Oral Presentation NCGRT/IAH Australasian Groundwater Conference 2019

Application of uncertainty analysis to salinity modelling (303)

Greg Hoxley 1 , Brian Barnett 1 , Luk Peeters 2
  1. Jacobs, Melbourne, VIC, Australia
  2. CSIRO, Adelaide, SA, Australia

A series of three numerical groundwater models that cover the irrigated area in the Victorian Mallee were refined to improve the confidence in estimates of the salinity impact of irrigation in the region. Groundwater recharge from irrigation causes mounding on the regional aquifer(s) which, in turn, drives increased groundwater discharge to the Murray River. This region of Victoria has experienced orders of magnitude increase in irrigated area and consequential recharge over the past 15 years. Increased groundwater flow delivers increased salt to the Murray River which contributes to economic and environmental damage. The refined numerical models included a quantitative uncertainty analysis which defined parameter uncertainty of groundwater flow to the river for the first time in a model for salinity impact.

The approach adopted involved a preliminary qualitative uncertainty analysis that defined the key parameters of the numerical model and how they were defined and constrained.  This was followed by quantitative uncertainty analysis using deterministic modelling with linear uncertainty analysis. The predictive uncertainty is calculated from the sensitivities of each model outcome to each parameter, the pre-calibration uncertainty of the parameters, the observation uncertainty and the mismatch between observation and simulated values. Linear uncertainty analysis starts with calculating the Jacobian matrix, the matrix containing the sensitivities of each model outcome to each parameter. In this study, this is achieved by running the model at least twice for each parameter included in the uncertainty analysis.

The assessment found that the river conductance term was a key contributor to model uncertainty. Groundwater flux estimates were found to vary by up to +/- 15% with 95% confidence based on the acceptable calibration. This for the first time puts the degree of uncertainty in salt load estimates and hence on river salinity impacts in context and is an important indication of the way that uncertainty can be used in salinity assessment in the Murray Basin.