Generating High Spatiotemporal Resolution LAI Using MODIS And GF-1 Data

Leaf Area Index(LAI) is a process model to describe the various terrestrial matter and energy exchange between the land surface and plant boundary layer. It is an extremely important status parameters so accurate estimates LAI is an important prerequisite for the assessment of vegetation growth and carbon sequestration capacity of maintaining ecological balance, and improving the living standards of human production. Thus, by means of remote se nsing to estimate large canopy, LAI is now the primary means of research in the field of ecological environment-related work. However, at this stage LAI standard product can not be taking into account time and spatial resolution at the same time, largely limiting the development of LAI-related research. Therefore, this in-depth discussion and research starts on this issue.Taking middle suburb in Huangpu District of Guangzhou C ity, explores Kriging interpolation method based on the assimilation of filtering fusion GF-1 and MODIS LAI high spatial resolution and high temporal resolution to generate high spatial and temporal resolution of LAI products. Meanwhile, in order to improve the accuracy of LAI products, adding landscape exponential factor in the integration process, describes LAI curve changes at different times from the macroscopic more aptly. The main content of the work as follows:(1)To investigate the reasonableness of Kriging interpolation method applied to time series of LAI productsCarrying out remote sensing image scaling K riging interpolation method, you first need to verify their compliance with spatial autocorrelation and meets the requirements of a normal distribution, otherwise it will reduce the accuracy of the scale after the conversion. Moran coefficients for MODIS products utilizing red band reflectance, near infrared reflectance, Normalized Difference Vegetation Index(NDVI and LAI of spatial autocorrelation analysis were 0.722, 0.703, 0.774 and 0.655, respectively, reflecting the study area vegetation types tend integrity, and meet the requirements of Kriging interpolation method. Moreover, the use of Kolmogorov-Smirnov test of MODIS products were subject to the normal distribution. Similarly, GF-1 red band reflectance, near- infrared reflectance and NDVI meet premise Kriging assumptions.MODIS product point of Kriging scaling to the appropriate 30 m resolution, GF-1 image by block Kriging scaling to the same spatial resolution verified MODIS downscaling indirect accuracy. The results showed that the red band reflectance near- infrared reflectance and NDVI error coefficients were less than 20.62%; Meanwhile, the use of the ground measured 50 groups LAI and MODIS LAI after scaling horizontal comparison, the average error factor was 25.48%, the results showed that after Kriging downscaling MODIS standard products was reasonable.(2)Construction of GF-1 NDVI and LAI estimation modelsBased on GF-1 remote sensing image data, ground LAI measured data were built for regression models with differe nt vegetation between LAI and NDVI. C lassified by vegetation, farmland, grassland and broad- leaved forests were built for equation exponential model to estimate the respective NDVI, R2 coefficients were greater than 0.833, so fitting accuracy was good enough to meet the large-scale estimates of different vegetation types canopy LAI needs.(3)LAI distribution based on fusion assimilation between MODIS LAI products and GF-1The GF-1 LAI profile image inversion integration into the MODIS LAI products need to use filtering assimilation method and also joined landscape index factor in the fusion process, and at last generated the annual study area high spatial resolution of LAI distribution. LAI measured after verification, the average prediction accuracy of LAI distribution time series was 72.80%, the results showed that filtered through assimilation method LAI time series was feasible.