Analysis of parameter uncertainty in model simulations of irrigated and rainfed agroecosystems

Abstract Crop water production functions (quantifying crop yield as a function of irrigation rate) can help in the design of management systems that reduce the water footprint. We examined the role of parameter uncertainties in characterizing production functions using the DayCent agroecosystem model. A global sensitivity analysis was conducted to identify the model parameters associated with the greatest uncertainties in model responses. Under both irrigated and non-irrigated conditions, growth/production-related parameters had relatively more impact on grain yield than did soil-related parameters. Under non-irrigated conditions, there was greater sensitivity to evapotranspiration related parameters. We then used the DREAM method, a Markov Chain-Monte Carlo (MCMC) Bayesian approach, to determine the posterior distributions of the selected parameters. The DREAM method produced good estimates for the posterior distribution of the critical parameters. The utility of water production functions as predictive tools to guide water management decisions is greatly enhanced by incorporating rigorous estimates of uncertainty.