Exploring Sources of Multi-year Predictability of Terrestrial Ecosystem

The demand for decision-relevant and evidence-based near-term climate information is increasing. This includes understanding and explaining the variability and changes in ecosystems to support disaster management and adaptation choices. As climate prediction from seasonal to decadal (S2D) expands to encompass Earth system dimensions, including terrestrial and marine ecosystems, it is crucial to deepen our scientific understanding of the long-term predictability sources for ecosystem variability and change. Here we explore to what extent terrestrial ecosystem variables are driven by large-scale - potentially predictable -climate modes of variability and external forcings or whether regional random environmental factors are dominant. To address these issues, we utilize a multi-year prediction system based on Community Earth System Model version 2 (CESM2).  The system consists of 50-member uninitialized historical simulations, 20-member ocean assimilations, and 20-member hindcast initiated from every January 1^st integrating for 5 years from 1961 to 2021. The key variables assessed are surface temperature, precipitation, soil moisture, wildfire occurrence, and Gross Primary Productivity. Our results suggest that land surface processes and ecosystem variables over many parts of the globe can be potentially predictable 1 to 3 years ahead originating from anthropogenic forced signals and modes of climate variability, particularly El Nino and Southern Oscillation and Atlantic Multi-decadal variability. These global modes of climate variability shift regional temperature and precipitation patterns, leading to changes in soil moisture, wildfire occurrence, and terrestrial productivity.