| Aboveground biomass（AGB）is an important index of sustainable utilization of grassland resources and livestock carrying capacity of grassland.Rapid and accurate estimation of aboveground biomass in grassland is a hot issue in grassland ecosystem research.In the era of "big data" explosion,the inversion of above-ground biomass of grassland using only a single remote sensing data source has shortcomings in inversion accuracy and richness of plant information extraction.Moreover,traditional multispectral remote sensing is prone to saturation in areas with large vegetation coverage,which is also one of the uncertainties in retrieving biomass.Based on the integration of multispectral,hyperspectral,and UAV imaging spectral technologies,this study conducts an inversion study of above-ground biomass of grasslands in typical grasslands in Inner Mongolia based on ground field observation data,and conducts a comparative analysis with traditional techniques,so as to determine that the integration of multi-source data can compensate for the shortcomings of traditional remote sensing data sources considering only a single multispectral remote sensing,and provide a theoretical basis for using multi-source remote sensing technology for grassland canopy This study will provide a theoretical basis for using multi-source remote sensing technology to monitor biophysical parameters of grassland canopy and provide theoretical support for the conservation and scientific and rational use of grassland resources in Inner Mongolia.The main research results are as follows.（1）The Spatial and temporal dynamics of AGB in typical grasslands of Inner Mongolia and its relationship with hydrothermal factors.MODIS data were used to analyze the spatial distribution pattern and dynamics of AGB during the growing season from 2009.5 to 2015.10,and to study the relationship between AGB and hydrothermal factors.The results showed that the inversion accuracy of the logarithmic function model with RVI as the independent variable was the highest and better than NDVI/EVI vegetation index.7 years,The average AGB of typical grasslands in Inner Mongolia during the seven years was 97.61 g·m-2,the highest value was 151.43g·m-2 in August,and the lowest value was 45.34g·m-2 in October.the spatial distribution characteristics of AGB gradually decreased from northeast to southwest,from May～August,and aboveground biomass between August and October with The spatial distribution characteristics of AGB gradually decreased from northeast to southwest,and from May～August and August to October,the aboveground biomass alternately advanced from south to north and from north to south with the change of seasonal phase.The final ANOVA results showed that the contribution of precipitation was 7%,the contribution of temperature was 13%,and the joint contribution was 63%.（2）Hyperspectral characteristics of typical grasslands in Inner Mongolia.The differences in raw spectra,first-order derivatives,second-order derivatives and pseudoabsorption coefficients and red-edge characteristic parameters of grassland canopy in different months were identified.It is found that July and August have obvious vegetation spectral features,and the spectral curves in May,June and September do not have green light reflection peaks or red light absorption valleys.So July and August are the best time for typical grassland remote sensing inversion biomass.Secondly,after the first-order derivative spectral treatment,the "double peaks" of the canopy spectra in July and August are more obvious,both have "red edge" and "blue edge",and the first-order derivative spectra can well eliminate the influence of soil background on the canopy spectra.After the second-order differential transformation,the spectral curves in July and August were close to each other in this spectral range,and there were obvious differences around 719 nm.After the first-order derivative spectral transformation of the red edge,the reflectance in July and August showed a tendency to increase first and then decrease after reaching a maximum,and a "double peak" phenomenon appeared.The red edge is the most obvious spectral feature of the vegetation and a significant marker to distinguish other features.（3）Hyperspectral remote sensing estimation of above-ground biomass of typical grassland in Inner Mongolia.The experimental data from 2017 to 2018 with hyperspectral feature variables were used to establish the estimation model of aboveground extant biomass（fresh weight/dry weight）.The results showed that the above-ground extant biomass（fresh weight/dry weight）was correlated with the original spectral reflectance,first-order derivative spectrum,hyperspectral feature Variables including Dy,λr,SDy,SDr,SDr/RDb,SDr/SDy,（SDr-SDb）/（SDr+SDb）,and（SDr-SDy）/（SDr+SDy）allreachedthe significance test;the above correlation analysis was integrated,a generalized linear model was established,the coefficient of determination and F-test values of the model fit were compared,and the optimal estimation models for different variables were initially screened out,and 22 test data from 2017 were substituted into the model to test the prediction accuracy.The final best-fit models for aboveground extant biomass（fresh weight/dry weight）were selected as:y=-3.7953x2+60.065x-78.455,（x is SDr/SDb）.with a fitted R2 of 0.662,predicted R2 of 0.302,RMSE and REE of 33.03 and 0.36,respectively;y=7.744 e3.4349x（（SDr-SDb）/（SDr+SDb））,whose fitted R2is 0.559,the predicted R2 is 0.304,and the RMSE and REE are 18.24 and 0.33,respectively.（4）UAV imaging spectral data inversion of above-ground biomass and mapping of typical grasslands in Inner Mongolia.The ground and hyperspectral data in August 2021 were used to calculate narrow-band vegetation indices by filtering sensitive bands through correlation analysis of grassland canopy spectra and hyperspectral feature variables,and to establish ground and airborne hyperspectral inversion models.By comparing the measured R2 with the predicted R2,the model with SAVI as the independent variable appears to be the best among the four narrow-band vegetation indices analyzed by ground-based and airborne hyperspectral,and therefore the SAVI vegetation index is more suitable for hyperspectral remote sensing estimation of typical grasslands in Inner Mongolia.Among them,the airborne hyperspectral data fit best with the form of dry weight index of above-ground extant biomass,i.e.,y=17.962e4.672x,and its fitted R2 was 0.542,the predicted R2 was 0.424,and the RMSE and REE were 57.03 and 0.65,respectively.and the accuracy of hyperspectral vegetation indices fitting were all higher than that of the fitted models constructed with hyperspectral characteristic variable parameters.Compared with the multispectral MODIS data inversion of biomass,the small-scale range is more suitable for the combination of ground and UAV imaging spectral techniques,which can invert the grassland canopy biomass data quickly and efficiently,and has good reference value for the remote sensing inversion of natural grassland biomass. |
PDF Full Text Request