Divergent patterns of forest carbon uptake and loss stabilize net carbon balance as disturbance severity increases

Disturbance is broadly reshaping forests, with uncertain implications for the global carbon (C) cycle. Relative to severe disturbances that cause complete tree mortality, the biogeochemical consequences of less severe disturbances from patchy insects, pathogens, and extreme weather-related events are less understood. We implemented the Forest Resilience Threshold Experiment (FoRTE) in 2019 via the stem girdling of >3,700 trees to examine how disturbance severity from zero to 85 % gross defoliation affects forest compositional and structural change, and alters net C balance, the difference between C uptake and loss. We hypothesized that in the three years following disturbance: net C balance would decline non-linearly with increasing severity as net primary production (NPP, C uptake through biomass accumulation) decreased and soil respiratory C losses declined to a lesser extent and only temporarily; and changes in canopy composition and structure would be more strongly correlated with NPP rather than with soil respiratory C losses. Counter to our hypothesis, C uptake as NPP was stable and soil respiratory C losses declined as disturbance severity increased, which supported net C balance stability, even at the highest level of 85% gross defoliation. While forest structure and composition changed with increasing disturbance severity, these shifts correlated with soil respiratory C losses rather than NPP. The significance of these unexpected results is two-fold. First, they demonstrate that net ecosystem C balance can remain remarkably stable immediately after disturbance, but the integrative systems-level response cannot be inferred from analysis of individual C cycling processes, which displayed opposing patterns. Secondly, the stability of NPP and contrasting sensitivity of canopy composition and structure to increasing disturbance severity indicate canopy structure and C uptake responses were decoupled, suggesting common metrics of severity, such as leaf area losses, may not mirror C cycling responses to disturbance. We conclude that predicting how net C balance responds to varying levels of disturbance severity requires integrative, mechanistic understanding of the dynamic, unsynchronized and sometimes opposing changes in C uptake and loss.