ALBERT

Description: It has been advocated that history matching numerical models to a diverse range of observation data types, particularly including environmental tracer concentrations and their interpretations and derivatives (e.g., mean age), constitutes an effective and appropriate means to improve model forecast reliability. This study presents two regional-scale modeling case studies that directly and rigorously assess the value of discrete tritium concentration observations and tritium-derived mean residence time (MRT) estimates in two decision-support contexts; “value” is measured herein as both the improvement (or otherwise) in the reliability of forecasts through uncertainty variance reduction and bias minimization as a result of assimilating tritium or tritium-derived MRT observations. The first case study (Heretaunga Plains, New Zealand) utilizes a suite of steady-state and transient flow models and an advection-only particle-tracking model to evaluate the worth of tritium-derived MRT estimates relative to hydraulic potential, spring discharge and river–aquifer exchange flux observations. The worth of MRT observations is quantified in terms of the change in the uncertainty surrounding ecologically sensitive spring discharge forecasts via first-order second-moment (FOSM) analyses. The second case study (Hauraki Plains, New Zealand) employs paired simple–complex transient flow and transport models to evaluate the potential for assimilation-induced bias in simulated surface-water nitrate discharge to an ecologically sensitive estuary system; formal data assimilation of tritium observations is undertaken using an iterative ensemble smoother. The results of these case studies indicate that, for the decision-relevant forecasts considered, tritium observations are of variable benefit and may induce damaging bias in forecasts; these biases are a result of an imperfect model's inability to properly and directly assimilate the rich information content of the tritium observations. The findings of this study challenge the advocacy of the increasing use of tracers, and of diverse data types more generally, whenever environmental model data assimilation is undertaken with imperfect models. This study also highlights the need for improved imperfect-model data assimilation strategies. While these strategies will likely require increased model complexity (including advanced discretization, processes and parameterization) to allow for appropriate assimilation of rich and diverse data types that operate across a range of spatial and temporal scales commensurate with a forecast of management interest, it is critical that increased model complexity does not preclude the application of formal data assimilation and uncertainty quantification techniques due to model instability and excessive run times.

Description: Effective decision making for resource management is often supported by combining predictive models with uncertainty analyses. This combination allows quantitative assessment of management strategy effectiveness and risk. Typically, history matching is undertaken to increase the reliability of model forecasts. However, the question of whether the potential benefit of history matching will be realized, or outweigh its cost, is seldom asked. History matching adds complexity to the modeling effort, as information from historical system observations must be appropriately blended with the prior characterization of the system. Consequently, the cost of history matching is often significant. When it is not implemented appropriately, history matching can corrupt model forecasts. Additionally, the available data may offer little decision-relevant information, particularly where data and forecasts are of different types, or represent very different stress regimes. In this paper, we present a decision support modeling workflow where early quantification of model uncertainty guides ongoing model design and deployment decisions. This includes providing justification for undertaking (or forgoing) history matching, so that unnecessary modeling costs can be avoided and model performance can be improved. The workflow is demonstrated using a regional-scale modeling case study in the Wairarapa Valley (New Zealand), where assessments of stream depletion and nitrate-nitrogen contamination risks are used to support water-use and land-use management decisions. The probability of management success/failure is assessed by comparing the proximity of model forecast probability distributions to ecologically motivated decision thresholds. This study highlights several important insights that can be gained by undertaking early uncertainty quantification, including: i) validation of the prior numerical characterization of the system, in terms of its consistency with historical observations; ii) validation of model design or indication of areas of model shortcomings; iii) evaluation of the relative proximity of management decision thresholds to forecast probability distributions, providing a justifiable basis for stopping modeling.

Description: This paper explores the adequacy of steady-state-only calibration as a precursor to use of a groundwater model for decision-support. First, it reviews metrics by which a decision-support model should be judged. On the basis of these metrics, it establishes the shortcomings that a decision-support model may incur through foregoing transient calibration. These are 1) failure to reduce the uncertainties of management-salient model predictions to the extent that available data allows, and 2) creation of unquantifiable bias in management-salient predictions. Two methodologies for quantification of these deficiencies are proposed. The first of these addresses uncertainty reduction. This is relatively easy to implement, as it requires only that sensitivities of pertinent model outputs to a model’s parameters be calculated. The second methodology addresses predictive bias. Implementation of this second methodology is more expensive as it requires repeated calibration of a steady state model against stochastic realizations of a transient model.These methods are demonstrated using a synthetic case which explores the viability of steady-state-only calibration of models deployed to examine the impacts of pumping on stream flows and groundwater levels. It is demonstrated that, for some predictions of management interest, steady-state-only calibration is more than sufficient for this kind of decision-support modelling.