Research And Application Of Long Time-series NDVI Reconstruction Method

The change of vegetation index can reflect the change of vegetation cover in the area,and it is of great significance to the development of the region by monitoring the change of vegetation in the time series.The Normalized Differential Vegetation Index(NDVI)performs well in a variety of vegetation indexes and is widely used.This thesis taking a place with many clouds and fog as the research area,such as Chongqing.Taking Google Earth Engine(GEE)as the main platform,and take Moderateresolution Imaging Spectroradiometer(MODIS)NDVI data as an example,aiming at some missing value,abnormal low values,and over-fitting phenomenon that easily occurs in the image reconstruction.And an improved Savtzky-Golay filter reconstruction method with weight update(Weight-Savtzky-Golay,WSG)is proposed.Secondly,using the Maximum-Value Composite,resample,and WSG method to generate a long timeseries NDVI dataset in 40 years.And cross-validation is performed using the adjacent time phase images of Landsat-8 and Sentinel-2.Then this thesis estimates the vegetation coverage based on the dataset and analyzes the law of temporal and spatial evolution.The specific research content and conclusions of this thesis mainly include the following parts:1.Aiming at the SG filter reconstruction algorithm's inability to reconstruct data at different time intervals and prone to over-fitting,the WSG method uses kriging interpolation to fill in the missing part of the image data and the weight update mechanism to suppress over-fitting.Secondly,the comprehensive performance of the method is discussed based on sample points in different quality areas and simulated noise experiments: the correlation coefficients of the new method in the sample point experiments are 0.72,0.76,and 0.82,which are all higher than those of the SG method;in the simulated noise experiment,the new method is better than the SG method.The correlation coefficients of 0.87 and 0.94 are also higher than the SG method,and the comparison of the difference found that the difference between the reconstruction result of the new method and the selected reference image is mostly within 0.1,a small part is between 0.1 and 0.2,and the maximum is not more than 0.3;The results of SG reconstruction are relatively poor,most of which are between 0.1 and 0.3,and the maximum value is even close to 0.5.Therefore,overall,the restoration and reconstruction effect of the new method is better than that of SG.2.Generate a long time-series NDVI data set.The resampling method of cubic convolution interpolation is used to improve the spatial resolution and combined with the maximum value synthesis method to ensure the temporal and spatial consistency of the multi-source data,and then use the WSG method to generate a 40-year NDVI data set based on the multi-source data.Finally,the NDVI values of Landsat-8 and Sentinel-2 in similar time phases were used to cross-validate the generated data sets to analyze the credibility of the data.The original MODIS images and reconstruction results have a high correlation with Landsat-8,and the correlation coefficient is higher than 0.6;while the correlation with Sentinel-2 is slightly insufficient,both are only higher than 0.3.3.Based on the pixel binary model,this thesis estimated the vegetation coverage in the study area on the new NDVI dataset.Combining landcover data,digital elevation data,and the annual maximum vegetation coverage to illustrate the law of temporal and spatial evolution in Chongqing.And finally based on trend analysis to explain the vegetation changes in the short term.The results show that the vegetation coverage in Chongqing is high in the east and low in the west;the overall coverage is relatively high,but it fluctuates slightly over time;however,areas with low vegetation coverage in cities and towns have increased year by year.The changes in the short term are as follows: The vegetation coverage is mostly unchanged or slightly improved,with less degradation;the obvious improvement areas are concentrated in the city center and the banks of the Jialing and Yangtze river,and the obvious degraded areas are concentrated in the periphery of the city.