Property:CSDMS meeting abstract presentation
From CSDMS
This is a property of type Text.
P
Hydrology is a science of extremes; droughts and floods. In either case, the hydrologic response arises from the combination of many factors, such as terrain, land cover, land use, infrastructure, etc. Each has different, overlapping spatial domains. Superimposed upon these are temporal variations, driven by stochastic weather events that follow seasonal climatic regimes. To calculate risk (expected loss) requires a loss function (damage) and a response domain (flood depths) over which that loss is integrated. The watershed provides the spatial domain that collects all these factors. This talk will discuss the data used to characterize hydrologic response. +
I will discuss an application of the Migration, Intensification, and Diversification as Adaptive Strategies (MIDAS) agent-based modeling framework to modeling labor migration across Bangladesh under the stressor of sea-level rise (SLR). With this example, I hope to highlight some hard-to-resolve challenges in representing adaptive decision-making under as-yet unexperienced stressors in models. Drawing together what is more and what is less known in projections for future adaptation, I will discuss strategies for ‘responsible’ presentation and dissemination of model findings. +
If one system comes to (my) mind where the human element is intertwined with the environment, it is the Louisiana coastal area in the Southern United States. Often referred to as the working coast, coastal Louisiana supports large industries with its ports, navigation channels, oil, and productive fisheries. In addition to that, Louisianians have a significant cultural connection to the coastal wetlands and their natural resources. Unfortunately, the land is disappearing into the sea with coastal erosion rates higher than anywhere else in the US. Due to these high rates of land loss, this system needs rigorous protection and restoration. While the restoration plans are mostly focused on building land, the effects on, for example, fisheries of proposed strategies should be estimated as well before decisions can be made on how to move forward. Through several projects I have been involved in, from small modeling projects to bold coastal design programs, I present how coupled models play a key role in science-based coastal management that considers the natural processes as well as the human element. +
Image recognition is a powerful application of machine learning (ML) where computers can learn to automatically identify objects, patterns, and more. Meanwhile, there are enormous volumes of satellite imagery being collected every day with a variety of important landscape features readily visible. Though the name "image recognition" sounds like it's just based on visual data, modern ML methods allow many types of data to be included in the "image" - including full multispectral raster stacks and digital elevation models. If a data type can be converted to a raster, then ML image recognition can learn from it and recognize patterns in it. In this clinic, we will cover how to get started using ML to detect interesting landscape features in remotely sensed imagery using beaver dam identification as a case study. +
In dry regions, escarpments are key landforms for exploring landform-rainfall interactions. Here we present a modeling approach for arid cliffs and sub-cliff slopes evolution incorporating rainfall forcing at the scale of individual rainstorms. We used numerical experiments to mechanistically test how arid cliffs and sub-cliff slopes evolve according to different geomorphic characteristics and variations in rainstorm properties. +
In formulating tectono-geomorphic models of landscape evolution, Earth is typically divided into two domains; the surface domain in which “geomorphic” processes are solved for and a tectonic domain of earth deformation driven generally by differential plate movements. Here we present a single mechanical framework, Failure Earth Response Model (FERM), that unifies the physical description of dynamics within and between the two domains. FERM is constructed on the two, basic assumptions about the three-dimensional stress state and rheological memory: I) Material displacement, whether tectonic or geomorphic in origin, at or below Earth’s surface, is driven by local forces overcoming local resistance, and II) Large displacements, whether tectonic or geomorphic in origin, irreversibly alter Earth material properties enhancing a long term strain memory mapped into the topography. In addition to the gathering of stresses arising from far field tectonic processes, topographic relief, and the inertial surface processes into a single stress state for every point, the FERM formulation allows explicit consideration of the contributions to the evolving landscape of pore pressure fluctuations, seismic accelerations, and fault damage. Incorporation of these in the FERM model significantly influences the tempo of landscape evolution and leads to highly heterogeneous and anisotropic stress and strength patterns, largely predictable from knowledge of mantle kinematics. The resulting unified description permits exploration of surface-tectonic interactions from outcrop to orogen scales and allows elucidation of the high fidelity orogenic strain and climate memory contained in topography. +
In landscape evolution models, climate change is often assumed to be synonymous with changes in rainfall. In many climate changes, however, the dominant driver of landscape evolution is changes in vegetation cover. In this talk I review case studies that attempt to quantify the impact of vegetation changes on landscape evolution, including examples from hillslope/colluvial, fluvial, and aolian environments, spatial scales of ~10 m to whole continents, and time scales from decadal to millennial. Particular attention is paid to how to parameterize models using paleoclimatic and remote sensing data. +
In response to the CSDMS community’s interest, the Human Dimensions group is excited to host a virtual Coffee Hour on community engagement in earth systems science and policy projects. Please join us for our first Coffee Hour, which will include an engaging panel on the topic: “Engaging diverse stakeholders in earth-systems modeling projects.” We recognize the importance of working collaboratively with stakeholders in scientific projects (e.g., for knowledge co-creation, for guidance, and for implementation of solutions derived from the research), but we are not traditionally trained to do so. Rigorous scientific practices can sometimes be alienating and extractive, eroding the trust between the scientific community and the public that is necessary for the advancement of science, policy, and human wellbeing. We discuss here the challenges involved in community engagement and possible ways to overcome them.
Our panelists are Leilah Lyons, NSF, Laura Schmitt Olabisi, Michigan State University and Mehana Vaughan, University of Hawaii. The Coffee Hour will begin with a short introduction by each panelist, followed by a set of questions by the facilitators, and concluding with a period of open questions and discussion with the audience. +
In response to the CSDMS community’s interest, the Human Dimensions group is excited to host a virtual Coffee Hour on community engagement in earth systems science and policy projects. Please join us for our first Coffee Hour, which will include an engaging panel on the topic: “Engaging diverse stakeholders in earth-systems modeling projects.” We recognize the importance of working collaboratively with stakeholders in scientific projects (e.g., for knowledge co-creation, for guidance, and for implementation of solutions derived from the research), but we are not traditionally trained to do so. Rigorous scientific practices can sometimes be alienating and extractive, eroding the trust between the scientific community and the public that is necessary for the advancement of science, policy, and human wellbeing. We discuss here the challenges involved in community engagement and possible ways to overcome them.
Our panelists are Gabriela Garcia, Northeastern University, Miyuki Hino, University of North Carolina at Chapel Hill, and Theo Lim, University of British Columbia. The Coffee Hour will begin with a short introduction by each panelist, followed by a set of questions by the facilitators, and concluding with a period of open questions and discussion with the audience +
In software engineering, an interface is a group of functions with prescribed names, argument types, and return types. When a developer implements an interface for a piece of software, they fill out the details for each function while keeping the signatures intact. CSDMS has developed the Basic Model Interface (BMI) for facilitating the conversion of a model written in C, C++, Fortran, Python, or Java into a reusable, plug-and-play component. By design, BMI functions are simple. However, when trying to implement them, the devil is often in the details.
In this hands-on clinic, we'll take a simple model of the two-dimensional heat equation, written in Python, and together we'll write the BMI functions to wrap it, preparing it for transformation into a component. As we develop, we’ll explore how to use the wrapped model with a Jupyter Notebook.
To get the most out of this clinic, come prepared to code! We'll have a lot to write in the time allotted for the clinic. We recommend that clinic attendees have a laptop with the Anaconda Python distribution installed. We also request that you review the
* BMI description (http://csdms.colorado.edu/wiki/BMI_Description), and the
* BMI documentation (https://bmi-spec.readthedocs.io)
before the start of the clinic. +
In software engineering, an interface is a set of functions with prescribed names, argument types, and return types. When a developer implements an interface for a piece of software, they fill out the details for each function while keeping the signatures intact. CSDMS has developed the Basic Model Interface (BMI) for facilitating the conversion of an existing model written in C, C++, Fortran, Python or Java into a reusable, plug-and-play component. By design, BMI functions are straightforward to implement. However, when trying to match BMI functions to model behaviors, the devil is often in the details.<br>In this hands-on clinic, we'll take a simple model--an implementation of the two-dimensional heat equation in Python--and together, we'll write the BMI functions to wrap it, preparing it for transformation into a component. As we develop, we’ll explore how to use the wrapped model with a Jupyter Notebook.<br>To get the most out of this clinic, come prepared to code! We'll have a lot to write in the time allotted for the clinic. We recommend that clinic attendees have a laptop with the Anaconda Python distribution installed. We also request that you read over:<br>BMI description (https://csdms.colorado.edu/wiki/BMI_Description)<br>BMI documentation (http://bmi-python.readthedocs.io)<br>before participating in the clinic. +
In soil-covered landscapes, the genesis, weathering, and erosion of soils are an integral part of both biogeochemical cycles and landscapes’ morphologic evolution. However, the intimate feedback between landscape biogeochemistry and morphology through soils has not been frequently described and predicted through mathematical models. Nonetheless, a growing number of critical earth science questions, which are central to understanding climate change, ecosystem dynamics, and agricultural sustainability, hinge on integrating the two aspects of soil-covered landscapes. Here, we present three studies that explicitly, quantitatively, and mechanistically consider the connections. The first case demonstrates how the modeling of soil erosion and soil production can contribute to re-interpreting radiocarbon data. Once physical movements of soil constituents – a topic of soil geomorphology – is incorporated into interpreting radiocarbon age – a topic of soil carbon - a new insight on the biological sensitivity of soil carbon decomposition is gained. The second ongoing study zooms into the feedback between soil carbon, bulk density, and soil faunal activities, illustrating their co-impacts on landscape morphology and hydrology. This example utilizes the data from the earthworm invasion transects in Minnesota sugar maple forests. Lastly, the significance of explicitly incorporating agricultural activities in geomorphic modeling of rugged but soil-mantled landscapes is acknowledged. Using our mapping exercise from the Upper Mississippi Valley, where its steep slopes and deep valleys sharply contrast with the surrounding landscapes in Minnesota, Wisconsin, Iowa, and Illinois and where smallholder farmers are concentrated, we show how agricultural land uses reflect geomorphic process domains determined by landscape morphology and associated soil properties. These examples stress that disciplines of soil biogeochemistry and geomorphology are mature enough to be quantitatively and mechanistically integrated and that such integration can help us better understand the pressing issues related to climate change and water resources and improve the relevance of earth surface science to the broader groups around the world.
In the modeler community, hindcasting (a way to test models based on knowledge of past events) is required for all computer models before providing reliable results to users. CSDMS 2.0 “Moving forward” has proposed to incorporate benchmarking data into its modeling framework. Data collection in natural systems has been significantly advanced, but is still behind the resolution in time and space and includes natural variability beyond our understanding, which makes thorough testing of computer models difficult.<br><br>In the experimentalist community, research in Earth-surface processes and subsurface stratal development is in a data-rich era with rapid expansion of high-resolution, digitally based data sets that were not available even a few years ago. Millions of dollars has been spent to build and renovate flume laboratories. Advanced technologies and methodologies in experiment allow more number of sophisticated experiments in large scales at fine details. Joint effort between modelers and experimentalists is a natural step toward a great synergy between both communities.<br><br>Time for a coherent effort for building a strong global research network for these two communities is now. First, the both communities should initiate an effort to figure out a best practice, metadata for standardized data collection. Sediment experimentalists are an example community in the “long tail”, meaning that their data are often collected in one-of-a-kind experimental set-ups and isolated from other experiments. Second, there should be a centralized knowledge base (web-based repository for data and technology) easily accessible to modelers and experimentalists. Experimentalists also have a lot of “dark data,” data that are difficult or impossible to access through the Internet. This effort will result in tremendous opportunities for productive collaborations.<br><br>The new experimentalist and modeler network will be able to achieve the CSDMS current goal by providing high quality benchmark datasets that are well documented and easily accessible.
In this clinic I will give an overview of lsdtopotools so that, by the end of the session, you will be able to run and visualise topographic analyses using lsdtopotools and lsdviztools. I will show how to start an lsdtopotools session in google colab in under 4 minutes, and will also give a brief overview for more advanced users of how to use our docker container if you want access to local files. I will then use jupyter notebooks to give example analyses including simple data fetching and hillshading, basin selection, simple topographic metrics and channel extraction. Depending on the audience I will show examples of a) channel steepness analysis for applications in tectonic geomorphology b) calculation of inferred erosion rates based on detrital CRN concentrations c) terrace and valley extraction d) channel-hillslope coupling. In addition I will show our simple visualisation scripts that allow you to generate publication-ready images. All you need prior to the session is a google account that allows you to access colab, and an opentopography account so you can obtain an API key. The latter is not required but will make the session more fun as you can use data from anywhere rather than example datasets. If you are not an advanced user please do not read the next sentence, as you don’t need it and it is nerdy compu-jargon that will put you off the session. If you are an advanced user and wish to try the docker container you should install the docker client for your operating system and use the command “docker pull lsdtopotools/lsdtt_pytools_docker” when you have access to a fast internet connection. +
In this clinic we will explore how to use the new cloud-based remote sensing platform from Google. Our hands-on clinic will teach you the basics of loading and visualizing data in Earth Engine, sorting through data, and creating different types of composite images. These techniques are a good starting point for more detailed investigations that monitor changes on earth’s surface. Prerequisites:<br>1) Bring your own laptop.<br>2) Chrome installed on your system: It will work with Firefox but has issues.<br>3) An active Google account - Register for an account with Google Earth Engine (https://earthengine.google.com/signup/) +
In this clinic we will explore how to use the cloud-based remote sensing platform from Google. Our hands-on clinic will teach you the basics of loading and visualizing data in Earth Engine, sorting through data, and creating different types of composite images. These techniques are a good starting point for more detailed investigations that monitor changes on earth’s surface. Prerequisites include having Chrome installed on your system: It will work with Firefox but has issues and an active Google account. Once you have those please register for an account with Google Earth Engine (https://earthengine.google.com/signup/) +
In this clinic we will first review concepts of glacial isostatic adjustment and the algorithm that is used to solve the sea level equation. We will then provide an overview of the sea level code, which calculates the viscoelastic response of the solid Earth, Earth’s gravity field, and rotation axis to changes in surface load while conserving water between ice sheets and oceans. Participants will run the code, explore manipulating the input ice changes, and investigate its effect on the predicted changes in sea level, solid Earth deformation, and gravity field. +
In this clinic, we will explore RivGraph, a Python package for extracting and analyzing fluvial channel networks from binary masks. We will first look at some background and motivation for RivGraph's development, including some examples demonstrating how RivGraph provides the required information for building models, developing new metrics, analyzing model outputs, and testing hypotheses about river network structure. We will then cover--at a high level--some of the logic behind RivGraph's functions. The final portion of this clinic will be spent working through examples showing how to process a delta and a braided river with RivGraph and visualizing results.
Please note: This clinic is designed to be accessible to novice Python users, but those with no Python experience may also find value. If you'd like to work through the examples during the workshop, please install RivGraph beforehand, preferably to a fresh Anaconda environment. Instructions can be found here: https://github.com/jonschwenk/RivGraph. It is also recommended that you have a GIS (e.g. QGIS) available for use for easy display/interrogation of results. +
In this clinic, we will first demonstrate existing interactive computer-based activities used for teaching concepts in sedimentology and stratigraphy. This will be followed by a hands-on session for creating different modules based on the participants’ teaching and research interests. Active learning strategies improve student exam performance, engagement, attitudes, thinking, writing, self-reported participation and interest, and help students become better acquainted with one another (Prince, 2004). Specifically, computer-based active learning is an attractive educational approach for post-secondary educators, because developing these activities takes advantage of existing knowledge and skills the educator is likely to already have.
The demonstration portion of the clinic will focus on the existing rivers2stratigraphy (https://github.com/sededu/rivers2stratigraphy) activity, which illustrates basin-scale development of fluvial stratigraphy through adjustments in system kinematics including sandy channel migration and subsidence rates. The activity allows users to change these system properties, so as to drive changing depositional patterns. The module utilizes a rules based model, which produces realistic channel patterns, but simplifies the simulation to run efficiently, in real-time. The clinic will couple rivers2stratigraphy to a conventional laboratory activity which interprets an outcrop photograph of fluvial stratigraphy, and discuss logistics of using the module in the classroom.
For the second part of the clinic, familiarity with Python will be beneficial (but is not required); we will utilize existing graphical user interface (GUI) frameworks in developing new activities, aimed to provide a user-friendly means for students to interact with model codes while engaging in geological learning. Participants should plan to have Python installed on their personal computers prior to the workshop, and a sample module will be emailed beforehand to let participants begin exploring the syllabus.
''Prince, M. (2004). Does Active Learning Work? A Review of the Research. Journal of Engineering Education, 93(3), 223-231. doi: 10.1002/j.2168-9830.2004.tb00809.x''.
In this clinic, we will introduce and experiment with open-source tools designed to promote rapid hypothesis testing for river delta studies. We will show how pyDeltaRCM, a flexible Python model for simulating river delta evolution, can be extended to incorporate any arbitrary processes or forcings. We will highlight how object-oriented model design enables community-driven model development, and how this promotes reproducible science. Our clinic will develop an extended model to simulate deltaic evolution into receiving basins with different slopes. Then, the clinic will step through some basic analyses of the model runs, interrogating both surface processes and subsurface structure. Our overall goal is to familiarize you with the tools we are developing and introduce our approach to software design, so that you may adopt these tools or strategies in your research.
Please note that familiarity with Python will be beneficial for this clinic, but is not required. Hands-on examples will be made available via an online programming environment (Google CoLab or similar); instructions for local installation on personal computers will be provided prior to the workshop as well. +
