Objective In recent years, forest frost damage has occurred frequently in China, causing significant impacts on forest growth and ecosystems. This study aims to explore the application of GF-6 WFV data in forest frost damage monitoring and provide a scientific basis for precise forest regulation.

Method Taking the Guandishan National Forestry Bureau in Shanxi Province of northern China as the research area, this study analyzed the spatiotemporal changes of NDVI in response to the development process of forest frost damage caused by spring cold snaps based on the GF-6 WFV time-series data during the 2019 forest vegetation growing season. Through the Net5 deep learning model of the ENVI software, the optimal model was selected to evaluate frost damage.

Result The coastal blue band is highly sensitive to the yellowing information of forest vegetation, which helps improve the remote sensing monitoring accuracy of sub-healthy forest vegetation. The user accuracy of deep learning model training composed of this band was 81.0%, and the accuracy of spatial distribution verification of frost damage remote sensing classification was about 90.7%. The time-series NDVI data of GF-6 WFV can effectively monitor the frost damage in the study area during spring and had been verified by landsat-8 OLI and field survey data. During the study period, the area proportions of severely damaged, moderately damaged, slightly damaged, and healthy forest stands were 17.4%, 36.3%, 30.9%, and 15.4%, respectively. Frost damage was mainly distributed in low-altitude areas below 1 600 m.

Conclusion The study reveals the potential application of GF-6 WFV time-series data in early warning monitoring of tree crown yellowing or growth stress caused by frost damage and provides scientific technical means for forest damage monitoring and precise forest management.