15 minutes about Predicting Extinction

[Music] this one right okay so hi so as a as a paleontologist i study life on this planet as it used to be because everything i study is already dead so according okay i'm sorry can you all see the screen okay because i'm a mathematician so i can either show formulas or pictures so seeing the pictures is really really important to my talk if you want to get the most of my talk please come close enough so you can actually see the screen okay and as a mathematician i should accordingly to hollywood spend most of my time staring at numbers solving impossible equations and probably also develop a serious mental health issue the thing is this is all horrible cliche and the reality is very very different from that and what i spent my time with is actually looking at the question can we predict extinction risk so can we determine the probability that any species on this planet will vanish forever in the near future and from my paleontology site i use the fact that well everything i look at is already extinct so there has been six major mass extinction on this planet and i can take the fossils from all these extinctions and use them as a basis for my study whereas from the mathematical part i can use all the framework that is usually used to predict the weather or the stock market and trying to get the prediction part of predicting extinction risk right and this is not only a purely academic question it is actually a very very interesting and important question for conservation because we're not doing this for fun we're doing this to identify species that are very vulnerable and will most likely go extinct in the near future so we can then protect them from going extinct and one of the major reasons to protect them is why they might go extinct is actually climate change so what you can hopefully see here is the predicted temperature change on earth until the year 2100 and if we continue as we do at the moment we will have a temperature change of 4 degrees on average until the year 2100 and that has really really strong and strong effect on all ecosystems and will put a lot of species at risk just because droughts are more severe temperatures are higher and a lot of the ecosystems will actually most likely collapse so there's different approaches to trying to assign extinction risk to a species and one approach is what i just called this has happened before approach and it's pretty straightforward the idea is really simple you just look at come on you just look at what is happening right now so what we are doing is putting a lot of co2 into the atmosphere which leads to a global warming and to ocean acidification and then as a paleontologist we go back in life's history and try to identify time intervals where this has happened before and there are actually really good analogues and these are time intervals where there was a lot of volcanism that also put a lot of co2 into the atmosphere and then we're looking at the species that went extinct during these intervals of a lot of volcanism and then we try to find the analogues of these species in the current situation and then conclude that they will have a elevated extinction risk another approach that i will now continue talking about in this talk is the so-called life cycle approach and i will focus on this just because this is my my research topic so this is what i do all the time and it is based on a very simple empirical pattern oh come on and that is that the abundance of a species so how common a species is is very low when the species first appears and then it slowly increases until it reaches a maximum at the middle of its life span and then it slowly decreases again until it's going extinct and what's really striking is that this pattern is really really symmetrical so you could say the way to success of a species is always the is also the way of downfall of a species and a common hypothesis is that this pattern actually reflects the life cycle of a species just like humans have a life cycle this species also has a life cycle starting with the teenage years close to the origination where the species is young and dynamic so it will spread out become very abundant then the adulthood where it will become very established and very abundant and then it will slowly approach retirement there will be less and less of that species until it's going extinct it's pretty much like humans and using this pattern to predict extinction risk is also pretty straightforward as an example here's a curve of a species that still lives so it's still around in the present and we trace how common it was in the past so we see it's had it's had its teenage years it had its adulthood and it's slowly approaching retirement and with the knowledge of that common pattern of the life cycle we can then predict okay it will most likely go extinct in the near future so it has a high extinction risk as a counterpart come on here the blue species is constantly increasing so it's no sign of adulthood still in its teenage years and if we make a projection into the future we assume it will still constantly increase so this species has a low extinction risk and this is a pretty straightforward approach you can do this for every species for which we have fossils and that are still around the thing is this is based on the hypothesis that this is actually a life cycle and when i looked at this i found a second hypothesis which we can use to explain this pattern this very very symmetrical pattern and that argues for that this is actually not a life cycle but it's rather an expression of purely noise and an observation bias that we introduce simply because well everything we look at is already dead of course it has to drop before it's going extinct right and i will run you through this for uh then the rest of my talk and the effect of noise is uh best explained by the brownian motion which was uh discovered by a very wealthy british guy that looked through his microscope and realized that the poland in the water he observed are not still but they're actually moving and if you if he traced his their movement they showed a very very distinct zig zaggy pattern so they moved away from the original position and came back and moved away again and he first thought this is actually well the pollen are obviously alive they're moving right but after a while people fought into this and looked at the polling more closely and it turns out this is not that the pollen are alive but it's that they're like randomly shuffled around by the water molecules just because at some point more water molecules are bumping into the pollen from one side it will move in the other and at some other point randomly more pollen will bump from the other side so it will just always shuffle around and this brownian motion is very common in all systems that are governed by a lot of background noise for example if you look at the news from the stock market you will always find curves that are very very similar to this and there's people arguing that the stock market is on short time intervals actually behaving like a poland under a microscope but how does this actually connect to our extinction and just assume there's a very simple example assume there's a group of individuals of one species that is living somewhere on this planet and it's relatively stable so it's moving it's not changing its abundance and then at some point it's hitting by a catastrophe and you have this sudden drop and then slowly get get a recovery and then we add another group of individuals from the species that is also hit by a catastrophe but just at some other point because they're living somewhere else and when we combine these two groups just because we don't have the resolution to we just want to combine them to get more information something very interesting happens and that is the catastrophe does not look that catastrophic at all and they look a lot more smoother and a lot more regular than the original than the original curves and you can continue this for example if i average 10 groups from all around the globe that are independent of each other it looks a lot more smoother and if i add more you will always end up with a brownian motion so in any system that develops through time and you add a lot of noise to it you will always end up with a brownian motion so in this case there's a strong argument for the abundance of a species to actually follow a brownian motion and now comes the second part which is the observation bias so we know this is a brownian motion so this is suppose this this is how the abundance of a species is supposed to look like if there's a lot of random noise from all over the globe contributing to it we know two very important pieces of information and the first is well at some point there is an origination and at some point there will also be an extinction so it will originate so this is where we set up the curve to zero and it will also go extinct at the moment the abundance drops to zero and so now we have a curve that looks surprisingly symmetrical and we can retrospectively assign it teenage years adulthood and retirement so we we tend to see these life cycles even in random patterns just because well it starts at zero and ends at zero there's not that many different patterns then an increase and then a drop to zero again and the cool thing is this is actually pretty independent where i choose the the origination so i can place it up here and it's always symmetrical or i can place it up here and it's still symmetrical it will always look symmetrical just by introducing the additional information of the extinction and this is kind of like a bummer as a scientist because we have this very beautiful empirical pattern um this is the data there's nothing we can do about the data the data is like unchangeable it's like pristine and there's nothing we can do about we have two very very different hypotheses that explain this pattern one is well this is a life cycle and there's this teenage year adulthood and retirement and this is borrowed from a metaphor that species are in a way like humans on the other hand we have this other hypothesis that is well it we're just seeing background noise and like retrospectively attributing these life cycles into the data that are not really there and so what do we do in science there's a very fundamental principle in science that's called occam's razor and occam's razor pretty much says if you have two hypotheses then the simpler one is most likely the better one and here is simple explicitly means not easier to understand but thus it requires less assumptions so now the question is which one of those hypothesis is simpler okay okay everybody who thinks that the life cycle hypothesis is simpler please raise their hands okay everybody who thinks that the noise hypothesis is simpler raise their hands this is perfect because actually this is not a rhetorical question like we simply don't know so this is this is my research it's an ongoing debate there's just two weeks ago a new paper has been published arguing for the life cycle hypothesis but i can't tell maybe in five years and this is actually for me the most exciting part about science it's not finding the answer the moment you have the answer it's kind of boring because you know it it's actually that you actually get to work on a really hard problem thank you nicholas thank you for your talk i really hope we don't get extinct otherwise it might get rough so who of the audience has any questions nicholas please raise your hands the lady on the first row sorry for that by the way are there any contemporary species that you consider with that life cycle thing is it useful or is it just useful for things where you can already see the whole bow the thing is so this is um this is commonly used but it's always arguable whether it's like retrospectively assigned or whether it's a real life cycle so nobody really knows because you will always find like a maximum and then assign okay this is adulthood but it's the the problem with this metaphor about humans is a bit that for a human you see whether they're adult nobody has an idea what it looks like that the species is adult so that that's the main difference and so we cannot like go to a species and say this is in its teenage years it doesn't work like that there is one more question from the guy over there where are human beings as a specie on that curve are we have we reached the top or are we continuing to continue to destroy the planet very good question i think we're probably at all of these stages because as a species we're extremely young from a paleontological perspective so for example the data i'm working here with the temporal resolution is 10 million years so that's like the finest i can achieve um and if you're trying to apply these methods to human they just don't work they don't have the resolution but i would personally i think we're pretty close to extinction that's some bad news any more positive questions there is one over there yeah hi um i was wondering like when you talk about the species here um to what extent um is like evolution a part of that like um the very same thing might change enough that you might recognize it as a different thing and count it differently maybe or um some other animal is taking over the same role basically and then it doesn't matter if there's an extinction because the same thing keeps getting done or that's that's a very good question the the thing is that it's um so this is what you would call macro evolution so it takes usually evolution takes one individual and looks at its success what we're doing is we look at one species and look at its success the question you're asking is a ecological question that is most of the time at a much lower temporal resolution and for the species we also these are not like strict species in what like biologists use as species these are morpho species so they're only distinct by their morphology and most of the time we don't have a species resolutions on but we do it on a higher taxonomic taxonomic level okay we'll have time for one more question any willing ones there is one race hand over there do you see taxonomic consistency when you come and predict the life cycles so perhaps at i don't know um what's above gmos family levels um so everything is symmetric at all taxonomic levels but it gets worse with the higher levels because the resolution gets so the samples get smaller that's the problem what's probably more interesting is that the pattern does not persist through mass extinctions so the broad if the moment you have a very big impulse a very big destruction the symmetry collapses which is also an argument for the the random noise uh argument i think to translate answer species go extinct by chance yeah great thank you very much for this talk if you have any further questions to him you'll probably find him at the ice cream shop after the talk so thank you very much [Music] you