Science activity

Spectral indices such as NDVI have long been found to be good predictors of plant productivity at many spatial scales from the canopy to the landscape. Spectral indices are an important tool for upscaling GPP fluxes we measure at the ecosystem scale through Eddy Covariance up to larger spatial scales. Other indices, such as NIRv (expressed as NDVI * total NIR) have also been shown to be potentially more accurate predictors of GPP using in-situ spectral measurements than NDVI alone. Additionally, associating spectral signals within modeled flux footprint areas has been shown to improve the predictive capability of spectral indices compared to estimates using remotely sensed data centered directly on top of flux towers. Most if not all of these spatially explicit footprint analyses have been done by aggregating footprints into polygons based on their 50%-90% estimated flux contributions, and then associating those polygons with fluxes and spectral signals within them. This approach has been necessary largely because of the spatial scales involved with satellite remote sensing products, reaching a practical minimum of 3m, downsampled from 4.8m imagery by Planet Labs. By combining pixel-weighted flux footprint contributions with ultra-high resolution (3cm) spectral drone data, we will examine and compare how different spatial scales and indices affect the capability of spectral data to predict fluxes which are not directly measured.