Workshop Details - CUAHSI Water Science Exchange, Fall 2026

Date:
November 12^th, 2026

Time:
3:30 pm | 2 hours

Audience Type:
Open to all attendees. No AI or coding experience is required.

Capacity:
35

Description:
Artificial intelligence, particularly large language models (LLMs), is rapidly becoming part of hydrologic research, environmental analysis, scientific communication, and water-resource decision-making. These tools can accelerate literature synthesis, data interpretation, coding, hypothesis development, and model conceptualization. However, their effective use in water science requires more than familiarity with AI platforms. Credible applications depend on domain expertise, structured problem formulation, transparent reasoning, critical evaluation, and meaningful human oversight. This interactive workshop will introduce participants to a human-centered approach to AI in water science. Rather than treating LLMs as tools that simply generate answers, participants will learn to use them as collaborative systems for exploring ideas, identifying assumptions, developing hypotheses, organizing scientific workflows, and critically examining alternative interpretations.

Participants will work individually and in small groups on realistic water-science problems such as water-quality assessment, drought and flood analysis, watershed management, nutrient transport, groundwater sustainability, and conservation planning. Guided exercises will demonstrate how prompt structure, scientific context, iterative questioning, and expert judgment influence the quality and reliability of AI-generated outputs. The workshop will also examine common limitations of LLMs, including hallucination, unsupported citations, overconfidence, bias, loss of scientific context, and difficulties with causal interpretation. Particular attention will be given to identifying situations in which AI-generated information must be independently validated using observations, established scientific knowledge, models, or peer-reviewed evidence.

Date:
November 12^th, 2026

Time:
3:30 pm | 2 hours

Audience Type:
Open to graduate students, postdoctoral researchers, and early-career professionals with hydrologic or environmental training and limited formal remote sensing background. No prior experience with foundation models, Harmonized Landsat Sentinel-2, or Google Earth Engine is expected.

Prerequisites:
Working familiarity with Python (able to read and modify a provided notebook) and basic raster and geospatial concepts.
Free HydroShare, GitHub, and NASA Earthdata Login accounts, plus a Google account with Google Earth Engine enabled on a Cloud project registered for noncommercial use. This registration must be completed before the session; setup instructions and an optional setup clinic will be offered.

Capacity:
25

Description:
Land cover is the variable through which working landscapes act on water. Crop type, planting date, tillage practice, and cover crop presence govern infiltration, evapotranspiration, runoff timing, and nutrient export, yet water scientists often treat land cover as a fixed input rather than a derived and uncertain product. This session makes that layer explicit and puts its construction in the participants’ hands. Working with Harmonized Landsat Sentinel-2 imagery over a Midwest agricultural landscape, participants run inference with an open geospatial foundation model from the IBM and NASA Prithvi family, then evaluate the resulting classification against USDA Cropland Data Layer reference labels. The session closes by ingesting the finished map into Google Earth Engine and publishing it as an interactive web map that opens in any browser without an Earth Engine account. All computation runs in the cloud, and all notebooks and datasets are published openly on HydroShare. The intended outcome is concrete: each participant leaves with a crop map they produced themselves and a public link to it, suitable for a thesis chapter, an outreach conversation, or a stakeholder meeting.

Session flow:

Model input and pre-staged data. Harmonized Landsat Sentinel-2 imagery as foundation model input: band composition, image chips, and growing-season time steps. Reference labels from the USDA Cropland Data Layer, and their known limitations.
Inference and evaluation. The IBM and NASA Prithvi model family: pretraining on harmonized imagery, available checkpoints, and the multi-temporal crop classification fine-tune. Running inference over a Midwest scene and assembling a georeferenced output. Accuracy assessment against held-out labels, reported per class.
Publication. Exporting a cloud optimized GeoTIFF. Ingesting the result as a Google Earth Engine asset and setting sharing permissions. Publishing the map from a supplied application template, with layer styling, legend, and inspector. What a public map URL means for reproducibility and outreach.

Date:
November 12^th, 2026

Time:
3:30 pm | 2 hours

Prerequisites: A HydroShare account. If you don’t have one, create one here.

Managing research data well is a skill that pays off long before publication and long after it. This workshop walks researchers through how HydroShare supports the full research data lifecycle, from initial planning and collection through description, preservation, discovery, and integration.
Participants will learn how to collaborate with co-authors and team members on shared HydroShare resources, adopt best practices for file naming and data organization, and grant reviewers access to data and code during the publication process. Whether you're preparing to share a dataset for the first time or looking to streamline how your team manages files together, this session offers practical tools for making collaborative data management easier.

Date:
November 14^th, 2026

Time:
Module 1
8:30 am | 2 hours

Module 2
10:40 am | 2 hours

Module 3
1:30 pm | 2 hours

Audience Type:
Open to graduate students, postdoctoral researchers, and early-career professionals in hydrology, meteorology, or a related field. Basic familiarity with Python and Jupyter notebooks is required.

Capacity:
30

Description:
This workshop guides participants through a complete flash flood prediction workflow, beginning with the estimation of precipitation forcing, continuing through the distributed simulation of rapid runoff response, and concluding with the quantitative evaluation of forecast skill. Flash floods are the deadliest form of flooding in the United States, and they are particularly consequential in working landscapes, where tile drainage, terraces, grazed uplands, and managed channels reshape the timing and magnitude of runoff at spatial scales that continental-scale operational models resolve poorly. Each module pairs a short conceptual presentation with a guided computational exercise. All three modules operate on a common set of documented flash flood events, so that decisions made in the precipitation module propagate visibly into the simulated hydrographs and, in turn, into the verification statistics computed in the final module. Every notebook, configuration file, and dataset is published as an open HydroShare resource, so participants can reproduce the full workflow independently after the meeting.

Module 1. Quantitative Precipitation Estimation from Radar and Satellite (2 hours). Retrieval principles and sampling assumptions behind ground-based radar and spaceborne precipitation estimates; the structure, latency, and product tiers of the Multi-Radar Multi-Sensor system (MRMS) and the NASA Integrated Multi-satellitE Retrievals for GPM (IMERG); accessing and subsetting both archives in the cloud; regridding and temporal aggregation onto a common analysis grid; comparison against rain gauge observations; and diagnosis of the detection, systematic, and random error components that each product carries into downstream simulation.

Module 2. Distributed Hydrologic Simulation of Flash Flood Events (2 hours). Formulation of distributed rainfall-runoff models designed for rapidly responding basins; preparation of terrain, flow direction, flow accumulation, and parameter grids; configuration of the Ensemble Framework For Flash Flood Forecasting (EF5), including its water balance components and kinematic wave routing; execution of event simulations forced alternately by radar and satellite precipitation; and interpretation of gridded discharge, unit streamflow, and return-period exceedance fields as flash flood indicators.
Module 3. Event-Based Model Evaluation and Forecast Skill (2 hours). Evaluation design for short-duration, spatially localized events; continuous hydrograph metrics, including Nash-Sutcliffe efficiency and the Kling-Gupta efficiency with its decomposition into correlation, bias, and variability terms; peak magnitude and peak timing errors; contingency-table verification of threshold exceedance using probability of detection, false alarm ratio, critical success index, and frequency bias; construction of reference baselines and interpretation of the resulting skill scores; and attribution of simulation error among precipitation forcing, model structure, and parameter selection.

Date:
November 14^th, 2026

Time:
8:30 am | 3.5 hours

Audience Type:
Open to all attendees, no coding experience is required.

Capacity:
25

Description:
This training highlights the GRIME-AI system developed at the University of Nebraska–Lincoln for analyzing imagery collected from distributed camera networks. Participants will learn how GRIME-AI supports hydrologic research by using AI to extract water-related information from time-lapse imagery datasets. By the end of the Boosting Environmental Monitoring Using Camera Networks session, attendees will be familiar with GRIME AI capabilities that support imagery and other data acquisition, data cleaning and management steps, image annotation, and fine-tuning segmentation models to support environmental monitoring projects. The session will include demonstrations relevant to using USGS imagery from gauging stations, while highlighting how GRIME AI facilitates the use of imagery from trail cameras and other public resources such as PhenoCam, National Ecological Observatory Network (NEON). To maximize the value of the session, attendees who are interested in using GRIME AI will be invited to join post-conference virtual office hours to support installation and application of GRIME AI.

Date:
November 14^th, 2026

Time:
8:30 am | 2 hours

Audience Type:
Open to individuals who have written Python code with guidance or examples, understand core concepts, and have some experience with libraries. Familiarity with scientific data management principles and creating HydroShare data resources.

Capacity:
25

Description:
This workshop will equip participants with foundational data and coding skills essential for conducting research that advances water science. Through hands-on, interactive coding exercises, participants will gain practical experience with tools and capabilities central to water research, while working through science use cases centered around high-value water datasets. Participants will be exposed to topics including accessing and using JupyterHub cloud computing resources; commonly used scientific Python libraries and analytical techniques; and the creation of scalable, shareable cloud-based analyses that leverage large datasets. By the conclusion of the workshop, participants will have a comprehensive understanding of available community tools, techniques, and resources to effectively engage in modern research activities. The workshop is designed to ensure attendees gain meaningful exposure to, and practical experience with, CUAHSI cloud computing tools to maximize their research productivity and impact.

Date:
November 14^th, 2026

Time:
10:40 am | 2 hours

Audience Type:
Open to all attendees.

Capacity:
25

Description:
The field of Participatory GIS includes a diverse range of methodologies to integrate the community into social-ecological research by leveraging stakeholders’ localized spatial expertise with quantitative and qualitative components. The workshop will consist of two parts. First, the instructors will deliver a brief presentation on Participatory GIS methods and how to develop a community engagement strategy with approval from a university Institutional Review Board. We will present examples drawing on our recent work using this method in the Blue River watershed of Kansas/Missouri. Then, attendees will be invited to participate in mock focus groups where they will engage with participatory GIS methods as both facilitators and participants. These focus groups will involve a sketch-mapping exercise where participants use physical maps to draw places of importance, and a digital activity where participants will complete a map using their personal devices.

Date:
November 14^th, 2026

Time:
1:30 pm | 1.5 hours

Audience Type:
Open to all attendees, no prior experience with programming, machine learning, or image processing is required.

Capacity:
25

Description:
This workshop will introduce participants to the fundamentals of artificial intelligence (AI), machine learning (ML), computer vision, and basic image processing, with an emphasis on practical applications in water resources, hydrology, and environmental monitoring.

We will explore how images and videos can be transformed into meaningful quantitative information using image-processing techniques and AI-based approaches. Through examples from ongoing research, participants will learn about applications such as water level estimation, surface water velocity measurement, rainfall and snowfall detection, and agricultural monitoring.

The workshop will also highlight opportunities for using low-cost cameras and AI-based methods to develop affordable and accessible monitoring systems.

Date:
November 14^th, 2026

Time:
1:30 pm | 2 hours

Audience Type:
Open to all attendees.

Capacity:
30

Description:
Water science researchers rely on datasets from federal agencies, but few have been formally trained on using these datasets. During this workshop, participants will “meet the datasets” surrounding key river data sets: NASA Surface Water and Ocean Topography (SWOT), NOAA National Water Model (NWM), USGS Water Data for the Nation (WDFN). Participants will gain a better understanding of the strengths of each in water research applications. At the end of the workshop students will be able to characterize the river data products from NASA SWOT, NOAA NWM, and USGS WDFN, including relevant variables, derivation, spatial and temporal resolution, and known limitations.

The content taught in this workshop is a sample of a new pilot certificate program that teaches programmatic data access patterns for reproducible workflows that has been developed with the recommendations of the data providers themselves. The pilot offering of this new program is launching this fall, read more at CUAHSI Skills Certificate Pilot - Reproducible Workflows with Open Water Data: NASA, NOAA, USGS River Data Access.

Date:
November 14^th, 2026

Time:
3:10 pm | 2 hours

Audience Type:
Open to undergraduate and graduate students in Civil Engineering, Environmental Engineering, Water Resources Engineering, or related fields, as well as students and researchers interested in hydrology, flood-frequency analysis, flood hazard assessment, flood mapping, and hydraulic infrastructure design.

Capacity:
15

Description:
This workshop will explore how assumptions of stationarity and non-stationarity in flood-frequency analysis can influence hydraulic infrastructure design. Using an annual peak-flow record from a U.S. Geological Survey (USGS) stream gauge, participants will estimate and compare stationary and time-varying 1% annual exceedance probability (AEP) discharges. These design discharges will then be applied in a practical culvert-design exercise using HY-8.

Participants will evaluate how differences in the estimated 1% AEP discharge affect key hydraulic design outcomes, including headwater depth, flow-control conditions, outlet velocity, overtopping risk, and required culvert size. Through this comparison, participants will assess the potential engineering consequences of assuming stationarity when flood behavior is changing over time and consider how non-stationarity may affect infrastructure design and performance.

The workshop will conclude with a discussion of the potential physical drivers of flood non-stationarity. Participants will examine relationships between changes in annual peak flows and relevant watershed covariates, such as precipitation and imperviousness. This final discussion will connect statistical evidence of non-stationarity with the hydrologic and watershed processes that may be contributing to changes in flood behavior.