Abstract:
In this talk, we demonstrate an ability to forecast carbon fluxes accurately at multi-month leads
in the Northern Hemisphere boreal region, an ability that appears to be linked in part to
snowpack initialization in the forecast model. Using 20 years of forecasted meteorology from
NASA GMAO’s S2S ensemble forecast system (from forecasts initialized in December each
year) to drive offline runs of the Catchment-CN model, we evaluate the degree to which we can
forecast greening onset date and Gross Primary Production (GPP) relative to a fully
independent observation dataset. We find that skillful forecasts of the greening onset date
largely occur where we have skillful forecasts of snow cover removal date in western North
America and Europe, and this snow cover removal date is naturally tied to the amount of snow
at the start of the forecast. The ability to predict greening onset date in turn leads to an ability to
predict annual (or more specifically, January-September) GPP, presumably because an earlier
greening start date implies a longer time period for carbon uptake. We also, however, identify
some regions for which GPP is accurately predicted without the snow mechanism; in these
areas, it appears to be the initialization of the carbon/vegetation reservoirs in December that
leads to the skill. Overall, this study demonstrates the significance of accurate land initialization
for S2S carbon forecasts.