The distinction between those making policy and consent decisions and those making environmental models to support these decisions seems obvious. Yet what is often neither obvious nor prioritised, is how a modeller can quantify a model’s limitations and uncertainty and present this in a manner that decision makers can readily access to make truly informed decisions.
To take steps to address this lack, a 1.5 day workshop was held targeted at managers and decision-makers who need to understand the reasons why predictions made by environmental models are often accompanied by wide margins of uncertainty. The workshop was attended by 30 water resource scientists and managers from regional authorities, universities, research organisations and consultancies. The workshop on Uncertainty in Water Resource Management was held at the Commodore Hotel in Christchurch, April 7-8, 2016 and was jointly delivered by research programmes: Ground Water Assimilative Capacity (GWAC); Smart Aquifer Characterisation (SAC); Smart Aquifer Models (SAM); and Tracer Validation of Hydrological Systems (TVH).
Workshop presentations were given by Chris Daughney (GNS Science), John Doherty (Watermark Numerical Computing), Linda Lilburn (Landcare Research), Graham McBride (NIWA), Stephen McNeill (Landcare Research), Catherine Moore (GNS Science/ESR), David Scott (ESR), Shailesh Kumar Singh (NIWA) and Simon Woodward (Lincoln Agritech). A panel discussion chaired by Murray Close (ESR) was held at the end of the workshop.
It’s crucial to acknowledge that all models are simplifications of reality and as such there is no such thing as a model that can simulate reality, only those that can history-match against recorded data, i.e., a model can never be right: the best model that can be created is one that it is minimally wrong. This requires the uncertainty of a model to be clearly stated. Model uncertainty stems from a number of sources: measurement uncertainty, interpretation of data, weighting applied to data, the relationshipbetween data availability and real world complexity, uncertainty in prior information/expert knowledge, and model conceptualisation/structure (e.g. discretisation and interpolation method).
A significant number of talks focused on research around model simplification questions, for example: how simple can we go and still have a beneficial model and how can we check that such a model is suitable; what are the implications of simplification and what methods can be used to provide simplified models that don’t introduce bias.
Authors: Zara Rawlinson (GNS Science), Catherine Moore (GNS Science/ESR)
About the SAM Programme:
The Smart Aquifer Models (SAM) programme is investigating methods of significantly simplifying integrated groundwater-surface water flow and transport models in a manner that allows for shorter model run times (consistent with time-frames required for risk based decision optimisation and policy making), without resulting in unacceptable loss of prediction reliability, to better support the limit setting decision making process. Included in this work are the development of guidelines for the selection of appropriate simplified modelling methods and for determining the most cost effective data in the limit setting decision making context. Integration with economics and ecological models ensures the full causal pathways to impacts are considered when assessing the performance of simplified models.
About the TVH Programme:
The Tracer Validation of Hydrological Systems (TVH) programme is investigating methods for merging two existing, widely applied methods for characterisation of groundwater systems: 1) hydrological tracers, which provide vital information about water source and age that cannot be directly obtained by any other means; and 2) numerical groundwater models, which can be used to predict the response of the groundwater system to pressures such as increased water use or change in climate or land use. In the past, hydrological tracer methods and numerical groundwater models have usually been used separately, which leaves resource managers with disparate, ambiguous and uncertain information. However, these two methods are highly complementary, and so this programme aims to develop “tracer-validated numerical models” and “model-validated tracer interpretations” that will overcome weaknesses that exist when these two methods are used in isolation.
About the GWAC Programme:
The Ground Water Assimilative Capacity (GWAC) programme is developing prototype methods for incorporating uncertainty in the decision making process (via a decision making framework), including up-scaling techniques that allow for accurate descriptions of groundwater quality at the management unit scale.