| In order to improve the estimation accuracy of fractional vegetation cover and master the changes of fractional vegetation cover and its development trends in time.Qinghai province was chosen as the study area,remote images with 30 m spatial solution were chosen as data source,studies included estimation,prediction and change monitoring of fractional vegetation cover were carried out in this thesis.The final aim is to provide data and dicision support for ecological enviroment management and ecological security early warning in Qinghai.The main research contents and conclusions are as follows:(1)Study on the estimation of fractional vegetation cover using endmember spectral interpolation.The method of deciding the (1 and (1 parameters of dimidiate pixel model was improved in this study,a new method based on endmember interpolation was proposed and clarified it could increase the evaluation accuracy of FVC.During the FVC estimation process based on dimidiate pixel model,the influence of local environmental difference on endmember spectral characteristics is taken into account,and according to the theory of spatial autocorrelation,(1 and (1 were interpolated based on the (1 of pure soil samples and pure vegetation samples by Ordinary Kriging method.The results showed that,this method of deciding the (1 and (1 improved the FVC estimation accuracy,compared with the common statistics and thresholds method,RMSE decreased from 0.170 to 0.156(decreased 8.24% relatively),MAE decreased from 0.137 to 0.124(decreased 9.49% relatively);further analysis showed that the estimation accuracy of edge verification points and non-edge verification points were both increased,but due to the impacts of registration errors and diffuse reflection,the accuracy of edge verification points was lower than the accuracy of non-edge verification points.(2)Study on the prediction of vegetation cover.Vegetation cover prediction model was instituted based on Cellular automata and Markov,the accuracy prediction was tested,each grade condition in the Yellow River Basin in 2018 was predicted by this prediction model.In this paper,the vegetation Yellow River Basin in 2015 was firstly predicted by the prediction model based on vegetation ranked data from 2009 to 2012 to test the prediction accuracy,the results showed that,the Kappa coefficient between estimated results and predicted results was 0.83.Then the ranked data in 2015 were taken as the original state of cells,transition probability matrix of each vegetation grade from 2012 to 2015 was taken as the transition rule,each vegetation grade of the Yellow River Basin in 2018 was predicted by the predicted model.The results showed that high vegetation cover dominated in 5 grades and the condition of each grade is stable repectively.(3)Study on the change monitoring of fractional vegetation cover in Qinghai.In this study,we first estimated the fractional vegetation cover in Qinghai annually,then the characteristics of vegetation cover in Qinghai in 2015 were analyzed.The FVC stability in time series was rated by coefficient of variation,the FVC change trends were revealed by Slope function.The results showed that,the FVC was higher in the east and lower in the west,and higher in the south and lower in the north,generally,FVC represented a trend that it was decreasing from southeast to northwest.The average FVC in Qinghai grass areas was 0.52,the CV was 0.29.The CV in Sanjiangyuan was 0.35,the stability in Yellow River Basin was highest of the three river basins which was 0.26.In 2009-2015 years,the proportion of no change trend was 68.29%,which occupied most in 7 change trends.Three kinds of improving trends occupied more than three kinds of deteoriating trends,and slight improving trend and slight deteoriating trend dominate the six change trends.The proportion of no change trend and three kinds of improving trends were totaly 87.95%,it meant that FVC in the most areas in Qinghai province represented improving and stable trend. |
PDF Full Text Request