CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: A Bayesian non-linear hierarchical framework for crop models based on big data outputs Authors:  Muhammad Mahmudul Hasan - Durham University (United Kingdom) [presenting]
Jonathan Cumming - Durham University (United Kingdom)
Abstract: Due to the increasing trend of world population, proper fertilization is very crucial for crop productivity to maintain the levels of food which will be required. We base our analysis on the big data output from the Environmental Policy Integrated Climate (EPIC) model, which provides time series output of the crop yield (among other outputs such as pollution indicators) in response to changes in inputs such as fertilizer levels, weather, and other environmental covariates. At the initial stage of our research, we investigate the results of the simulation of a full factorial design in nitrogen and phosphorus fertilizer levels for 3 different crops and crop rotations. We apply a non-linear Bayesian hierarchical model based on established yield models in order to make inferences about the response of crop yield with respect to fertilizers by using EPIC outputs. We use Markov Chain Monte Carlo to obtain samples from the posterior distributions, to validate and illustrate the results, and to carry out model selection. The results highlight a strong response of yield to nitrogen, but surprisingly a weak response to phosphorus for this particular simulator configuration and catchment.