Aggregation of parameters for the land surface model CLASS

Abstract

Land surface schemes are used in climate and weather forecasting models at various resolutions requiring the use of effective or aggregated parameters to adequately represent each grid square. In this study we investigate the rules for aggregating the surface parameters for the Canadian Land Surface Scheme (CLASS). The method consists of running a one‐dimensional version of CLASS over a period of 105 days in summer using meteorological data observed at an agricultural site near Quebec City. The aggregation of parameters is tested by successively running the model with two homogeneous values (usually a small one and a large one) of a chosen parameter and then with a mean or aggregated value of that parameter; the results of the latter run are then compared with the area‐averaged results of the two homogeneous runs. Vegetation coverage, rooting depth, soil texture and roughness lengths are the input parameters thus tested. Heterogeneity of soil moisture content due to uneven distribution of precipitation is also discussed. The results indicate that the sub‐areas of CLASS must have their own soil variables, that roots must occupy full soil layers and not part of a layer, and that the aggregating rule for the roughness lengths (z_o for momentum and z_ot for heat and moisture) should be changed from the current logarithmic averages to the blending height method for z_o and to a new formula involving both roughness lengths for z_ot. The surface‐layer scheme in CLASS was found inadequate and replaced. Results for soil texture aggregation are not as clear; it seems difficult to obtain simultaneously a good averaging of atmospheric energy fluxes and a good averaging of soil moisture contents, runoff and drainage. Horizontal variability of soil moisture due to uneven distribution of rainfall generates an overestimate of evapotranspiration and an underestimate of runoff in an aggregated model lacking this effect. Preliminary results indicate that, in order to effectively parametrize this effect in CLASS, both surface ponding capacity and ground infiltration rate must be reduced over the grid square when convective precipitation occurs.