Scaling Phenocam GCC, NDVI, and EVI2 with Harmonized Landsat-Sentinel using Gaussian Processes

Abstract

The use of near-surface remote sensing for monitoring vegetation phenology has advanced greatly over the past decade. The Phenocam Network has deployed more than 500 web-enabled cameras across the globe that use digital repeat photography to capture color information and measure changes in vegetation phenology across diverse ecosystems. Vegetation indices (VIs) such as the Green Chromatic Coordinate (GCC), Normalized Difference Vegetation Index (NDVI), and two-band Enhanced Vegetation Index (EVI2) have been derived from phenocam imagery. However, it is often necessary to scale these metrics to align them with satellite imagery since phenocam data are not standardised to surface reflectance. We developed a method to convert phenocam digital numbers (DNs) to align with Harmonized Landsat-8 and Sentinel-2 surface (HLS) reflectance products using a Gaussian process. We applied our method across six grassland phenocam sites. The Gaussian Process regression was on average able to account for 77.4 percent of the variation in the HLS surface reflectance, and our resulting phenocam VIs had a high level of agreement with the modeled HLS VIs with an R2 of 0.811. This technique provides a novel method for standardising phenocam imagery, easing comparison between multiple phenocam locations and satellite or other sensors that have a standardised surface reflectance product.