Development of the fine resolution fractional vegetation cover product and its application in urban area

Abstract

The rapid pace of urbanization has accentuated the need for precise monitoring and assessment of vegetation cover changes within urban areas. Fractional Vegetation Cover (FVC) is an important parameter used to quantify the proportion of land covered by vegetation, providing essential insights into urban vegetation change and phenological dynamics. Traditional methods for estimating FVC, often hampered by limited spatial resolution, struggle to accurately capture the intricate dynamics of urban greenspace. Addressing this gap, our study introduces an innovative approach for generating high-resolution FVC products, aiming to provide a more detailed analysis of urban vegetation cover and quantify the extent to which urban residents are exposed to greenspace. In this study, we established a framework for estimating vegetation cover at a fine spatial resolution by combining high-resolution GF-2 images with 30 m Landsat 8 images. We generated 30 m FVC training data by aggregating 1 m GF-2 FVC data collected from various regions across China. Random Forest (RF) regression, an ensemble machine learning method used for classification and regression, was employed to construct multiple models with different combinations of Landsat spectral and angular features to test the performance of various variables. The model utilizing Landsat surface reflectance and solar zenith angle as inputs proved to be the most accurate, with an R² of 0.979 and an RMSE of 0.041 and was selected for estimating the 30 m vegetation cover. The results showed that when validated using 346 independent ground measurements from 14 sites around the globe, the R² was 0.814 and the RMSE was 0.170. By comparing the Landsat vegetation cover with three satellite-based vegetation cover products over a period of five years for different vegetation types, it was further proven that Landsat vegetation cover has the highest consistency and temporal continuity with other vegetation cover products in coniferous forests, shrublands, and grasslands. Finally, through the instantaneous estimation of vegetation cover in Wuhan City, Hubei Province, the annual average of vegetation cover was obtained. The spatial distribution of vegetation cover was derived, the phenological indicators of urban vegetation were extracted, and the trends in vegetation and its phenology at a 30 m resolution were presented. Our conclusion is that, during 2000 and 2020, urban greenspace coverage has significantly increased, the length of the vegetation growing season has noticeably extended, and human exposure to urban greenspace has risen. In urban environments, the method and application proposed in this study contribute to providing detailed insights into changes in urban greenspace, thereby enabling accurate monitoring and assessment of vegetation cover in urban environments, which is crucial for urban planning and environmental management.