CSDMS meeting 2026 by Maya Stokes

All right. Um, hello and um, I'm really pleased to be here to share some of my work um, and uh, this may be a little bit of a different flavor. I'm going to talk a little bit about um, biology and um, the approach that I've taken for integrating biological and uh, earth surface process models. So, um, I'll start with this map of uh, species diversity across earth's continents and there's a few important patterns to note. Uh, there's more species located near the equator um, and there's also uh, of interest to us as geomorphologists a lot of species distributed in mountain ranges and I've highlighted some of these places that kind of pop out on this map here. This shows that the residual of a correlation between um, uh, climate variables and species richness um, again across the world and again you can see that some of these mountain ranges like the Andes, the Himalaya, Central African Rift Valley are some notable mountain ranges that kind of pop out as red. So, climate local climate alone cannot explain the high species richness of mountain ranges. Um, and this suggests that there there might be a link between geologic processes, mountain building, topographic evolution and um, the diversification of life which is exciting. So, how would this work? Uh, so today and in my work I'm um, generally focusing on uh, what's called allopatric speciation um, and so a phase of allopatric speciation is thought to accompany or initiate at least most speciation events. Um, so let's say we have some bunnies on a landscape and um, uh, they can migrate to cross this flat landscape. Uh, a mountain range grows and they find it more challenging to migrate across this landscape. They accumulate genetic differences and um then we can resolve the relative um uh ordering of how lineages may become separated from each other and perhaps we may be able to actually infer information about absolute time as well. Um although we may have some more work to do before we get there. Uh so this would suggest that, you know, okay, in places with more topographic change, we should find higher species diversity. So this is a map of mammal diversity across North America and um there's generally more mammal species in the western um half of the uh continent where you have more topographic complexity, but for freshwater fishes, uh this is not true. You have your highest species diversity on the eastern or tectonically inactive margin. And this is true not just for fishes, but also crayfish, salamanders, mussels. The southeastern US is an exceptional biodiversity hotspot when compared to other temperate regions across the world. I wanted to understand this on a global scale, so I categorized the world's mountains as tectonically active or inactive. And if you don't like my categorizations, I'm open to feedback. Um but um then I uh counted the number of species per unit area in these different uh geologic settings. And I'm showing a histogram of these results for mammals on the left and freshwater fishes on the right. And so the point is that there uh mammals there's more species per unit area for mammals in tectonically active mountain ranges as compared to tectonically inactive mountain ranges, but that difference collapses for freshwater fishes. Um I also repeated this analysis for squamates, so these are uh include reptiles like lizards, snakes, turtles, crocodiles, etc. And um for terrestrial taxa, uh they are more diverse in tectonically active mountain ranges, but this is not true for the um freshwater taxa. So, are tectonically active mountain ranges more biodiverse? Perhaps yes for terrestrial taxa, but no for these freshwater taxa. And this motivates my work. Um when we think about uh you know, I drew a polygon for this landscape where the bunnies lived, but if we think about the biogeographic arenas for freshwater fishes, these are uh river networks. Um and we know that river networks can change their connections to each other over time, and that this can split and merge lineages of freshwater organisms and may present opportunities to initiate allopatric speciation. Um so, let's say we have a uh river flowing across this sort of plateau. We have a a neighboring river across the drainage divide um flowing down an escarpment type situation, and there's a fish that lives on this plateau side. Say there's a river capture, so the headwaters of this plateau draining river get rerouted into this escarpment draining river, and this um uh fish population may find itself in the wrong place, on the opposite side of a drainage divide, and it can accumulate genetic differences. And then again, we can um infer uh the the pace or geographic arrangements of these the history of this escarpment and these river captures using molecular data and um things like phylogenetic inferences derived from molecular data sets. And something that's exciting um I think is that we can um interrogate the phylogenetic record at different time scales. And um today I'll mostly be talking about the relatively recent time scale, but I'm excited to think about how we can um detect things in deeper time that may not be immediately apparent in uh topographic data sets. So, when I started to work on this, um I was not working with an evolutionary biologist, and so I decided uh we decided that we should approach this using some simulations. Um so, I'm going to walk through those simulations, um and then I'll move on to some sort of ongoing work in my research group at Florida State that is exploring these things in an empirical setting. So, um we coupled a landscape evolution model with what's called a metacommunity model, and I'll talk more about that in a second, but basically this is a um stream power incision model, and um to induce river network reorganization into this model, we introduced a um dipping fault. So, this is just a plane of more erodible rock, and it's dipping towards the lower right-hand corner of the domain. So, as the surface of the landscape lowers, it intersects with this fault, and the fault trace migrates across the landscape. Um then we seeded this model with um this is This is like an an agent-based model, um and so we seeded it with these organisms, and these organisms can be born, die, move throughout the river net- network, and then we also have speciation that occurs. So, I'll just play the landscape evolution model for you. For fun. Perhaps. Space bar. Okay, thank you. Okay, so you can see that as the um rivers are eroding into this plane of more erodible rock, they expand, and then these little blobs are some small river captures that happen. This is a typical time series from one of these simulations, and so this is uh model time on the x-axis, and then the relative size of these river captures. Um so, you have these little pulses of river captures throughout the simulation. Um this is showing uh the speciation rate of the uh the average speciation rate of the organisms throughout this model. And I'm comparing a dynamic model, that's the black line here, and this is um, basically a the coupled model, so where we have the changing landscape and we have the metacommunity model running on top of that. The gray line is a static landscape, so we just have the metacommunity model running on a landscape that isn't changing. So, we get these bursts of speciation associated with these uh, river captures. But, we also see these little pulses of extinction, and that's because these uh, basins shrink out of existence and it causes some of these lineages to go extinct. So, we can uh, difference these two and get a diversification rate, and on the whole, we have an elevated diversification rate despite these little blips below zero as these basins shrink out of existence. Um, but the, you know, the net impact of these river captures, um, is uh, depends on both the size and the frequency of these river captures. So, if we have bigger captures, we have more um, that we have a bigger biological impact, as well as rates of molecular evolution and the dispersal capacity of the organisms. And um, this model um, uh, produced similar results to that of Lines et al. 2020 um, Species Evolve, which was incorporated into the into the lab land, sorry, land lab uh, framework. Um, so, if as if we can inform on um, the landscapes in which there are frequent river captures, how big these are, maybe we can start to better constrain the impact that this may have on species um, and explain some of these patterns of biodiversity. So, with that, I'll move on to an empirical setting. Um, so, I've been working here for a while. Uh, this is the Eastern Continental Divide. It separates rivers that flow into the Atlantic Ocean from those that flow into the Gulf. And um, it is in motion and particularly along the Blue Ridge Escarpment, um, which goes from Georgia up to uh, Virginia. There is asymmetric topographic uh, relief and slopes across this drainage divide. You can see this escarpment here, which is a steep landform on the Atlantic side of the um, escarpment and uh, people have been describing this in the literature for a long time. Davis described the contest for water across this drainage divide as the most stubborn contest ever waged. Um, we can apply uh, different topographic proxies for inferring uh, drainage divide instability. This is a chi map, um, which uh, indicates again that the drainage divide is moving towards the northwest. So, we uh, wanted to test these different topographic proxies, chi, chi or Gilbert proxies, etc. Um, and so I measured uh, paired brilliant tin derived erosion rates from catchments across the drainage divide. And I won't go into too much of the details, but here I'm showing a delta chi versus an inferred horizontal migration rate and then this is a difference in channel head slope and um, anyway, they both generally uh, suggest the correct uh, relative uh, direction of divide migration. But another take home is that the rate of horizontal divide migration is very slow in this landscape. So, like, you know, tens of meters per million years and perhaps this is how we maintain these topograph these passive margin escarpments for a very long time. But we also have evidence for river captures. So, that diagram was inspired by this setting. So, we have nick points and capturing rivers. We have barbed tributaries. We have sedimentary deposits with provenance um, that suggests, uh, an origin from outside modern catchments, and we have wind gaps. Um, so we might expect that there should be slower erosion rates in tributaries upstream of the nick zones and faster higher erosion rates within the nick zone. So we again collected collected brilliant 10 derived erosion rates to test this. So this is the Dan River. It flows to the Atlantic and you see it's taken this huge bite out of the escarpment. Um, there's this uh, somewhat uh, impressive for the East Coast, um, uh, gorge here or nick zone. And the idea is that this river used to flow sort of parallel to the escarpment and go into the Ohio River system. So we do see that there's there's about three times difference in erosion rate for a tributary within the nick zone as compared to upstream of the nick zone. Um, so I wanted to see if we could try and date this river capture using, um, uh, uh, 1D stream power incision model and try and match it basically to our observations of the position of this nick point as well as the, uh, measured erosion rates. So, um, this is how the model is set up. There's some initial condition and there's uncertainty about where this, uh, river capture point should be, but, uh, so the upper part here, um, we basically allow that to flow into this paleo upper Dan River at the start of the simulation. And then, um, the, uh, nick point migrates upstream. And we also track the erosion rate in this basin where we this catchment where we measured, um, this, uh, this elevated erosion rate. And, uh, I went through a whole sort of, uh, explored a wide parameter space and did a model misfit sort of exercise and try to minimize the misfit. And this is showing the frequency of local minima in, uh, in this exercise, um, versus the age of the river capture. And, uh the answer is rather messy, but it seems like either the capture happened relatively recently. Um we have a peak here, or perhaps it was older, around 7 million years ago. Um and a lot of this uncertainty has to do with where the capture position happened as well as we varied the value of uh the slope exponent n, and that n is why there's a bimodal distribution for the answer. So, let me move on to the biology. Um so, in addition to these nick points, etc., we have zoogeographic evidence for river rearrangements across the Blue Ridge escarpment. So, there are some species that are um widespread west of the drainage divide. So, here I'm showing the Tennessee River um in purple. So, there are many species that are widespread in the Tennessee River, which drains into the Ohio, and then they have these sort of odd disjunct distributions across the drainage divide. My student Alexis Coley is working on one of these species. Uh this is called the saffron shiner, Hydroflux rubriclus. And it is distributed again in the Tennessee River with these few disjunct distributions in headwaters on the Atlantic coast. And um these places where it's found have these nick points that are consistent with the history of river capture, including in the Savannah River and the uh Linville River. So, we um uh used tissues collected from across this range. We sequenced the DNA from these um individuals, and um I'll show you some of these results now. So, this is showing a phylogeny. Each of the tips on this tree represent an individual, and then the lines show how they're related to each other. And I'll uh call out the things that you should pay attention to. So, the Linville, uh which is on the Atlantic coast, is sister to the North Toe, which is on the uh western side of the drainage divide. We see the same relationship for the French Broad western side of the drainage divide in Savannah, eastern side of the drainage divide. Um This is a sort of clustering analysis. It's comparable to 23andMe. So, each of the bars is an individual, and we determine an optimal number of clusters and then assign the figure out the proportion of ancestry that would explain the um uh observations. And so, we the the French Broad and Savannah come out as two separate clusters, which suggests there's a significant amount of genetic differentiation between these two rivers. The Linville and North Toe come out as one cluster, which indicates there's comparably less genetic differentiation between these two rivers. But, something that we were surprised by was this um admixture. So, you can see there's yellow bars in the Linville between the Linville and the Watauga River, which uh we weren't quite expecting. I'll dig into that in a minute. So, let's move on to the Linville River. So, here is a topographic map of the Linville River Gorge. Uh the Linville River flows to the Atlantic coast, and again, it has this taken this big bite out of the um uh plateau side of this environment. It's a really cool place. It's one of the few wilderness areas east of the Mississippi River in the US. Um there's old growth forest in the gorge because it was too steep to log. Um and there's nice waterfalls, hiking trails, etc. So, I recommend it as a visit a visit. Uh Brad Johnson um published a paper in 2020, kind of describing an idea for the relative chronology of of a river rearrangement in the Linville Gorge. So, the North Toe Nolichucky, this flows into the Ohio River system, flows through this valley here. And um so, the idea is that this Linville used to flow into the North Toe Nolichucky and then was captured by this Atlantic slope river. And again, you have this um uh nick point in the river that supports this. Um recall that the Linville and the North Toe populations of the Saffron Shiner are sister to each other, which supports this sort of um uh T-bone rearrangement between these um river systems. Um but we saw this admixture between the Linville and the Watauga, which was unexpected until we zoomed into the DEM. And in the headwaters of the Linville River, um here's the Watauga, which again, this flows to the Ohio. Linville flows to the Atlantic. You can see it's a very unimpressive drainage divide. It's rather flat. There's in fact a swamp right here. Um and it looks like maybe there's some sort of um wind gap migration occurring. Um in this analysis, um you can instead of assuming that evolution must occur through these bifurcation events, you can allow for migration between different lineages. And so Lexi's analysis um resolved this migration event between the Watauga and the Linville, which is notable cuz the Watauga is rather distantly related from this Linville lineage. Um so Linville and North Toe remain sister to each other, but we have this sort of second migration event from another population. And um I'm excited to think about whether this genetic data and analysis can inform on um when this migration event occurred. And I'll note that these uh lineages are too recently um re- they're too recently diverged to accurately use molecular clocks. You may have seen the use of molecular clocks where you assume some rate of molecular evolution to date um uh things in absolute time. Um but these this is this isn't appropriate for the study system. So, I want to introduce you to um, rather new a simulation approach that we're using. Um, and so we're trying to employ this uh, simulation framework called Slim, which I took this from the manual, but uh, they originally were calling the selection on linked mutations. They're studying a specific evolutionary process, but now they want to call it simulating life in machines, which I think is some I thought might appeal to this audience. Um, and basically this is a flexible open source um, framework for population genetic simulations and it's it's a really impressive modeling system. Um, and um, base I I won't go into the details. You can do all sorts of things including spatially resolved models, but basically individual organisms and their genetic information is resolved. So, it's appropriate for comparing to these population genetic analyses. And so, uh, these results are rather preliminary, but I wanted to explore what uh, the population genetic differentiation would look like in a scenario where we basically have repeated river captures. So, the Blue Escarpment has been continuously migrating inland for hundreds of millions of years and so you you know, have this repeated movement of organisms over the drainage divide. What does that do? So, I set up a really sort of simple framework where I have a parent population that sends a daughter lineage over the drainage divide. It becomes genetically different. Then at a later time we have an another uh, lineage that another river capture and a new set of organisms gets moved across the drainage divide and then you have bidirectional migration between these daughter lineages. Um, so in these preliminary results, what I'm showing on the Y axis is called FST, which is a metric of genetic differentiation. Um, and here I have a scenario where there's just one river capture, two river captures, and three river captures, and then I vary the size of these river captures. Um and the point is that you as you have more river captures, you lower the degree of genetic differentiation across the drainage divide. And anyway, so I'm still working sort of relative space and thinking about how to calibrate this to the real world, but I'm excited about um I I wanted to share it with you because this is a a modeling conference and I think this is an exciting tool. So, I'll wrap up um fluvial environments in tectonically inactive landscapes like the Eastern North America remain active. Perhaps this could explain um the uh this observation that these freshwater fishes uh are just as diverse in tectonically inactive mountain ranges as they are in active mountain ranges. Um drainage basin boundaries are dynamic and the movement of these drainage divides can provide opportunities for allopatric speciation. And biological data both uh phylogenetic information in deeper time as well as populate population genetic information in more contemporary time scales could inform on the frequency and age of river rearrangements. Um special thanks to my research team um who have helped move this research forward since I started my position at Florida State. Uh my students Lexi, Kat, and Malia in particular. And with that, um thank you and I'm happy to take questions. >> Thank you, Maia. Um any questions? Jeff. >> How often does an animal like the small uh capture and watch the the little guy >> Uh I forget the question. Uh the question was about what the impact of range size and the relative importance of these river rearrangements for range size. One thing that we learned from the simulations I showed early in the um uh talk. So, basically when we reduce the dispersal capacity of the organisms, we get small range sizes. And something that happened is that they went extinct all the time. And and so, um and so, I hear what you're saying. You may impact more things, right, if you have a larger range size, but I think you do run into the risk of risk of extinction as well, and you might have to think about both processes. So, yeah, good question. Yeah, so the question was about whether we could use different species and leverage differences in, for example, rates of molecular evolution or something to sort of uh um uh learn about different components of the landscape. Um So, Lexi is working with a few different species. I think that the rate of molecular evolution is probably similar across them, but something that is different is um their uh range sizes, their population sizes, and then uh their life history. So, she's working with minnows and also darters. So, minnows have large populations, they live in the water column, and darters live are benthic and have smaller population sizes. And so, that's one thing we're considering. I think also we're um we're trying to start a crayfish project as well because those are aquatic but can also move over land. So we may see some interesting differences there but these crayfish is also have similar disjunct distributions across the drainage divide so