Sexton, Justin David (2015) Bayesian statistical calibration of variety parameters in a sugarcane crop model. Masters (Research) thesis, James Cook University.

Preview

PDF (Thesis) Download (1MB) | Preview

Abstract

Process‐based agricultural systems models capable of simulating crop growth, management decisions and varietal differences in productivity allow researchers to investigate the interactions between varieties, production environments and management decisions. The Agricultural Production Systems sIMulator (APSIM) currently includes variety parameters that represent physiological traits for 14 sugarcane varieties. Unfortunately most of these 14 sugarcane varieties are no longer grown commercially. This makes it difficult for industry decision makers to trust the outputs from the model and thereby incorporate model outputs into the decision making process. To overcome this weakness in the APSIM crop model for Australian sugarcane systems the following thesis objectives were developed:

1. investigate the capability of the APSIM‐Sugar model to simulate yield differences between sugarcane varieties under different climatic conditions;

2. investigate the sensitivity of model outputs such as biomass and sucrose yields to key model input parameters; and

3. evaluate the use of two Bayesian approaches to calibrate variety parameters in the APSIM‐Sugar model.

The APSIM‐Sugar model was used to simulate biomass and sucrose yields of four sugarcane varieties grown under well irrigated and water stressed conditions in a breeding trial conducted at Home Hill, Queensland, Australia. Comparisons were made between observed and simulated varietal differences in yield and yield response to water stress. Bayesian Analysis of Computer Code Output (BACCO) was then used to perform a global sensitivity analysis of model outputs (biomass and sucrose yields) to key variety parameters under well irrigated and water stressed conditions. Finally, Generalized Likelihood Uncertainty Estimation (GLUE) and Markov Chain Monte Carlo (MCMC) techniques were used to calibrate APSIM‐Sugar influential variety parameters. GLUE and MCMC were evaluated based on a theoretical and real world calibration. APSIM‐Sugar was able to accurately reproduce the average biomass and sucrose yields of the four sugarcane varieties grown in the Home Hill trial when effects of weeds, lodging and stalk death were implemented in the simulation. However, APSIM‐Sugar had limited skill in simulating yield differences between varieties and varietal yield responses to water stress. Global sensitivity analysis identified how key APSIM‐Sugar input parameters affected model outputs. Parameters representing radiation use efficiency (rue), transpiration efficiency (transp_eff_cf), number of green leaves (green_leaf_no) and the leaf size profile (leaf_size) were found to strongly influence simulated biomass and sucrose yields. In a real world application, the MCMC calibration of Australian variety Q117 was better able to reproduce observed yields than the GLUE calibration and was able to estimate realistic parameter values for difficult to measure traits such as transpiration efficiency, using readily available field data.

Results from this thesis clearly show that updated variety definitions are needed for APSIM-Sugar. This thesis developed and tested a methodological framework which include performing a global sensitivity analysis and a Bayesian approach to calibrate variety parameters in APSIM-Sugar. The methodological framework provided a validated strategy for improving and updating variety definitions.

Several avenues for future research into the simulation of variety, environment and management interactions in sugarcane systems were identified in this thesis. The comparison of simulated and observed differences between sugarcane varieties highlighted a clear and pressing need for improved and updated variety definitions in current sugarcane models such as APSIM‐Sugar. Variety parameters in the APSIM‐Sugar module can now be routinely updated as new varieties are released using a limited amount of data which is collected in breeding programs and the methodological framework implemented in this thesis. Updating the model to include variety definitions for current commercial varieties will allow industry decision makers to have greater confidence in the model outputs.

Item ID:	41338
Item Type:	Thesis (Masters (Research))
Keywords:	Agricultural Production Systems sIMulator; applied statistics; APSIM; BACCO; Bayesian statistics; breeding; crop variation; cultivar; farm production quotas; Gaussian process; Home Hill; Queensland; sugar; sugarcane; TE; trait modelling; variation; water stress; yields
Related URLs:	http://researchonline.jcu.edu.au/33970/ http://researchonline.jcu.edu.au/33968/
Additional Information:	Publications arising from this thesis are available from the Related URLs field. The publications are: Chapter 2: Sexton, J., Inman-Bamber, N.G., Everingham, Y., Basnayake, J., Lakshmanan, P., and Jackson, P. (2014) Detailed trait characterization is needed for simulation of cultivar responses to water stress. In: Proceedings of the Australian Society of Sugar Cane Technologists (36), pp. 82-92. From: ASSCT 2014: 36th Annual Conference of the Australian Society of Sugar Cane Technologists, 28 April - 1 May 2014, Broadbeach, QLD, Australia. Chapter 3: Sexton, Justin, and Everingham, Yvette (2014) Global sensitivity analysis of key parameters in a process-based sugarcane growth model: a Bayesian approach. In: Proceedings of the 7th International Congress on Environmental Modelling and Software. From: IEMSs 2014: 7th International Congress on Environmental Modelling and Software, 15-19 June 2014, San Diego, CA, USA.
Date Deposited:	02 Dec 2015 04:37
FoR Codes:	07 AGRICULTURAL AND VETERINARY SCIENCES > 0701 Agriculture, Land and Farm Management > 070103 Agricultural Production Systems Simulation @ 50% 01 MATHEMATICAL SCIENCES > 0104 Statistics > 010401 Applied Statistics @ 50%
SEO Codes:	82 PLANT PRODUCTION AND PLANT PRIMARY PRODUCTS > 8203 Industrial Crops > 820304 Sugar @ 100%
Downloads:	Total: 590 Last 12 Months: 9
	More Statistics