Remote Sensing Yield Estimation Of Meadow Steppe Based On Hyperspectral Data

Grassland,as an important ecosystem regulating carbon balance,is the main carrier of animal husbandry development.Measuring grassland productivity can protect grassland ecological environment,ensure the sustainable development of animal husbandry,and play an important role in the rational utilization of grassland resources.Based on the reflectance of the ground ASD non-imaging spectrometer and the UAV Resonon imaging spectrometer,the yield estimation model can quickly,comprehensively and accurately obtain the grassland productivity.In this paper,two meadow steppes treated by different measures in Bayanhua Town,Xiwuzhumuqin Banner,Xilinguole League,Inner Mongolia were selected as the study area.The spectral reflectance of vegetation was measured from July to August 2019,and the aboveground biomass data were collected by sample method.After analyzing the correlation between aboveground fresh biomass and spectral reflectance,single-band and hyperspectral variables were selected to construct the yield estimation model.Conclusions are as follows:1.Comparing the ASD hyperspectral reflectance curves of meadow steppe with different treatments,the two curves are consistent with the spectral response characteristics of green plants.Due to differences in the species and dominance of the dominant plants within the sample,meadow canopy reflectance values in the mowing treatment were higher than those in the restoration measures treatment at all wavelengths.2.By comparing the correlation coefficient between aboveground biomass and ASD canopy spectral reflectance in the same year,it was found that the correlation between canopy reflectance and aboveground fresh biomass was greater than that between aboveground dry biomass,indicating that the yield estimation modeling of meadow steppe was more suitable for fitting regression curve with fresh biomass.3.The ASD hyperspectral reflectance was significantly correlated with the Resonon hyperspectral reflectance at the 0.01 level,and the Pearson correlation coefficient was 0.988,which confirmed the reliability of the yield estimation model constructed by the reflectance data measured by the Resonon hyperspectral spectrometer equipped with UAV.Compared with ASD hyperspectral reflectance,Resonon hyperspectral reflectance had higher correlation with aboveground fresh biomass.4.Single-band,hyperspectral characteristic variables and narrow-band vegetation index estimation modeling based on ASD hyperspectral data.The 377nm reflectivity,quadratic polynomial model fitted by Rr and logarithmic model fitted by TCI are better in the yield estimation model of mowing plots.（R~2 0.517,0.412,and 0.327,respectively）;In the yield estimation model of the restoration sample plots,the quadratic polynomial models fitted by the first derivative of 624 nm reflectance,VI2and h NDVI were better（R~2 was 0.457,0.253 and 0.236,respectively）.5.In the yield estimation model of mowing meadow steppe constructed by single-band,hyperspectral characteristic variables and narrow-band vegetation index based on Resonon hyperspectral data,the first-order derivative of 729 nm reflectance,the quadratic polynomial model of VI1 and SAVI were better（R~2 was 0.541,0.408 and0.553,respectively）.It was found that the R2 model based on Resonon hyperspectral data was higher than that based on ASD hyperspectral data,which was more suitable for estimating the yield of mowed meadow steppe.