Revealing the spatial distribution of crown base height across China based on close-range Lidar data

Crown base height (CBH) is essential for characterizing forest vertical structure over time for sustainable forest management and serves as a key input in fire model and growth model. At plot level, the average CBH (CBH_a) is mainly used to assess tree growth and construct biomass models while the minimum CBH (CBH_min) can indicate the fire risk and fire behaviour. However, there are currently few CBH products available at a national or global scale. Close-range light detection and ranging (Lidar) has shown great potential in collecting plot-level forest structure parameters and can be easily scaled up to national or global scale. But considering the integrity of point clouds, CBH estimation utilizing airborne Lidar data would be always overestimated compared with other close-range Lidar data such as TLS data or backpack data. This is mainly because of the significant difference in density between the upper and lower point clouds, as well as lacking considering the tree shape. By filling the point clouds with the same density and lowering the CBH condition which considers the tree shape, we proposed an improved CBH estimation method to reduce the overestimation when using airborne Lidar data. Verified by field-measured data in six plots, the proposed method improved the root-mean-square error (RMSE) by nearly 50 % compared with the original method. The mean absolute error (MAE) was 0.694 m, R² was 0.777 and the RMSE was 1.039 m for the validation trees. Facing different sensors and point densities, this method generally generates stable CBH estimation results. Then, we developed a newly tree-based framework that uses machine learning and multiple source remote sensing data for generating CBH products across China. We collected over 1117 km² close-range Lidar data and used the proposed method for estimating CBH. The CBH estimation results were converted to average value and minimum value in a 1 km × 1 km plot and served as training data to generate CBH maps across China at 1 km resolution. To our best knowledge, this is the first CBH map across China, and also the first national-scale average and minimum CBH maps around the world. The results showed that the average CBH_a and CBH_min were 6.76 m and 2.70 m with standard deviations of 1.59 m and 0.85 m. The methods and maps would provide a new dimension in monitoring changes in forest structure, assessing fire risk and constructing biomass models in future studies.