Simulated Vegetation Response to Climate Change in California: The Importance of Seasonal Production Patterns

MC1 dynamic global vegetation model simulates vegetation response to climate change by simulating vegetation production, soil biogeochemistry, plant biogeography and fire. It has been applied at a wide range of spatial scales, yet the spatio-temporal patterns of simulated vegetation production, which drives the model's response to climate change, has not been examined in detail. We ran MC1 for California at a relatively fine scale, 30 arc-seconds, for the historical period (1895-2006) and for the future (2007-2100), using downscaled data from four CMIP3-based climate projections: A2 and B1 GHG emissions scenarios simulated by PCM and GFDL GCMs. The use of these four climate projections aligns our work with a body of climate change research work commissioned by the California Public Interest Energy Research (PIER) Program. The four climate projections vary not only in terms of changes in their annual means, but in the seasonality of projected climate change. We calibrated MC1 using MODIS NPP data for 2000-2011 as a guide, and adapting a published technique for adjusting simulated vegetation production by increasing the simulated plant rooting depths. We evaluated the simulation results by comparing the model output for the historical period with several benchmark datasets, summarizing by EPA Level 3 Ecoregions. Multi-year summary statistics of model predictions compare moderately well with Kuchler's potential natural vegetation map, National Biomass and Carbon Dataset, Leenhouts' compilation of fire return intervals, and, of course, the MODIS NPP data for 2000-2011. When we compared MC1's monthly NPP values with MODIS monthly GPP data (2000-2011), however, the seasonal patterns compared very poorly, with NPP/GPP ratio for spring (Mar-Apr-May) often exceeding 1, and the NPP/GPP ratio for summer (Jun-Jul-Aug) often flattening to zero. This suggests MC1's vegetation production algorithms are overly biased for spring production at the cost of summer production. We summarized MC1 outputs under the four climate change projections also by computing statistics by EPA Level 3 Ecoregions. MC1 predicts the greatest shift of vegetation types for the drier biomes of California, including the Sonoran Desert and the Eastern Cascade Slopes and Foothills. Strong shifts are predicted for forest biomes, including the Sierra Nevada (42-74%), and the Coast Range (24-40%). The simulated shifts in vegetation types are mediated by simultaneous changes in live vegetation carbon, which in turn are driven by simulated net primary production (NPP) at a monthly time step. MC1's poor skill in reproducing seasonal variations of NPP when compared to MODIS monthly GPP values suggests that the resulting vegetation type shifts are likely inaccurately predicted by MC1. Improving MC1's algorithm for NPP to more closely track seasonal variations in MODIS GPP should result in more reliable predictions of vegetation response to climate change.