Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Environmental Microbial & Food Safety Laboratory » Research » Publications at this Location » Publication #392635
Research Project: Improving Pre-harvest Produce Safety through Reduction of Pathogen Levels in Agricultural Environments and Development and Validation of Farm-Scale Microbial Quality Model for Irrigation Water Sources

Location: Environmental Microbial & Food Safety Laboratory

Title: Machine learning in vadose zone hydrology: a flashback

Author
item GHANBAIAN, BEHSAD - KANSAS STATE UNIVERSITY
item Pachepsky, Yakov

Submitted to: Vadose Zone Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/24/2022
Publication Date: 6/30/2022
Citation: Ghanbaian, B., Pachepsky, Y.A. 2022. Machine learning in vadose zone hydrology: a flashback. Vadose Zone Journal. https://doi.org/10.1002/vzj2.20212.
DOI: https://doi.org/10.1002/vzj2.20212

Interpretive Summary: The drastic increase in artificial intelligence and machine learning applications in geosciences obscures the need and slow development of research on basic matters such as the understanding and quantifying reliability and transferability of machine learning results, and issues of delivery research machine learning results to the developer and user community. These topics were noted and addressed in the beginning of machine learning applications, and this short communication is intended to serve as a reminder of major thought-provoking findings made in the beginning of the machine learning applications in geoscience.

Technical Abstract: Artificial intelligence (AI) and machine learning (ML) have been recently applied extensively in various disciplines of geosciences. However, not much attention has been paid to their database-dependent accuracy and uncertainty, reproducibility, and delivery, which undermines their applications to real world problems. We discuss lessons from the past and emphasize the need in and lack of fundamental protocols i.e., detailed clarification on data processing, ML models accessibility, and a clear path for reproducing results.