| Bamboo is an important subtropical forest resource in China, which plays an important role in forest ecosystem. As an essential factor of forest resource inventory, crown closure(CC) can reflect the degree of of canopy density and space utilization of trees, and also can indicate stand density. Nowadays, there are two methods to measure CC, namely traditional field inventory method and remote sensing method. The traditional field inventory methods has the disadvantage of small area, time-consuming, labor-intensive, and easily affected by human factor. By contrast, remote sensing technique has the characteristics of real-time, continuous and large area, which makes it has unique advantages in CC quantitative reversion. With the improvement of remote sensing, it has become a new trend to monitor forest based on multi-source remote sensing data, and a hot topic to estimate forest parameters at home and abroad.the information that remote sensing image concluded is becoming richer, which provides a new way to realize forest resource monitor and CC estimation and makes it possible to retrieve multi-scale forest parameters. Based on these study, taking moso bamboo(Phyllostachys heterocycla var. pubescens) as study object, and Anji county(Shanchuan town included), which is located in northwest Zhejiang province as study area, this study will retrieve CC of moso bamboo using UAV imagery, SPOT5 data and Landsat TM data in multi-scale based on field survey.It mainly contains the following aspects:1、CC retrieval of moso bamboo using LSGM based on UAV data. Taking UAV imagery as data resource, the endmembers were selected using minimum noise fraction(MNF) method. Then, the proportion of sunlit background(Kg) could be obtained using the methods of fully constrained and unconstrained linear spectral mixture analysis(SMA), respectively, so that CC could be retrieved by taking Kg into LSGM; Finally, the retrieval accuracy of these two SMA methods was compared.2、CC retrieval of moso bamboo using Erf-BP model based on SPOT5 imagery. Based on CC retrieved by UAV imagery and taking SPOT5 image as data resource, CC of moso bamboo for Shanchuan town was retrieved using Erf-BP model, among which the optimal model structure was explored by setting and selecting variables, setting the neurons number of hidden layer and training goal.3、CC retrieval of moso bamboo using Erf-BP model based on Landsat TM imagery. Base on CC retrieved by SPOT5 image and taking Landsat TM image as data resource, CC of moso bamboo for Anji county was retrieved using Erf-BP model.Through this study, conclusions can be drawn as follows:1、The combination of UAV remotely sensed imagery and geometric-optical model can, to some degrees, achieve the estimation of crown closure. But CC of moso bamboo in study area using unconstrained SMA method shows some disadvantages, with large RMSE, seriously underestimation. In contrast, CC using fully constrained SMA method has significant improvement in retrieval accuracy, with R of 0.7933 and low RMSE of 0.04 at a significance level of 0.01, which is closer to the actual condition of moso bamboo forest.2、CC of moso bamboo retrieved by SPOT5 data using Erf-BP model, to some degrees, obtained satisfactory results. The results shows that there is a significant correlation between estimated CC and field-measured CC(p < 0.01), with R of 0.7414 and RMSE of 0.0223, which demonstrate a high retrieval accuracy with this method.3、It is feasible to estimate CC of moso bamboo by combining Erf-BP model and Landsat TM data. The results shows that the correlation coefficient(R) and RMSE between retrieval CC and field-measured CC is 0.7378 and 0.0156, respectively, which indicates that there is a better fitting between them. |
PDF Full Text Request