Diagnosing post-fire recovery with satellite-based indicators: divergent trajectories and phase-dependent predictor associations after a boreal megafire in Northeast China

Wildfire alters forest attributes that may recover at different rates, yet their temporal divergence and phase-dependent associations remain poorly understood. We developed a multi-indicator diagnostic framework to assess two decades of recovery following the 2003 megafire across 1116 km 2 of boreal forest in Northeast China. The framework integrated the Normalized Difference Vegetation Index (NDVI), Gross Primary Productivity (GPP), Leaf Area Index (LAI), and percent tree cover (PTC), representing vegetation greenness, photosynthetic carbon uptake, leaf area, and tree cover, respectively. Long-term trajectories and predictor associations were derived from satellite products; ground-based and drone surveys were used to evaluate the 2023 vegetation classification. Indicator-specific phases were identified from control-corrected data series, and pixel-level mismatch between indicator- and PTC-based recovery was quantified using explicit thresholds. NDVI, GPP, and LAI approached 0.95 of their product-specific pre-fire values within a decade, whereas PTC remained below 0.80 in 2023. Approximately 43–50% of the valid area reached the corresponding NDVI, GPP, or LAI threshold while remaining below the PTC threshold. Predictor importance and model-based associations shifted among phases, with fire severity and climate ranking higher early and pre-fire vegetation conditions becoming more prominent later. Retrogressive transitions were more frequent in higher-severity classes. Divergence among indicator trajectories was consistent across alternative data sources, temporal-processing methods, and spatial resolutions. By integrating explicit thresholds with spatial disagreement among indicators, this framework advances post-fire assessment toward an ecologically explicit diagnosis of vegetation resilience that distinguishes attributes and identifies persistent constraints, thereby directing management attention to where it is most needed.