Advancing river corridor science beyond disciplinary boundaries with an inductive approach to hypothesis generation

This is a Preprint and has not been peer reviewed. This is version 1 of this Preprint.

Downloads

Download Preprint

Authors

Adam Scott Ward , Jennifer Drummond, Angang Li, Anna Lupon, Marie Kurz, Jay Zarnetske, James Stegen, Eugenia Marti, Valerie Ouellet, Nicolai Brekenfeld, Feng Mao, Emily Graham, Susana Bernal, Stefan Krause, David Hannah, Aaron Packman, Todd Royer, Megan Klaar

Abstract

Traditional, deductive approaches have generated a large body of site-, scale-, process-, or method-specific understanding of the physical, chemical, and biological processes that occur within river corridors. However, this body of facts does not until itself constitute a predictive understanding of river corridors in their full complexity. We contend a new paradigm is required to synthesize existing knowledge with the goal of linking internal dynamics, external forcing, and historical contingencies to the emergent spatial structure, temporal dynamics, and ecosystem services that are derived from river corridors. Here, we prototype an inductive approach to synthesis, using machine learning as a hypothesis generator to identify potential couplings or feedbacks that would not necessarily arise from classical, deductive, disciplinary approaches. This approach generated a network of 672 relationships linking a suite of 157 variables each collected at 62 locations in a 5th order river network, 84% of which have not been previously co-investigated in the literature. We document the critically important role of collecting data beyond disciplinary norms (89% of predictive models required out-of-group data for optimal prediction), and both the emergence and shredding of spatial structure as variables combine to explain observed patters in the network. This study demonstrates the value of a hypothesis generation approach that is agnostic to disciplinary boundaries and pre-existing conceptual models as a compliment to traditional, deductive models of inquiry. Ultimately, the network of multi-scale, cross-disciplinary relationships generated here may catalyze new ideas and conceptualizations that would not be obvious starting from pre-existing conceptual models and approaches.

DOI

https://doi.org/10.31223/X54W44

Subjects

Hydrology

Keywords

machine learning, stream corridor, inductive

Dates

Published: 2021-05-08 18:07

License

CC BY Attribution 4.0 International

Additional Metadata

Conflict of interest statement:
None

Data Availability (Reason not available):
http://www.hydroshare.org/resource/de6d92d314354ea6819157818669fc59

Add a Comment

You must log in to post a comment.


Comments

There are no comments or no comments have been made public for this article.