Impacts of Land Use/Land Cover Distributions and Vegetation Amount on Land Surface Temperature Simulation in East China

Land surface temperature (LST) plays a crucial role in the earth system because its heterogeneous spatial distributions can trigger local circulations through land surface-atmosphere interactions. An accurate simulation of the LST spatial distribution heavily rely on the accuracy of the land surface characteristics, such as LULC (land use/land cover). To investigate the impact of uncertainty in LULC maps on LST simulation, the Community Land Model, version 4.5 (CLM4.5), was used in this study with four LULC products as overlying vegetation characteristics. East China, with its complex land surface characteristics, was employed as the study area. The simulation results were compared to the observations at nine China Meteorological Administration (CMA) stations and to Moderate Resolution Imaging Spectroradiometer data (MOIDIS_LST) over the whole study region. Based on the comparison at CMA stations, CLM4.5 can properly simulate ground temperature. However, large differences are found between MODIS_LST and the simulated LST, particularly over crop areas. Comparison of simulated results using different LULC products showed a large dissimilarity over forest areas, mainly due to the different identification methods for forest types. Large biases in the prescribed leaf area index (LAI) of the model are also found compared with MODIS_LAI. Then, the LAI in the model default data was replaced by the MODIS_LAI product, which greatly reduces the LST simulation biases. These findings provide insights into improving the simulation of LST and land-atmosphere interactions in regional weather models or global climate models.

Automated summary

This study investigated the impact of uncertainty in LULC maps on LST simulation using the CLM4.5 model, revealing that the model can accurately simulate ground temperature but shows large differences compared to MODIS_LST, especially over crop areas. Different LULC products also led to significant dissimilarity in simulated results over forest areas, primarily due to the different identification methods for forest types. Replacing the LAI in the model default data with the MODIS_LAI product greatly reduced the LST simulation biases.