A1040
Title: Estimating changing crop-low yield risks using non-stationary generalized Pareto distributions
Authors: Akio Onogi - Ryukoku University (Japan) [presenting]
Abstract: Estimating changing low-yield risks is important for crop breeding, as breeding efforts should prioritize regions where such risks are increasing. Based on extreme value theory, the probability of rare events can be approximated using generalized Pareto distributions (GPD). Non-stationary GPD models are applied to estimate trends in low-yield risks over time. The models are applied to global yield data of maize, wheat, rice, and soybean from 1961 to 2022, as well as local yield data for wheat, rice, and soybean in Japan from 1948/1958 to 2020. Results revealed increasing low-yield risks for maize and wheat in many global regions, while only wheat in Japan showed a mitigating trend. Model validation using simulations confirmed that the proposed methods can reasonably estimate changing risk levels, with estimation precision depending on data size. While the models can be further improved, the feasibility of a data-driven approach is demonstrated for assessing low-yield risks without relying on assumptions about climate or crop physiology.
