Sensitivity of WRF Land Surface Schemes to Land Cover Classification over Complex Alpine Terrain

Presently, Limited Area or High-Resolution Mesoscale (HRM) models with grid spacing on the
order of 1 km are used for numerical weather forecasting. Mountainous terrain is, however,
characterized by large surface heterogeneity since steep topography, urban areas, and different land-
cover types co-exist on small spatial scales. Because of this surface heterogeneity, local small-scale
processes occur within the Mountain Boundary Layer (MoBL) that cannot be explicitly resolved
with a 1-km grid spacing and thus need to be parameterized by the Land Surface Model (LSM) and
the Planetary Boundary Layer (PBL) schemes. The large surface heterogeneity can be poorly
represented in the Land-Use Classification (LUC) and can further lead to errors within the model.
Correct land-use classification is, however, crucial to provide accurate surface characteristics (e.g.,
albedo, roughness length, thermal inertia, emissivity, and soil moisture availability) to correctly
calculate near-surface exchange processes in the LSM. A careful evaluation of the LUCs, the
associated surface characteristics, and their impact on the modeled land-atmosphere exchange
against observations is thus a key to a better understanding of the model’s performance.
We will present Weather and Research Forecasting Model (WRF) simulations with a grid spacing
down to 1 km over the steep Alpine terrain of the Inn Valley, Austria. Focusing on convective
summer conditions, simulations are performed for individual undisturbed valley-wind days.
Various LSMs are tested with four LUCs, that is, the Corine Land Cover 2012 and the updated 2018
(CLC12 and CLC18) datasets and the WRF built-in MODIS and USGS datasets. Initial and
boundary conditions come from the ERA-5 reanalysis. The model simulations are evaluated against
high-quality observations from the i-Box measurement platform, which includes a temperature and
a humidity profiler and six eddy-covariance towers (including four full energy-balance stations),
which are located at various locations throughout the valley covering different surface
characteristics (e.g., slope aspect, slope angle, land cover, and elevation.) Automatic weather
stations in the Inn Valley and its surroundings increase the spatial coverage of observations
available for model evaluation.
Both standard meteorological variables (e.g., temperature, humidity, pressure, wind speed and
direction) and the full surface-energy balance (e.g., heat fluxes and radiation) will be evaluated
against observations for all the simulations to determine the impact of differences in LUC on
surface exchange in the LSMs. Because of the large spatial heterogeneity of the topography and the
land cover, an optimized grid-point selection is developed for evaluating the model against these
point measurements in addition to correcting for differences in elevation and height above ground
between the model and real topography. Surface fluxes integrated over the whole valley are further
analyzed to determine the impact of the LUC on the MoBL, such as the thermal structure and the
valley-wind circulation.