Predictions of hydrologic function based on aquatic DNA fragments (NSF-EAR 1836768)

Project Leads: Stephen Good, Byron Crump
Graduate Students: Dawn URycki
Undergraduate Students: Jessica Chadwick, Marilee Hoyle, Lindsay Paige, Lindsey Spencer

Overview: In watersheds where discharge observations are limited or non-existent, the collection of some other type of information-dense dataset during short field campaigns can be remarkably useful in understanding watershed function. This proposal suggests that the collection of stream-water DNA presents an opportunity to rapidly gather a large, digitally encoded, dataset that contains information about hydrologic function at the multiple scales. This project is motivated by recent studies in microbiology that documented strong connections between the composition of aquatic microbiomes and the discharge of rivers where these microorganisms are found. By collecting, sequencing, and classifying the relative abundance of different members of aquatic microbiomes and then relating variations in microbiome community to discharge patterns, this proposal will develop a new hydrologic tool useful for quantitative inference about watershed function.

Intellectual Merit: The overarching hypothesis is that hydrologic discharge characteristics at multiple scales can be predicted based on the unique composition of DNA material within water samples. To test this hypothesis, this project will use aquatic DNA collected at USGS stream gauges throughout Oregon by the PIs in 2017, as well as additional samples collected regionally and nationally. Machine learning techniques will be used to link the relative abundance of different microbial groups to discharge patterns based on methods recently developed by the PIs. These methods were published in a pilot study that predicted seasonal flows and return intervals in six arctic rivers based only on collected DNA with a Nash Sutcliffe efficiency of 0.84 and 0.67, respectively, which were considerable improvements over predictions based only on area-scaled mean specific discharge similar rivers. This project will apply these methods to a large and diverse set of rivers throughout North America. This project will evaluate the utility of DNA based hydrology predictions using (1) summer samples, (2) seasonally collected samples, and (3) region-and continent-wide samples. A leave-one-out jackknifing approach will be used to determine accuracy and bias in the estimated discharge values. Estimated discharge values will also be compared with current methods for prediction at ungauged locations to quantify the added hydrology information contained in aquatic DNA. Connections between microbial community composition and other macrosystem properties such as topography, land cover, and infrastructure development will also be examined. At the conclusion of this project the biogeography of microbial communities will be quantitavely linked to their hydrology and a new tool for prediction of discharge in poorly instrumented catchments will be made available to the community.

Broader Impacts: This project will integrate students and faculty from two diverse backgrounds: microbiology and hydrology, in a unique interdisciplinary undertaking, and will also develop new open source interactive learning materials for hydrology students. This project will include training in science, technology, engineering, and mathematics (STEM) for students from high school to the PhD level. By partnering with The Saturday Academy, a local organization focused on connecting under-represented communities with STEM professionals, this project will engage two high-school summer interns in a STEM learning experience. At the university level, four undergraduate students and a PhD level graduate student will receive training in modern microbiology and hydrology methods. The PIs have a demonstrated history of working with students from a diverse set backgrounds and will strive for inclusivity throughout the duration of this project. As part of this project, an interactive textbook-like learning resource focused on physical hydrology will be created within the Jupyter Notebook platform. Jupyter is a free, web-based, open source approach for creating documents called “Notebooks” that contain rich text and live code which can be executed together in a single document. This interactive document will describe the fundamental equations used in physical hydrology, as would be found in a traditional textbook, but will also include snippets of executable code so that students can examine the influence of different conditions on predicted hydrologic function. The PIs have partnered with the Open Oregon State initiative to develop and make widely available the proposed learning resource.

Publications

Good, S. P., D. R. URycki, and B. C. Crump. 2018. Predicting hydrologic function with aquatic gene fragments. Water Resour. Res. 54: 2424–2435. doi:10.1002/2017WR021974