B1419
Title: Where should we grow them? Deep learning for agricultural management in Canada
Authors: Amanjot Bhullar - University of GUelph (Canada) [presenting]
Khurram Nadeem - University of Guelph (Canada)
Ayesha Ali - University of Guelph (Canada)
Abstract: Land suitability is used in agricultural management to understand which lands are able to grow a given crop. Deep$S^3$ is presented, a multilayer perceptron-based simultaneous land suitability scoring model that can accommodate both high- and low-resolution data. Farm-level locations are combined with district-level crop yields and soil-climate-landscape variables from Google Earth Engine to predict the land suitability of several crops in Canada simultaneously. The trained model is then used to project the future land suitability under RCP scenarios 4.5 and 8.5. Land suitability of peas, spring wheat, canola, and soy is expected to become less suitable in the Prairie provinces, central British Columbia, and parts of Ontario and Quebec. In contrast, the future land suitability in southern British Columbia, southern Ontario, and the Maritime provinces is predicted to remain consistent with current times. These findings suggest that it may be advantageous to investigate crop diversification or cultivation of novel crops that can survive higher diurnal temperatures in order to maximize yield in current agricultural lands. Regardless, the development of sustainable agricultural management strategies informed by land suitability models and increasing crop biodiversity to adapt to changing climatic conditions will be critical for maintaining food security and economic stability.