Substituting Sentinel-3 OLCI for MODIS NDVI, Converting Historical MODIS into Synthetic Sentinel-3, and Generating National Days Early/Late Phenology Timing Maps
Abstract
ForWarn (https://forwarn.forestthreats.org) has been re-designed to utilize input data from the Sentinel-3 OLCI sensors on board the S3A and S3B satellites. The S3 OLCI production process requires downloading nearly 1,500 images from two servers, mosaicking, and applying ForWarn algorithms every 8 days.
Interpretation of the impact of forest disturbances requires the perspective of a lengthy historical archive going back 10 years or more, but the Sentinel 3 satellites were only launched about four years ago. We developed a series of statistical transfer functions that can transform historical MODIS NDVI maps into synthetic "Sentinel-3-like" NDVI maps, even for time periods before the Sentinel-3 satellites were first launched. We developed and used a set of special, customized statistical ecoregions, called "phenoregions," to spatially stratify vegetation types across the USA. A multivariate geographic clustering procedure was used statistically delineate 500 phenoregions across CONUS whose vegetation shows similar annual profile behavior in phenological timing. We then developed an independent linear regression between MODIS and S3 NDVI within each of the 500 phenoregions. Use of these independent, phenoregion specific regressions successfully removed any geographic pattern in the residuals, including high-NDVI conifer forest differences. A new Adaptive Baseline disturbance detection method uses the phenoregion zonal median current percent greenup progress as a basis to select the most-appropriate percentile of NDVI greenness from the historical phenoregion profile values, irrespective of the date when that level of greenness was historically observed. The constructed adaptive baseline map is a temporal mosaic of selected NDVI values that may have occurred at different historical times, and may be different from the current observation date. The temporal difference between the current time and the normal time when this best-matched greenness is usually seen shows the current earliness/lateness of phenological timing in each location.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.B36C..01H