Photo / File
Opinion: The reaction to the farm nutrient model Overseer is out of kilter, especially when compared to society's acceptance of Covid-19 modelling. Dr Jacqueline Rowarth explains.
All models are wrong, but some are useful.
This isn't a new statement. It was first coined by statistician George Box in the mid-1970s and statisticians and modellers have quoted him ever since.
It seems, however, that some people using those models have forgotten the problem.
Models contain constraints and assumptions.
Constraints are placed to restrict the model to a minimum and maximum range of limits considered logical or are as far as the known science or knowledge can explain.
Assumptions, whether implied or clearly stated, are never exactly true.
Covid-19 has shown us where models can be helpful. Media has been full of trajectories showing how the virus might spread, based on models under constant revision as more information becomes available.
The models have been accepted as an indication of trend. They project what might happen under different scenarios, such as the number of people at an event, or the wearing of masks.
At the same time economic models have indicated what might happen at the various levels of lockdown.
In all cases the success in actual prediction has been variable, but society accepts that we are in uncertain times.
Modellers have been invited back for their latest pronouncements, based on latest information, whether what they said in the past has come true or not.
Economists gave dire warnings about how bad things would be during and after the 2020 national lockdown.
As we now know, it didn't crash: incorrect assumptions had been made about the impact of Government support and the primary sector export earnings.
But nobody has suggested taking any of the economists to court because they got it wrong.
Given all this understanding by society in general, the reaction to the farm nutrient model Overseer is out of kilter.
Some industry players have paused its use while awaiting government decision.
Local government has put consents on hold for a time while trying to explain to government the importance and utility of the tool.
Some farmers have expressed their frustration… while others have stated that its use is vital.
Confusion reigns.
Overseer is a model. It was designed to help farmers make decisions about changes in management.
It was never designed to give an absolute figure that could be used in a regulatory framework.
Media articles proclaiming "not fit for purpose" have not always made this clear.
The purpose was to give users an indication of trend upon which decisions could be made that would have a likely outcome (such as reduced nitrogen loss). And Overseer needs ongoing investment and refinement, just like the Covid models, as more information becomes available.
The fundamental problem, just as with Covid, is that Overseer deals with biology as well as physics and chemistry.
Some people show no Covid symptoms, others have been prostrated. Some people are super-spreaders, others keep it to themselves. The models work on "likely".
Beyond Overseer is Essential Freshwater and attempts by regional government to meet standards.
Models are the obvious tool to assist with deciding what action needs to be taken, but there does need to be a direct link between cause and effect.
Dissolved Inorganic Nitrogen (DIN) has been proposed as an indicator of potential algae growth, but science has shown that the link is weak.
Stream flow, temperature, shade and waterway bottom (stony or soft) all affect whether algae will grow or not.
This means that regulation based on DIN is likely to have little effect on the presence of algae.
More research is required before regulation is imposed – the sort of research that involves data and real-life experience.
Most of society appears to trust the Covid modellers to know what they are talking about and incorporate new technologies and robust information whenever they can.
This is what farmers do. In addition, they test their possible actions using Overseer to examine what might happen – more or less grass production, more or less nitrogen loss.
Unlike some of the simple models attempting, and failing, to relate DIN to algal growth, Overseer is complex.
It has been developed over several decades for soil type, rainfall, temperature, stock type, shelter… working with industry and with farmers to ensure usability. To ensure, in fact, that it is fit for purpose.
Listen to Jamie Mackay interview Dr Jacqueline Rowarth on The Country below:
It assists farmers to adjust their operation according to fluctuations in weather, markets and development of new technologies. It enables innovation in a way that no other country has achieved.
The same as Covid management.
This time, because of the knowledge that was generated last year, the economy isn't being predicted to tank.
Economists know that the primary sector will keep the export revenue flowing and they've modified their models.
All models need regular investment in upgrading, checking the constraints and assumptions as knowledge develops and technologies change allowing greater refinement.
George Box also said that "Since all models are wrong the scientist must be alert to what is importantly wrong. It is inappropriate to be concerned about mice when there are tigers abroad."
New Zealanders should be very concerned about poor relationships being used to create regulations. The tiger is the economy.
• Dr Jacqueline Rowarth, Adjunct Professor Lincoln University, is a farmer-elected Director of DairyNZ and Ravensdown. The analysis and conclusions are her own. jsrowarth@gmail.com