Scientists will simulate a robust replica – what they call a ‘digital twin’ – of the Earth.
This model will predict, among other things, the effects of natural phenomenon on humans and visa versa. What’s more, it’s meant to guide decision-makers in their climate change mitigation efforts.
We talked to Dr. Peter Bauer, co-initiator of Destination Earth and deputy director of research at the European Center for Medium-Range Weather Forecasts (ECMWF) about the project. We spoke about the model’s workings and accuracy, the climate movement, and the mechanisms behind weather and human behavior prediction.
Can you briefly describe how weather modeling works, specifically in terms of a digital twin?
At the core of a digital twin is a physics-based model. Compared to what we have for weather today, it’s like going from a Toyota Prius to a Formula One car. That’s the gap we want to jump.
Weather models are based on the physics of the system that we understand. And ‘the system’ means atmosphere (including everything that happens in the atmosphere), sea ice, ocean, land, surface, vegetation, hydrology of the land surface, and these things — everything that matters on weather timescales.
Climate timescales become slightly more complicated because you have things like atmospheric chemistry included; like CO2 and methane, a closed carbon cycle, and the nitrogen cycle.
So, for weather, we at the ECMWF try to replicate the physics that happen out there with equations we know in a simulation model. It’s basically like virtual reality. If you play a virtual reality computer game, you walk through the Earth's system. And if you visualize what we do in our weather model, it's just like that: with realistic clouds, with realistic land surfaces, and realistic wind. You can visualize that in four dimensions, just like a computer game.
The better you know your processes, the better your spatial resolution is, the more processes you have, and the more realistic it looks. It’s like moving from a poor rendering of a computer game 20 years ago to something like today on a PlayStation 5.
The more accurate we are, the more realistic and reliable our predictions of the future will be.
This new model will include influences humans have on water, food, energy management, etc. Can you explain some of the challenges inherent in modeling human behavior at such a level?
In a very nonlinear, chaotic system, there are still some simple things.
I sometimes give this example of human intervention that you can test in a system like this. Take New Orleans: It's below the sea surface, and it’s very exposed to tropical storms and storm surges. So, if you look at climate predictions 30 years ahead or 50 years ahead, should they heighten their dikes or not?
This kind of thing is something that you would like to look at with the digital twin. You would like to play through it and say, “If I take my much better predictions of climate, sea level height, and extremes like storm surges in the future, should Louisiana or New Orleans do something like this?”
Other things are really hard to integrate.
If you take the connection between extremes, impact on food production, and potential social unrest…If you have food limitations and water resource limitations in places, outside immigration pressure could cause that. It's chains — logical chains — that are part of the Earth’s system that are, of course, very hard to predict. But fundamentally you want to have that capability, because it is connected.
And at least you want to have the ability to run through scenarios, and say, “Look. Europe needs to invest now in other countries outside Europe to enhance its sustainability to climate change, to kind of avoid some of these effects that we certainly expect in the future.” Only then can you have a less reactive and more proactive dealing with environmental change that is certainly going to happen.
Why would such a tool be useful, especially in terms of climate change? What are specific use cases for the digital twin that may come up?
We have quite good ideas of what climate change means. We have a very good observation record. We have pretty decent climate models. We know what's going on up there. It's measurable, it's observable, and it's never been like this before.
We understand from basic physics where [this change is] coming from. A great deal of it is man-made or human-made; no question about that. The community agrees to 99 percent on that.
But there is still a great lack of skill in terms of climate prediction when it comes to regional detail, or when it comes to things that are more complicated than temperature.
If you take the leading ten climate models, they agree quite well about where temperature's going to change and by how much. If you then look at what they conclude on how general circulation changes as an effect of the weather regimes, for example, they disagree actually a lot.
And if you talk decision-making, this is the level of [information] you need. In Germany versus France, or the UK, or the Mediterranean, in the East Coast versus West Coast [of the US]; you need to know with some reliability what the regional effects of climate change will be. And those are hard nuts to crack.
That’s where the digital twins come in.
It’s really a step change in decision-making because right now, at this level, the models do not agree sufficiently well. It's really hard for a policymaker to say, “Do this or that.” So, you need to really upgrade — and not just by little, by a big step — your skill in order to support that.
We expect a big deal coming out of this in terms of decision-making.
The press release states that the goal here is “to map climate development in extreme events as accurately as possible.” How accurately are we talking?
Well, that depends on who you ask. We will do our best to produce good, reliable estimates.
[But the Earth] is a fundamentally chaotic system.
So sometimes,one prediction is simply easier than in others. If you have stable winter conditions with a high-pressure system, nothing changes much and your prediction for next week is good. It's easy. If you have summer conditions and you have frontal systems coming through, storms building up, tornadoes, and large instabilities; predictive skill is not very good no matter how good your model is.
"Only through science can we adapt to change and mitigate the consequences in the best possible way" ~ Dr. Bauer.
So, how good is good enough? That depends on whom you ask.
If you do predictions for ocean currents, using present day models to predict for a month ahead or a season ahead is probably good enough. [It’s] certainly good enough in terms of the Earth system representation of these currents and how ocean and atmosphere interact with the energy exchanges and all that.
But is it good enough for a shipping company to decide whether they route their ships this way or that way? Probably not.
And this is why this close integration of the application sectors is so important. We have to make design decisions for the digital twins based on that.
Is it too late to start making policy decisions about climate change based on science, such as the tool that you're developing here?
I don't think so. We're on the clock — that's clear. Increasingly we've seen that; the last eight to 10 years have been the warmest on record.
And there's an acceleration gradient; it's not just a constant increase. There's an acceleration factor in there. And if you have an acceleration factor of something in a very chaotic system, you risk reaching points where irreversible things are being kicked off, and some of them may have happened already.
But that doesn't mean you resign. There is enormous potential at the regional level. We need to understand what these implications are with more detail. Even though we know potential catastrophes are going to happen, you need to quantify it, because we have measures we can take — we can adapt.
Only through science can we adapt to change and mitigate the consequences in the best possible way.
Maybe one more thing I usually mention is — because Destination Earth is a European effort and we talked about American infrastructures and these things — regardless of who's paying right now for Destination Earth, it doesn't matter. In the end, it's a global problem and requires global collaboration.
And ideally, you would have matching efforts in Asia, in the US, in all kinds of countries. And if we bring them together, then the overarching result [will be] better.