Title: The Survey Science of Food Security: Astronomical Methods for Earth Data
Global agriculture is monitored by what an astronomer would recognize as a heterogeneous, multi-epoch, ground-based survey. Governments report subnational crop production with inconsistent survey instruments, changing footprints, and unknown selection functions. Meanwhile, machine learning models trained on satellite imagery now produce global yield predictions with no agreed-upon validation sample. I will discuss how methods from survey astronomy map onto food security science: catalog cross-matching, selection-function modeling, and photometric calibration of satellite observations to “spectroscopic” ground reports. I will present HarvestStat, the first globally harmonized dataset of subnational crop statistics, and StableBound, a framework for establishing a consistent coordinate system for these statistics over time. Together, these efforts represent the first consistent, global calibration datasets for Earth observations of agriculture. I trained as an ISM and time-domain astrophysicist, and this talk is an invitation to discuss how methods from survey astronomy can improve famine early warning systems and models of global agricultural production.
For Zoom information, please contact the seminar coordinators.
