Should we average rain gauge values to estimate throughfall? A research of canopy structure and throughfall spatial variability using LiDAR technology

Abstract

Throughfall ($\mathrm{TF}$) is a significant component in hydrological studies of forest ecosystems that characterize subcanopy precipitation ($P_g$) inputs. Previous studies averaged the $\mathrm{TF}$ observations from rain gauges placed uniformly under the canopy. However, variations in canopy structure and the spatial distribution of canopy density led to measurement variability across locations, even in single-species plantation forests. 3D forest point clouds were reconstructed using a drone LiDAR system and a voxelization method was employed to compute the volume of leaves at rain gauges and individual tree scales to describe the canopy structure. Canopy saturation was assessed by observing changes in the proportion of direct $\mathrm{TF}$ from high-temporal-resolution rain gauge data and correlating these patterns with quantitative canopy structure, a volume-$\mathrm{TF}$ model was built to characterize the challenge of raindrops passing through the leaves. In addition, the $\mathrm{TF}$ conversion rate was corrected by considering the amount of canopy saturation and $\mathrm{TF}$ proportion changes before and after saturation. After inputting the volume and $P_g$ parameters, the accuracy of the model was verified at the rain gauge scale with R² = 0.8239 – 0.9906 (n = 70). The modeled results showed that $\mathrm{TF}/P_g$ during the observation period was 53.24 %, which was slightly lower than the average of the 20 rain gauges over 265 rainfall events (63.21 %). Spatial variability significantly affected $\mathrm{TF}$ generation during changes in rain gauge location (coefficient of variation ($\mathrm{Cv}$) = 0.64) compared with Pg ($\mathrm{Cv}$ = 0.3). The incorporation of saturation improved the model accuracy (RMSE; 3.69 mm > 3.28 mm,). This suggests that the method of averaging the 20 rain gauges to estimate $\mathrm{TF}$ results in limited confidence owing to the spatial variability of the canopies. Our model quantifies the spatial variability of canopy and $\mathrm{TF}$, contributing to the construction of more robust and precise hydrological models, effectively capturing stand heterogeneity and enabling targeted forest management practices.