GSA Annual Meeting in Seattle, Washington, USA - 2017

Paper No. 366-7
Presentation Time: 9:00 AM-6:30 PM

PREDICTION OF TOTAL PHOSPHORUS CONCENTRATION IN NEW YORK CITY USING ARTIFICIAL NEURAL NETWORK


JAHAN, Khurshid and PRADHANANG, Soni, Department of Geoscience, University of Rhode Island, 45 Upper College Road, Kingston, RI 02881, khurshidjahan@uri.edu

Excessive and lower limit of total phosphorus would degrade water quality and harm ecosystems. In water bodies having total phosphorus concentrations less than 10 parts per billion, waters will be nutrient-poor and will not support large quantities of algae and aquatic plants. At the other extreme, total phosphorus levels of 100 or more ppb categorize lakes as highly eutrophic, with high nutrient and algae levels. The purpose of this study is to forecast the total phosphorus applying artificial neural network (ANN) using eight water quality variables such as Total Nitrogen (TN), Total Suspended Sediment (SS), Air Temperature, Precipitation, Total Dissolved Phosphorus (TDP), Total Soluble Phosphate (TSP), Soluble Reactive Phosphorus (SRP), and Stream flow were applied as inputs to the network. Lately, the neural networks method has been applied to many branches of science. The approach is becoming a reliable application for assessing the forecast, and real scenario of the events or situation. This study applied on Cannonsville watershed (891 km2), one of the major watersheds in the New York City water supply system. Both land use change (mostly decline in agriculture) and watershed management created significant decreases in P loading. In Cannonsville, loading reduction ~18% for dissolved P produced due to land use change and the combination of land use change and watershed management produced reductions of ~55% for dissolved P. If the load reduction is following this trend, the water quality and the ecosystem are going to be affected severely. For this study, two different types of artificial neural network (BPNN—static neural network; NARX network—dynamic neural network) are constructed in modeling the dynamic system. Sensitivity analyses also did here to determine the influence of input variables on the dependent variable due to the accuracy of the study result. Long term historical time series daily data (DEC has collected the samples and calculated the nutrient Load) was used to analyses, and other meteorological data were collected from the Northeast Regional Climate Center Meteorology, and USGS Streamflow. The findings of this study will be beneficial to the management of the water resources of NYC water supply and for the entire ecosystem.