LAUSR

Abstract Remote sensing of solar-induced chlorophyll fluorescence (SIF) shows promise for monitoring the productivity of global agricultural systems. SIF-based primary productivity metrics have demonstrated higher fidelity to large-scale patterns of crop productivity than reflectance-based vegetation indices when averaged across the growing season. In-season crop yield forecasting typically relies upon reflectance-based vegetation indices, raising the question of whether in-season monitoring could be improved by utilizing SIF. Here, we analyze patterns of US agricultural productivity from USDA surveys and their in-season relationships with coarse-resolution GOME-2 SIF, high-resolution downscaled SIF, SIF-based primary productivity metrics, MODIS NDVI, and MODIS GPP. We find that coarse-resolution SIF-based metrics and NDVI exhibit similar out-of-sample in-season (April–July and April–August) predictive ability, even when spatially filtering higher-resolution NDVI data to cropland areas. The downscaled SIF product performed more poorly than the coarse-resolution SIF, and MODIS GPP performed more poorly than MODIS NDVI. All forecasts are improved by incorporating county fixed effects to control for cross-sectional differences between counties. NDVI-based metrics allow for significantly better yield predictions during drought conditions than SIF-based metrics, suggesting limited added value of SIF for early warning of drought impacts. The benefits of SIF for crop monitoring should be continually evaluated as the frequency and quality of SIF measurements continue to improve.