 One one is fine right because they're only ever going to talk Hey, we're gonna we're gonna do the breakout Recaps against gonna be exact same format as yesterday. So the online group will go first everyone will have four minutes I'll struggle with the timer again, and we'll go from group five four three two one after Online session goes so it's on and Julianne and then Erica and Sam And so on hopefully you guys all remember Okay Sharing the screen you can start now Okay, can can everyone here? Okay cool Yeah, so I'm Going to present the results of what we talked about on our online Breakout session and just like to say a quick. Thanks to everyone who helped set up an online side to this It's very nice to be able to participate from across the world So the the first question what challenges are facing the tectonic models Something we raised was what benchmarks could we come up with? for our large-scale tectonic models and On what scale both temporal and spatial Could these benchmarks be applied? And tying in with this. There's are there good representations in the real world for these Processes, how can we test our models against something that actually exists and how do we deal with when? We have non unique solutions and How do we compare between? between data and Models between observations and models when they sometimes come from very different things Do we need models that can? Deal with different timescales both short and very long timescales And how would these work and how much can we just parametrizing different scales? And something important we raised was that these type of models require links between the observation the the field observations and data collection side and the modeling community and It's very cough. It's very common to find field geologists who just don't trust models at all or the other way round modelers who don't necessarily see the value in Detailed field evidence and we raised this as something that you need more discussion between these two groups as well for for the models to work and a few specific points so weakening on faults is critical, but the exact physics are not clearly known right now, so that's something to work on and Something that could help deal with some of the problems is looking at tectonics of places that aren't earth Looking at other planets as well in particular Venus maybe the may have more active tectonic processes and could give a good environment to compare our tectonic models from earth to Another planet with very different conditions and maybe less blurring factors less erosion and all that as well Moving on to the second question. It was maybe something that was less our our field of expertise raised a few issues We weren't sure if there was More of an emphasis on detail and rather than the bigger picture But something that needed thinking about and also how accurate term surface processes might be when extrapolated over long million tens of million year geological timescales as well and Where are the offshore studies in this what's going on in depositional environments below the ocean and And What coupling do we require with Climate and ocean circulation models as well for these surface models to work properly And you try to wrap in about a minute shop just coming to the end And to come to the last question We What do we need to tie this all together between the two? but better understanding of the other community's science Help constrain what properties are important and feedback into couples models. I guess that's what we're doing right now and Important point. How do 3d processors influence the interpretation of 2d models? how important 3d models compared to 2d models and how we come to terms with the uncertainties and models and And a few specific points To what extent the the surface and deep earth systems are coupled which was something that was coming up in a few talks and Finally down to computation What techniques can be used to make computation cheaper? So that's more or less what we talked about Wonderful. Thank you so much. I appreciate that Google that too participating Okay, it's on Julian. All right, so I'll start with the points that came up in our group I'll start with the points that came up in our group about the challenges that long-term tectonic models face and so one point that came up was that not many or like 3d models is just something that people start to do and they are sometimes not well constrained and it's difficult to validate them and Also, in particular, they are not enough for constrained data sets to do that and Some things like earthquakes are not well understood. And so we have to improve that Then there's this issue of strain localization. So as soon as there's brittle deformation in the model there's a resolution dependence and There we are just starting on how to resolve that Then there are large uncertainties in the input parameters for these long-term tectonic models in particular the rheology Where it's not clear how exactly should we choose that? And we at the moment we don't deal at all with the interaction of fluids and brittle deformation like melts and water in terms of the surface processes and some of the things that they were already said also is set in our group in one of the challenges that came up is how well defined and physically based equations are and Related to that is should we aspire for a globally applicable Multiple setting that they attempt to model How can we better constrain Initial conditions in specific settings where they really influence the outcome of the model in terms of our ability to evaluate our model performance in The same issue that came up With the geodynamics component is a resolution dependence on results and what are the metrics need to Look at in this the influence of these different resolutions those there were also a few points and that were joint challenge to the communities regardless of coupling and One was is how to compute uncertainty in the models in a consistent way The second one was what is the degree of complexity? This is very much dependent question guides the model and How do we make results reproducible? Yeah, and then we also talked about what the challenges are for coupling and So one point that came up was which components do we want to exchange between the models and can we Like we need to produce some kind of platform where code can be shared and where we can define some kind of interface What do we want to exchange? Then we talked about that a lot there are different times and length scales in both of these models And how do we define the interaction between these different time and length scales and like how to do that best? Then there are these different parameters in the surface evolution models And it would be good if we could find the way to link that to rock types or to the inputs in the tectonic model so that these parameters are consistent between the two models and then One point that also came up was that there is some communication issue or that we just do not know enough about the other community So people should maybe like someone suggested we should come up with a list of things that we know about our community or think we know things that we think we don't know in our community and Things that we don't know about the other community or that we would like to know about the other community And one suggestion was can we maybe in some models some processes use a stochastic representation of them instead of a like representing representation based on physical processes like for example falls they are just like a statistical feature where they show up and And Yeah, then that came up in the presentation today. How do we parallelize surface models? Can we do that more and Yeah, how do we develop the right software and the right modeling tools to implement all of these things? Sam and Erica and on deck is Adina and Timo Okay, so I think we already hit a lot of the other groups already hit on some of these things But with regards to tectonic processes With regards to tectonic processes a big thing that came out was uncertainty With regards to lithospheric rheology How accurately is accurate? How accurate is accurate enough? We also talked about complexity. How complex do the models need to be for the tectonic processes? What's the question you're trying to answer? Do you know what level of complexity is necessary and which parameters to test and for which purposes? We had talked a lot about simplification Simplifying the models to understand what level of complexity is needed and for what reason Compare complex models to simplified ones and then we talked about initial conditions What are the initial conditions and how do they influence the final results? Okay, so for surface processes Starting off with uncertainty in physics We need a better understanding of the physics We need a better understanding of the physics within surface processes, particularly the Coefficient K because everything is lumped into that tunable parameter and also I'd like to Reflect what the remote group said about 3d processes and 2d processes. What do we lose by integrating into fewer and fewer dimensions? Complexity what level of complexity is important and where should we put our resources our time and effort? How can we incorporate important? Complexity to the models. How can we determine which types of complexity which parameters require our greatest amounts of attention? And how to bring other needed Processes that can be something that can be represented in simplified models Into our own surface processes models such as groundwater climate and vegetation dynamics In cases where we can trade off some complexity for faster models Taking an ensemble approach to parameter calibration or Tuning for model fitness is also useful and We also spoke about it the issue of equifinality So we're calibrating our models based on what is easily observable a lot of times that is Topographic shape But the parameter values that are generated by that calibration may not accurately reflect the the actual physical conditions in the landscapes So you might be getting the fit for the wrong reasons And we need more checks in our model for that In terms of metrics in addition to topography is defining the fitness of the model We can also begin to use stratigraphic records to help us understand the temporal variation in surface processes And environmental commit conditions that influence them So that Yeah, so question three In terms of scale there's disparate temporal and spatial Scales are we too biased by the processes that we see right now and can we assume that they are the same over millions of years? So the major scale difference how best can we use our sub modeling or nested modeling methods to link processes that are more Relevant at finer scales, but we'd still like to see their Their contribution to larger scale process tectonic processes Accessibility one of the best practices for sharing of data tools Such as models and knowledge always a model responsibly make model results and uncertainty comprehensible to larger audiences But you also of course want to avoid the black box method of sharing Sharing your model and also that's important to come up with compelling ways to share your methodology and your results results to assist with collaborations between surface processes modelers and tectonic modelers All right one last point definition of goal Should we always integrate into one model or should we just interface between the tectonics and surface processes models? Do we need high fidelity output or don't we and can we accomplish that goal using simpler models? Thank you guys. Okay and Benoit and Renee are on deck Okay, this is these are the results from group three question one long-term tectonic challenges Regarding numerical methods and physics is the question about How can we deal with the really large computational cost and the robustness of solvers, etc? There was the discussion that there's need for better transient formulations instead of instantaneous models There's the question of how to deal with the fact that they are discrete and continuum models and how how do they compare? How do they integrate with each other? and lastly Even the question of governing equations and how to incorporate rheology is not not really clear and needs to be tackled Then regarding verification We feel there's a lack lack of understanding of the differences between the different codes models and numerical methods used by the different groups There's different time scales between the observations that are accessible to us and the response and the computations that we do So that's a that's an issue to verify these computations and then finally regarding verification is What what are even the observations that that we want that we could get in the future? How do we deal with that? Lastly Regarding the community I think this was brought up already earlier today There's a lack of training support documentation access to to tools to models to software etc and it would be good to have some way to enable that I'll talk about surface process challenges So I think our group agreed that the biggest challenge is related to the validation of models with observational data So and that is because we have non-unique interpretation of data due to uncertainty And noise and we'll never get a perfect fit So how well models deal with this do we need to improve the physics of the different types of processes? Or the long short term large small scale response And also what is the model response to different parameter inputs? For example the input of high dimensional parameter data So some suggestions some people propose that we need to use smarter techniques Like the adjoint gradient and homogenization techniques And other challenges Relate to mesh dependency and also how flexible our models to address the questions Okay, and question three is about coupling On the physics and data side What what we discussed early on is that all of the problems regarding the coupling Depend on the question that we are trying to address Why do we want to do the coupling and that everything else leads from there? Especially what time length scale do we want to look at? And of course there is different time and length scales between them So we need to scale one of the processes up or down And depending on what question we are trying to answer There's a big issue with respective validation and tuning of the models And we assume this is a lot more challenging because you now have two models with twice as many parameters And yeah, we'll have a difficult time doing that Another question we asked is Do we need different data to verify that coupling of these processes is correct or make sense? And if yes, how do we get that? From the numerical point of view There's of course when you couple two different codes or methods There's the question of robustness stability accuracy, etc There's a big issue if you want a couple existing codes is that there are different numerical schemes that are used By elements finite differences, etc, etc So you have to worry about interpolation different resolutions and all these kinds of problems And this was also already brought up by earlier groups is we need to communicate data So the question is what do we communicate and well Data might be represented in different ways for example And lastly the Issues about the community is we need better communication between the disciplines We need to find questions that actually motivate both communities so that they are people Going to come together and work on these issues And then when you're presenting your results The question is how do you communicate and describe the methods the results the limitations of your experiments, etc So that the other communities understand what's going on And there's certainly also the need to incorporate computational scientists That's all we have last but not least will be on so and Nathan guys after these guys Okay, so like the other groups we discussed the challenges for long-term tectonic models and had some numerical challenges like efficient inverse modeling schemes And stable multigrid servers for large viscosity variations Um, but also like the other groups we had a lot of challenges that came from accessibility of codes of models Like how do you even get to a geodynamic modeling code? If you for example, not a model of a profession Um, how do how do you learn how to use it? What is the role of community organizations like cag and cstms in this project? Like how can they help? Making codes the codes accessible um put Regular webinars or video tutorials helping that Um, how do you design codes so that they are accessible with interfaces and documentation? And also, how do you make the computational resources available that you need to run these models? How do you make it easy for? Also smaller universities or smaller working groups to access these resources Okay, so um second question about the challenges landscape motion models So actually a lot of in our session a lot of challenges came from dynamic community and the big challenges actually where to start which model Pick up because there is no single model that everyone agrees on Well, we still Use a lot stream power and diffusion models, but there is still ongoing debates on whether or not These models are accurate and in which cases So the problem is that there is many different models and codes for different natural environments This is something that is confusing for tectonic people to choose the model Another challenge is scaling up the processes to origin and continents obviously and there was also some ongoing debates on the mesh sensitivity, so Model parameters that are sensitive to the scale So some said It is an issue others said that it's not an issue. So it needs definitely so verification and so then Last question about coupling different models. So there were two main Categories of challenges that arise The first is the social challenges. We kind of all agree that we both both communities Need some communication effort to make so to ensure that models can be understood by other people outside of the community Um, it's also important to work together. So not only reuse the work that has been done in another community for Or studied but actually if you want to couple models, then we need to work full of researchers from both community So there is actually some work that has been done on coupled models, but there are still some technical challenges And one of the main challenge for a geodynamic is to have parallelized Landscape evolution models that can run on this system So how to pass the information between Different kinds of models. They have different kinds of Schemes and so on so it's it's a big issue too Choosing the the time steps so So this can be an issue in some cases, maybe other cases. This is not really an issue and last Some that have tried to couple models issues of how Having a model setup like initial and boundary conditions that are consistent models Thank you. And last but not least Nathan and Anso First challenges facing long-term tectonic models. So we need coordination between tectonic and surface process models in terms of What model output and input are needed for coupled models and then identifying metrics for validating models And we need a higher resolution and therefore good pattern scaling And we need more detailed strain localization And we need to incorporate fluids in long-term between models and have to tackle associate challenges with coupling between fluid and solid Need more physics based knowledge On hydrothermal processes need to decide on how much details are needed to Be incorporated in models and how much averaging or upscaling is appropriate and it to improve Stress fields and how they can be compared with available stress observations and petroleum model What is the important on time and space scales? and What are the then good science questions we need to answer then we Well, lack of constraints over long time scales They are very difficult to Validate or acquire in the first place and therefore What is the what part of the primary space should be explored? Those are the challenges And for the challenges of surface process models, these were some of our primary talking points And one that came up in a few different ways was incorporating What are good practices for incorporating climate or weather variability and how can climate be implemented without relying on climate models that have several non essential components Also, what are best practices for incorporating uncertainties and sediment transport materials and other process parameters How can we tackle or should we should we tackle Uh, implementing more complex upload fields in from that we can get from tectonic models in order to model topography A challenge that remains is constraining uh paleo elevation As important implications of um relic landscapes And um string one is how is energy of water dissipated through domains of earth surface, uh system, ecosystem, vegetation, hemisphere, et cetera Similar to another group, we discussed the challenges of equifanality and model models and how to distinguish Different processes that led to similar model outcomes and getting to the Technical side We discussed the importance to not only develop but also to make paralyzed code that is user friendly and doesn't fall into a black box We discussed about applicability the carefully applying models. So for example, all streams are not bedrock streams So to to consider where the stream power model is best suited Someone similarly three quarters of earth is underwater. So perhaps we should learn how that works as well And we also emphasize the importance of uh depositional systems And asked ourselves how can archives and depositional systems be used for models Last question. Uh, what are the key questions or key challenges in coupling long-term and surface process models? And uh, some of the key ones were models driven by should continue to construct our models and fine tune them to the scientific questions that we ask And um, we also asked do Fully coupled models need to be are they necessary in many applications? And, uh, we we envisioned a technique to Create modular to to follow a modular approach to couple tectonic and surface models than to Couple complete models. Thank you everyone. Great upper