A1097
Title: A deep learning framework for predicting multi-crop land suitability with satellite data
Authors: Amanjot Bhullar - University of Guelph (Canada) [presenting]
Abstract: Assessing land suitability is key to agriculture management, as it helps identify the capability of different lands to cultivate different crops. $DeepS^3$ is presented, a multilayer perceptron framework that simultaneously predicts the land suitability for multiple crops based on satellite imagery, crop-specific farm location data, and district-level crop production census data. The method can be viewed as a multi-task learning method that enables a multivariate model in the presence of unobserved responses. The multi-crop response is exploited during training, and the interdependencies among different crops are captured, thereby facilitating extrapolation. This framework easily generalizes to other spatial problems that incorporate data obtained at diverse spatial resolutions. Applying $DeepS^3$ to predict crop land suitability for Canada under climate change projects diminishing suitability for canola, peas, wheat, and soy in the Prairie Provinces, mainly driven by increased heat stress.
