The Southwest Sichuan Mountain Broad-leaved Forest Leaf Area Index Of Remote Sensing Inversion Methods Based On Different Measurement Ground

Leaf area index (LAI) as an important parameter to reflect the forest canopy structure, directly related to vegetation in the biological and physical activities. The broad-leaved forest is widely distributed in the mountains of the southwestern Sichuan province. It is one of the typical top type successions of the forests, and plays an important role of keeping the stability of the whole ecological system. The study of Leaf area index of the broad-leaved forest in the southwestern Sichuan province can provide scientific basis for the study of maintaining the stability of the forest ecosystem in this area and the sustainable forest management planning. However the mountain forest canopy structure is very complex. Taking different ground measurement methods has a greater influence on the remote sensing for estimating the LAI. The study of this paper selects the town of Shangli in Ya’an city of Sichuan province as the research area. We use the allometric growth equation method, LAI-2000 canopy analyzer and hemispherical photography method for acquiring the LAI of the broad-leaved forest, and combine with RVI, NDVI and other five commonly vegetation index, Moreover we use multiple stepwise regression and partial least squares regression for establishing the effective predicting models of LAI in the study. Then we use different methods to obtain the model data, and finish the comparative analysis, choose better precision of model, generates the leaf area index distribution of the study area ot the broad-leaved forest. The conclusions are as follows:(1) The range of LAIe is 2.78±0.72m2/m2 by using the hemispherical photography method,; the range of LAIe by LAI-2000 canopy analyzer is 3.64±1.06m2/m2; The range of LAI by the allometric growth equation method is 6.29±0.99m2/m2. In the Person correlation analysis of the seven vegetation indexs and LAI, the RVI, NDVI, SAVI, TNDVI, and SLAVI show a good correlation to the LAI obtained by the allometric growth equation method, the hemispherical photography method and the LAI-2000(|p|>0.3).(2) We use the multiple linear regression method to establish the estimation model of the broad-leaved forest LAI, and in the allometric growth equation method model we get Y =.3.089*NDVI+0.89*SLAVI+1.15*VI3+2.749 and verify the accuracy of the model, the determination coefficient R2 is 0.714, RMSE value is 0.45, the model accuracy reaches 79%. In the multiple regression model" of the hemispherical photography method, we get Y = 2.721*NDVI+0.577*SLAVI+0.587, through precision test, the determination coefficient R2 is 0.651, RMSE value of 0.49, the model accuracy reaches 72%. In the the multivariate regression model of the LAI-2000 canopy analyzer method, we get Y= 1.733*RVI-18.593*NDVI+6.900, then we verify the accuracy. The determination coefficient R2 is 0.625, RMSE value is 0.51, the model precision is 69%. The allometric growth equation model on the determination coefficient and precision are higher than hemispherical photography method and LAI-2000 model.(3) Using partial least squares regression method, we establish the forest LAI estimation model in the study area, and verify the model. The model got by the allometric growth equation methodl:Y=3.2372*NDVI+2.1248*SAVI+0.2449*RVI+0.2328SLAVI+0.2453*VI3-5.48. In the test results, the determination coefficient R2 is 0.718, RMSE is 0.31, model accuracy is 86%.The model got by the hemispherical photography method:Y=0.1608*NDVI+0.1568*SAVI+0.1106*TNDVI+0.1545*RVI. In the test results, the determination coefficient R2 is 0.685, RMSE is 0.34, model accuracy is 83%.The model got by the LAI-2000 canopy analyzer:Y=0.2863*RVI+2.125*NDVI+1.51*SAVI-0.003*DVI-1.1473, and the test results show that the determination coefficient R2 is 0.647, RMSE is 0.37, model accuracy is 80%.(4) Compared the hemispherical photography method with the LAI-2000 canopy analyzer, whether adopting the multiple linear regression or the partial least squares regression, use the hemisphere method to establish the regression model on the stability and the accuracy are better than the LAI-2000 measurement data fitting model. Visible in the study area, hemispherical photography method is more accurate than the LAI-2000 measurement on obtaining the leaf area index.(5) The real LAI obtained by the allometric growth equation method has very high correlation to the real LAI got by the hemispherical photography method and the LAI-2000 canopy analyzer. The determine coefficients of the regression equations are to above 0.7. Bringing it to LAI estimation model, we can get real LAI estimation model of broad-leaved forest in the study area:Multiple stepwise regressionThe hemispherical photography method:Y=1.8516*NDVI+0.3926*SLAVI+3.9503The LAI-2000 canopy analyzer:Y=1.052*RVI-11.2878*NDVI+7.7447The partial least squares regressionThe hemispherical photography method:Y=1.7098*NDVI+1.1315*SAVI+2.3818*TNDVI+0.1261*RVI+0.2773The LAI-2000 canopy analyzer:Y=0.1738*RVI+1.29*NDVI+0.9167*SAVI-0.0018*DVI+2.8592(6) The results show that the highest precision partial least squares regression equation method Allometric model, its stability is best. use the model-generated broad-leaved forest in the study area real leaf area index distribution.