Solar and sensor geometry, not vegetation response, drive satellite NDVI phenology in widespread ecosystems of the western United States

Abstract:

Satellite-derived phenology metrics are valuable tools for understanding broad-scale patterns and changes in vegetated landscapes over time. However, the extraction and interpretation of phenology in ecosystems with subtle growth dynamics can be challenging. US National Park Service monitoring of pinyon-juniper ecosystems in the western US revealed an unexpected winter-peaking pattern in normalized difference vegetation index (NDVI) time-series derived from Moderate Resolution Imaging Spectroradiometer (MODIS) imagery. We assessed the validity of the winter peaks through ground-based observation of phenology and examination of solar and satellite geometry effects. To test the premise of a true vegetation response, we analyzed NDVI values extracted from a time-series of ground-based digital camera (‘phenocam’) images collected September 2017 to December 2018 in a pinyon-juniper woodland in Arizona, US. Results show pinyon and juniper growth peaked in the warm season, as did the other species in the phenocam field of view. Examination of NDVI time-series (2003-2018) derived from daily 250-m MODIS data in the broader pinyon-juniper ecosystem revealed that solar-to-sensor angle, sensor zenith angle, and forward/back-scatter reflectance explained >80% of intra-annual variability. Solar-to-sensor angle, which exerted the greatest influence, is controlled seasonally by solar zenith angle and daily by variations in satellite overpass geometry. Mapping winter peaks across the western US in Google Earth Engine using 500-m MODIS MCD43A4 data revealed that winter peaks are consistent (>= 14 years, 2003 to 2018) in pinyon-juniper and non-pinyon-juniper conifer ecosystems and common (>= 5 years, 2003 to 2018) in areas of shrubland. We attribute winter peaks to the positive correlation of NDVI with solar-to-sensor angle and solar zenith angle in combination with sparse, vertically oriented evergreen vegetation canopies. These findings are significant for the appropriate interpretation of phenology signals from satellite time-series for assessments of ecological health, status, and trends. Future research will investigate the susceptibility of alternate vegetation indices to solar-sensor geometry effects to determine the most accurate method of representing vegetation dynamics.