Sensitivity analysis of three-dimensional steady-state and transient spray irrigation models

We conducted a sensitivity analysis of ground water flow on a spray irrigation site to evaluate key parameters affecting use of spray irrigation for sewage disposal. Spray irrigation is considered successful when water recharges the aquifer (low runoff) and mounding of the water table is below set Limits. It can be difficult to predict success and allowable irrigation rates given uncertainty in parameters. The U.S. Geological Survey (USGS) code MODFLOW was used to compare two- and three-dimensional steady-state simulations, and three-dimensional transient simulations. We compared Latin hypercube sampling and systematic adjustment of parameters as methods of sensitivity analysis, transient versus steady-state sensitivity, and variations in layer configurations. These models were not calibrated because they were created during planning stages before field data became available, The Latin hypercube sampling effectively tested a large number of parameters simultaneously, and can be used as a screening toot for systematic adjustment of parameters. Transient and steady-state simulations predicted similar sensitivity to hydraulic conductivity, but steady-state modeling produced higher peak heads and discharges. Again, we recommend that the steady-state simulations could be used as a screening tool to evaluate trends. Transient modeling is likely to provide more realistic predictions of heads and discharges. When different layer configurations were compared, we found that two-dimensional modeling overpredicted mounding and drain discharge because flow in a permeable soil laver is neglected. Two-dimensional modeling would not be effective for predicting allowable irrigation rates. These results provide modelers an indication of the effects of model simplification, which can be used to plan modeling strategies.