Study On The Estimation Method Of Forest Volume Using Airbrone LiDAR And Polarimetric SAR Over Hilly Region

As the major component of the biosphere, forest plays one very important role in the equilibrium of ecosystem, as well as in the carbon sinks. As it can provide three-dimensional information for the objects with high-precision, Light Detection And Ranging (LiDAR), a kind of laser detection and ranging system, is one of the most up to date and hot research field in current earth observation science. Consequently, it is imperative challenge for us to further study the data processing technologies and its applications in earth observation. Since Synthetic Aperture Radar (SAR) has the penetrating ability free of the weather, cloud, sunshine effects, it takes the irreplaceable advantages to optical remote sensing, Indeed, quantitative remote sensing is the development orientation of current earth observation study. Nevertheless, many sensors cannot satisfy the requirement of the application at a large-scale but with high accuracy, so that, it recurs to collaborating multi-source dataset to the goal of quantitative inversion.The purpose of this study is focusing on the quantitative inversion of the forest stem volume using multi-source remote sensing data for large scale mapping in mountainous region. Firstly, forest stand mean height and volume were estimated based on low-density airborne LiDAR (0.39 points / m2) using the improved crown recognition algorithm. Secondly, by using L-band Synthetic Aperture Radar (SAR) , the potential of polarimetric SAR data for forest volume estimation is addressed. Finally, forest stand stem volume is estimated based on LiDAR and polarimetric SAR data.The main study contents are as follows:(1) Classifing of airborne LiDAR point cloud data and extracting surface modelsThe digital elevation model (DEM) is obtained by interpolating the classified ground points, and the digital surface model (DSM) is generated by interpolating all points. The normalized crown height model (Canpoy Height Model, CHM) is the difference between DSM and DEM. (2) Retrieving forest parameters based on airborne LiDAR dataAt stand scale, the mean height in mountain forest is generated by improved recognition algorithm using the CHM information, and the overall accuracy is 75%. For Black Locust forest, the correlation coefficient between predicted stand stem volume and ground measurements is 0.696, and RMSE is 21.056m3/hm2 For Chinese Pine forest, the correlation coefficient between predicted stand stem volume and ground measurements is 0.453 and RMSE is 21.866m3/hm2, which is lower than the former estimation.(3) Estimating forest parameters using polarimetric SAR dataThis study analyzed the response characteristics of PALSAR signal to the forest volume at stand scale. The result shows that the model established by the polarization ratio is a relatively better way in stand scale than others. The heterogeneity of the forest has great influence on the SAR signal. By comparison between the estimated volume of Black Locust forest and Chinese Pine forest, the correlation coefficient between predicted Black Locust stand volume and SAR parameter of the regression model is 0.787, and the correlation coefficient between predicted Chinese Pine stand volume and radar parameter of the regression model is 0.743. There is no much difference between the correlation coefficients, but when comparing the RMSE, Black Locust forest stand is of 17.893m3/hm2, Chinese Pine forest stand is higher with RMSE of 29.3m3/hm2.(4) Estimating forest volume by integrating multi-sources of dataThis study investigated the potential of estimating forest volume using the multi-sources data, the airborne LiDAR, CCD and the space-borne ALOS PALSAR. The result shows that the regression correlation coefficient for estimating the Black Locust stand stem volume is 0.864, and that for Chinese Pine stand stem volume inversion is 0.813. It is also concluded that, at stand scale, this method can represent well the relationship between the forest stem volumes with the estimates from the multi-source dataset, with RMSE of 20.064m3/hm2 for Black Locust and 24.730m3/hm2 for Chinese Pine.