Uncertainty Analysis of CROPGRO-Cotton Model
Abstract
An application of crop simulation models have become an inherent part of research and decision making process. As many decision making processes solely rely on the results obtained from simulation models, consideration of model uncertainties along with model accuracy in decision making processes have also become increasingly important. Newly developed crop model, CROPGRO - Cotton model is complex simulation model that has been heavily parameterized. The values of those parameters were obtained from literature which also carries uncertainties. True uncertainty associated with important model parameters were not known. The objective of this study was to estimate uncertainties associated with model parameters and associated uncertainties in model outputs. The uncertainty assessment was carried out using widely accepted Geenralized Likelihood Uncertainty Estimation (GLUE technique. Dataset on this analysis was collected from four different experiments at three geographic locations. Primary results show that the amount of uncertainties in model input parameters were narrowed down significantly from the priori knowledge of selected parameters. The expected means of parameters obtained from their posterior distributions were not considerably different from their prior means and default values in the model. However, importantly the coefficient of variation of those parameters were reduced considerably. Maximum likelihood estimates of selected parameter improved the model performance. The fitting of the model to measured LAI, and biomass components was reasonably well with R-squared values for total above ground biomass for all four sites ranging between 0.86 and 0.98. Approximate reduction of uncertainties in input parameters ranged between 25%-85% and corresponding model output uncertainties reductions ranged between 62%-76%. Most of the measurements were covered within the 95% confidence interval estimated from 2.5% and 97.5% quantiles of cumulative distributions of model outputs generated from posterior distribution of model parameters. The study demonstrated an efficient prediction of uncertainties in model input and outputs using a widely accepted GLUE methodology.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2009
- Bibcode:
- 2009AGUFMGC41B0775P
- Keywords:
-
- 1615 GLOBAL CHANGE / Biogeochemical cycles;
- processes;
- and modeling;
- 1847 HYDROLOGY / Modeling;
- 1990 INFORMATICS / Uncertainty