 Good morning everybody, my name is Andrew Moody, I'm happy to be here this morning talking to you a bit about some of my research, as well as this numerical delta model that we're developing that we hope becomes a real community tool for delta modeling. So this is a satellite image of the Mississippi River Delta, it's the largest delta in the US and home to more than 2 million people. It's looked by many lobes switching emulsions over the last several thousand years, and anthropogenic restrictions to the upper reaches of the river are preventing another natural evolution from occurring towards the west here which would be an act that would distribute sediment widely across the landscape. These restrictions in the form of levy construction have cut off the natural connection between the main river and much of the surrounding wetland environment because sediments naturally compact over time. This has led to extensive land loss, along much of the Louisiana coast over the last half century or so. There have been natural detrimental impacts for the society that lives on the delta. For example, retreating coastlines can force human migration, as we heard about changing water depths and salinity effect animal habitats and important industries like shipping and others depend on these stable landscapes. The Louisiana government has devoted $50 billion over the next 50 years to at least in part implement what we call nature based solutions. So these nature based solutions take inspiration from natural land building processes and try to recreate some of this in an engineered way. For example, is sediment diversion projects. The midbury terrier Bay sediment diversion project is in a final comment review period and could could begin construction in the near future. The sediment diversion projects work by connecting the main river to the surrounding wetland environment by an engineered conduit, and then downstream of this engineered conduit natural delta land building processes such as channel migration and channel diversion projects are expected to take over and are responsible for distributing sediment widely over the landscape to build land. But these diversion projects don't happen in isolation. There's a large number of external extreme events that will affect the ability or the natural land building system to distribute sediment widely downstream of sediment diversion structures. And we need to understand how all of these different processes interact in the same numerical framework if we want to be able to predict land building downstream of these diversion sites. So that's what we've tried to develop is a flexible delta model. It's able to incorporate different processes and easily compare across studies. And so I'm going to talk a bit about our our model implementation and show two projects that we're working on now that have used this model to try to incorporate some of these processes. So the first one I want to talk about is faulting induced subsidence on the delta. So we have a map here of interpreted faults in the subsurface in all of these dotted lines. And these are large listric growth faults that are sort of slumping off into the Gulf of Mexico. And some of the larger ones make expression at the surface and they create these scarps in the marshland that affect the distance that water has to travel from some point to reach base level. So it makes steeper gradients in the landscapes in certain directions as a result of slippage on these faults. So because the sediment diversion projects will, will this one and many others likely sit right on top of some of these possible faults. And to understand whether faulting induced subsidence events that we expect on the Mississippi River Delta are large enough that they could reorganize channels and potentially limit land building or affect sort of when and how land is built downstream of these diversion structures. The numerical Delta model that we're using is called Delta RCM. You've probably heard a bit about it. This meeting already. It's a rules based cellular morphodynamic model that routes parcels of water and sediment across the landscape to build a land form over time. I think it's gotten really popular because it is a really nice balance between computational cost realism and interpretability of different processes in the model. A lot of people, including some in this room have added components, tested the morphology or compared it with real deposits, as well as the dynamics so not really going to talk a lot about the details of the model today. I want to talk a little bit more about our implementation, which is a Python implementation of the original Delta RCM model that we simply call PI Delta RCM. We really tried to embrace a lot of the fair principles that Neil mentioned on our first day here in developing implementing this model. So I'll give a little highlight of that. And then we have a companion sort of analysis package that that goes along with the numerical model but also works for physical experiments and so I'll mention that as well. The original Delta RCM model was implemented in MATLAB, which made it open source but not free. It was also implemented as around 1000 lines of code written in a single script, not with a lot of documentation, parcel stepping was implemented in a serial fashion so it's sort of step by step in a series of nested for loops. So I like this analogy of Delta RCM being a really attractive or delicious bowl of fruit, but it was sort of hard to tell where you would add in different behavior or different processes. If you wanted to modify, for example, hydrodynamic aspects of the model. So you might see some banana sort of over there, as well as over there. The high delta RCM implementation is fully in Python, and it ends up being around 10 times faster than the original MATLAB version. But as you heard, it's still not super fast. We've also broken out all of the different components of the model into separate modules, so that it's now extremely clear sort of where hydrodynamics exist within the model, as well as that the spoon is the input and output of the bowl. All right, so we've we've made a huge effort to do to provide documentation commenting and testing, which could still be improved further. But but we've really tried to do that well. We've made it so that model runs are fully reproducible. We've implemented checkpointing features, you can run simulations in parallel, as well as configure complex suites of simulations that sort of explore parameter spaces as well as replicates and ensemble simulations. It is compliant with the basic model interface and supports time varying inputs and all of these things natively. In order to achieve this flexibility that we're after we take advantage of the object oriented approach of our model, which allows us to use this thing called subclassing. So what we're doing here is basically in order to simulate faulting and do substance that I'm interested in for this scientific question. We simply create a custom class that incorporates the behavior of faulting and do subsidence that separate from the underlying delta model. This means that when I publish the code for my faulting and do substance model, the logic of all the core model components that have been unchanged is completely separate from the components that we've changed, which makes it really easy for scientists to understand what we changed decide what they don't like and move forward from there. Alright, so some results then we're interested in understanding land building in response to faulting and do substance over a range of different displacement lengths. So we simulated a range of different displacement lengths. You can see on the on the right here, three different simulations, the one on the left is a control. The one in the middle is sort of a moderate slip length of about 10 centimeters, and then on the right is much more extreme of two meters. At the beginning of this cycle, which will start over again in a minute. You can see all three simulations will move exactly the same pattern because they're starting from the same initial condition. There's this vertical blue line, which is when the displacement event happens and the simulation will pause. You can see the red which represents the sediment flux through the model is sort of redirected in this extreme case on the right into the subsided area, and all sediment or most of it is being delivered into that subsided area, whereas the middle simulation is not so strongly reorganized. So this is sort of a qualitative interpretation though, and we'd like to quantify that. So the quantification that we do in both numerical Delta models, as well as physical experiments is really very routine. We do a lot of land masking and finding channels and all of these things, simulating or accumulating land area and computing stratigraphy. So we've developed this Delta metrics package, which tries to streamline a lot of that routine analysis and provide helpful tools. This is a Python package as well. Who's here is going to lead a clinic this afternoon. So if you're interested in some of these topics, I encourage you to come and check it out. So to quantify that distributory networks network change, we're going to look at the amount of sediment that's going into the subsided area, divided by the total sediment flux that's coming into the Delta. This is a simple ratio but lets us quantify sort of the amount of reorganization that's happened right after the displacement event. So the x axis here is time vertical black line is the displacement event, and then the colors from dark, the light represent larger displacement events. And there's a whole lot of variability because we assess this over many different sort of spin up conditions as well as different locations over like 115 simulations. But the overall trend here is that for larger displacement events there's a swift and extreme reorganization of sediment flowing towards the subsided area. And then there's a gradual recovery and the rate of that recovery depends on the initial displacement. But I want to just again then even boil this down further to just look at like a delta value so sort of an immediate change following the displacement for for each one of these different displacement lengths. This allows us to compare model results with real world simulation or real world, or maybe contextualize our model results in a real world system, which is really valuable. So we developed just this simple non dimensional number on the left here that a displacement magnitude, which is basically just a ratio of characteristics of the displacement or the fault relative to the channel dimensions so channel size as well as some characteristics of the work. When we plot all of our simulations on the right in this space, you can see that the amount of distributary reorganization increases with displacement magnitude over many orders of magnitude. So let's take a look at the Mississippi River Delta, look at the cubits gap sub delta, which sort of scales or gives us a sense of the scale of distributary networks that will, that might evolve downstream of sediment diversion projects quantify some of the characteristics there and use data from faulting on the Mississippi River Delta to get a range for what we think is sort of a realistic bound for faulting induced subsidence or displacement magnitudes on the x axis, and a possible response of distributary networks downstream of sediment diversions there. So the other example is that faulting induced subsidence can delay land building can reorganize distributary networks sort of at the scale of distributary networks and faults on the Mississippi River Delta. But what we don't have yet really is a good understanding of the probability of these events at different locations on the delta. So we kind of place a call for better geotechnical data in that regard. So the example that I wanted to show is just really preliminary but I think it's a nice example of how flexible this model is and how easy it is to add in new components. You've also seen this from the participant of the summer Institute participants the other day who were able to do this really incredible work and just a few short days with with this model as well as another kind of version of delta or cm that's floating around. So the question here that we're really interested in is, can manual seating of marsh crash like Spartana, or maybe mangroves. Can this bolster land building downstream of sediment diversion projects. And if it can, where within the sediment diversion efflux, should we place these, these marsh seating projects to maximize land building. So here yet. The first thing to do is just to understand how marsh accretion. So the addition of organic matter effects land building downstream, or during land building on these deltas. We took a super simple approach here, which is just to use this equation on the left here which is a parabolic equation describing how the amount of organic matter that's produced as a function of depth below water level. With just a few lines of additional code in a subclass of the delta RCM model, we're able to add in this behavior that at every time step, we just create the marsh land by some small amount based on the water depth. Hopefully you can see in the movie. So the right here that the size of the deltas appears quite different. There's also a large area of the delta in the marsh accretion experiment that's has has actively accreting marsh, and it's sort of this value in the orange of about point seven. This simulation sea level rise is set to seven millimeters to point seven centimeters here, which indicates that the marsh over this whole delta area is keeping up with sea level rise. We think this is probably a little bit over optimistic, and it's a little bit probably due to our simple simplistic equation here, as well as maybe the relative choice of sea level rise. And the marsh accretion rate, but it's a nice demonstration that this simple behavior can really change the morphology of the delta. So I'm showing you the first plot is land area. That's just the number of pixels that are above sea level. And then the delta area is sort of the amount of land that's within the within sea level so that it acts as sort of meaningful marsh land that could dissipate something like hurricane energy. Or is useful for fisheries or something like that. So we see that there's a really huge area of this land that's near sea level created in the marsh experiment. So, we're interested in continuing this work, but hopefully it gives a nice demonstration of the flexibility of this numerical delta model and how we think it can become a community tool to add new processes in and test a lot of different ideas. So this is my last slide then I dealt RCMs a community tool for Delta modeling, please please get involved with us if you're interested. I've put a couple links here where everything is online all the time GitHub active development, we don't do anything secretly and then post it later. So, please join us if you're interested at all. And so faulting into subsidence may delay land building at diversion sites, due to reorganization, we need better data on the probability of these faults slipping and then early just preliminary marsh creation can really significantly impact elevation distributions on the delta. So that's it thank you. And the marsh accretion addition, if I understood right that is just pure and extra accretion, rather than putting any of the other potential effects that vegetation could have that. So the answer was yes for online people. That's cool that that seeing how much can come from just one simple addition with this tool. So, I like it. Hi, great talk. Yes, I have a few quick questions. So first, what is the time scale of this model and I guess my second sub question that goes into that. Are you incorporating vegeta, in terms of like salt marsh vegetation growth and death or is it a constant, like, are they at a certain level already when this model starts. So for the time scales question, the pie delta RCM delta RCM are operated on kind of like an intermittency idea. So the scaling between model time and real world time is kind of an arbitrary choice. But you can think of these simulations as like a few hundred years for most of what I showed today. And the second question are our marsh accretion process is absolutely as simple as it can be anywhere that's within this marsh window. We just look at the elevation, figure out what the rate of accretion is based on that equation. And that's what we do. There's no patch dynamics. There's no effects on flow. It's just simple vertical accretion. Thank you. Just a quick question. You said that there was a math lab version and originally that you can use sped it up by 10 times. But those languages have a potential to run about the same if you do things right so can you explain about what you did to get the speed up. Yeah, it's a really good question the. It's sort of an unfair comparison for these reasons. The main acceleration that we've gotten in the Python model is using number. So this is a just in time compiler that converts Python code to bytecode machine code. And it's, I'll say it's like remarkably effective and it's incredibly easy to use. So that alone has has really contributed to this like 10x speed increase but probably like creating mechs functions or something like that in math lab would would lead to an increase as well and we didn't go down that avenue. So it was an interesting talk. So I don't know much about this model but I was curious to know if it will be applicable in a lagoon kind of setting where you have a barrier and you have a marsh land growing on the backside of the barrier. So because somewhat what I've read about the coastal dynamics is that one important factor, which determines how your delta evolves beyond the barrier island kind of thing is this connectivity between seawater and the backwaters. So, does this model do those kinds of things or it's not for lagoon kind of settings. Yeah, it's another good question. Maybe I should have included I probably should have included a bit more about the model. There are no. So the representation of hydrodynamics is really simplified. There's no representation of tides water mass is not really conserved. I could maybe represent things like tides or other other forcings that might be important in the goons, but the current configuration or setup of the model doesn't include these processes at all. I was curious how, how thick is your model, like how deep does it extend and those faults, whenever you impose the faults are they only in the thick sediments or do they actually go into the crust. Yeah, another good question. Again, simple as possible. The models are usually on the order of like five meters to 10 meters water depth or less. And the way we've implemented faulting that I've shown here is just instantaneous vertical displacement so we're not actually even doing any kind of a strict movement or horizontal displacement it's just instantaneous vertical of, you know, one millimeter up to five meters. So nothing crustal really about it.