CSDMS 2025 Annual Meeting: Exploring Earth's Surface with Models, Data & AI

Registration

Registration & abstract submission will open early 2025

Introduction

The CSDMS 2025 Annual Meeting will be broad in scope, bringing together CSDMS members to present new scientific insights in the modeling of surface dynamics and the impact of time and process scales, new advances in cyber-infrastructure, examples on coupling models, how social and ecological models can inform management, and more. Also, this is the 3^nd year that you can submit Electronic Publications (Epubs); Jupyter notebooks that contain e.g. a scientific hypotheses description, a numerical solution, and some findings that are investigated by numerical algorithms or model, see also: Form:Annualmeeting2025#Electronic_publications_(Epubs). We reserved time during one of the plenary sessions for presentations of Epubs, so don't hold back and submit your Epubs! The meeting will further include:

State-of-the art keynote presentations in earth-surface dynamics
Hands-on clinics related to community models, tools and approaches
Transformative software products and approaches designed to be accessible, easy to use, and relevant
Breakout sessions
Poster and Epubs Sessions

Agenda

A draft agenda will be posted closer to date.

Keynote presentations

Volker Grimm
Helmholtz Centre for Environmental Research – UFZ, Leipzig

The Open Modeling Foundation: operationalizing FAIR principles The FAIR principles for scientific data and software have not yet been extended to scientific modeling. However, making models findable, accessible, interoperable and reusable is urgently needed to make the modeling of social and natural systems and beyond more coherent and efficient. The Open Modeling Foundation (OMF) is a recently formed alliance of modeling organisations that coordinates and administers a common, community developed body of standards and best practices among diverse communities of modeling scientists. Several working groups cover all relevant aspects of standards development and adoption. First results in 2025 will be (1) an ontology for modeling standards that refines the FAIR principles for modeling, in particular interoperability, and links these principles to existing standards, (2) a tiered fit-for-purpose approach that makes the effort required to follow good practices proportional to the expected real-world impact of a model, (3) a general, unifying language for documenting models that is based on the ODD protocol from agent-based modeling, follows the tiered approach, and supports the I and R principles of FAIR. I will describe who we are, how we work and what OMF’s main challenges are at the moment. OMF's working groups are open to anyone interested in meeting new colleagues, learning to ask new questions, and helping to produce much-needed articles.

Frederik Kratzert
Google Research

Long Short-Term Memory networks for rainfall-runoff modeling Long Short-Term Memory networks (LSTMs) have been around since the early 90’s but only in the last few years, LSTMs gained increasing popularity in hydrological sciences. Publication counts see exponential growth and LSTMs power some of the largest-scale operational flood forecasting systems.In this presentation, we look at some of the research from the past few years and try to understand why the LSTM is a particularly well suited architecture for the application as rainfall-runoff model but also discuss limitations and open research questions.

Jaap Nienhuis
Utrecht University

Two worlds on a single planet: long-term coastal geomorphological data versus models Accurate projections of future coastal change require a tight integration of coastal geomorphological models with data. For long-term projections, beyond the year 2100 for example, one particular difficulty is that our typical validation timeseries are short (often <30 years) compared to our projection horizon. In this talk I will discuss two model-data integration methods that we have used to circumvent this issue. The first method is to apply generic simple models to many different coastal systems, and do space-for-time substitution. For example, we test the effect of sea-level rise on wetland change by comparing wetland change between rapidly subsiding and rapidly uplifting coasts. The second method also employs simple models but tests them using long timeseries extracted from sediment core data. For example, we test and apply our barrier island models using paleogeographic reconstructions going back ~5000 years. In this case it is key to use simple models to not overdimensionalize the model fitting. For both methods there is an added benefit: we learn something about coastal geomorphology along the way.

Allison Reilly
University of Maryland

Extending research on the burden of sea-level rise on the built environment beyond housing Many rural communities are on the front lines of climate change. The need to understand who may be affected, and the ways they may be affected is widely acknowledged as needed for inclusive and cost-effective adaptation planning. In this work, I will explore the potential threat of sea-level rise on non-housing infrastructure (e.g., roads, septic systems) in rural, coastal areas to help inform the impact of their loss on local communities. I will present work that evaluates local accessibility loss during high tides for the entire US coastline (termed “risk of isolation”) for various climate change scenarios, and preliminary evidence from research that suggests that the risk of septic system failures is quite significant in many coastal communities. I will use the discussion to simultaneously highlight under-explored research areas within both engineering and social science that could support more inclusive and efficient adaptation policy.

Yalan Song
Penn State University

High-Resolution Differentiable Models for Operational National and Global Water Modeling and Assessment Continental and global water models have long been trapped in slow growth and inadequate predictive power, as they are not able to effectively assimilate information from big data. While Artificial Intelligence (AI) models greatly improve performance, purely data-driven approaches do not provide strong enough interpretability and generalization. One promising avenue is “differentiable” modeling that seamlessly connects neural networks with physical modules and trains them together to deliver real-world benefits in operational systems. Differentiable modeling (DM) can efficiently learn from big data to reach state-of-the-art accuracy while preserving interpretability and physical constraints, promising superior generalization ability, predictions of untrained intermediate variables, and the potential for knowledge discovery. Here we demonstrate the practical relevance of a high-resolution, multiscale water model for operational continental-scale and global-scale water resources assessment. (https://bit.ly/3NnqDNB). Not only does it achieve significant improvements in streamflow simulation compared to the established national- and global water models, but it also produces much more reliable depictions of interannual changes in large river streamflow, freshwater inputs to estuaries, and groundwater recharge. As a related topic, we also showcase the value of foundation AI for global environmental change and its benefits for resource management.

Kathe Todd-Brown
University of Florida

Global soil carbon potential – What if everyone is right and where do we go from here? There has been an explosion of interest in soil carbon sequestration as a natural carbon reduction strategy. Soil carbon stocks are an appealing reservoir for sequestering anthropogenic carbon dioxide due to their relatively low risk, low technological barrier, and potential for long residence time. But how much carbon sequestration potential is there globally? Where are the places soils currently accumulating anthropogenic carbon dioxide and would these locations lend themselves to more active management interventions? Two very contrasting approaches are being taken in soil science: digital soil mapping approaches rooted in soil carbon stock surveys and machine learning, as well as process models rooted in soil carbon flux studies and differential equations. In this talk we’ll explore where these representations deviate, how they could be reconciled, and how we can use modeling as a tool to expand our understanding of soil carbon potential for carbon dioxide draw down in the future.

Manzhu Yu
Penn State University

Spatiotemporal methodologies for the analysis and prediction of extreme weather events A better understanding of drivers and processes will improve the prediction of extreme weather events and will support process-based representation of weather and climate extremes in climate model simulations. The increasing availability of observational, simulation, and user-generated (e.g., social media or crowdsourced) datasets, along with the rapid progress of computing technologies, has provided us the unprecedented opportunity to enhance the understanding and predictability of extreme weather events. My research centers on developing spatiotemporal methodologies for the analysis and prediction of extreme weather events, such as dust storms, hurricanes, and extreme heat. In this talk, I will demonstrate several case studies from my research to 1) understand the spatiotemporal dynamics of extreme weather events, 2) explore the relationship of these events with other physical and social factors, and 3) integrate heterogeneous data to enhance the predictability, response, and mitigation of extreme weather events.

Clinics

Wolfgang Bangerth
Colorado State University

The finite element method (FEM) There are in essence three comprehensive frameworks to numerically approximate the solution of (i.e., "solve") partial differential equations: The finite difference method (FDM), the finite volume method (FVM), and the finite element method (FEM). There have been clinics about the FDM and FVM at past CSDMS Annual Meetings; this clinic will introduce the FEM. Specifically, I will (i) outline the philosophical idea behind the FEM, (ii) how that differs from the FDM and the FVM, (iii) how the FEM is described mathematically, and (iv) how this translates into implementations.For the FEM, there exist a number of large, widely used, well documented, and extensively tested open source libraries that have been used for essentially all partial differential equations one can think of. I will base this clinic on lectures I have been using to teach the FEM for the past 25 years, utilizing in part the open source library deal.II (https://www.dealii.org/) of which I am one of the principal authors.

Caitlin Haedrich
North Carolina State University

An Introduction to GRASS GIS and Tangible Landscape This hands-on clinic will introduce participants to GRASS GIS, an open-source geospatial processing engine, and Tangible Landscape, a tangible user interface for GRASS GIS. We will explain and practice GRASS GIS concepts, and work through example Python-based workflows for topics such as hydrology, flood modeling, and viewshed analysis. These workflows will be implemented as a series of computational notebooks. Then, we will show how these workflows can be configured as activities on Tangible Landscape. Using GRASS GIS as a backend, Tangible Landscape is an interactive, open-source platform that integrates physical sand models of landscapes with digital simulations by using a scanner (xBox Kinect) and projector. It allows users to interact in real-time with models by, for example, carving the sand and seeing the resulting water flow pattern. By the end of the clinic, participants will have hands-on experience with: - Setting up GRASS projects and importing data - Visibility analysis - Configuring and running overland flow models - Creating timeseries of inundation flooding - Building Tangible Landscape activities

Allen Lee
Arizona State University

Get lazy with LLMs Large Language Models (LLMs) have been rolling on the hype train since ChatGPT-3.5 was first introduced in November 2022. Hop onboard with an interactive hands-on clinic where we’ll collectively explore how to use LLMs to be more efficient (aka lazy) in our research software development and computational modeling work.Participants will engage in interactive prompt engineering with commercial LLMs (e.g., ChatGPT, perplexity, Google NotebookLM). We will explore how to fine tune prompts to learn about new topics, summarize, assess, or evaluate texts, and generate possibly useful research software artifacts: documentation, tests, containerization recipes, shell scripts, etc. If all goes well we hope to develop a curated LLM prompts / recipes repository tailored to the CSDMS / CoMSES communities with all of your properly credited contributions.All are welcome from LLM novices to seasoned prompt engineer pros. Bring the burning questions you haven’t had the energy to figure out and we’ll see what the allegedly hitherto sum of human knowledge has to statistically say about it.

Andrew Moodie
Texas A&M University

Using the collaborative sandpiper toolchain to support interoperability in geomorphology research Current software engineering and data management practices amongst different research teams impede collaboration in geomorphology. For example, researchers who create software tools often do not document them, so the tools do not port easily to new systems. Often, tools go unmaintained after publication, so other teams that want to use the tool or conduct the same analysis will rewrite the software rather than reuse the existing code. This clinic will demonstrate several advances of the recently-launched sandpiper toolchain initiative that facilitate data reuse and research team collaboration, and reduce research effort duplication. sandpiper is forging a data standard for regularly-gridded three dimensional data (i.e., time and two spatial dimensions), and building a software package for data analysis in Earth surface processes research. In this clinic, we will show the features of the data standard, how to create datasets that are compliant with the standard, and how existing datasets can be “rescued” and made findable and reusable. We will also demonstrate the analysis software package, and how it is being used for research. Importantly, sandpiper is a growing community of users, and we want you to join. Bring your data problems, and help us build solutions that work for the whole community.

Irina Overeem
University of Colorado, Boulder

Accelerating Glacier and Surface Processes Modeling with Machine Learning and New Python Libraries Join us for a hands-on clinic exploring the intersection of glacier mass balance, glacier dynamics, and surface processes modeling. We will discuss recent python libraries to model glacier processes for surface processes applications. We will introduce the Instructed Glacier Model (IGM), a machine learning-based glacier dynamics emulator, it provides significant speed-up while maintaining high process accuracy. IGM opens up new research possibilities for longterm, landscape scale simulations. We will demonstrate several applications of combined glacier and surface processes modeling. We will then proceed to run a tutorial on running combined models of glacier and sedimentary processes using these existing python libraries. This session is targeted to researchers interested in glacier impacts on downstream landscapes, glacial geomorphology, and integrating new python libraries/glacial models into their earth surface processes modeling research.

Allison Pfeiffer
Western Washington University, Geology

Landlab’s NetworkSedimentTransporter: A Lagrangian model for riverbed material transport dynamics This hands-on clinic will introduce Landlab’s NetworkSedimentTransporter (NST) component, giving participants an overview of the workings of the model and a chance to explore a simple use case. The NST is a 1D morphodynamic model for gravel riverbed evolution that allows for the Lagrangian tracking of sediment. The NST operates on a network model grid, tracking the transport of individual ‘parcels’ of sediment as they pass downstream according to sediment transport equations. As these parcels move from one link (reach) to another, the grid topographic elevation and bed surface grain size distributions evolve. By the end of the clinic, participants should be able to: - Explain how sediment is represented as discrete parcels that are tracked as they transport through the network model grid- Explain how sediment parcels are transported within a link and how the model selects which parcels to transport or store- Explain how grid topographic elevations evolve as a function of the volume of parcels present on the neighboring linksUnderstand the purpose of each of the essential sections of the code needed to run the NST - Make minor edits to an example NST model script to explore model sensitivity to input parameters- Propose several possible use cases of the NSTThis clinic is appropriate for beginners with little to no experience using the Landlab library. Participants will run the model on the OpenEarthscape JupyterHub (on the cloud), so there are no specialized software requirements. Prior experience with Python programming and some knowledge of sediment transport equations is helpful but not necessary.

Steve Railsback
Lang Railsback & Associates

Pattern-oriented agent-based modeling to achieve structural realism and testable predictions Both too simple and too complex models have limited payoff in understanding real systems and making reliable inferences. Pattern-oriented modeling (POM) is a strategy to find the right intermediate level of complexity. It is based on the goal of making a model simultaneously reproduce multiple patterns that have been observed at different scales and levels of organization. The patterns are used as multiple criteria for model design, selection, and parameterization. POM was developed for agent-based models, but can be used for any model type. We will introduce POM using examples and conceptual exercises, and participants will conduct an exercise, using a NetLogo model we provide, to test how well alternative versions of a key submodel cause the model to reproduce observed patterns. Participants should bring a computer with NetLogo (version 6.0 or later; http://ccl.northwestern.edu/netlogo) and software for graphical and statistical analysis of results (e.g., Excel, R).

Mohamed Said
Florida Gulf Coast University

Integrating CNN with LSTM Models for Capturing Spatiotemporal Dynamics of Flood Modeling This clinic will introduce the concept and implementation of a hybrid Deep Learning (DL) framework, that integrates: Convolutional Neural Network (CNN) and Long-short Term Model (LSTM), to simulate two-dimensional flood scenarios. This advanced DL framework enables spatiotemporal predictions of hydrodynamic parameters, with a particular focus on predicting water depths of flood events. Participants will first acquire a brief introduction to both models and their integration concept, followed by hands-on experience in developing a simple hybrid model using the PyTorch library. The training process will utilize a small-scale 2D flume as a practical and time-efficient exercise; however, this technique can be scaled up and applied to large stream segments in real-world scenarios. The workshop will highlight the capabilities, applications, and best practices of the developed model within the water resources domain. Notably, it will showcase the DL models’ ability to generate predictions significantly faster than traditional hydrodynamic models like HEC-RAS, which face substantial computational challenges in simulating 2D flood scenarios, especially for large-scale or long-term simulations. This workshop will equip the participants with the necessary background and technical skills for various spatiotemporal applications, encompassing geomorphological processes, groundwater dynamics, and wave-driven simulations. For this workshop, a basic understanding of machine learning in Python is recommended to maximize the benefits of this session.

Daniel Shapero
University of Washington

Simulating glacier flow with ICEPACK This clinic will describe a bit about glacier physics and then go on to show how to use the software package ICEPACK to simulate the flow of glaciers and ice sheets. We'll run through a few use cases, first on synthetic and then on real data. Then I'll show a few other useful features, like the ability to easily alter parts of the model physics like the sliding law. Finally I'll talk about some open problems and other things we're working on that might be of relevance to other disciplines within the earth sciences.

Moira Zellner
Northeastern University

Fora.ai: Reshaping collaboration for climate and social impact Fora.ai is an intuitive digital environment that enables diverse stakeholder groups to collaboratively interact with embedded simulation models to understand real world socio-environmental problems and create novel and impactful solutions. Stakeholders interact with this digital representation and with each other, iteratively creating, revising and testing solutions until diverse needs are addressed. Workshop participants will use fora.ai’s interactive game-board to collectively build green infrastructure solutions to flooding in a neighborhood in Chelsea, Massachusetts. The virtual environment allows for participation in a facilitated process in which users will: 1) input their individual priorities, 2) collaboratively run simulations to understand flooding issues in the neighborhood, 3) co-design green infrastructure scenarios to address these problems, 4) see how their changes affect the simulation, and 5) deliberate on the tradeoffs that arise from each solution due to competing priorities. Participants will be introduced to the flooding model and, with facilitator assistance, engage in multiple iterations of the process of prioritization, solution-building, and reflection on results. This process will allow them to refine their proposed solutions towards a design they would jointly support for implementation, with an understanding of its benefits and drawbacks. The workshop will end with a focus group debrief. Laptops or tablets required.

Interested in providing a clinic during the next annual meeting? Contact CSDMS@Colorado.EDU.

Participants

Who is registered as of 01/21/2025?

Conference Venue

This year the conference will be held at SEEC at the University of Colorado, Boulder, Colorado.

Address:
University of Colorado
SEEC Building
C120, Auditorium
4001 Discovery Drive
Boulder, CO 80303

Conference Lodging

Poster guidelines

The poster boards are configured for up to 46" wide by 60" tall (portrait orientation) posters (116 cm wide by 152 cm tall). Anything larger than these dimensions will reduce the space of your colleagues so please be respectfull of these poster dimensions.

Electronic publications (Epubs)

We're excited to announce that we will also offer Epub submissions for this year's annual meeting. Epubs are Jupyter notebooks that contain a scientific problem description and some findings that are investigated by numerical algorithms or models, that walks the reader through the science by executing the algorithms or models. The Epubs will be reviewed and can be part of a poster or oral presentation or separately submitted. Guidelines on what the notebook should include can be found here.

Travel Scholarships

Applications due by February 9, 2025
This year CSDMS is offering a limited number of travel scholarships for graduate students, post-docs, early career faculty, and faculty from minority-serving institutions to attend the CSDMS annual meeting. A number of these scholarships will be offered for the purpose of increasing participation of underrepresented students. To be eligible, applicants need to meet the following requirements:

Attend the whole meeting (May 13-15, 2025) at the University of Colorado-Boulder, Colorado
Submit an abstract for and provide a poster presentation at the meeting (this requirement may be waived under limited conditions, i.e. 1st year graduate student that has not started their research, etc.)
Submit a letter of motivation that states why you wish to participate in the meeting and explain how/if your participation would enhance diversity in the field of surface dynamics modeling.

The CSDMS travel scholarships will cover:

Registration costs (to be reimbursed after attending the meeting)
Travel (for US participants airfare and local transport up to $600, for international participants up to $1,200 of transportation costs will be reimbursed)
Per diem to help reimburse the cost of meals from 13-15 May 2025 not offered in the conference schedule
Shared lodging in the conference hotel for the evenings of May 12th, 13th and 14th.

Please submit your letter of motivation and contact information to csdms@colorado.edu by February 9, 2025. Applicants will be notified of the decision by the end of February.

Transportation to and from Boulder

We encourage all who are able, to take advantage of public transportation or ride-share services between the Denver International Airport and Boulder.

RTD bus to Boulder takes less then 1.5 hours and costs $10.00 each way. Trip planner here: https://www.rtd-denver.com/app/plan
Eight Black Airport shuttle service takes about 45 minutes and costs about $64.00 each way. Reserve here: https://eightblackairportshuttle.com/

Land Acknowledgement

We acknowledge that the land, on which we will hold our meeting in Boulder, Colorado, is part of the land within the Traditional Territories of the Arapaho, Cheyenne, and Ute peoples. Further, we acknowledge that 48 contemporary tribal nations are historically tied to the lands that make up the state of Colorado. For more information, see the official CU-Boulder Acknowledgement.

Code of Conduct

CSDMS is committed to fostering a professional, respectful, and inclusive environment at the annual meeting, such that all participants can participate to the fullest in a welcoming, respectful, inclusive, and collaborative environment that is free of harassment and discrimination. CSDMS expects all participants and staff to comply with this code of conduct, as outlined at CSDMS code of conduct.

Important dates

January 24: Application deadline Student Modeler Award 2025
February 9: Application deadline travel scholarships
April 1: Abstract submission deadline
April 1: Meeting registration deadline
May 13-15: CSDMS annual meeting
May 16, 9AM to 11AM: CSDMS Executive committee meeting (by invitation only)
May 16, 12PM to 2PM: CSDMS Steering committee meeting (by invitation only)