[A-CG31] Aircraft and UAV Observations for Earth-planetary sciences

[ACG31-06] Understory biomass estimation based on Structure from Motion data by Multi-layered forest drone survey

*Yupan Zhang¹, Yuichi Onda¹, Hiroaki Kato¹, Takashi Gomi² (1.University of Tsukuba, 2.Tokyo University of Agriculture and Technology)

Keywords:Structure from Motion, Plantation, Understory biomass, Drone survey

Understory vegetation is an important part of evapotranspiration from forest floor. Forest management changes the forest structure and then affects the understory vegetation biomass (UVB). Quantitative measurement and estimation of UVB is a step cannot be ignored in the study of forest ecology and forest evapotranspiration. However, large-scale biomass measurement and estimation is challenging. In this study, Structure from Motion (SfM) was adopted simultaneously at two different layers in a plantation forest made by Japanese cedar and Japanese cypress to reconstruct forest structure from understory to above canopy: i) understory drone survey in a 1.1h sub-catchment to generate canopy height model (CHM) based on dense point clouds data derived from a manual low-flying drone under the canopy; ii) Above-canopy drone survey in whole catchment (33.2 ha) to compute canopy openness data based on point clouds of canopy derived from an autonomous flying drone above the canopy. Combined with actual biomass data from field harvesting to develop regression models between the CHM and UVB, which was then used to map spatial distribution of UVB in sub-catchment. The relationship between UVB and canopy openness data was then developed by overlap analysis. This approach yielded high resolution understory over catchment scale with a point cloud density of more than 20 points/cm². Strong coefficients of determination (R-squared = 0.75) of the cubic model supported prediction of UVB from CHM, the average UVB was 0.82kg/m² and dominated by low ferns. The corresponding forest canopy openness in this area was 42.48% on average. Overlap analysis show no significant interactions between them in a cubic model with weak predictive power (R-squared < 0.46). Overall, we reconstructed the multi-layered structure of the forest and provided models of UVB. Understory survey has high accuracy for biomass measurement, but it’s inherently difficult to estimate UVB only based on canopy openness result.