A novel re-compositing approach to create continuous and consistent cross-sensor/cross-production global NDVI datasets

The longest Normalized Difference Vegetation Index (NDVI) time series produced from the National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) has ended in 2017. At some point in the near future, all AVHRR sensors will be retired. To maintain continuity and consistency of this global data set, it is imperative to extend NDVI from other sensors, especially the operational Visible Infrared Imaging Radiometer Suite (VIIRS), which planned to maintain continuity at least through 2038. NDVI could be de-composited into two components: (1) the multi-year climatology and (2) the vegetation condition index (VCI). The former contains climate information and a majority of sensor noise, and the latter contains weather information and residual sensor noise. With the assumption that VCI from different sensors are similar, we re-composited the cross-sensor/cross-production NDVI with original VCI and the cross-sensor/cross-production climatology, and compared various cross-converted datasets with the three base NDVI datasets: two NDVI productions derived from AVHRR observation and another from VIIRS observation. As a result, the re-composited NDVI agrees well with the target base NDVI spatially and temporally, with an accuracy of 0.02 NDVI unit at a global scale. The comparison with several regression approaches distinguish the superiority of the new re-compositing approach.

Automated summary

Extending the NDVI dataset from NOAA AVHRR sensors to other sensors like VIIRS is crucial for maintaining global data continuity. By decomposing NDVI into components and re-compositing cross-sensor NDVI, the study demonstrates the superiority of the new approach in maintaining spatial and temporal consistency with high accuracy.