Geostatistical space/time estimation of water quality along the Raritan River basin in New Jersey

The assessment of the river water quality across space and time is a considerable public health concern and it is an important issue for the efficient management of our natural water resources. The state of New Jersey is mandated by the federal Clean Water Act to assess water quality along all streams and rivers in the state, which is critical to designate use attainment and to direct total maximum daily load (TMDL) development. However due to budget and scientific limitations less than 30% of the state's non-tidal stream miles have been assessed. Therefore there is a need to develop a method that can use the partial monitoring information available to estimate water quality along the unmonitored network of streams and rivers. However the high natural variability of water quality over space and time, the limited number of water samples, and the varying levels of measurement errors between samples introduce major sources of uncertainty in the estimation of water quality along rivers and over time. In this work we present the Bayesian Maximum Entropy (BME) framework to rigorously process information about the space/time variability of water quality in its aquatic environment, the uncertainty and scarcity of the monitoring data, and the relevant flow and transport governing laws, in order to obtain statistical estimate of water quality at unmonitored reaches.,We implement the BME method for a case study involving the estimation of phosphate along the Raritan river basin from 1990 to 2002. and we find through cross validation that the BME space/time analysis is a substantial improvement over a purely spatial analysis.