Building a Spatially-Explicit Phenology Model for Colorado Potato Beetle

Incentive programs to promote climate mitigation and soil health often resort to incentivizing practice adoption and crudely estimated benefits, rather than the actual soil carbon accrual. A transition to incentivizing the benefit itself, aligned with the principles of true cost accounting, is critical. Our interdisciplinary team will tackle this aspect in the Pacific Northwest dryland systems. As part of this seed grant, we will develop and evaluate a machine learning model for predicting soil organic carbon in the 0-30cm of soils by integrating satellite imagery with environmental and management covariates. Two hypotheses will be tested: (a) we can successfully predict SOC in the 0-30cm of soils with satellite observations, and (b) a gaussian-processes-based machine learning approach is better than traditional generative imputation approaches for small datasets. We believe the proposed outputs will strengthen an extramural resubmission to USDA NIFA.