Surface Water Nutrient Estimation With Bayesian Maximum Entropy Space/Time Geostatistics
Jamie Smedsmo, University of North Carolina
The objective of this research is to develop an improved method to estimate spatially and temporally varying nutrient loading to surface waters. Excess nutrient loading causes water quality problems such as eutrophication of inland lakes, low dissolved oxygen along river reaches and eutrophication in downstream estuarine systems. We will develop a method to map nitrogen loading along the Cape Fear River in North Carolina to better understand the impact of various point and non-point sources on nutrient loads along the river.
Nutrient loading along rivers is generally estimated using mechanistic watershed models (e.g. SWAT, LSPC), which use watershed characteristics along with forcing data to estimate flow and constituent concentrations along a river. However, these models are quite resource-intensive to implement, particularly at the scale of a major river basin. As a result, predictive uncertainty estimates, which require multiple model runs, usually are not provided. A structurally simple statistical model, in contrast, may be sufficient for estimating nutrient loading with the benefits of being easier to implement and providing uncertainty associated with estimates.
Our model uses the Bayesian Maximum Entropy (BME) method of modern geostatistics to combine two knowledge bases. First, a land-use regression model provides an estimate of the mean trend of nitrogen along the river. Second, nitrogen concentration measurements provide accurate data but samples are limited in space and time. BME combines these two sources of information to produce a continuous map of nitrogen along the river. This map will provide a tool for managers to understand the combined impact of point sources, land use, and groundwater leaching on nitrogen loads as a function of space and time.
Abstract Author(s): Jamie L. Smedsmo, Marc L. Serre