Estimation Of Forest Canopy Closure Based On LiDAR And Landsat 8 OLI Data

Forest is the largest ecosystem on land,which plays an important role in maintaining the ecological balance of the whole earth.The change of forest has a great impact on the ecosystem and human.Therefore,dynamic monitoring of forest resources can better grasp the situation of forest,which is conducive to the development of sustainable forest management.However,the traditional measurement method of canopy closure is time-consuming and laborious,which is only suitable for the measurement of small-scale canopy closure and unable to observe the change of large-scale canopy closure.Therefore,the estimation of large-scale forest canopy closure using remote sensing data has attracted much attention in recent years.The study area,with 63 plots of 0.09hm~2(30m×30m),was located in the Mao’ershan Forest Farm.Airborne laser radar(ALS)and Landsat 8 OLI data were used to extract point cloud features.Three feature selection methods(Pearson correlation coefficient,Random Forest(RF)and Boruta method)and three models(Partial Least Squares Regression(PLSR),Random Forest Regression(RFR)and Support Vector Regression,SVR))were used to estimate the forest canopy closure and the accuracy evaluation and analysis of variance(ANOVA)were carried out,which was to explore the influence and analysis of different feature selection methods and estimation models using ALS and Landsat 8 OLI data on the estimation of forest canopy closure.The results were showed that:(1)From the feature selection,the effect of using Pearson method and Boruta for feature selection is better than the RF feature selection.And the Boruta method and SVR for estimating canopy closure has the highest accuracy.The value of R~2 is 0.4893 and RMSE is 0.0718.(2)Among the three regression models(PLSR,RFR and SVR),the estimation accuracy of PLSR and SVR is higher than that of RFR.From the three different data,the ALS and Landsat 8 OLI data is the most accurate data,the Landsat 8 OLI data is the lowest accurate data(3)From the analysis of variance(ANOVA)results,the regression model has a significant impact on the estimation of canopy closure using the ALS and Landsat 8 OLI data,and the others have no significant impact.And the other data had no significant effect on the estimation of canopy closure.(4)The forest vertical structure information of extracted by ALS data(i.e.,height and height percentile of point cloud)can effectively estimate forest canopy closure,which provides a reliable basis for large-scale estimation of forest canopy closure.