Evaluation of NPP VIIRS Vegetation Index EDR performance using MODIS and AVHRR data records

Vegetation Index (VI) is one key parameter to specify the boundary condition in global climate models, weather forecasting models and numerous remote sensing applications for monitoring environmental state and its change. The VI Environmental Data Record (EDR), which includes the Top of Atmosphere Normalized Difference Vegetation Index (TOA NDVI) and the Top of Canopy Enhanced Vegetation Index (TOC EVI), is currently operationally generated from data delivered by the Visible Infrared Imaging radiometer Suite (VIIRS) instrument onboard the National Polar-orbiting Partnership (NPP) platform launched in October 2011. The VI EDR was implemented to provide continuity for 30+ years of historical VI records provided by MODIS and AVHRR sensors. This presentation reports on the results of the analysis performed by the JPSS VI group at NOAA-NESDIS-STAR on two major aspects of performance of the VI EDR in the early phase of the NPP mission: (1) assessment of accuracy of the VIIRS VI EDR product with respect to input data including Surface Reflectances, Cloud and Aerosol masks as function of vegetation (biome) types; (2) temporal and spatial consistency of VIIRS VI EDR with respect to heritage MODIS and AVHRR VI products. This analysis is based on data from VIIRS (daily TOA NDVI and TOC EVI, and daily surface reflectances), Terra MODIS (16 days composites of TOC EVI and TOC NDVI, and daily TOA radiances) and NOAA-18 AVHRR (7-days composites of TOA NDVI). MODIS 8-biome landcover mask was used to quantify variations in VI product performance as function of vegetation type. Best overall agreement is achieved between VIIRS and MODIS data (TOC EVI and TOC NDVI) in terms of minimum systematic discrepancy (minimum bias and STD) and highest correlation of spatial patterns (highest r^2). The agreement is highest for biomes with low vegetation cover, but degrades with increased foliage density. VIIRS cloud mask provides a fair screening of daily data over the globe. While performance of TOC EVI is reasonable over the globe, anomalously high/low values are observed spot-wise at northern high latitudes at snow-covered regions at high values of solar zenith angle. Substantial systematic bias was found between VIIRS and AVHRR TOA NDVI, especially at high values of NDVI, where AVHRR data are lower. This could be due to difference in channel bandwidth and sensor calibration. VIIRS VI and surface reflectance algorithms are currently undergoing the Cal/Val phase and improvements in the algorithms are expected to occur and enhance the EDR performance.