Sensitivity of surface climate to land surface parameters: A case study using the simple biosphere model SiB2

[1] Significant uncertainties may result from numerical models if fed with inappropriate input data. Biosphere-atmosphere transfer models are sensitive to input vegetation parameters, and the degree to which parameter estimates rely on the definition of the land cover type varies with models. In this study we use the simple biosphere model (SiB2) to evaluate uncertainties associated with misclassification of the land cover type and how they propagate to surface climate variables. We estimate that in regions with heterogeneous landscapes, the aggregation of land cover types from 1 x 1 km to 100 x 100 km horizontal resolution overestimates the area of the dominant type by up to 70%. The largest uncertainties associated with land cover misclassification are found in leaf area index and roughness length both of which have significant impact on the fluxes of carbon, water and energy at the earth surface. Other important uncertainties occur when the misclassification confuses plants with different carbon pathways. An assessment of the uncertainties is obtained comparing outcomes resulting from a choice of a dominant type in a 100 x 100 km area to those obtained using a mosaic of land cover composition weighted by its fractional cover. The difference shows the choice of the dominant type to be cooler by 0.6 degrees C than the average of the mosaic at local noon, while at night it is warmer by 1.7 degrees C. Our results indicate that the diurnal temperature range (dtr) varies from 13 degrees C for the dominant type to 15 degrees C for the weighted average. The difference in the dtr is due to higher minimum temperature simulated with the dominant type. The choice of a dominant type also results in a daily carbon assimilation loss of 28,000 gC compared to the average.