Modelling the Impossible? 🤯🧠🦊🐮🦆🐛🐍🦀🐳🌴🌾🌺🌍☀️🌧🍎🥦🍞🚗✈️🚀🏖🏢💻💰🦠🧬🚰🌡🗑📆📚❤️☮️♻️

Well that was quite a fun title to come up with... 🤪 There can't be many topics to cover for which all of those emojis (and indeed many more) are relevant. A few weeks I did a workshop that inspired this particular blog post (written the same evening but idiot here forgot to actually publish the thing!), and it all started with this particular TED talk: https://www.ted.com/talks/tom_wujec_got_a_wicked_problem_first_tell_me_how_you_make_toast.

Look at the title "Got a Wicked Problem, First tell me how you make toast" and I know what you're thinking -- what on Earth does modelling have to do with toast? Well I highly recommend you watch the talk because its a fascinating insight into the workings of the human mind and how we collaborate and develop ideas. Essentially, by drawing out models as a series of events and the links between them that make a sequence, we can develop causal models for use in statistical and practical modelling of how systems work. This then helps us to predict how a system will progress over time, and the impacts of our actions on the system. For example, say we need to decide whether, for the sake of increasing food production, we introduce a new pesticide which is currently banned due to health concerns. We need to look at the impacts of that. Where to start? The obvious one would be health, given the current reason for its lack of use. OK -- farmers might be affected by the continuous exposure to any volatile compounds within the pesticide, and the consumer would be exposed to traces of it in their food. This could lead to an increase in hospital admissions and a shorter life expectancy. One very depressing chain of events. On the other hand, it would reduce the pest impact on the crops, so increase the yield, reducing the price of food to the consumer, so overall people could receive better nourishment. This could have knock on effects of improved health and individual productivity, leading to increased national profit and a boost to the economy. Brilliant, maybe it is a good decision then. Flip that on its head and it could mean that a boost in local cheap food production makes it more difficult for other countries to produce competitively-priced food, increasing the economic divide between countries. Further, other countries may change where they import a product from, because now the food is cheaper where we use the pesticide. This then leads to an increase in food miles and packaging for transport, increasing landfill waste and accelerating climate change. OK, not so good... Maybe we should think about it from a different angle? Reducing the number of insects means reducing the number of pollinators for native flora. This leads to a reduction in the diversity and population size of producers, so reduces the availability of natural fodder for herbivorous wildlife and reduces local biodiversity. This in turn removes the food source for carnivores, which either starve as well or move into urban areas to feed on litter and waste. This can lead to increased human-wildlife conflict when bins get broken into, and potentially raise the risk of zoonosis or disease spread when wildlife and humans/pets come into close contact. Oh dear - this isn't sounding good. It gets worse: the presence of the pesticide in the soil can leach into local water ways, poisoning fish and freshwater invertebrates, creating dead zones and causing the pesticide to be spread far beyond where it was originally applied. Similarly the wind can blow them far away from their site of application, causing them to poison the environment over a much greater area than initially intended. But then again on the plus side, what if this pesticide only affects a very specific species of invasive pest which is also harming native wildlife and disrupting the food chain? Then maybe actually the spread of it could lead to an improvement in local biodiversity but removing the invasive competitor and creating space for native fauna. Those native fauna then overall improve the air and soil quality of the local area, making it a healthier and more pleasant space to live in, so human wellbeing improves. With an improvement in wellbeing, we see better health, contradicting the original reason for the ban. Maybe it doesn't have to be so bad after all.

Now... why am I talking about all this? This is an entirely hypothetical scenario, and how did I go from the best way to make toast to the many impacts of pesticides? Think again about the reason why we want to know about the effects that the new pesticide may have: we want to be able to predict what will happen if we lift the ban on it. How do we do that? We make a causal model. We actually physically draw out and visualise all of the events and scenarios that we can possibly think of for how the pesticide might affect people and the environment, and all of the knock on effects and chains that this induces. Suddenly we have a crazily large and complex model with arrows going in every direction, human health being affected from every angle, biodiversity being likely reduced but potentially improved, and one very confused researcher desperately trying to work out how to code a computer programme to bring the model to life and actually produce any meaningful predictions at all. Swiftly, the models become so complex that they almost lose their meaning as everything starts to become circular. Believe it or not, this is actually a relatively simple scenario compared to many of the systems that biologists would like to be able to measure, model and predict outcomes. Not just biologists, but every scientific discipline. Consider climate change: an incredibly well-researched subject with thousands of man-hours going into building models and collecting data every day and yet the projection models can at best give us a set of probabilities for different scenarios. Nothing is certain, because there are so many variables in the models that can all shift at the finest change in emissions or natural events. We can be fairly confident that in general the global temperature is going to increase, and there are many consequences to that, but even then, there are some scenarios in which some areas get colder... or Yellowstone goes off and we get plunged into a nuclear winter... again - a cheerful thought for the day. Sorry.

Not often I talk about bugs or have reason to put insect photos into these posts, so I'll take the opportunity on the day I rant about pesticides!

What it all boils down to is this: all models are wrong, but some are useful. This is an incredibly famous quote by statistician George Box. We can model and model and model, but models cannot be perfect and therefore they will always produce results which are to some level incorrect. But they can still be useful in some ways. I said today that I was inspired by a workshop. In this workshop we were asked to draw a causal model for what impacts biodiversity. Everything! In some way or another, everything that every person does on this planet, impacts biodiversity. Some for the better, some for the worse, some with no measurable effect but still having an impact. This is an example of an impossible model. We can't ever build a model that incorporates every single thing that impacts biodiversity, and everything within it would continuously be affecting something else and creating infinite loops of causality. We can however model parts of it. We can model the impact of increasing ocean acidity on coral reefs, caused by rising atmospheric carbon dioxide. We can from that predict how shark populations on those reefs may be affected, by following through the food chain from the fish that depend on the reef. This doesn't mean we have to think about Australian reef sharks when modelling the impact of a UK land-based pesticide, but it makes for a great exercise in challenging the way we think about science, biology, and computational modelling.

Apologies if that one was a bit heavy for a Monday night, and indeed if it made no sense at all and was just a series of random sentences apparently not following on from one another in the slightest... maybe I should draw it out next time!

I think it's time for some toast, don't you?