Session: 08-06-01 non-presentations

Paper Number: 127644

127644 - Prediction of Nutrient Loads Across the Mississippi-Atchafalaya River Basin (Marb) Through Various Deep Learning Methods 

Anoxic zones or dead zones with oxygen concentrations below 2 mg/L cannot sustain the survival of most marine life (such as fish and shrimp). The anoxic zone forms west of the Mississippi Delta on the continental shelf near Louisiana and sometimes extends as far as Texas. The cycle of anoxic zone formation typically begins in late spring, increases in summer, and ends in autumn. Hypoxic zones occur naturally. But in recent decades it has occurred more frequently, forming in shallow water columns, sometimes even in winter. These unusual anoxic behaviors are associated with nutrient-rich discharges during eutrophication in the Mississippi-Atchafalaya River Basin (MARB). When excess nutrients are introduced into the water, the eutrophication process is activated, that is, excess nutrients stimulate algae growth, excess algae blocks the sunlight needed by benthic organisms, and the decomposition of dead algae consumes oxygen.

The focus of this paper is to predict the nutrient loading at United States Geological Survey (USGS) monitoring stations in the MARB using different deep learning approaches, and to better understand the nature of excess nutrient loading delivered to the Gulf of Mexico. Three methods will be applied and was developed to obtain temporal patterns of nutrient yield changes in the Mississippi-Atchafalaya River Basin (MARB), including Multivariate Regression, Random Forest Regression methods, and a fully connected recurrent neural network (RNN).

Some preliminary results have shown that these deep learning approaches can be an effective tool to investigate the temporal and spatial properties of nutrients yields, pinpoint the distribution location of excess nutrient, and provide comprehensive understanding of distribution of nutrient loads across MARB. The statistical results implied that the variations in nutrient loads are jointly affected by the agricultural activities, poultry industry, municipal wastewater, industrial activities, and precipitation.

Presenting Author: Xiaochuan Yu University of New Orleans

Presenting Author Biography: Dr. "Vincent" Xiaochuan Yu is an Associate Professor at Boysie Bollinger School of Naval Architecture and Marine Engineering, University of New Orleans (UNO). Before joining UNO, he worked for SBM Offshore, IntecSea, Technip as senior mooring engineer, senior naval architect, and riser engineer, respectively. Dr. Yu earned his PhD in Ocean Engineering from Texas A&M University (TAMU), MS degree in Civil Engineering from University of Hawaii at Manoa and MS & BS degree in Naval Architcture & Ocean Engineering from Shanghai Jiao Tong University.

Authors:

Xiaochuan Yu University of New Orleans
Hui Zhou UNO
Hanqi Yu UNO
Tianjiu Zhou UNO
Yi Zhen Southern University at New Orleans