| 【Objective】Studying the relationships between the fraction of photosynthetically active radiation (FPAR) and the difference of hyperspectral reflectance rates from Seeding stage to Vomiting stage in cotton,we conducted a comparable experiment with four waters and densities.Based on the China environment and disaster reduction satellite data from shihezi 121 groups study area in xinjiang Province, the survey data and integrated the synchronization observation data, monitoring the FPAR of cotton.This paper used the equation NPP=(SOL×FPAR×O.5)×εto calculate the cotton net primary production. Finally, estimation yield data through converting the calculated NPP to drymatter and then cotton yield were obtained.【Method】The data which from near ground hyper-spectrum, satellite image and field research were analysized using original reflectance,first derivative spectrum,spectral continuum removal method and anova analysis in 2009-2010 years. The characteristics of spectrum, the estimating models of FPAR (Fraction of Photosynthetically Active Radiation) were established in cotton on the base of relation analysis and regression analysis, and the accuration of estimation models were assessed using RMSE(Root Mean Squar Error),RE(Relative Error)and r (relatation coeffectance). In addition, the images were preprocessed and analyzed by ENVI 4.7 software in study area. Finally, monitoring the FPAR changes of cotton, the CASA (Carnegie Ames Stanford Approach) models was used for estimating cotton yield.【Result】Through monitoring cotton FPAR using multi-platform remote sensing, this paper got progress as follows:(1) The fraction of photosynthetically active radiation changes and estimation by spectrum parametersThe results indicated that fourteen vegetation indices were compared with corresponding FPAR.The FPAR had significantly positive correlation to GREENNDVI and GMI in those vegetation indexes,and theirs Correlation coefficient(r) arrived at 0.794 and 0.765, respectively.The coefficients of determination(R2) of two models were 0.6566 and 0.6346 by erecting model with the two indexes and FPAR,Root mean square error(RMSE) were 0.089 and 0.093, respectively. The overall results suggested that fraction of photosynthetically active radiation(FPAR) in cotton have stable relationships with some vegetation indices,especially indexes of GREENNDVI and GMI.(2) Monitoring cotton FPAR based on HJ-1 CCDThe three vegetations (NDVI, RVI and SAVI) were calculated. The estimation precision of the three vegetation indexes model for the cotton was compared by the regression analysis of the vegetation indexes with the survey data and integrated the synchronization observation data. The results indicated those vegetation indexes with high relation to FPAR; the NDVI was the top inversion precision, the optimum model to estimate the cotton FPAR, the model validation mean error was 1.84%. That is, the model with high precision and using the model to inverse the June to August FPAR of 121 groups research region in xinjiang Province.(3)Prediction of cotton yield based on HJ satellite imageThrough Time series of the environment satellite (HJ) dataset obtained Fraction of absorbed Photosynthetically active radiation, Photosynthetically active radiation and NPP, they were sequenced as: August,July,June. Average NPP of the August,July and June was proportion of total cumulation NPP is 45%,30% and 25%. Using the NPP model to estimate cotton yield, relative error between the predicted cotton yield and the actual yield was-18.00%.【Conclusion】The overall results suggested that spectral parameters can be effectively used to estimate the FPAR during growth stage of cotton. That is, the model with high precision and using the nodel to inverse the June to August cotton FPAR of the 121 groups of shihezi research region in xinjiang Province. It suggesting that it was feasible to predict cotton yield by using the NPP parameter to estimate cotton yield based on remote sensing data. |
PDF Full Text Request