-series

Full abstract

Fire disturbance severely modifies ecosystem structure and functioning, and therefore predicting post-fire responses is pivotal to improve land management. Indicators that efficiently link post-fire recovery with a timely decision on landscape management can play a key role in the governance of fire risk. We describe a framework to evaluate post-fire recovery based on remotely-sensed measures of relative vegetation recovery, calculated from satellite NDVI time-series. Three indicators are proposed: the novel Cumulative Relative Recovery Index (CRRI), measuring the (mid-long term) extent and completeness of recovery; the Recovery Trend Index (RTI), measuring the steepness of the mid-term post-fire recovery trend; and the Half Recovery Time index (HRT), a measure of the short-term recovery rate. We used Random Forest (RF) models to predict the observed recovery patterns and ranked the predictive importance of several candidate explanatory factors. The performance of RF models ranged from good (CRRI, RTI) to moderate (HRT). Three sets of predictive variables consistently ranked higher: fire traits, landscape composition, and post-fire climatic conditions. The relative contribution of individual variables was different across recovery indicators. These results show that proposed indicators seem to capture different facets of the post-fire recovery process. The short-term recovery indicator (HRT) was linked to landscape composition and post-fire climate. Thus, HRT expresses the speed of initial recovery, related to differences in fire-response traits of vegetation and to climatic conditions immediately following fire. The mid-term recovery indicator (RTI) was mainly influenced by fire traits and post-fire climatic conditions. This indicator captures multiple interacting effects that shape the recovery process related to fire severity, vegetation type and post-fire conditions. Finally, the long-term recovery indicator (CRRI) was clearly more influenced by fire attributes related to severity than by vegetation type and structure or by post-fire climatic conditions. Overall, our results suggest that a combination of biotic processes (driven by plant life-history traits) and abiotic filters (e.g., post-fire climate) determine the early post-fire recovery process. Conversely, the mid to long-term recovery response (expressing its completeness) is driven by the depletion of resilience capacity and by the amount of change in vegetation structure and functioning modulated by spatial differences in fire severity. Our results strongly suggest that an indicator-based approach grounded on satellite time-series of vegetation indices can effectively cover various facets of post-fire recovery. This will improve the monitoring and prediction of post-fire recovery dynamics, with valuable applications in fire hazard management and post-fire ecosystem restoration and monitoring.