Legacies of model initialization on predictions of future ecosystem dynamics in California's Sierra Nevada: insights from GEDI

Summary Accurate estimates of aboveground vegetation structure are essential for making reliable predictions of terrestrial ecosystem responses to climate change. However, traditional small‐scale ground‐based inventory methods cannot easily be scaled up to comprehensive, large‐scale estimates of ecosystem structure. We assimilate remotely‐sensed Light Detection and Ranging measurements of vegetation structure and corresponding imaging‐spectrometry‐derived estimates of canopy composition into the ecosystem demography (ED2.2) terrestrial biosphere model across an elevational transect in California's Sierra Nevada. We then used the model to assess: how incorporating observed ecosystem structure and composition influences predictions of ecosystem change over the coming century as compared to simulations initialized with long‐term potential vegetation; and how ecosystems are predicted to respond differently to future climate change. Our analyses show multi‐decadal impacts of initialization on predictions of ecosystem composition and structure, emphasizing long‐term legacies of climate and disturbance history in predictions of ecosystem responses to climate change that are not captured when models are initialized with outputs from long‐term historical simulations. The remote sensing‐initialized simulations predict increases in aboveground biomass and leaf area index, and pronounced elevation‐dependent changes in canopy composition. The differences among initialization methods, climate scenarios, and elevational gradients have important implications for improving ecosystem modeling and informing land management strategies.