Land‐surface interactions with the atmosphere over the Iberian Semi‐arid Environment (LIAISE): First mesoscale modelling intercomparison

Abstract Land‐surface–atmosphere interactions determine atmospheric boundary‐layer (ABL) features, and in the case of semi‐arid regions the water availability in the root‐zone layer is a major factor. To explore this issue, the Land‐surface Interactions with the Atmosphere over the Iberian Semi‐arid Environment (LIAISE) initiative organized an observational campaign in the Eastern Ebro River sub‐basin in summer 2021, focusing on the effect of surface heterogeneities on the ABL in a semi‐arid environment enclosing a large irrigated area. With the aim of understanding the performance of numerical models for the region studied, a mesoscale modelling intercomparison was undertaken for a three‐day case study in July 2016. The intercomparison consisted of three models, Meso‐NH, the Unified Model (UM), and Weather Research and Forecasting (WRF), run with two embedded domains, with corresponding horizontal resolutions of 2 km and 400 m and similar vertical grids. This is a case dominated by well‐developed thermally driven circulations in the Ebro Basin, allowing an investigation of the representation of surface features in the models and its impact on the organization of the flow at lower levels. Other effects are also explored, such as using different initial and lateral boundary conditions, changing horizontal resolution, or modifying some surface features. The mesoscale circulations generated by the three models are similar and accurate when compared with the surface observations, variations being found at the ABL scales related to the representation of surface processes in each model, which provide different grid values of the surface fluxes. It is found that the models do not perform as well over the irrigated areas as in the rest of the domain. The challenge at this point is to relate the model biases to the particularities of the parameterizations and physiographic databases used by each model.