Assessments Of Different Kernel-Driven Models For Modeling Urban Daytime Thermal Anisotropy Over Simulation And Satellite Data

Lu Jiang,Wenfeng Zhan

DOI: https://doi.org/10.1109/IGARSS.2019.8898724

2019-01-01

Abstract:Land surface temperature (LST) is one of the key parameters of land-atmosphere energy exchange. The LST measured by thermal infrared remote sensing has a large directional difference between different viewing angles over urban areas with complex 3-dimensional urban surface structure; this phenomenon is the urban thermal anisotropy. In recent years, some kernel-driven models have been used to model the thermal anisotropy of urban surfaces, but a comprehensive and systematic comparison of the published kernel-driven models is lacking. Consequently, based on the simulation and the satellite datasets, three one-kernel models and eight dual-kernel models are comprehensively evaluated. The results show that: (1) the dual-kernel models are better than the one-kernel models. The average RMSE of the dual-kernel models is smaller than that of the one-kernel model by about 0.30 K. (2) The dual-kernel models with the solar kernel K-rou are better than those with solar kernel K-vs, because the solar kernel K-rou can simulate hotspots more accurately. (3) Urban geometric structure and solar radiation will affect the modeling error of the kernel-driven models, but basically do not affect their performance ranking. The optimal kernel-driven model is the RVI model. The modeling error of this model is minimal (RMSE is about 0.49 K) and its performance is stable. The optimal kernel-driven model recommended in this paper is expected to be used for the angular normalization of urban satellite LST products and remote estimation of complete urban surface temperature, which improves the thermal remote sensing of urban climates and environment.