‘The future’s uncertain and the end is always near’ sang Jim Morrison of The Doors – admittedly after he had started the day with a beer (in ‘Roadhouse Blues’) – but what do crystal ball gazers like economists, meteorologists and volcanologists rely on and should we expect them to be right more often?
Economist Shamubeel Eaqub happily quotes the joke that his profession has correctly predicted nine out of the previous five recessions.
His advice is, while accurate, it misses the point.
“Forecasting is a very, very tiny part of economics. In fact most serious economists don’t do forecasting.”
As a serious one himself, he can laugh now and throw in the odd jab.
“I think our view of economics is coloured by the d***heads on TV because in reality that’s not real economics. The economists that you see are essentially bank economists like I used to be.
“Most economists are actually looking to explain how the world works which is trying to look backwards.”
Talking head economists may be high visibility but Eaqub diminishes the value of their return to the public in a specific sense.
Forecasting is a really tiny part of economics, he says, and a tiny part of macroeconomics – the preserve of banks, central banks and treasury.
The purpose is not to get it exactly right: “It’s to ask ‘what’s the broad range of outcomes that are possible?’”
They are a starting point.
“Forecasts are useful for normal economic conditions but not good at predicting things that are unusual, recessions, pandemics, global financial crises.”
“Here are some big drivers that are affecting the economy and on balance this is what we think will happen.”
So we need to steel ourselves to the irony of forecaster as historical augur.
“When things are turning very quickly that’s when forecasters can be useful, to give you better context around how this might affect the economy unless we change behaviour.”
New Zealand’s response to Covid-19 was a case of forecast as cautionary tale, rather than prophecy to fulfil.
“When the pandemic first hit, there was a real risk that we would have this extreme recession, but we didn’t because we’ve had the biggest possible fiscal response ever seen in our lifetimes. The government expects to borrow something like $100 billion to support the economy…
“So the prediction was, in the absence of concerted massive action we would have a deep and terrible recession. As a result of those forecasts we saw unprecedented policy responses from both the government and the central bank.”
Predictions may be dressed up in a fancy ‘economics’ label, but underneath the models are wearing drab historical data.
But, says Eaqub, “history rhymes, it doesn’t repeat – so there’s a lot of judgement.”
And some of that will be informed by talking to people whether taxi driver, café owner, manufacturer or logistics operator.
“Quite often it’s those little anecdotes that give you an insight into what might be coming because some industries and some people will experience it before others.”
So don’t shoot the messenger for a view that’s too broad?
“The use of the forecast shouldn’t be ‘this is the only possible future’.” Rather the economist should be seen as a weaver of scenarios.
More than any other year, 2020 has shown that scenario planning is important, he says.
And what an economist divines from the micro-economy of a business or household plays into the bigger picture – and prediction.
“For example we might say that retail spending in the supermarket is going to go up but what’s driving that? Well, it’s people not going out to spend so much and people not only trading down a bit, because they do, but also getting more treats from the supermarket. You might be getting a slightly more expensive bottle of wine because you’re not going to the pub…
“When you see us forecasting things, that’s what you don’t see. We are talking at this big level whereas the lived reality of people is actually quite different and a lot more variable.”
Eaqub isn’t looking for pity – maybe understanding.
“Feel free to slam the economist. I mean we can’t get the direction or the magnitude right so when we are out there in front of you telling you what’s going to happen to the economy very confidently, understand that our confidence is completely misplaced.
“And I will put my hand up and say I mostly get it wrong when it comes to forecasting but that’s not the point of it, you have to understand the purpose…
“We are deserving of that mockery because we don’t explain what the forecasts are for, how to use them well and what the caveats are which all becomes too complicated and no-one can fit into a sound bite.
“I don’t know that the appetite’s not there, I think with Covid what we have seen is people have some capability to digest pretty complicated stuff and understand risk and distribution of risk. But I think you have to help people make that happen.”
At least it’s not looking like rain…
When it comes to the weather MetService communications meteorologist Lewis Ferris says the tools available to forecasters are leaps and bounds ahead of 10 or 20 years ago.
“But they do have their limitations. There are very small scale things that are going to be near impossible for models to be able to calculate and represent in an accurate manner. Even on day one let alone on day six or 10.”
As there is no way to input every piece of data that could have an impact, errors will build up as time progresses – the butterfly effect.
“The computations that go into forecasting the weather are done by some of the biggest supercomputers in the world and some traditional equations were derived from when people were doing these things by hand – when it took longer to do the computations than it took for the weather to occur.”
There’s a trade off to how much extra tech would help.
“We can’t have a weather station every kilometre. It would be great if we did but even then you wouldn’t be capturing absolutely everything… trees, cars, buildings all have an impact at a local scale. What we’re doing as meteorologists is looking at these broad scale models and then applying those to the local effects.”
Forecasters will check one simulation against another – the more in agreement, the more confidence they have, but the reality can still differ and a human element is crucial to picking outliers.
Prior to July’s intense thunderstorms in Northland, which dumped 220mm of rain in a day, Ferris says forecasters looked at what was happening in the morning and realised “the fresh models just had no idea.”
The event, an unusual interaction between two low-pressure zones, was dubbed a once-in-500-year storm.
Working beyond the computer model meant the service could issue thunderstorm and heavy rain warnings for the upper North Island.
“So that was quite a success story for us to be in a reactive space. It’s not like ‘oh the computer model has no idea what’s going on, what do we do now?’”
The severity of Napier’s November floods though was unpredictable.
In the days before, MetService had issued heavy rain warnings for the region, but convergent winds corralled and enhanced the rain into isolated regions “dumping huge amounts in a very short time.”
“That level of detail just cannot be modelled, it’s just such a small scale thing… Five kilometres away there was half the amount of rainfall over the 24 hours.”
So there are always going to be limits to humankind’s predictive ability.
“I’m not sure what instrumentation would have helped that. Some super mega high resolution modelling but that comes at an extreme cost.”
While anthropomorphic climate change is in the back of MetService’s thinking, its focus is from one to six days out.
“The climatic trend doesn’t really come into that exactly. It’s more when we’re seeing the severe end of the weather scale and research is being done into how much those have been altered by the changing climate.
“If we get a heavy rain event, what would that have looked like 50 or 100 years ago? Those things don’t inherently help us in the weather that is occurring now but it helps in telling a story and informing the public.
“You just can’t know what the future’s going to hold, that’s beyond the realms of science I feel.”
The shaky isles
Professor of Volcanology at Auckland University Shane Cronin says the country is not short of good equipment – seismometers, permanent GPS stations, access to satellite techniques – to help understand changes in land motion.
The biggest problem is a geographic one, he says – it’s hard to triangulate measurements.
“When you’re trying to use any type of geophysical sensors to understand something on the whole of New Zealand you’ve got lots of stations roughly northeast to southwest in a long line and then you’ve got all this bloody ocean east and west.
“Ideally if New Zealand was a big round blodge we would have a lot better idea of how things work.”
Offshore sensors, like ocean bottom seismometers are being rolled out but are expensive and not as easy to get real time data from.
The same tools used to predict potentially dangerous volcanic activity – which is mainly seismic – are the same sensors for earthquake detection.
As these seismic stations are used to calculate whether magma is rising up through the crust and how deep it is, more tech that could detect smaller scale events could help.
“But again it’s an expensive business drilling a hole – so it’s all about cost and benefit.”
Observing surface features are also in the toolkit.
For Mount Ruapehu: “if there have been small eruptions at the bottom of the crater lake then the colour changes to a steely grey.”
Elements in the water chemistry are also indicators.
“When you’ve got a rise in some of these elements you can actually detect magma getting closer to the surface or you’re getting an indication that magmatic gases are getting trapped below. Both of which are of concern.”
“With a place like Taupō, holey moley where do you start when a volcano has got a lake the size of Singapore… if you’re measuring a change in some thermal features from one end of the lake what does that mean for the whole volcano? It’s a very complex setting.”
The more we learn more about how volcanoes behave the more we realise how intricately volcanoes and earthquakes are interrelated, he says.
In the geological past there has sometimes been more earthquake, and less volcanic, activity in some places, but in others it appears large earthquakes could lead to more gas, triggering hydrothermal eruptions.
“So there is a real relationship, though we don’t always know what that relationship is.”
A classic example was the 1995 Ruapehu eruption. Seismic activity meant it wasn’t a complete surprise but in the last decade scientists have recognised that the main precursory activity happened earlier than previously thought.
And it wasn’t underneath Ruapehu, it was under Lake Moawhango on the other side of the Desert Rd.
“So we’re looking out for that now. Off-axis seismicity that could be related to a volcano but not actually directly underneath it.”
Fault lines do not necessarily descend vertically, Cronin says.
“It could be on a fault line that leads towards the volcano but on an angle of 45 degrees.
“We’re learning more and more about the way distance conceptually works. We are continuously surprised – it’s not as if we don’t know, or that they’re not theorised about, but there’s a bit of scientific debate about some of these processes because they may work in one setting but not in another.”
Does having a detailed picture make things any more predictable?
“I think it does… We’re learning about patterns, recognising patterns, how to apply that pattern recognition and machine learning to understanding seismic signals.
“Earthquakes are always going to be, let’s just say, near impossible to predict. We can forecast areas where we think there might be stress building and so on through looking at geodynamical models, stress models and recognising in the past there have been chains of earthquakes.
“But it’s not a hard and fast rule because you have got so much going on.
“In terms of volcanoes we have better potential to be able to forecast things. Primarily because magma has to make its way to the surface before you have a significant eruption.”
Magma rising through the crust breaks rock, causing earthquakes, it also releases a lot of gas – carbon dioxide and the especially recognisable sulphur dioxide.
More tech would help reduce risk, but Cronin says there’s a margin of unpredictability that we just can’t get below.
“We are in a very good state for monitoring systems, largely due to the big investments in GeoNet, you could always double that investment and make a measurable improvement to how much we know.”
But eventually the law of diminishing returns cuts in.
On probability – “that there hasn’t been an eruption for a gazillion years, so we’re due one” – Cronin says it is an important secondary piece of information. It tells us how much we should be worried.
“And this is why we start getting worried about Taranaki because it last erupted between 1780 and 1800 AD and since then it’s been one of the longest quiet periods on record.”
“The scary thing is that after a quiet period it usually kicks off into a really busy period where eruptions are only five to 10 years apart.
“So we’re thinking if Taranaki does erupt again we’ll be into multi-decade eruptions.”
Data is king, he says.
“I’d say we’d reached saturation point in that type of study though… Now it’s trying to figure out what’s causing those variations in activity.”
And yes the records show our volcanoes have different personalities – Tongariro the tectonic, Ruapehu the random, Taranaki with its pulses and Auckland, Taupō and Okataina which flare up and then stay more or less quiet.
“Once we figure out the different flavour of these volcanoes – it’s up to us to figure out why and also to figure out how would we know where we are in the cycle and are we staring down the barrel of a whole lot of eruptions, say from Taranaki, or just one.”