Science and environmental policy: models without data
are empty, but data without models are blind

In your last Farm Institute Insights newsletter, you wrote a piece arguing that there is an ‘increasing reliance of policy-makers on modelling, rather than actual science, in making decisions in response to complex environmental challenges’, and that this is an undesirable state of affairs.

Modelling is not a replacement for science: it is a part of science. All scientific progress is built from the interplay between theory and data (as Albert Einstein knew well). In environmental science, models are where theory lives, and so they are a vital way of ensuring that scientific progress continues.

Mathematical modelling has become more prevalent over the last 10–15 years in policy advice about environmental issues. You’re right to point out the risks involved in this trend, but I believe that it is wrong to think that there is any viable alternative. The solution is not to abandon modelling, but to do our best to ensure that environmental models – and datasets – are built well and reinforce one another.

There is no alternative

As you pointed out in your article, governments must face increasingly complex issues: salinity, water management, coal seam gas, climate change. Models have to be used to understand these ‘wicked’ problems, for at least three reasons:

  • Interactions are always important. Take the example of coal seam gas, where the economic trade-offs are driven by a complicated mixture of hydrological, ecological and agronomic changes. Data from different disciplines need to be drawn together to understand the multiple dimensions of these problems, and models are the only viable approach.
  • Data are always insufficient. To pick one example of many: you criticised the Sustainable Rivers Audit for working with model-derived ‘reference conditions’ in the catchments of the Murray-Darling Basin. What, though, was the alternative? Any approach driven by the patchy available data would have been open to – justified – criticism for assessing the different catchments inconsistently.
  • Policy-makers always ask ‘what if?’ Datasets can tell us about the past and present; if we’re lucky, they can forecast ‘business as usual’ a short way into the future. Once questions are asked about alternative futures, though, modelling becomes essential.

Environmental modelling is a rapidly evolving field and the Australian Dryland Salinity Assessment was done in 2000, or about eight iterations of Moore’s Law ago.

Every scientist has a model

It is a commonplace of scientific practice that researchers base their work on ‘mental models’ of how the world functions. These mental models influence the questions that are asked, the data that are collected – or not collected – and the ways in which results are analysed. The choice for policy-makers, therefore, is not between ‘science’ and ‘models’, but between models that are written down explicitly and models that remain in the heads of the people using them.

I strongly disagree that ‘it is... more difficult to scrutinise the results of modelling’ than the results of ‘actual science.’ Exactly the opposite is true. A formal model lays out its assumptions and opens them to critique; the assumptions of an informal ‘mental model’ are seldom made so clear.

Policy processes don’t deal well with uncertainty

You argued that ‘models... are subject to modification or assumptions that may bias results in a particular direction’, but your own examples tell a rather different story. In both the Australian Dryland Salinity Assessment process and in the Sustainable Rivers Audit, the scientists who did the research took pains to point out the limits of their models and the resulting uncertainties. As Prof David Pannell has already pointed out, the process of simplification that happened in the ADSA as the scientific results were synthesised tells us more about the ways that policy advisers and politicians approach risk and uncertainty than it does about the use or misuse of modelling.

Quality assurance in environmental policy-making

I agree with you and David Pannell that the models used in environmental policy advice should be open to scrutiny: that is why the main Australian models for hydrology and agricultural production make their science publically available.

As the experience of the climate scientists shows, however, it can take years to establish public confidence in research results when the problem is a ‘wicked’ one. Trying to address water, salinity, coal seam gas, etc one-at-a-time prevents this confidence from being built. What is needed is a strategic approach in which:

  • data collection in the landscape is informed by – and challenges – models of the underlying physical and biological processes 
  • both the models and the data are open to scrutiny and to creative reuse by government, researchers and the wider public
  • uncertainties in the science are assessed carefully and communicated clearly by scientists, and faced squarely by policy-makers.

Building such a ‘knowledge system for sustainability’ will be a major undertaking that will challenge the ways that both landscape science and environmental policy-making are currently practised. It is, however, the best way to ensure that the transparent and iterative input of scientific knowledge into policy that you are calling for actually happens.

Regards, Dr Andrew Moore
CSIRO Sustainable Agriculture Flagship

Back to November 2012 Insights contents page. 

Back to newsletter archive.