10:45 〜 12:15
[AHW19-P07] Estimation of rainfall interception from merged drone and terrestrial LiDAR data by modeling 3D canopy structure in plantation forest
キーワード：人工林、樹幹遮断、ドローンLiDAR、クラスタリング、樹幹物理モデル
The multidimensional arrangement of upper canopy features is a physical driver of energy and water balance under various canopies, and standard modeling approaches integrate leaf area index (LAI) and canopy closure (CC) to describe canopies.
However, it is unclear how the canopy affects the component and interception of rainfall within the forest system.
We generated multi-layered forest point clouds from trunk to canopy using fusion of drone and terrestrial LiDAR data then classified wood and foliage elements using a clustering algorithm to build a high precision physical model for describing throughfall, stemflow and interception.
The experiment was conducted in the thinning plantation forest located in Tochigi prefecture, Japan. Rainfall observation for the three components is important for model development. Throughfall was computed from 20 rain gauges distributed on a grid under the forest canopy, 3 stemflow collectors was set up around the tree trunks connected to a bucket with water level sensor.
We developed a capacity model to describe canopy saturation with foliage points, a voxel-based method was used to create 3D representations of forest canopies, and an analysis of these point-derived canopy structures and volume were performed to assess the canopy's capacity to contain rainfall.
For stemflow modeling, we define a new parameter, Branch Area Index (BAI) to describe the proportion of branches area to the total canopy area and simulate the additional rainfall accumulates to the tree trunk through branches when the tree canopy is saturated.
Preliminary simulation results show that: (1) fusion and registration of drone and terrestrial LiDAR data can greatly improve the point cloud accuracy and enrich the information contents such as coordinate geo-reference and filling of missing structures; (2) a strong correlation between the rainfall observed canopy interception results and the estimated canopy volume, and the volume-based interception prediction model has a high accuracy, with an R2 from 0.84 to 0.91 compared to past observations. (3) BAI has strong correlation with stemflow, with an R2 at 0.9796. BAI also affects the stemflow significantly than canopy area.
However, it is unclear how the canopy affects the component and interception of rainfall within the forest system.
We generated multi-layered forest point clouds from trunk to canopy using fusion of drone and terrestrial LiDAR data then classified wood and foliage elements using a clustering algorithm to build a high precision physical model for describing throughfall, stemflow and interception.
The experiment was conducted in the thinning plantation forest located in Tochigi prefecture, Japan. Rainfall observation for the three components is important for model development. Throughfall was computed from 20 rain gauges distributed on a grid under the forest canopy, 3 stemflow collectors was set up around the tree trunks connected to a bucket with water level sensor.
We developed a capacity model to describe canopy saturation with foliage points, a voxel-based method was used to create 3D representations of forest canopies, and an analysis of these point-derived canopy structures and volume were performed to assess the canopy's capacity to contain rainfall.
For stemflow modeling, we define a new parameter, Branch Area Index (BAI) to describe the proportion of branches area to the total canopy area and simulate the additional rainfall accumulates to the tree trunk through branches when the tree canopy is saturated.
Preliminary simulation results show that: (1) fusion and registration of drone and terrestrial LiDAR data can greatly improve the point cloud accuracy and enrich the information contents such as coordinate geo-reference and filling of missing structures; (2) a strong correlation between the rainfall observed canopy interception results and the estimated canopy volume, and the volume-based interception prediction model has a high accuracy, with an R2 from 0.84 to 0.91 compared to past observations. (3) BAI has strong correlation with stemflow, with an R2 at 0.9796. BAI also affects the stemflow significantly than canopy area.