What is the importance of climate model bias when projecting the impacts of climate change on land surface processes?

[摘要] Regional climate change impact (CCI) studies have widely involveddownscaling and bias correcting (BC) global climate model (GCM)-projectedclimate for driving land surface models. However, BC may cause uncertaintiesin projecting hydrologic and biogeochemical responses to future climate dueto the impaired spatiotemporal covariance of climate variables and abreakdown of physical conservation principles. Here we quantify the impactof BC on simulated climate-driven changes in water variables(evapotranspiration (ET), runoff, snow water equivalent (SWE), and waterdemand for irrigation), crop yield, biogenic volatile organic compounds(BVOC), nitric oxide (NO) emissions, and dissolved inorganic nitrogen (DIN)export over the Pacific Northwest (PNW) region. We also quantify the impactson net primary production (NPP) over a small watershed in the region (HJ-Andrews). Simulation results from the coupled ECHAM5–MPI-OM model with A1Bemission scenario were first dynamically downscaled to 12 km resolution withthe WRF model. Then a quantile-mapping-based statistical downscaling modelwas used to downscale them into 1/16° resolution daily climate dataover historical and future periods. Two climate data series were generated,with bias correction (BC) and without bias correction (NBC). Impact modelswere then applied to estimate hydrologic and biogeochemical responses toboth BC and NBC meteorological data sets. These impact models include amacroscale hydrologic model (VIC), a coupled cropping system model(VIC-CropSyst), an ecohydrological model (RHESSys), a biogenic emissions model(MEGAN), and a nutrient export model (Global-NEWS).

Results demonstrate that the BC and NBC climate data provide consistentestimates of the climate-driven changes in water fluxes (ET, runoff, andwater demand), VOCs (isoprene and monoterpenes) and NO emissions, mean cropyield, and river DIN export over the PNW domain. However, significantdifferences rise from projected SWE, crop yield from dry lands, and HJ-Andrews's ET between BC and NBC data. Even though BC post-processing has nosignificant impacts on most of the studied variables when taking PNW as awhole, their effects have large spatial variations and some local areas aresubstantially influenced. In addition, there are months during which BC andNBC post-processing produces significant differences in projected changes,such as summer runoff. Factor-controlled simulations indicate that BCpost-processing of precipitation and temperature both substantiallycontribute to these differences at regional scales.

We conclude that there are trade-offs between using BC climate data foroffline CCI studies versus directly modeled climate data. These trade-offsshould be considered when designing integrated modeling frameworks forspecific applications; for example, BC may be more important when consideringimpacts on reservoir operations in mountainous watersheds than wheninvestigating impacts on biogenic emissions and air quality, for which VOCsare a primary indicator.

[学科分类] 地球化学与岩石