Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: WATER QUALITY IMPROVEMENT FROM MANAGEMENT PRACTICES IN AGRICULTURAL WATERSHEDS Title: Inverse modeling with RZWQM2 to predict water quality

Authors
item Nolan, B -
item Malone, Robert
item Ma, Liwang
item Green, C -
item Fienen, M -
item Jaynes, Dan

Submitted to: Agronomy Society of America, Crop Science Society of America, Soil Science Society of America Meeting
Publication Type: Book / Chapter
Publication Acceptance Date: April 14, 2011
Publication Date: September 1, 2011
Citation: Nolan, B.T., Malone, R.W., Ma, L., Green, C.T., Fienen, M.N., Jaynes, D.B. 2011. Inverse modeling with RZWQM2 to predict water quality. In: Ahuja, L.R., Ma, L., editors. Methods of Introducing System Models into Agricultural Research. Madison, WI: Agronomy Society of America, Crop Science Society of America, Soil Science Society of America Meeting. p. 327-363.

Interpretive Summary: Agricultural systems models such as Root Zone Water Quality Model (RZWQM2) are complex and have numerous parameters that are unknown and difficult to estimate. Inverse modeling provides an objective means of automatically calibrating such models based on statistical criteria. In this chapter we describe application of parameter estimation software (PEST) to RZWQM2 models for prediction of water and N fluxes at two sites: an almond orchard in the San Joaquin Valley, California (CA), and the Walnut Creek watershed in central Iowa (IA) in a corn-soybean rotation. Inverse modeling provided reasonable fits to observed water and N fluxes and directly benefitted the modeling through: (1) simultaneous adjustment of multiple parameters, versus one-at-a-time adjustment in manual approaches; (2) clear indication of when model calibration is complete; (3) straightforward detection of parameters that are highly interdependent or insensitive to the modeling, conditions which can cause model instability; and (4) generation of confidence intervals for parameters and model predictions. Sensitivity statistics indicated that most of the RZWQM2 parameters at CA and IA could be reliably estimated from the available data. Prediction confidence intervals at CA were compared with independently measured annual water flux (groundwater recharge) and median nitrate concentration in an adjacent monitoring well as part of model evaluation. Both the observed recharge (42.3 cm/yr) and nitrate concentration (24.3 mg/L) were within the 90% confidence intervals estimated as part of the PEST optimized predictions, indicating that overall model error was within acceptable limits. This research will help model developers, model users, and agricultural scientists more efficiently, accurately, and objectively calibrate agricultural system models.

Technical Abstract: Agricultural systems models such as RZWQM2 are complex and have numerous parameters that are unknown and difficult to estimate. Inverse modeling provides an objective statistical basis for calibration that involves simultaneous adjustment of model parameters and yields parameter confidence intervals and sensitivities. This chapter presents guidelines for autocalibration of RZWQM2 by inverse modeling using PEST parameter estimation software. We describe operation of PEST in both parameter estimation and predictive analysis modes. The goal of parameter estimation is to identify a unique set of parameters that minimizes a weighted least squares objective function, and the goal of predictive analysis is to construct a nonlinear confidence interval for a prediction of interest by finding a set of parameters that maximizes or minimizes the prediction while maintaining the model in a calibrated state. Two sites with diverse climate and management were considered for simulation of N losses by leaching and in drain flow: an almond orchard in the San Joaquin Valley, California (CA), and the Walnut Creek watershed in central Iowa (IA), which is predominantly in corn-soybean rotation. Inverse modeling provided reasonable fits to observed water and N fluxes and directly benefitted the modeling through: (1) simultaneous adjustment of multiple parameters, versus one-at-a-time adjustment in manual approaches; (2) clear indication by convergence criteria of when model calibration is complete; (3) straightforward detection of nonunique and insensitive parameters, which can affect the stability of PEST and RZWQM2; (4) generation of confidence intervals for uncertainty analysis of parameters and model predictions. Composite scaled sensitivities, which reflect the total information provided by the observations for a parameter, indicated that most of the RZWQM2 parameters at CA and IA could be reliably estimated by regression. Correlations obtained in the CA case indicated that all model parameters could be uniquely estimated by inverse modeling. Parameter confidence intervals (CIs) at CA indicated that parameters were reliably estimated with the possible exception of an organic pool transfer coefficient (R45), which had a comparatively wide CI. However, the 90% confidence interval for R45 (0.03–0.35) is mostly within the range of values reported for this parameter in the literature. Predictive analysis at CA generated prediction confidence intervals that were compared with independently measured annual water flux (groundwater recharge) and median nitrate concentration in a co-located monitoring well. Both the observed recharge (42.3 cm/yr) and nitrate concentration (24.3 mg/L) were within their respective 90% confidence intervals, indicating that overall model error was within acceptable limits.

Last Modified: 10/22/2014
Footer Content Back to Top of Page