UC Berkeley

Discontinuous permafrost environments exhibit strong spatial heterogeneity at scales too small to be driven by weather forcing or captured by Earth System Models. Here we analyze effects of observed spatial heterogeneity in soil and vegetation properties, hydrology, and thermal dynamics on ecosystem carbon dynamics in a watershed on the Seward Peninsula in Alaska. We apply a Morris global sensitivity analysis to a process-rich, successfully tested terrestrial ecosystem model (TEM), ecosys, varying soil properties, boundary conditions, and weather forcing. We show that landscape heterogeneity strongly impacts soil temperatures and vegetation composition. Snow depth, O-horizon thickness, and near-surface water content, which vary at scales of O(m), control the soil thermal regime more than an air temperature gradient corresponding to a 140 km north–south distance. High shrub productivity is simulated only in talik (perennially unfrozen) soils with high nitrogen availability. Through these effects on plant and permafrost dynamics, landscape heterogeneity impacts ecosystem productivity. Simulations with near-surface taliks have higher microbial respiration (by 78.0 gC m−2 yr−1) and higher net primary productivity (by 104.9 gC m−2 yr−1) compared to runs with near-surface permafrost, and simulations with high shrub productivity have outlying values of net carbon uptake. We explored the prediction uncertainty associated with ignoring observed landscape heterogeneity, and found that watershed net carbon uptake is 60% larger when heterogeneity is accounted for. Our results highlight the complexity inherent in discontinuous permafrost environments and demonstrate that missing representation of subgrid heterogeneity in TEMs could bias predictions of high-latitude carbon budget.