Using Deep Learning to Fill in Gaps in Temporal Data Presented on Day 4 of AGU22

Big DataBig Data

Posted: June 18, 2024

Using Deep Learning to Fill in Gaps in Temporal Data Presented on Day 4 of AGU22

“Mixed Success” from current water quality policy and management strategies, research shows.

Phosphorus is a tricky thing. Without that chemical, things won’t grow. Too much of that chemical and the wrong things grow. Hydrologists, soil scientists, and other Critical Zone researchers keep a close eye on phosphorus levels in the places they study.

Big Data Cluster member Li Li is part of a team of researchers from Penn State University who went to find out more about the effectiveness of water quality improvement policies they discovered large temporal gaps in the data they would rely on to find some answers.

So they turned to computer and data science. The team reports that they trained “a continental-scale deep learning model for 513 basins across the U.S. and reconstructing their continuous daily total phosphorus (TP) records over the past 40 years.”

The team presented their latest findings at the American Geophysical Union (AGU) Annual Fall meeting in Chicago in December of 2022. Their analysis revealed that the passage of the Clean Water Act in 1972, designed to reduce the phosphorus runoff from urban areas and management measures to control agricultural runoff in the 1990s shows different levels of effectiveness.

In urban areas, they describe, total phosphorus has declined. In agricultural areas the phosphorus levels have increased, “with a noticeable deterioration of water quality in the Midwest,” the paper presented at AGU goes on to describe.

While the water quality aspects of their work had mixed results, the team maintains that these “results highlight the promise of deep learning techniques to improve predictability of riverine water quality.”