Water Observing Technology Forum Speakers

Authors: Jacob Ahlstrom – Florida Gulf Coast University; Robert Galletta – Stevens Institute of Technology; Briar Pierce – University of Florida

Presentation Type: Poster

Title: Low-Cost DIY Radar Sensing for Water Level Monitoring: A Comparison with Commercial Off-the-Shelf Technology

Abstract: Non-contact data collection has become increasingly necessary in hydrological studies. While some commercial off-the-shelf sensors can serve these purposes, they are costly and may be limited in functionality when used outside of their intended purpose. Do-It-Yourself (DIY) sensors are a low-cost alternative that allow for more specialized possibilities in hydrological data collection. This study evaluates the functionality, applicability, and cost of a DIY radar sensor against the Pyxis PRL-100, a comparable off-the-shelf radar level sensor designed for monitoring liquid levels in chemical tanks, which was recently tested by the USGS Hydrological Instrumentation Facility (HIF) for measuring water levels in rivers. The primary components of the DIY radar sensor include an 80 GHz liquid-level radar sensor, a Raspberry Pi Pico 2 W with MicroPython, and a Meshtastic-capable board. Both the PRL-100 and the DIY radar sensor were tested side-by-side at the HIF, assessing measurement accuracy and precision at various heights and water surface roughness conditions, both in controlled environments and in the field. Additional tests assessed instrument performance at maximum and minimum operating temperatures and maximum Long Range Radio (LoRa) communication distance. Results of this study will inform the viability of the DIY radar sensor as a cost-effective alternative to commercial solutions for hydrological monitoring.

---

Authors: Daniel Alguindigue – University of Oklahoma, Jordan Rose – California State University Northridge, Sayem Ul Alam -University of Alabama

Presentation Type: Lightning Talk and Poster

Title: A Collaborative Validation Experiment Using Bubble Surface Tracers for Low-Light Particle Image Velocimetry

Abstract: Particle Image Velocimetry (PIV) is a non-contact technique for estimating river surface velocity through the frame-by-frame tracking of visible surface tracers. Conventional PIV methods exhibit reduced reliability under smooth, glossy flow conditions and diminished lighting; circumstances frequently encountered in shaded channels, overcast weather, and evening monitoring windows. This study conducted at the USGS Hydrologic Instrumentation Facility as part of the 2026 FLOW Academy, examined whether a low-cost compressed-air bubbler system could produce consistent artificial surface tracers suitable for detection by the Setup PIV Edge Camera (SPEC), a Raspberry Pi-based edge computing platform developed by Deep Analytics with support from the USGS Next Generation Water Observing System (NGWOS) program. A bubbler apparatus consisting of acrylic pipping lines attached to aquarium bubblers and adjustable pressure regulators was designed and evaluated across a series of increasingly representative test environments. Testing progressed from bench-scale optics experiments to tow tank and small-scale flume trials. These results then informed the design of a full-scale setup for the facility’s large flume. Across each stage of testing, camera height above the water surface emerged as a critical variable. In the large flume, positioning the camera below the 1-meter threshold that was identified as limit for pixel-size resolution of surface tracers, produced an approximate 60 percent reduction in tracking error and enabling detection of substantially more tracers per frame. This suggests that the originally assumed 1-meter pixel-resolution limit does not directly predict tracking performance, and that camera height interacts with bubbler tracer visibility in ways that merit further investigation. These results, along with earlier observations that larger aerators create a more uniform and less disruptive bubble field while higher regulator pressure shortens bubble travel distance before dissipation, demonstrate that compressed-air bubble seeding can serve as a viable substitute for natural surface tracers when conventional PIV performance is degraded. This study establishes a repeatable, low-cost protocol for validating non-contact streamflow monitoring systems and identifies camera height optimization as a key direction from extending bubble seeding PIV to low-light field deployments.

---

Authors: Jingyang Cui¹, Charles Dubrule³, Natalie Restrepo¹˒⁴, Richard Iyun¹˒², Christopher G. Wilson⁵, Andrew J. Stokes⁵, and Marian V. Muste¹˒⁶

(1) U.S. Geological Survey FLOW Academy, United States

(2) The University of Alabama, Tuscaloosa, Alabama

(3) The University of Vermont, Burlington, Vermont

(4) Florida State University, Tallahassee, Florida

(5) U.S. Geological Survey Hydrologic Instrumentation Facility, Tuscaloosa, Alabama

(6) The University of Iowa, Iowa City, Iowa

Presentation Type: Poster

Title: Evaluation of Non-Contact Radar Sensors for Discharge Measurements in Open Channel Environments

Abstract: Measuring discharge is critical to water resource management and flood prevention. Historically, surface velocity and water depth have been collected using manual in-situ methods, e.g. stream gauges, which can be laborious to maintain and inaccessible during flood conditions. Advancements in noncontact sensing technology have provided an alternative to measuring water flow, utilizing devices emitting radar waves to measure water. This poster examines the capabilities of two different sensors in a controlled environment and how they respond to different parameter changes, including velocity, slope, sensor height, and roughness. Non-contact sensors allow operators to collect data safe, stable platforms at a record time.  Preliminary tests were conducted with these conditions, and further research is required to gauge the reliability of noncontact radar sensors.

---

Authors: Jingyang Cui¹, Charles Dubrule², Natalie Restrepo³, Richard Iyun¹, Christopher G. Wilson⁴, Andrew J. Stokes⁴, and Marian V. Muste⁵

(1) The University of Alabama, Tuscaloosa, Alabama

(2) The University of Vermont, Burlington, Vermont

(3) Florida State University, Tallahassee, Florida

(4) U.S. Geological Survey Hydrologic Instrumentation Facility, Tuscaloosa, Alabama

(5) The University of Iowa, Iowa City, Iowa

Presentation Type: Poster

Title: TrustFLOW: Vision-Based Water Surface Roughness Assessment for Evaluating Non-Contact Flow Sensors

Abstract: Non-contact radar sensors offer important advantages for streamflow monitoring because they can measure water-surface velocity and water level without direct contact with the flow. However, their measurement performance is strongly affected by hydraulic conditions, sensor configuration, and the characteristics of the water surface. In particular, insufficient or highly irregular surface roughness can weaken radar backscatter and contribute to unstable, biased, or missing velocity measurements. Current sensor evaluation methods typically compare radar measurements with reference instruments, but they do not continuously quantify the water-surface conditions associated with measurement performance.

This poster presents TrustFLOW, a vision-based framework for assessing water-surface roughness and supporting the reliability evaluation of non-contact flow sensors. The system uses video imagery to extract spatial gradients, temporal frame differences, and optical-flow features that describe the visible texture and motion of the water surface. These features are combined into a Surface Roughness Index and classified into multiple roughness levels representing smooth, moderately textured, and highly disturbed surface conditions. The roughness assessment is synchronized with velocity and water-level measurements from non-contact radar sensors, Acoustic Doppler Velocimeter observations, and flume reference data.

Controlled experiments were conducted at the U.S. Geological Survey Hydrologic Instrumentation Facility under different flow velocities, water depths, and surface conditions. Preliminary results show that radar measurement performance varies substantially with water-surface state. Smooth surfaces were frequently associated with underestimated or unavailable velocity measurements, while moderate surface texture generally produced more stable measurements. Highly disturbed surfaces could introduce fluctuations and occasional velocity spikes. 

TrustFLOW provides a low-cost and interpretable approach for documenting hydraulic conditions during non-contact sensor testing. The framework can support sensor calibration, field deployment assessment, quality control, and the future development of trust-aware water observation and digital-twin systems.

---

Authors: Jitae Do; Andrew Weber; Binbin Wang – University of Missouri 1

Presentation Type: Poster

Title: Practical considerations for image-based stream surface velocimetry: PIV, PTV, and STIV

Abstract: Image-based stream surface velocimetry has become an increasingly popular non-contact alternative to traditional in-stream gauging and velocity measurements. Particle Image Velocimetry (PIV), Particle Tracking Velocimetry (PTV), and Space-Time Image Velocimetry (STIV) are the three dominant methods used in the community. Practically, each method has known strengths and known failure modes, which limits their universal applicability. In this study, we apply all three methods to a single bridge-mounted consumer-webcam dataset from Hinkson Creek (Columbia, MO) and perform a head-to-head comparison. We focus on three different surface regimes recorded at the same site: a flood event with abundant debris and foam, a post-flood condition with well-distributed surface tracers, and a low-flow condition with sparse tracers and localized reverse flow near the channel margins. We document the preprocessing and parameter-tuning choices required by each method, identify the surface conditions under which each one degrades, and report the surface velocity fields and cross-section profiles produced under each combination. The results showcase practical considerations for selecting and tuning image-velocimetry methods for natural streams using fixed low-cost cameras.

Speaker Bios (for website): Binbin Wang is an Associate Professor and William Andrew Davidson Professor in the Department of Civil and Environmental Engineering at the University of Missouri. His research focuses on environmental fluid dynamics, utilizing image-based measurement techniques in laboratory experiments, field applications, and UAV-based remote sensing. Binbin has developed several in-situ underwater PIV systems and stereo imaging systems for flow and turbulence measurements. He is currently a member of the Fluid Dynamics Technical Committee of ASCE-EMI and the Hydraulic Measurements and Experimental Methods (HMEM) Committee of ASCE-EWRI.

---

Authors: Yousef ElSayed, Songyang Zhang, Mohamed ElSaadani, Emad Habib – University of Louisiana at Lafayette

Presentation Type: Poster

Title: Neighborhood-Scale Flood Monitoring Using Machine-Learning Enabled Sensors and Cloud-Based GIS Solutions

Abstract: Current flood monitoring systems are often constrained by high installation costs and site-specific limitations, creating a growing demand for more intelligent and cost-effective solutions. Recent advances in information technologies, particularly artificial intelligence (AI) and the Internet-of-Things (IoT), have opened new opportunities for the development of next-generation AI-enabled flood monitoring systems. This poster presents a novel intelligent IoT-based system for flood risk management that aims to improve flood prediction accuracy, reduce communication overhead, and protect data privacy in distributed flood monitoring environments. Specifically, we first introduce a vision-based object segmentation and water level prediction framework, benchmarking the AI-based flood monitoring. Secondly, by leveraging the emerging technology of semantic communications, we present an efficient data transmission framework for flood monitoring sensors that reduces communication overhead while preserving task-relevant information. Finally, we introduce privacy-preserving and resource-aware federated learning frameworks for distributed AI-native flood monitoring systems. The simulation results demonstrate the effectiveness of the proposed approaches in enabling accurate, communication-efficient, and privacy-aware next-generation AI-native flood risk management.

---

Authors: Troy Gilmore, John E. Stranzl. Jr., Mary Harner, Adam Caprez, Razin Bin Issa, Aaron Mittelstet, Kenneth Chapman, and the GRIME Lab Team – University of Nebraska

Presentation Type: Training

Title: GRIME-AI: Boosting Environmental Monitoring Using Camera Networks

Abstract: 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.

Speaker Bio: Troy Gilmore is an Associate Professor in the Conservation and Survey Division – School of Natural Resources and the Biological Systems Engineering Department at the University of Nebraska – Lincoln. Troy earned his PhD in Biological Systems Engineering at North Carolina State University. His undergraduate research included a water level camera publication supervised by Dr. François Birgand. John E. Stranzl, Jr. is a PhD Candidate in the School of Natural Resources at the University of Nebraska – Lincoln. John has extensive software development, image processing, and ML/AI experience across many industrial applications. John is completing his dissertation, centered around GRIME AI software and ecohydrological applications, in Summer 2026. John and Troy are part of the larger GRIME Lab team that includes co-authors of this talk and many others who deserve acknowledgement for contributions to software testing and development.

---

Authors: Ismail Gul – Stevens Institute of Technology, Sara Gutierrez Diaz – Tuskegee University, Peter Udenze, Jack Bathasar, Lisa Davis, Dawen Li – University of Alabama, Christopher Wilson, Robert R. Lotspeich – United States Geological Survey (USGS)

Presentation Type: Poster

Title: A Low-Cost and Rapid Deployment Flood Level Observation System

Abstract: The USGS actively maintains ~8,500 quality-controlled stream gaging stations across the conterminous United States. Yet many rivers lack real-time monitoring capability because of cost constraints. Recent catastrophic floods (e.g., Guadalupe River in Texas in 2025; Hurricane Helene inland floods in 2024) highlight the need for affordable, community-level flood monitoring/warning systems to support localized decision-making. In this study, we present the design, sensitivity analysis, and initial deployment of a low-cost, easily installed, flood monitoring prototype (FLO-RED). FLO-RED uses an Arduino Mega microcontroller board and costs ~$700 to build. FLO-RED was field and lab tested over 8 weeks as part of the USGS FLOW Academy.

We improved usability of an existing prototype by adding waterproof connectors to the control box that integrates a JSN-SR04T waterproof ultrasonic distance sensor to measure river stage via the time-of-flight principle, with back up stage readings from a Seed Studio analog G1/4 non-vented pressure transducer and an MPRLS barometric sensor. Testing of the integrated system components at the USGS Hydrologic Instrumentation Facility included determining the pressure transducer’s output linearity across five pressure levels (11–55 PSI) at 0 °C, 20 °C, and 40 °C.  The results showed that the ultrasonic sensor had an error range of ±0.6 cm at 3 m height and 20 °C. We developed a prototype capable of assessing sensor performance without relying on the previously used custom-designed data logger and power boards. For timestamping, an off-the-shelf RTC clock module was integrated into the system. We also added a 3-D printed cone to cover the ultrasonic sensor to reduce interference with non-water surfaces. The whole flood observation system is powered by solar plus battery energy with one-week autonomy for remote deployment. Future work will focus on power optimization, reliable data transmission, and intuitive real-time visualization, with the long-term goal of deploying a statewide rapid-response flood warning network across Alabama.

---

Authors: Razin Bin Issa, Jeffery S. Horsburgh, Safran Kahn, Sajan Neupane, Sierra Young – Utah State University

Presentation Type: Poster

Title: From Camera to Cloud: Automated Hydrologic Monitoring Using HydroCamCollect and HydroCamCompute

Abstract: Camera-based monitoring has emerged as a promising non-contact approach for streamflow observation, offering advantages in safety, scalability, and cost compared to traditional in-stream sensing methods. However, operational deployment of camera systems at scale requires robust workflows for reliable data collection, cloud transfer, automated processing, and system monitoring. This poster presents two complementary technologies developed to address these challenges: HydroCamCollect and HydroCamCompute. HydroCamCollect is a fault-tolerant edge data acquisition workflow built using Raspberry Pi computers and IP cameras that automates image and video capture, verifies upload integrity, manages local storage, and supports reliable cloud synchronization under intermittent network conditions. HydroCamCompute extends this workflow into the cloud using a serverless computing architecture that automatically processes uploaded imagery to extract hydrologically relevant information using computer vision and machine learning techniques. The event-driven pipeline scales dynamically without requiring dedicated server infrastructure and supports near real-time hydrologic inference. Together, these technologies provide an end-to-end system for scalable camera-based hydrologic monitoring networks. Results from multi-month deployments at river monitoring sites in northern Utah demonstrate reliable autonomous operation, automated cloud processing, and the feasibility of scaling to larger distributed environmental sensing networks.

---

Authors: Crane Johnson – National Weather Service, Jamie Pierce – United States Geologic Survey, Brian Shumaker – Beadedstream, Johnse Ostman – National Weather Service

Presentation Type: Poster and Lightening Talk (if available)

Title: Adaptive Water Level Monitoring: Three Innovations for the Alaskan Environment

Abstract: Hydrologists in Alaska face unique water level monitoring challenges driven by seasonal river ice cover, expansive braided rivers, and glacial lakes that experience hundreds of feet of seasonal water level fluctuation. Compounding these environmental factors is a severe lack of infrastructure; Alaska’s limited road network and sparse telecommunication coverage make disseminating near real-time data exceedingly difficult. To address these issues, the Alaska-Pacific River Forecast Center has spent the past decade collaborating with federal agencies and commercial partners to innovate specialized monitoring tools for the far north. This poster presents three unique technological solutions developed in Alaska. Two were engineered with commercial partners through Cooperative Research and Development Agreements (CRADA), resulting in commercially available off-the-shelf products. These two solutions are small, low-cost, satellite-based water level sensors using acoustic and radar water level sensors. The third solution, developed alongside partner agencies, was designed using an oblique laser gage to monitor ice-covered rivers during spring breakup and highly volatile glacial-dammed lakes—extreme environments where traditional in-situ sensors are vulnerable and unreliable.

---

Authors: Yeonju Kim_¹, Tantu Mandal^2, Taylor Sullivan³, Mendbayasgalan Lkhagvadorj⁴ , Daniel West2, Bryce J Redinger⁵, Christopher Holmquist-Johnson ^5, Yuehan Lu²

University of Connecticut_¹, University of Alabama², Mississippi State University³, University of Utah⁴ , U.S. Geological Survey⁵

Presentation Type: Lightning Talk and Poster

Title: High-resolution UAV Infrared mapping of surface water temperature for river monitoring: A case study of the Black Warrior River, Alabama

Abstract: River temperature is a key variable controlling aquatic ecosystem health, species habitat, and biogeochemical processes. Conventional river temperature monitoring has relied heavily on point-scale measurements using data loggers, which provide limited spatial coverage. However, spatial variability in river temperature — modulated by surface water-groundwater interactions, tributary inflows, and wastewater discharges — requires comprehensive monitoring at high spatial resolutions. This study presents an uncrewed aerial system (UAS)-based approach to map river surface temperature along a 1,670 m reach of the Black Warrior River in Alabama using a Skydio X10 UAV with the VT300-L sensor package which includes a Teledyne FLIR Boson+ thermal sensor and 1” 50MP CMOS natural color camera. Thermal and RGB imagery were processed using Pix4D and Agisoft Metashape to generate spatially continuous thermal orthomosaics. The UAS-derived thermal data were calibrated using ground-truth temperatures collected concurrently with HOBO temperature loggers and YSI multiparameter sondes deployed at multiple locations across the study reach. An empirical calibration model was developed to relate drone-derived pixel temperatures to in-situ measurements, enabling bias correction of the thermal orthomosaic. This study tested and refined a workflow for acquiring and calibrating river surface temperature data using UAS-based thermal infrared imaging, providing a transferrable framework for high-spatial-resolution monitoring of streams and rivers.

---

Authors: J. Derek Loftis – Virginia Institute of Marine Science, William & Mary, Sridhar Katragadda – City of Virginia Beach, and Yash K. Sanap

Presentation Type: Training (2-hours)

Title: TS4 – Inundation Monitoring with Machine Learning (Image-Paired ML Model)

Abstract: This training introduces a workflow for developing a machine‑learning model that integrates edge detection and AI to estimate stage from imagery. Using paired images and model-building examples, participants will learn how to prepare data, train the ML algorithm, and evaluate the resulting inundation predictions. For more information: Inundation monitoring using a machine learning algorithm combining AI and edge detection – ScienceDirect

Speaker Bios (for website):

J. Derek Loftis, Ph.D., serves as a Research Assistant Professor with the Center for Coastal Resources Management and the Virginia Commonwealth Center for Recurrent Flooding Resiliency at the Virginia Institute of Marine Science (VIMS). Dr. Loftis’ research centers on advancing hydrodynamic modeling to better understand and mitigate flood risk in vulnerable coastal regions through enhanced forecasting and flood monitoring via sensors, drones, and community science. Dr. Loftis earned his Ph.D. at William & Mary while modeling and monitoring street-level inundation during Hurricane Sandy in New York City, and his community resiliency collaborations with his co-author, Sridhar, as a postdoc formed StormSense, which has deployed over 50 low-cost, low-energy, water level sensors throughout Coastal Virginia in the past decade.

Sridhar Katragadda is the Lead Data Scientist working in the City of Virginia Beach, Information Technology Department with a background in Civil and Geological Engineering fields. Sridhar has worked in federal, state, local government(s) and private industry for over 25 years. Sridhar plays a key role in developing innovative resilience strategies and solutions for the City, regionally, and nationally. Sridhar’s flooding resiliency efforts focus on developing capacity in multiple cloud environments that include spatial, business intelligence, real-time systems, Internet of Things (IoT), ML/AI, and collaborating within the region in multiple federal, state, regional projects, and has received “2017 & 2019 AWS City on a Cloud Innovation Challenge Awards” for StormSense, in collaboration with his co-author, Derek Loftis.

---

Authors: Jon Derek Loftis, Sridhar Katragadda, Yash K. Sanap, Varshin Bhaskaran, Hunter Harman, Andrew McGowan, and R. Russell Lotspeich

Presentation Type: Tuesday Morning Keynote

Title: Machine Learning River Stage Detection from Webcam Imagery

Abstract: Video-based inundation monitoring offers a scalable, low-cost complement to traditional stream gaging. Building on the USGS Next Generation Water Observing System (NGWOS) and the River Image SEnsing (RISE) project, we present Flood VISION-AI (Flood Visual Inference of Stage Identification by Observation Networks – Artificial Intelligence). This convolutional neural network framework infers water surface elevation from fixed-position webcams in the USGS’ Hydrologic Imagery Visualization and Information System (HIVIS). During this study period, site-specific EfficientNet-based regression models were trained and evaluated at ten unique camera sites spanning tidal creeks, rivers, reservoirs, and an urban storm sewer across Virginia, Texas, South Carolina, Wisconsin, and California. Models were validated against co-located Ka-band radar water-level gages. This presentation features recent methodological advances in AI/ML water monitoring applications, including: a Universal-image Quality Index (UQI) to flag imagery degraded by adverse atmospheric conditions such as rain, fog, or lens obstructions; a resampling strategy that balances training data across the observed stage distribution to mitigate overfitting toward normal conditions; and a Structural Similarity–based Image Similarity Index (ISI) that automatically filters misaligned frames from pan-tilt-zoom cameras, which improved R² at one PTZ site from 0.75 to 0.90. Performance was strongest at tidal creek sites with proximal staff-gage targets (R² up to 0.9923; MAE of 0.0405 ft) and weakest where benchmarks exceeded 100 ft from the camera. Across the ten sites reporting complete validation metrics, models achieved an average R² of 0.89, indicating that roughly 89% of observed stage variance was explained by image-based inference. Corresponding mean error statistics averaged 0.26 ft (MAE) and 0.41 ft (RMSE), demonstrating operational viability for near-real-time flood monitoring.

---

Authors: Jon Derek Loftis, Sridhar Katragadda, Hunter Harman, and R. Russell Lotspeich

Presentation Type: Poster

Title: River Stage Estimation via a Deep Neural Network Incorporating Edge Detection

Abstract: Autonomous monitoring of surface water elevations is essential for improving flood forecasting and hydrologic situational awareness. Recent advances in passive optical sensing, deep learning, and computer vision have enabled image-based approaches capable of supplementing or replacing conventional water-level instrumentation in environments where sensor deployment is challenging or cost-prohibitive. This study presents and validates a hybrid computer vision framework that integrates deep learning with edge-based feature extraction to estimate free-surface water elevations from oblique imagery acquired by web-enabled cameras deployed in tidal waterways. Field experiments were conducted over a three-month observation period using cameras configured to acquire images at 6-minute intervals. Multiple imaging platforms incorporating high-resolution optical sensors and integrated infrared illumination were evaluated to characterize system performance under both daytime and nighttime operating conditions. Water-surface elevations derived from image sequences were validated against A-style staff gages positioned within each camera’s field of view and independently compared with nearby Ka-band radar water-level sensors. The proposed methodology achieved a root mean square error (RMSE) of less than 1.25 cm when cameras were installed within approximately 10 m of the monitored cross section, demonstrating high agreement with conventional stage measurements. The results demonstrate that passive, web-connected imaging systems can provide accurate, continuous, and automated river stage observations using inexpensive commercial hardware and advanced computer vision algorithms. The proposed framework offers a scalable, low-cost alternative for real-time hydrologic monitoring and flood surveillance, with potential applications in distributed stream gaging networks, infrastructure resilience, and next-generation flood early warning systems.

---

Authors: Travis Loof, Robert Gill, Ch’Ree Essary, Matthew VanDyke – University of Alabama

Presentation Type: Poster

Title: Validating and Scaling a Multi-Model AI Pipeline for Operational Analysis of Weather Influencer Flood Risk Communication

Abstract: Social media weather influencers now reach audiences that rival or exceed those of local National Weather Service forecast offices, and their flood coverage shapes public risk perception and protective action in real time. Whether this content reinforces or undermines official guidance is unknown. The Entertainment Overcoming Resistance Model suggests the stakes run in both directions. For example,. entertainment value and parasocial connection lower audience resistance to persuasive messages (such as warning messages), so influencer content that aligns with NWS guidance could increase warning acceptance, while content that diverges could draw audiences away from official guidance when protective action matters most. No method currently exists to measure this alignment at scale, as manual content analysis of an event window across a dozen influencers requires months of coding effort. We propose to validate and scale an open-source, multi-model AI pipeline that would make this analysis fast, repeatable, and deployable for NWS situational awareness. The pipeline’s core components have been built and pilot-tested on general YouTube transcript data, but no flood-domain validation has been performed. The proposed project would validate pipeline outputs against a five-dimension manual coding framework covering temporal alignment, hazard accuracy, uncertainty communication, tone and framing, and protective action messaging, targeting Krippendorff’s alpha of 0.80 or higher; extend the pipeline with a multimodal vision layer. Expected outcomes include a validated open-source pipeline, a benchmark dataset, and an operational deployment guide.

---

Speaker Name: Ronan Lucey

Authors: Ronan Lucey – University of Alabama in Huntsville / NASA Disasters Program

Presentation Type: Plenary Technical Presentation 1

Title: The NASA Disasters Program – Connecting Remote Water Monitoring through Earth Observations to Informed Decision Making for Disaster Management

Abstract: When floods and hurricanes strike, utilizing remotely sensed data for situational awareness of water location and depth can feel like a distant option, especially when capacity to understand and work with new products is strained during a disaster. The NASA Disasters Program addresses this challenge with a dedicated team focused on supporting user needs across disaster preparedness, response, recovery, and resilience. While integrating AI-enhanced remote water monitoring products derived from Earth Observations (EO) into disaster management activities speeds decision-making, their true value relies on seamless integration into existing workflows. The Disasters Program delivers targeted capabilities that apply EO for water-relevant applications throughout the disaster management cycle: for preparedness to inform risk and hazard decision making; for response to improve situational awareness and decision support; and for recovery to build back better. Central to this support is the NASA Disasters PORTAL, a platform where users can interact and visualize data curated to meet individual needs, learn from data stories and priority training modules, and more. By co-developing a user-driven, dynamic disaster information portal, NASA is bringing the power of EO to users to support water-related disaster management activities, protect lives, and build more resilient communities.

Speaker Bios: Ronan Lucey is a Research Scientist with the University of Alabama in Huntsville positioned at NASA Marshall Space Flight Center. He has supported the NASA Disasters Program since 2019, including as an Associate Program Manager overseeing the Disasters PORTAL since 2025. Previously, he served as a Center Response Coordinator for the NASA Disasters Response Coordination System (DRCS), and has led numerous NASA Disasters Program event responses, including for the Hawaii Wildfires in 2023 and Hurricanes Eta and Iota in 2020. Additionally, he has led various Disasters Program stakeholder engagement efforts at local, state, and national levels, working to empower stakeholders with the tools and information that they need to reduce disaster risk, build resilience, improve response, and accelerate recovery in their communities. He enjoys connecting NASA’s data, science, and expertise with stakeholders to collaboratively solve problems, ensuring informed and impactful decision-making.

---

Authors: Ekaterina Miliutina, Jilin Men, Amanjit Premsagar, Anindya Palaparthi, Hongxing Liu, Yuehan Lu, Naveenkumar Purushothaman

Presentation Type: Poster

Title: Cross-Scale Water Quality Monitoring Through Integrated Satellite, UAV, and In-Situ Observations

Abstract: Water quality monitoring is critical for understanding aquatic ecosystem dynamics and supporting environmental management, yet single-platform observations are often limited in either spatial resolution, temporal frequency, or measurement accuracy. This study presents a multi-platform water quality observation framework integrating satellite remote sensing, unmanned aerial vehicle (UAV) observations, and in situ measurements for comprehensive aquatic environment monitoring. Satellite imagery provides large-scale and long-term observations, while UAV-based multispectral and hyperspectral sensors enable ultra-high-resolution mapping of localized water features. Field measurements were conducted using YSI multiparameter sondes, HyCAT systems, and WISP-Orca spectrometers to acquire synchronized water quality and above-water spectral observations for calibration and validation. The integrated system was applied to monitor key water quality parameters, including chlorophyll-a, turbidity, total suspended solids (TSS), and colored dissolved organic matter (CDOM), across inland and coastal waters. Results demonstrate that the combination of satellite, UAV, and field observations improves the characterization of complex aquatic optical properties and enhances the accuracy and robustness of water quality retrievals across multiple spatial scales.

---

Authors: Jim Mitchell, Reece Robinson, Nathan Wardwell, Mike Zieserl – JOA Surveys LLC

Crane Johnson – National Weather Service Alaska-Pacific River Forecast Center

Taylor Borgfeldt – Alaska Ocean Observing System

Joel Cusik – National Park Service

Mark Laker – U.S. Fish and Wildlife Service

Presentation Type: Oral or Poster

Title: Monitoring Ice and Water Levels on the Yukon River using GNSS Interferometric Reflectometry

Abstract: Communities along the Yukon River face serious flood risk each spring. Monitoring and forecasting these events is a challenge: infrastructure along the river is sparse, and traditional sensors with in-water components are compromised by ice. The result is a significant data gap that limits forecasters’ ability to communicate risk to vulnerable communities. A multi-agency collaboration funded by the Alaska Ocean Observing System is testing GNSS Interferometric Reflectometry (GNSS-IR) for year-round water and ice level monitoring in Galena, AK. GNSS-IR is a non-contact system that complements existing monitoring technologies. Geodetic-grade antennas function as both a continuously operating reference station and a water/ice level sensor, with measurements tied directly to a global reference frame, eliminating the need for additional sensor surveying. In 2025, the U.S. Fish and Wildlife Service installed a GNSS base station maintained by the National Park Service near the river bank. Data is processed by JOA Surveys to produce near real-time water levels, validated against a USGS stream gauge and a riverbank camera installed by the National Weather Service and USFWS. This presentation shares results from the Galena pilot project, discusses readiness for broader adoption, and explores how GNSS-IR could scale as a modern river observation network across Alaska.

Speaker Bios: Nathan Wardwell is Managing Partner of JOA Surveys. He has more than 20 years of experience measuring water levels and vertical datums. Nathan has an M.S. in Earth Science with an Ocean Mapping Emphasis (2008) from the University of New Hampshire’s Center for Coastal and Ocean Mapping and a B.S. in Environmental Science (2004) from Alaska Pacific University.  He is currently the Chair of the Hydrographic Services Review Panel, a member of the University of Alaska Anchorage Geomatics Advisory Board and a member of the Alaska Water Level Watch Steering Committee.

---

Authors: Amanjit Premsagar, Yuehan Lu, Jilin Men, Ekaterina Miliutina, Anindya Palaparthi, Hongxing Liu, Tantu Mandal- University of Alabama

Presentation Type: Poster

Title: Source-Resolved DOM Dynamics in Oligotrophic Inland Waters: Linking UAV Hyperspectral Remote Sensing with PARAFAC Spectroscopy

Abstract: Traditional water quality monitoring relies on bulk metrics such as total organic carbon, chlorophyll concentration, and turbidity. While useful for quantifying total concentration, these measurements fail to distinguish dissolved organic matter (DOM) composition and provenance-information critical for anticipating treatment challenges, as terrestrial (allochthonous) and algal/microbial (autochthonous) DOM pose different risks, from disinfection byproduct precursors to harmful algal bloom indicators. This research implements a multi-scale framework combining UAV hyperspectral imagery, dense in-situ sampling, and PARAFAC-resolved fluorescence/absorbance spectroscopy to capture DOM dynamics in oligotrophic river and lake systems. Rather than assuming a single universal reflectance-concentration relationship, we systematically screen spectral band-ratio combinations to evaluate which relationships hold across sampling sites and seasons-a validation step often overlooked in optically complex inland waters. By linking hyperspectral signatures to PARAFAC-derived source fractions (protein-like, terrestrial humic-like, and microbially reworked components), this approach is designed to resolve compositional shifts driven by runoff events, algal blooms, or in-stream microbial processing. Model development to date remains site-specific, with per-site sample sizes still limiting robustness; larger, geographically distributed datasets are needed to establish transferability. By targeting band configurations compatible with satellite sensors, this methodology lays the foundation for upscaling source-resolved DOM characterization from local UAV surveys to regional and satellite-scale water observation networks.

---

Authors: Amanjit Premsagar¹, Godwin Sunday¹, Ekaterina Miliutina², Gabrielle Justine Tapat³, Bryan Gutierrez², and Hongxing Liu²

(1) The University of Alabama, Department of Geological Sciences, Tuscaloosa, AL, 

(2) The University of Alabama, Department of Geography and the Environment, Tuscaloosa, AL 

(3) University of Hawai’i at Mānoa, Department of Natural Resources and Environmental Management, Honolulu, HI

Presentation Type: Poster

Title: Integrating UAV-Based Remote Sensing and In-Situ Monitoring for High-Resolution Water Quality Assessment in North River and Lake Tuscaloosa

Abstract: This study, supported by the USGS FLOW (Future Leaders in Observation of Water) Academy 2025, investigates the integration of in-situ, mobile, and remote sensing technologies for high-resolution water quality monitoring of freshwater rivers and lakes, focusing on the North River, a major tributary of Lake Tuscaloosa, Alabama. HyCAT Autonomous Surface Vehicles (ASVs) equipped with YSI EXO2 Multiparameter Sondes collected continuous, dense measurements of turbidity, fluorescent dissolved organic matter (fDOM), and chlorophyll-a. Simultaneously, 10-band multispectral imagery was acquired with a MicaSense Dual Camera mounted on a UAV, flown along predefined flight paths with evenly distributed Ground Control Points for accurate georeferencing. Drone imagery was processed using Agisoft to generate orthomosaics and reflectance maps, which were integrated with in-situ measurements to develop and validate water quality remote sensing models. These models produced fine-resolution (8 cm) water quality maps, revealing spatial variability linked to natural and anthropogenic influences, and enabling near-real-time, spatially continuous assessment of aquatic ecological health. By combining UAV imagery with robust in-situ measurements, this study demonstrates a scalable, cost-effective approach to inland water monitoring with broader applications for watershed-scale assessment, particularly in data-scarce or logistically challenging environments, supporting improved understanding of inland water dynamics and informed water resource management.

---

Authors: Jihee Seo, Ekaterina Miliutina, Basit Akinade, Brodie Alexander, Jack Balthasar, Aijun Song, Hongxing Liu – University of Alabama; Matthew C. Gyves, R. Russell Lotspeich – USGS

Presentation Type: Poster

Title: Integrating Aquatic Drone, Autonomous Surface Vehicle, and Unmanned Aerial Vehicle Platforms for 3D Flow and Water Quality Mapping in Riverine Systems

Abstract: Accurate spatiotemporal characterization of flow dynamics and water quality is essential for understanding riverine hydrology and aquatic ecosystem health. This study presents an integrated methodology to estimate vertical current velocity profiles, generate two-dimensional surface drift velocity fields, and map water quality parameters using autonomous platforms in inland waterways. Deployments were conducted in the Black Warrior River, Alabama utilizing a multi-platform approach: a JaiaBot aquatic drone was employed to capture vertical velocity profiles in addition to key water quality parameters such as dissolved oxygen, pH, conductivity, and chlorophyll; a Phantom 4 Unmanned Aerial Vehicle (UAV) was employed to collect surface velocity via Particle Image Velocimetry (PIV) to compare with the JaiaBot’s surface drift estimates; and a HyCAT Autonomous Surface Vehicle (ASV) was employed to collect Acoustic Doppler Current Profiler (ADCP) data to serve as the ground truth for flow measurements. Ascent/descent behavior and time–distance estimates were used to reconstruct the JaiaBot’s underwater path to compensate for the lack of GPS positioning while submerged, and interpolation techniques were applied to produce spatially continuous 2D velocity and water quality maps from discrete measurements. The results demonstrate the feasibility of low-cost, multi-platform autonomous sensing systems for high-resolution hydrodynamic and environmental monitoring in riverine environments. This approach offers a scalable solution for flow and water quality mapping in areas where traditional sampling methods are limited.

---

Poster Presenter Name: Taylor Sullivan, Mendbayasgalan Lkhagvadorj

Authors: Taylor Sullivan1, Mendbayasgalan Lkhagvadorj2, Yeonju Kim3, Tantu Mandal4 , Daniel West4, Yuehan Lu4

Mississippi State University1, University of Utah2, University of Connecticut3, University of Alabama4 

Presentation Type: Lightning Talk and Poster

Title: Evaluating Handheld Thermal Infrared Imaging as a Rapid Screening Tool for Groundwater Seepage Detection

Abstract: Groundwater discharge is an important control on river temperature, water quality, and aquatic habitat, yet locating groundwater seepage zones using conventional field methods is often labor-intensive and spatially limited. Thermal infrared imaging offers a rapid approach for detecting surface temperature anomalies associated with groundwater inflows, enabling efficient reconnaissance over large stream reaches.

This study evaluated the use of a handheld FLIR E4 thermal infrared camera as a rapid, low-cost tool for identifying groundwater seepage along the Little Cahaba River in Alabama. A field survey was conducted to capture localized thermal anomalies. Thermal observations were validated using water temperature measurements collected with a YSI multiparameter sonde. We identified several potential groundwater seepage locations that may be associated with beaver dam return flows and/or hydrogeological transitions across the Fall Line.  This study demonstrates that handheld thermal infrared imaging is a simple, low-cost screening technique that can rapidly survey large stream reaches, substantially reducing the time and effort required to locate groundwater discharge compared with conventional point-based hydrological surveys.

---

Speaker Name: Christine VanZomeren

Authors: Christine VanZomeren, Joe Wright, Chris Frans

Presentation Type: Poster

Title: Sensing the Future: Prize Competitions Driving Innovation in Water Availability

Abstract: This presentation will describe using prize competitions to tackle water availability challenges in the west. Prize competitions generally will be described in addition to describing three competitions on water availability: Counting Every Drop Challenge, Streamflow Forecast Rodeo, and the Divide and Conquer Challenge. The competition objective, winning solution, and subsequent assessment efforts will be described. The Counting Every Drop Prize Competition will be the focus and was created to spur development of precipitation‑measurement technologies that reduce maintenance needs while improving accuracy and reliability. The winning solution, the Precipitation Measurement with Advanced Solid‑state Sensors (PMASS) system, integrates a downward‑facing pulsed coherent radar, a camera‑based sensor, and a temperature sensor, with machine‑learning techniques used to estimate precipitation rates and accumulated depth. An ongoing collaboration between MIT Lincoln Laboratory and the Bureau of Reclamation includes capability‑gap assessments, technology roadmapping, and prototype development to support improved water‑resource management.