Statistical Emulation for Environmental Sustainability Analysis

Oyebamiji, Oluwole Kehinde (2015). Statistical Emulation for Environmental Sustainability Analysis. PhD thesis The Open University.



The potential effects of climate change on the environment and society are many. In order to effectively quantify the uncertainty associated with these effects, highly complex simulation models are run with detailed representations of ecosystem processes. These models are computationally expensive and can involve computer runs of several days for their outputs. Computationally cheaper models can be obtained from large ensembles of simulations using a statistical emulation.

The purpose of this thesis is to construct cheaper computational models (emulators) from simulation outputs of Lund-Potsdam-Jena-managed Land (LPJmL) which is a dynamic global vegetation and crop model. This research work is part of a project called ERMITAGE. The project links together several key component models into a common framework to better understand how the management and interaction of land, water and the earth’s climate system could be improved.

The thesis focuses specifically on emulation of major outputs from the LPJmL model; carbon fluxes (NPP, carbon loss due to heterotrophic respiration and fire carbon) and potential crop yields (cereal, rice, maize and oil crops). Future decadal changes in carbon fluxes and crop yields are modelled as linear functions of climate change and other relevant variables. The emulators are constructed using a combination of statistical techniques of stepwise least squares regression, principal component analysis, weighted least squares regression, censored regression and Gaussian process regression.

Further modelling involves sensitivity analyses to identify the relative contribution of each input variable to the total output variance. This used the Sobol global sensitivity method. The data cover the period 2001-2100 and comprise climate scenarios of several GCMs and RCPs. Under cross validation the percentage of variance explained ranges from 52-96% for carbon fluxes, 60-88% for the rainfed crops and 62-93% for the irrigated crops, averaged over climate scenarios.

Viewing alternatives

Download history


Public Attention

Altmetrics from Altmetric

Number of Citations

Citations from Dimensions

Item Actions