| In recent years, with the development of hyperspectral remote sensing technology and its own advantage, the applications of hyperspectral technology in vegetation growth information detection, yield prediction and nutrient diagnosis has been one of the hot research topics. This article studied the grassland’s spectral characteristics and the relationship between its agronomic parameters and its high spectrum in two ways of the artificial quantitative test and field plot test. The artificial quantitative test studied from the two aspects of the different nitrogen levels and the different soil texture by pot experiment. This research took the different types of grassland (the first year of the Annual bluegrass, Festuca arundinacea and Sorghum sudanense; the second year of Mexican teosinte, Pennisetum americanum, and Sorghum hybrid sudangrass) as the object, studied the spectral characteristics of grassland, and established the sensitive spectral parameters and prediction equation of the aboveground biomass, SPAD value and nutrient content (N, P, K) in different types of grassland. The field test was mainly based on NDVI data of MODIS, took the south grassland as the object, combined with the field measured data, to analyze the relationship between the grassland’s NPP and NDVI and construct the Southern grass NPP remote sensing estimation model based on the NDVI for the independent variable and hydrothermal conditions for adjustment factor. The model was validated, through the different years of independent experimental data. The main contents and results of the present study were summarized as follows:1. The grass yield of the three species (Annual bluegrass, Festuca arundinacea and Sorghum sudanense) was different in different nitrogen levels. In different nitrogen treatment, the canopy spectral reflectance of these three species had significant differences in the visible region. It had a strong absorption band at a wavelength of about680nm-700nm, the spectral reflectance change was decreased first and then increased basically with the increase of the nitrogen concentration.2."Red edge" was the reflectance of green plants increased fastest point in the680-760nm, there were "two peaks" phenomena for the red edge of canopy spectra of Sorghum sudanense. Red edge positions moved to longer wave bands with the amount of nitrogen applied increasing. This was because the leaves have the higher content of Chlorophyll under higher nitrogen level. With the arrival of the late growth period, the "Twin Peaks" phenomenon was gradually weakened. The phenomenon of" Red transference" of the Red edge was not obvious.3. In the two years of pot experiment, the normalized difference vegetation index (NDVI) and ratio vegetation index (RVI) had good correlation to grass yield, so it was feasible to establish grassland yields’spectral model based on the NDVI and RVI. But there were some difference in the correlation of different vegetation index and aboveground biomass. Different types of grassland had their optimal models in different periods.4. In different soil textures, we measured the hyperspectral reflectance and its corresponding leaf SPAD value of the three species in different periods. In these three types of grassland, the spectrum parameter for the best fitting regression relationship with leaf SPAD value was the peak area of red edge. And get best model to estimate the leaf SPAD value, Mexican teosinte was y=301.87x2-109.38x+54.774; Pennisetum americanum was y=504.36x2-127.29x+37.22; and Sorghum Hybrid Sudangrass was y=451.3x2-153.36x+50.374.5. In different soil textures, we measured the hyperspectral reflectance and its corresponding total nitrogen content of the plant. After the selection, verification, the result showed that reflectance value of the near-infrared peak was the best inversion parameter to estimate the total nitrogen content of the plant. And got best model to estimate the total nitrogen content of the plant, Mexican teosinte was y=-30.606x2+16.244x+0.5248; Pennisetum americanum was y=-13.789x2+0.7791x+4.2015; and Sorghum hybrid sudangrass was y=-30.392x2+14.859x+0.8592.6. In different soil textures, we measured the hyperspectral reflectance and its corresponding total phosphorus content of the plant. The canopy spectral reflectance in726,773,873nm wavelength and the first derivative spectra data of720nra were chosen to analyze. After the selection, verification, the result showed that reflectance value of773nm wavelength was the best inversion parameter to estimate the total phosphorus content of the plant. And got best model to estimate the total phosphorus content of the plant, Mexican teosinte was y=16.753x29.0759x+1.4304; Pennisetum americanum was y=-30.949x2+24.278x-4.4188; and Sorghum hybrid sudangrass wasy=0.1033e27825x.7. In different soil textures, we measured the hyperspectral reflectance and its corresponding total kalium content of the plant. In the Mexican teosinte and Sorghum Hybrid Sudangrass, almost no band had the significant correlation with the total kalium content of the plant in the350-1200nm band. The canopy spectral reflectance of Pennisetum americanum in the790-1119nm band showed a significant negative correlation with the total kalium content of the plant. The spectrum parameter for the best fitting regression relationship with total kalium content of the plant was the reflectance value of886nm wavelength, and got best model to estimate the total kalium content of Pennisetum americanum was y=48.952x2-43.342x+10.223.8. There were five kinds of correlative types between NPP and NDVI in Southern grassland. There was a good correlation and consistency between simulated values and measured values of NPP. R2values of the five types were0.9022,0.8266.0.8712,0.8877and0.8755, they all had reached the extremely significant level, and the root mean square error (RMSE) and relative root mean square deviation (RRMSE) were all small. It showed that the simulation result of the model was reliable, and it provided an effective method for the estimation of grassland NPP and the resource management in Southern grassland. |
PDF Full Text Request