Assessment of different approaches for estimating volume change in secondary forests using bi-temporal LiDAR data

Predicting the change of volume in secondary forests is critical for monitoring forest resources and understanding the carbon dynamics. The secondary forests in southeast China are carbon-dense, species-rich ecosystems with high productivity, play a key role in maintaining regional environment and mitigating climate change. In this study, we assessed the capacity of bi-temporal airborne Light Detection and Ranging (LiDAR) data to predict forest volume change in a secondary forest of southeast China over a 6-year period. To do so, we evaluated four different approaches for estimating the change in volume, specifically: (1) predicting volume for Time 1 by the model fitted by the field and LiDAR data in Time 2, volume change was then calculated as the difference between the predicted volume from the two times; (2) predicting change using the differences of LiDAR metrics; (3) predicting change as the difference between separate volume predictions; and (4) predicting change using the LiDAR metrics from each point in time. The results demonstrated that, in general, Method 2 and 4 have higher accuracy than Method 1 and 3. Compared with the results of the indirect approach (Method 3) for volume (CV-R 2 =0.62), the direct approach (Method 2) (CV-R 2 = 0.66) has a higher accuracy. Within the volume estimation models, the Δ volume model (R 2 = 0.72) using the direct approach explains lower variability in the response variable than volume models in Time 1 and Time 2 (R 2 = 0.76-0.80). The canopy height distribution metrics of median height (i.e., h50) and delta value of median height (Δh50) were most often selected by the models. A small number of canopy return density metrics from the middle and upper canopy (i.e., d5 and d7) were selected for estimating volume in Time 1 and Time 2.