Study On The Reconstruction Of NDVI Time Series Data In Chongqing

The normalized vegetation index is a commonly used vegetation index,which is widely used in many fields such as global climate change,vegetation dynamic monitoring,vegetation biomass inversion and so on.There are many clouds in Chongqing,and the average sunshine percentage is less than 35% for many years.The cloud weather affects the quality of NDVI remote sensing products in the area,which leads to the hindrance of relevant application research.Therefore,based on the quality analysis,the NDVI of MODIS and SPOT-VGT are used to study the characteristics and applicability of various reconstruction methods in Chongqing area,which provides some references to the related application.First,the MODIS/QA and SPOT-VGT/SM dates are analyzed by space-time.Then,eight methods(such as BISE,MEAN,S-G,A-G,D-L,WS,SWETS and HANTS)are introduced from the aspects of principle,model and parameter setting.NDVI data of MODIS and SPOT-VGT are reconstructed by the eight methods.Finally,The reconstruction results of noise image based on visual contrast,curve comparison based on different vegetation types,based on root mean square error(RMSE)and correlation coefficient(R)of the fidelity of the comparison,and based on the Akaike information criterion(AIC)and the Bias information criterion(BIC)of model comparison.To systematically evaluate and compare the applicability of different methods,and to reveal the characteristics of NDVI data of two kinds of data sources,and finally quantitative analysis of the results of the two kinds of data.The results show that:(1)In terms of the quality analysis of the two data sources NDVI: MODIS/QA and SPOT-VGT/SM data are affected by climatic conditions,which are best in spring and worst in winter.However,the quality of remote sensing data of two kinds of NDVI in the main city and surrounding areas is poor.In southeastern region of Chongqing,MODIS/QA data is good,but SPOT-VGT/SM is poor.(2)The 8 kinds of reconstruction methods have their own characteristics and application:(1)the effects of A-G's and D-L's reconstruction are stable in the application of two dates.The D-L method is more clutter to the noise,while the A-G method still has an uncompleted value.The two reconstructed curves are similar to each other,and they can effectively identify reconstructed noise values.At the same time,the two methods are at a better level in the fidelity and model evaluation,A-G method is better than the D-L method.(2)The parameters required by the method of WS are less,the performance of MODIS/NDVI is well,the smoothness of reconstruction curve is poor,and the reconstruction capability of continuous noise is the same.But can effectively fill the original image of the missing value,while it has a better performance in the fidelity and model evaluation.The results of the SPOT-VGT/NDVI fidelity and model evaluation are not as good as MODIS/NDVI.(3)The SWETS method is slightly better than WS method in MODIS/NDVI reconstruction,but it is better to maintain root mean square error when reconstructing SPOT-VGT/NDVI.(4)The HANTS can effectively fill the missing data to maintain the smoothness of the reconstruction curve,but the credibility of its reconstruction remains to be verified.It applies to the continuous absence of data quality space-time.(5)The S-G reconstruction curve is smoother and is in the upper envelope of the data,which is at a moderate level in the reconstruction of the two data.But it will appear to ignore the normal low value of the situation,after the reconstruction of the image can effectively fill the missing value.Its effect is slightly better in SPOT-VGT/NDVI,and it is at a moderate level in the MODIS/NDVI reconstruction.(6)BISE method is not suitable for large-scale data reconstruction,because its reconstruction curve is rough and a number of indicators of the evaluation results are poor.(7)MEAN method has better effect on data fidelity,which can effectively maintain the quality of high-quality data.But the ability of eliminate noise is poor,so it is more suitable for data noise less and discontinuous research area.(3)Compared with SPOT-VGT/NDVI,MODIS/NDVI data range is wide,which can clearly show the river and other information.The correlation coefficient between the two data is generally above 0.7 level,of which the highest in Northeast Chongqing,the northeast is low and around to improve.During the year,the MODIS/NDVI value of each month was higher than that of SPOT-VGT/NDVI,but the trend was similar.All of them were higher in summer,followed by spring and autumn,and lowest in winter.