Decomposing patterns of interannual variability in remote sensing vegetation data

Gross Primary Production (GPP) is the single largest flux in the global carbon cycle, but is difficult to measure directly and therefore contributes significant uncertainty in assessing climate-carbon feedbacks. We analyze the climate-driven interannual variability of several satellite datasets that provide insight into GPP on regional to global scales. We compile monthly averaged time series of Solar-Induced chlorophyll Fluorescence (SIF) from the Global Ozone Monitoring Experiment (GOME-2), EVI from the MODerate resolution Imaging Spectroradiometer (MODIS), and NDVI from MODIS and the Advanced Very High Resolution Radiometer (AVHRR) from 2007 through 2015. We select a few representative ecoregions, and apply a singular vector (SV) decomposition method to deconstruct the interannually varying component of the timeseries over the course of the growing season. Across all regions we analyze, this decomposition results in two dominant SVs to describe patterns of interannual variability: a 'shift' vector that represents a temporal redistribution of productivity within the growing season, and a 'scaling' vector that modulates the magnitude of the peak summer signal. We investigate climate drivers that result in either temporal shifts or changes to growing season GPP via the two SVs by correlating the timeseries of SV weights with interannual anomalies in liquid water equivalent (LWE) from the Gravity Recovery And Climate Experiment (GRACE) satellite and temperature from the Climate Research Unit Time Series (CRU-TS). Our results suggest that temperature variations tend to drive shifts in GPP, while moisture availability often controls the cumulative seasonal productivity. Further, our results demonstrate the contrasting characteristics of SIF and VI observations, and show that SIF variations are more evident during peak growing season than are those of the VIs.