Theodore (Ted)
Langhorst

This project develops a novel deep learning model to predict daily suspended sediment concentration, suspended sediment flux, and river discharge by combining hydrologic data, optical remote sensing, and watershed characteristics. Our approach addresses limitations of existing methods by using satellite and in-situ data without same-day matchups. Focusing on rivers in the conterminous United States, the research demonstrates improved performance through simultaneous modeling of multiple variables and the integration of water color data with hydrology-based predictions. This new methodology aims to enhance our understanding of river processes and provide more accurate, continuous estimates of sediment flux.

Our new hybrid model architecture.

Optical backscatter sensors are instruments commonly used to determine the turbidity (cloudiness) of water, or how much particulate matter is in the water. OpenOBS is an academic research project to develop an open source, affordable, easy to build, and accurate OBS. While commercial sensors cost upwards of $3000, OpenOBS can be built for less than $50 in parts. This low-cost alternative allows us to more extensively monitor natural systems, or use in high-risk environments, where deployment of a large number of commercial sensors would be cost-prohibitive.

Documentation site
Publication in Nature Water

An assemebled OBS with housing cut open for visualization.

Graphical interface for programming the sensor before deploying.

Full schematic of the data logger.

This study investigates how cloud cover affects the ability of optical satellite imagery to accurately capture river conditions globally. Using 22 years of daily cloud cover data and modeled river discharge for over 21,000 river reaches worldwide, we show significant biases in satellite observations, particularly in tropical and Arctic regions. The findings demonstrate that cloud-related biases are geographically organized, potentially impacting large-scale comparisons of river conditions derived from satellite data. By quantifying these biases and exploring their relationship with various environmental factors, this work provides insights for improving the interpretation and application of satellite-based river observations in hydrological studies.

Publication in GRL

Global map of normalized errorsaveraged to Pfafstetter level 2 basins. Eight regions are highlighted across environments to show the relationship between seasonal cloudiness and discharge with observation bias. The solid line indicates the regional median ND value, and the shaded area represents the 10th and 90th percentiles of ND. Some regions are named according to large, well-known rivers but do not exactly coincide with these basins.

Our global sample results of discharge biases.

Often we try to limit these biases by averaging our data to longer time scales. These boxplots show the efficacy of temporal averaging on the 1st, 50th, and 99th percentiles.

Remote sensing of river discharge and suspended sediment concentration has seen big improvements in the past decade. It's now possible to combine these algorithms to use satellite images to estimate the sediment flux of a river. This work is currently in preparation, but has been years in the making, with the first remote sensing work being completed in 2019 and field work in 2021 and 2022 for validation data. This project was the impetus for OpenOBS, after realizing I did not have the budget for even 1 deployable commercial turbidity sensor. We have used these sensors extensively in the Sag, deploying 10 each year in order to monitor months of turbidity changes in high resolution.

Ellie Friedmann installing sensors in the Sag.

The sensor setup for one site (2 OpenOBS and a pressure logger) before being installed in 2022.

The Sag at low flow in August 2021, showing off its gravel bed and headwaters in the Brooks Range Mountains.

Pulling rebar-mounted sensors at the end of the 2021 season.

Riverbank erosion presents a real threat to people who live near the banks of active rivers, yet understanding and predicting it at large scales is extremely difficult. One impediment to these predictions is a globally representative dataset of riverbank changes over long time scales. This project is attempting to fill that gap by processing global maps of surface water extent to extract river extents over the past two decades to calculate rates of erosion and accretion.

Publication in JGR:ES

Global map of riverbank erosion with insets showing the Amazon Basin and Purus River.

Methods for calculating riverbank erosion and accretion from maps of surface water.

Visualization of annual river maps of the Madre de Dios River in Bolivia.

IGEA is a unique class taught in three parts by Colin Gleason at UMass Amherst. In the Fall, undergraduate students learn how to plan and prepare for field science studying Arctic Hydrology. The students then follow through with their plans at Toolik Field Station on the North Slope of Alaska that Summer with the help of graduate students. Finally, the students analyze their data and write a report in the Spring. The science objectives were broadly defined in the NSF CAREER grant to Gleason which gives the students lots of freedom to follow their interests.

Likely explaining something tangential that I found interesting.

Group 2 from the 2021 IGEA class on a snowy day.

Group 1 from 2022 navigating to their sites, walking 2km over frozen tundra.