Land Use/Land Cover (LULC) and soil texture play a key role in meteorological models because they determine the vegetation and soil proprieties that interfere in the exchange of energy, moisture, and momentum between the land surface and the atmosphere. Additionally, LULC and soil texture are relevant input datasets in meteorological models affecting their results and future applicability as in weather researches and air quality modeling. Brazil has a complex and heterogeneous LU, and it has faced significant LULC changes in the past years. Therefore, this paper aims to update the LULC, using the national product MapBiomas, and soil texture data, by SoilGrids, to replace the default input data in the Weather Research and Forecast (WRF) model for São Paulo, Brazil. Aiming to evaluate the impact of those input data on WRF simulations, five cases were simulated using WRF v4.1.3 with 1 km of grid resolution, and combinations of “Default” and “Updated” input data. Sixty-days simulations from March 15th to May 15th of 2015, covering the transition of wet to dry season, were performed and evaluated with observational data over São Paulo State. The results showed significant differences in the classifications of LULC and soil texture in the entire domain between the default and updated data. The updated data is more realistic and coherent with local characteristics, being more representative, as an example over Santos city area being correctly classified as urban and built-in updated LULC and not water, as in the default. The comparison between the modeled results with observations data has shown a similar behavior for temperature and humidity for the five cases at the monitoring stations grid cells because the LULC changes were between classes with similar land parameters, such as albedo, roughness length, and soils moisture, although the Default classes are not accurate. However, the latent and sensible heat fluxes were ways more sensitive to the LULC/soil texture changes in the WRF model. Additionally, reasonable differences were observed over the entire modeling domain for these two variables. The updated land surface data provoked low temperatures at 9h and 17h UTC, less humidity at 9h UTC, and more humidity at 17h UTC, especially in the north part of the modeling domain, the area which has faced more LULC and soil changes. The PBL height was also affected by the updated data, probably caused by the impact at heat flux over the domain, causing a variation from 30% to 70% over the modeled grid cell, which may have a higher impact on air quality modeling. Thus, it is recommended to update the land surface data for Brazil to avoid misclassification of LULC and soil texture, even if the comparison at monitoring stations has shown similar behavior between the default and updated land surface data. Additionally, updates in the land parameter inside the model are required to represent each LULC/soil class better.
Read full abstract