Global Patterns of Post-Fire Vegetation Productivity Recovery

Fire is a major ecosystem disturbance impacting the global carbon cycle, with its frequency and severity projected to increase. The time required for ecosystems to recover productivity after fire (recovery time) is an important metric for resilience, yet its global patterns remain poorly quantified. Here, we conduct a global analysis using Moderate Resolution Imaging Spectroradiometer (MODIS) observations from 2001 to 2024. We employ satellite-derived burned area data and the near-infrared reflectance of vegetation (NIRv) as a robust proxy for Gross Primary Production (GPP) to track recovery, which is defined as the duration to recover 90% of pre-fire productivity. Our analysis focuses on single-fire events, filtering out areas with recurrent disturbances, and defines recovery as the point when at least 90% of pre-fire productivity is regained.

Our results reveal that the global mean post-fire recovery time is 3.9 ± 0.3 years. This average is masked by strong geographical disparities: recovery follows a pronounced latitudinal gradient, with boreal ecosystems (≥50°N) requiring nearly twice as long to recover (5.6 ± 0.5 years) compared to tropical regions (3.0 ± 0.2 years). Evergreen needleleaf forests exhibit the longest recovery times (6.3 ± 0.9 years), while savannas and grasslands recover fastest. Statistical machine learning modeling identifies the magnitude of the immediate fire-induced GPP loss as the dominant factor controlling recovery duration, with burn severity and pre-fire productivity acting as important secondary drivers.

We show that CMIP6 Earth System Models (ESMs) significantly underestimate these recovery periods (simulating a global mean of 1.8 ± 0.1 years) and fail to capture the observed spatial heterogeneity, particularly in high-latitude regions. This suggests that current models may overestimate the carbon sink capacity of regenerating post-fire landscapes and underestimate positive fire-vegetation feedbacks. Our findings provide a new observational benchmark for improving the representation of post-disturbance dynamics in land surface models and refining global carbon budget assessments.