Estimation Of Above-ground Biomass Of Grassland In Da'an Based On Multi-source Remote Sensing Data

Grassland is an important renewable resource and plays an important role in maintaining the ecological environment and sustainable development of human society in ecologically weak areas.Biomass is a significant index to evaluate the ecosystem structure and basic function of grassland and a basic parameter to study the primary productivity of grassland.Grassland biomass is the material source and energy basis of the whole grassland ecosystem,and is the basis of productivity research and net productivity research.Therefore,timely and accurate monitoring of the spatial and temporal changes of biomass in ecologically fragile areas has important scientific significance and application value for the scientific and reasonable utilization of resources in ecologically fragile areas and the protection of their ecological balance.In this study,Da 'an was taken as the study area.The radar data,optical data and field survey data of Da 'an were fused to form a comprehensive data source.By adopting the Object-oriented classification and model simulation methods,To extract the grassland distribution information and establish a collaborative biomass inversion method based on multi-source remote sensing data suitable for grasslands in ecologically fragile areas,and select part of the measured data to verify and evaluate the model accuracy.The main research results are as follows:(1)Based on image segmentation and object-oriented classification,this study extracts the grass distribution information.Combined with radar(Sentinel-1)and optical(Sentinel-2)remote sensing data to map grassland distribution.Establishment of interpretation marks according to the optical image features of land cover types in Da'an.This paper takes full advantages of object oriented segmentation and decision tree classification,to establish decision tree rule set by various types(waterbody and built-up land for radar image,vegetation and non-vegetation for optical image)on radar imageries and optical.After that,the spatial information of grassland is extracted from the interpretation of two types of images by comparing crossing.The spatial distribution range of the grassland in da 'an city was extracted with high accuracy,according to the field sample points and samples randomly selected from high-resolution images,the classification results were evaluated with accuracy as high as 90.7%,which basically met the requirements of grassland biomass inversion.Compared with the results of optical image(Sentinel-2)interpretation alone,the accuracy of this method is higher.The classification accuracy increased from 86.5% to 90.7%.(2)This study proposes a method to improve the water cloud model.Ai ming at the problem that the distribution of grassland vegetation in ecologically fragile areas is not uniform and the vegetation cover is sparse,the underlying surface soil has a strong influence on the radar backscattering coefficient;the gap information of vegetation is taken into account in the established herbace ous vegetation scattering model.Vegetation coverage calculated from optical dat a is used to distinguish different scattering mechanisms between vegetation and soil in order to eliminate the influence of underlying soil on backscattering co efficient.Finally,the grassland biomass of da 'an city was inversed by using t he herbaceous vegetation scattering model.There is a strong linear relationship between the biomass obtained by inversion and the measured value.(3)This paper compares the difference between the water cloud model and the improved water cloud model in the inversion results and accuracy,and analyzes the reasons for the difference in the results.The results showed that the simulation accuracy of the improved water cloud model was improved to a certain extent,with R2 increased from 0.7535 to 0.853 and RMSE reduced from 0.27kg/m2 to 0.25kg/m2.It is proved that the improved water cloud model can effectively estimate the biomass of grassland in ecologically fragile areas(4)This study predicted and analyzed the spatial and temporal distribution characteristics of grassland biomass in Da 'an city.In order to understand the spatial distribution of grassland in the study area and the change trend of grassland biomass in vegetation growing season from a macro perspective,the improved water cloud model was used to estimate the biomass of grassland in da 'an from May to October.Study time and space distribution rule and characteristics of da 'an grass,grass biomass years overall comply with the principle of grassland vegetation growth curve,the grass grow from May begin to accumulate,increase soon,in June to reach maximum at the end of August or early September,and then to the end of the growth of the grass,grass biomass reduced gradually from the end of September,in October to stabilize;From the perspective of spatial distribution,the maximum biomass of grassland was up to 2.28kg/m2,the minimum was 0.01kg/m2,and the average was 0.52kg/m2.The biomass of grassland was mainly distributed in the range of 0.5-1 kg/m2.The results were visualized and analyzed to provide reference for herdsmen's reasonable grazing.