Inversion Of Vegetation Structure Parameters From Multi-angle CHRIS Data In Changbai Mt.

Vegetation is an important factor affecting the terrestrial and atmospheric energy exchange and even the biodiversity. Vegetation structure parameters consist of ecological parameters and geometric parameters. Ecological parameters include total LAI, green vegetation cover, Leaf biomass, and total biomass while geometric parameters include average height of trees, maximum height of trees, minimum height of trees, etc. Currently, there are two main ways of obtaining vegetation structure parameters: field measurements and RS (remote sensing). Limited by external factors such as observing conditions, field measurements is hardly satisfying the regional scale study requirements. On the contrary, with fast development of technology, RS makes it possible to have a short-period, large-scaled, high-speed vegetation parameters acquisition. Multi-angle remote sensing can provide observations over different directions, which can obtain more detailed and reliable. Combining these three-dimensional structural information with vegetation reflectance got by satellites and ecological and geometric parameters inverted with radiative transfer model, would make it better to monitor the change of vegetation state and more detailed to understand the plant growth state.Hyperspectral reflectance data provided by sensor of CHRIS, has two different resolutions: 17m and 34m, and band range of 410nm-1050nm. CHRIS can obtain continuous image in different angles, including 0°,±36°and±55°. CHRIS has 5 different imaging modes, with different band settings and ground sampling densities according to different applications such as water or land. Especially the chlorophyll mode designed specifically for vegetation study, providing plenty of valuable data of BRDF of atmosphere, land and sea, more accurately simulates the RS physical properties of the Earth's surface, which is not only beneficial for evaluating biomass and monitoring bio health, but also important to identify vegetation or woodland canopy structure, density, vegetation and study the species of trees.Based on the discrete ordinates method and ray tracing methods, discrete radiative transfer model (DART), takes the anisotropy caused by different spatial distribution of target objects, which has never been considered before, into account. It can set a number of bands at the same time, and simulates the radiation transfer received by sensors or terrestrial landscape in any position on the earth with directions at any angle as well as design accurately to make calculation more convenient. All components of the scene are divided into unit cells composed of the rectangles as a collection, located according to the center of each unit in the three-dimensional scene. Units are used to simulate different types of scenes elements, such as vegetation, soil, buildings, water, etc. Energy enters the scene from the top, and scatters while passing down each cell. The types of units are identified according to the different scene elements and each unit is processed with ray tracing method, which complete the radiative transfer simulation of the whole scene and BRF values of the elements in the scene are obtained.Choosing multi-angle remote sensing data of CHRIS, vegetation structure parameters were inverted with DART. According to the characteristics of coniferous forest of Changbai Mountain region and situation of CHRIS data, solar zenith angle of incident direction was set into 39.00°, and azimuth was set into 141.77°, 18 bands between 484nm and 800 nm were simulated. The Reflective type of simulated branches and soil were set as Lambertian with type of leaves as coniferous forest, then corresponding spectral data were loaded. The size of the basic simulation scenario was set into 17m*17m*17m, convenient for the matching by pixel. Vegetation structure parameters consisting of crown height, crown base radius, height of bar under crown, crown and DBH, leaf area index, etc. was entered, BRF values was simulated with DART. The structural parameters were made into a table, so were the corresponding simulated BRF values. The lookup table of vegetation structure parameters of coniferous forest was made. After CHRIS data's preprocess consisting of denoising, atmospheric correction, geometric correction, parameters required for the inversion of the values of vegetation BRF were obtained. The 5 observation angles of CHRIS imaging were not included in the 61 directions computed with discrete simulation in the Look up table. In order to obtain the 5 corresponding angles and BRF values, the Interpolation of the 61directions was made, then the matching between the BRF value of CHRIS data and the corresponding interpolated values was made and the result showed there were 18 bands whose BRF value differences were minimum error satisfying the principle of least squares, thus the required vegetation parameters were obtained. An area of 150×150 from CHRIS image was used to obtain the inversion results map of structure parameters. Results are tested with a coupled LAI images reversed by ETM+ data. The distribution of vegetation structural parameters indicated that the parameters are closely interacting and crown base radius and crown base diameter under of tall trees are also larger than the other short trees, which consistent with the actual situation.Besides the vegetation structure parameters, the affect of cell size of scene and distribution of trees were also considered. When a particular parameter was simulated, the others were assumed fixed. The result showed that every selected parameter brought certain affect different from each other on BRDF. The affect brought by cell size was not regular while the others were distributed according to certain rules. For the visible and near infrared bands, affects and variations were different from each other. Bidirectional reflectance properties of vegetation were the result of interaction of multiple parameters.