Authors: Michael R Desjardins*, Johns Hopkins Bloomberg School of Public Health, Frank C Curriero, Johns Hopkins Bloomberg School of Public Health, Benjamin J.K. Davis, Exponent Inc
Topics: Quantitative Methods, Geographic Information Science and Systems, Water Resources and Hydrology
Keywords: Geostatistics, INLA, Space-time, Non-Euclidean, Water Quality
Session Type: Virtual Paper
Start / End Time: 3:05 PM / 4:20 PM
Room: Virtual 25
Presentation File: No File Uploaded
The Chesapeake Bay is one of the most widely studied bodies of water in the United States and around the world. Routine monitoring of water quality indicators (e.g. salinity) relies on fixed sampling stations throughout the Bay. Utilizing this rich monitoring data, various methods develop continuous surfaces of water quality indicators to further characterize the health of the Bay as well as to support wildlife and human health research studies. Statistical methods have successfully been applied in this arena mostly focused on development of spatial prediction approaches for generating continuous water quality surfaces. However, most studies often just focus on spatial prediction, not fulling utilizing the rich and temporally connected aspect of the water quality monitoring data. Considering both the space and time dimensions together allows models to be more flexible to account for longer term temporal trends as well as shorter term space-time interactions for improved predictions. For this study, we utilized a Bayesian space-time Integrated Nested Laplace Approximation (INLA) approach to produce continuous prediction surfaces of salinity in the Chesapeake Bay between 2010 and 2019 at the monthly level. Cross-validation was conducted to evaluate model performance based on prediction accuracy and precision. Our INLA models also showcase how non-Euclidean prediction techniques can outperform traditional Euclidean-based geostatistical methods in complex bodies of water. For example, kriging will not account for physical barriers (such as land between tributaries), which can result in inaccurate predictions of water quality. Future research will include other key water quality indicators (e.g. turbidity).