Urban forest canopy height modeling using spaceborne laser ICESAT-2 LIDAR

Abstract

Abstract Urban forests are an important part of urban ecosystems. Carbon sequestration in urban forests helps reduce the concentrations of greenhouse gases in the region where they are present. Forest height is an important structural parameter for calculating the forest carbon sequestration capacity. Based on this, our study proposes a space-borne laser fusion multi-source remote sensing inversion model of urban forest tree height based on urban space environmental characteristics. This paper mainly consists of three parts: (1) First, a variety of highly correlated tree feature factors were extracted from ICESat2 satellite-borne laser data, LandSat8 multi-spectral data, and spatial environment auxiliary data, and a feature database was constructed. (2) The importance of the feature factors in the feature base was analyzed, and a large-scale forest height inversion model of Shanghai was constructed using a support vector machine (SVM), random forest (RF), and backward propagation neural network (BP-ANN). (3) The accuracy of the urban forest height inversion model was improved by introducing urban spatial environmental features such as texture features. Ablation experiments show that the texture features considered in this study can improve the accuracy of each model to varying degrees, and the accuracy of the BP neural network can reach R2 =0.61, RMSE=3.6589. The accuracy of the urban tree height inversion model was R2 =0.6433, RMSE=1.0967, which proves the effectiveness of the space-borne laser fusion multi-source remote sensing urban forest height inversion model considering the characteristics of the space environment.