Integrating satellite and street-level images for local climate zone mapping

Timely and accurate local climate zone (LCZ) classification maps are valuable for urban climate studies. The integration of remote sensing and street-level images is promising to produce high-quality LCZ maps, since the former can efficiently capture the information of landscapes on a large-scale while the latter include ground-level details. However, due to their significant differences in spatial distributions and capture views, as well as existing sampling issues of street-level images, how to fuse them effectively is challenging and remains an uncharted research area. To address these issues and fill the gap, this study proposes an effective method to integrate satellite and street-level images for LCZ mapping. Additionally, a simple yet effective street-level image sampling method is proposed. Extensive experiments have been performed and the results demonstrate the effectiveness of the proposed data fusion method and also confirm the usefulness of fusing street-level images with satellite images in enhancing the performance of LCZ mapping. Moreover, the proposed sampling method can increase data representativeness and avoid data redundancy, thus significantly reducing the number of required images while retaining high classification accuracy. To the best of our knowledge, this study is the first attempt to integrate cross-view satellite and street-level images for LCZ mapping. The study and proposed methods can contribute to the development of multi-source data fusion for LCZ map production and further benefit urban climatic research.

Automated summary

This study proposes an effective method to integrate satellite and street-level images for local climate zone (LCZ) mapping, as well as a simple yet effective street-level image sampling method. Experimental results demonstrate the effectiveness of the proposed data fusion method in improving the performance of LCZ mapping, and the sampling method can increase data representativeness and reduce the number of required images while maintaining high classification accuracy. This study contributes to the development of multi-source data fusion for LCZ map production and benefits urban climatic research.