The Estimation Of Effective Leaf Area Index And Canopy Closure Using UAV-based LiDAR In Ginkgo Plantations

Ginkgo biloba L.is one of the most important tree species in China,forest canopy is an important place in physiological processes of plants,such as respiration,transpiration,photosynthesis,and the carbon-water cycle.Leaf area index(LAI)and canopy closure(CC)are two parameters to characterize forest canopy,they play the key role in vegetation growth.Real-time,quantitative and accurate estimation of LAI and CC has important reference value for ginkgo precision cultivation and high quality management.LiDAR(Light Detection and Ranging)has unique advantages in acquiring large area of forest structure as an active remote sensing technique.This study is based on the typical and typical ginkgo plantation in China(the ginkgo plantation in Tiefu town,Pizhou city,Jiangsu province).Combined with the high density point cloud obtained from the multi-rotor UAV-based LiDAR system and the ground measured effective LAI and the position of individual tree.First,used Gap-fraction modeling and statistical modeling to estimate the eLAI and cross validation based on the metrics extract from point cloud.Then,combined with the high density point cloud and the position of individual tree,compared and analyzed different individual tree segment results(watershed algorithm,polynomial fitting,individual tree crown segmentation and point cloud segmentation)of the ginkgo segmentation and the extraction effect of canopy.And performed sensitivity analysis of the segmentation methods by changing the CHM resolution and point cloud threshold.At last,based on the individual tree segmentation,analyzed and validated the results of canopy closure extract from CHM(direct approach)and estimation of statistical model(indirect approach),and performed sensitivity of analysis estimating canopy closure by changing the resolution and height threshold of CHM.The results of study showed that:(1)The estimation of eLAI using UAV-based LiDAR in ginkgo plantations,the results showed that:When estimated the eLAI using the statistical model,the estimated accuracy was R~2=0.38(rRMSE=54%)by LiDAR height metrics.After introducing several other sets of metrics(i.e.,density metrics,canopy volume variables and intensity metrics)step by step,the estimated accuracy increased to R~2=0.64(rRMSE=26%),R~2=0.61(rRMSE=28%)and R~2=0.74(rRMSE=23%)respectively.The estimated accuracy of the eLAI was R~2=0.71(rRMSE=32%)by the gap-fraction model.By designing the UAV flight(fixed height at 60m)and setting the parameters of the UAV-based LiDAR sensor,we found that the intensity metrics can be used effectively to improve the estimated accuracy of the eLAI(?R~2=0.36,?rRMSE=31%).(2)The accuracy comparison of 4 individual tree segmentation based on LiDAR point cloud,the result shows that:in the total plots,the PCS has the highest accuracy(overall accuracy F=0.83),then is ITCS(overall accuracy F=0.82),higher than the watershed algorithm(overall accuracy F=0.79),and the overall accuracy of the polynomial fitting is F=0.77.With the increase of trees density,the segmentation accuracy of the 4 methods is reduced.In the low density plots,the overall accuracy of 4 segmentation methods are F=0.87,0.87,0.9,0.91;in the medium density plots,the overall accuracy of 4 segmentation methods are F=0.79,0.78,0.81,0.83;in the high density,the overall accuracy of 4 segmentation methods are F=0.75,0.72,0.77,0.79.The sensitivity analysis results showed that when the CHM resolution is 0.5×0.5m,the watershed algorithm,the polynomial fitting and ITCS had the highest segmentation accuracy.When the distance threshold set to 2m,the segmentation accuracy of PCS had the highest segmentation accuracy.(3)The estimation of canopy closure in study area by using the crown from individual tree segmentation and metrics from UAV-LiDAR,the result shows that:the accuracy of canopy closure estimation by using CHM extraction(direct approach)was R~2=0.94(rRMSE=8.4%);when estimated canopy closure using the statistical model,the estimated accuracy was R~2=0.50(rRMSE=54%)by LiDAR height metrics.After introducing several other sets of metrics(i.e.,density metrics and canopy volume variables)step by step,the estimated accuracy increased to R~2=0.78(rRMSE=5.9%)and R~2=0.69(rRMSE=6.5%)respectively.The sensitivity analysis results showed that when using direct approach to estimate canopy closure,the resolution of CHM is0.5×0.5m had a higher accuracy than the resolution of CHM is 1×1m,the estimation accuracy extract from above 2m is the highest among above 1m and above 5m.