Data-Driven Models to Enhance Physically-Based Groundwater Model Predictions

Finally, the applicability of the methodologies and the validity of the complementary modeling framework are tested using both hypothetical and real-world groundwater flow problems of varying complexity. The results indicate that the complementary modeling framework presents a promising and viable alternative to improve groundwater flow predictions, especially, those related to long-term temporal predictions at observation wells and spatial predictions at arbitrary locations. For the real-world groundwater flow problem, the complementary modeling framework reduced MODFLOW's root-mean-square errors (RMSE) for temporal and spatial head predictions by about 78% and 67%, respectively. The uncertainty analysis techniques also significantly improve the estimated 95% confidence and predictions intervals. The percentage of data coverage by the intervals is improved by as much as 88%, while the width of the intervals is diminished.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Civil Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Demissie, Yonas Kassa

Contributors dc:contributor

• Valocchi, Albert J.

• Computer Science

Language dc:language

eng

Identifier

(MiAaPQ)AAI3314759

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/83370