A global dataset of noiseless time-series vegetation and water indices for farmland analysis

Time-series vegetation and water indices derived from remote sensing images are crucial materials for farmland analysis based on crop phenology and monitoring such as classification of cropping patterns and monitoring of crop growth and productivities. However, the optical remote-sensing images often contain problematic pixels, such as those affected by thick clouds and its shadows, resulting a noisy time-series data which is inadequate to use for farmland analysis. Lack of the high quality noiseless time-series data products have been a bottleneck for many studies on farmland analysis using remote-sensing data in both of global and local scales. Hence we aims in this study to produce remote-sensing data archives containing noiseless smoothed time-series Normalized Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Land Surface Water Index (LSWI), and RGB composite (true color) image, covering global wide with the adequately downscaled 250-m resolution in 8-days interval from the period of 2000 to 2014 and up-to-date, derived from based on the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor images. The processing system for generating the global noiseless time-series data in this study is consist of four modules; image acquisition, preprocessing, noise reduction, and data archiving, implemented in Interactive Data Language (IDL) with PDP (Parallel Distributed Processing). Data products are derived from two sets of Terra/MODIS surface-reflectance 8-day composite data (MOD09Q1 v005 (250-m resolution) and MOD09A1 v005 (500-m resolution)) acquired from LP DAAC (Land Processes Distributed Active Archive Center, USGS) on a regular basis automatically to update the latest products. To generate noiseless and smooth time-series data, we applied following well-established procedure: 1) Eliminate cloud and sensor noise pixels by referring the image quality information: 2) Reconstruct daily time-series data by referring the actual date of observation for each pixels: 3) Synthesize daily time-series Terra and Aqua: 4) Interpolate eliminated pixel value on time-series axis by liner interpolation method: 5) Apply the Savitzky-Golay filtering to time-series data to produce smoothed time-series data. Denoised true color image can be used for visually checking the quality of noise reduction. Data products are mapped by MODIS Sinusoidal Tiling System, and archived with Hierarchical Data Format-Earth Observing System (HDF-EOS) format containing various meta-information, which are the standard archive format for MODIS Land products distributed by USGS, so that users can use existing favorite tools such the MODIS Reprojection Tool (MRT) for handling the archives. The advanced noiseless time-series vegetation and water indices data archives produced in this study would contribute significantly to advanced analysis and monitoring of the crop production.