Modelling carbon responses of tundra ecosystems to historical and projected climate: sensitivity of pan‐Arctic carbon storage to temporal and spatial variation in climate

Historical and projected climate trends for high latitudes show substantial temporal and spatial variability. To identify uncertainties in simulating carbon (C) dynamics for pan‐Arctic tundra, we compare the historical and projected responses of tundra C storage from 1921 to 2100 between simulations by the Terrestrial Ecosystem Model (TEM) for the pan‐Arctic and the Kuparuk River Basin, which was the focus of an integrated study of C dynamics from 1994 to 1996. In the historical period from 1921 to 1994, the responses of net primary production (NPP) and heterotrophic respiration (RH) simulated for the Kuparuk River Basin and the pan‐Arctic are correlated with the same factors; NPP is positively correlated with net nitrogen mineralization (NMIN) and RH is negatively correlated with mean annual soil moisture. In comparison to the historical period, the spatially aggregated responses of NPP and RH for the Kuparuk River Basin and the pan‐Arctic in our simulations for the projected period have different sensitivities to temperature, soil moisture and NMIN. In addition to being sensitive to soil moisture during the projected period, RH is also sensitive to temperature and there is a significant correlation between RH and NMIN. We interpret the increases in NPP during the projected period as being driven primarily by increases in NMIN, and that the correlation between NPP and temperature in the projected period is a result primarily of the causal linkage between temperature, RH, and NMIN. Although similar factors appear to be controlling simulated regional‐and biome‐scale C dynamics, simulated C dynamics at the two scales differ in magnitude with higher increases in C storage simulated for the Kuparuk River Basin than for the pan‐Arctic at the end of the historical period and throughout the projected period. Also, the results of the simulations indicate that responses of C storage show different climate sensitivities at regional and pan‐Arctic spatial scales and that these sensitivities change across the temporal scope of the simulations. The results of the TEM simulations indicate that the scaling of C dynamics to a region of arctic tundra may not represent C dynamics of pan‐Arctic tundra because of the limited spatial variation in climate and vegetation within a region relative to the pan‐Arctic. For reducing uncertainties, our analyses highlight the importance of incorporating the understanding gained from process‐level studies of C dynamics in a region of arctic tundra into process‐based models that simulate C dynamics in a spatially explicit fashion across the spatial domain of pan‐Arctic tundra. Also, efforts to improve gridded datasets of historical climate for the pan‐Arctic would advance the ability to assess the responses of C dynamics for pan‐Arctic tundra in a more realistic fashion. A major challenge will be to incorporate topographic controls over soil moisture in assessing the response of C storage for pan‐Arctic tundra.