Comparison of Spatial Interpolation Methods for Water Quality Evaluation in the Chesapeake Bay

Abstract

Spatial interpolation methods are frequently used to estimate values of physical or chemical constituents in locations where they are not measured. Very little research has been conducted, however, to investigate the relative performance of different interpolation methods in surface waters. The study reported here uses archived water quality data from the Chesapeake Bay to compare three spatial interpolation methods: inverse distance weighting, ordinary kriging, and a universal kriging method that incorporates output from a process-based water quality model. Interpolations were performed on salinity, water temperature, and dissolved oxygen “snap shots” (cruise-based data sets) taken between 1985 and 1994 at 21 different depths for multiple locations in the mainstem Bay, using data compiled by the prototypical Chesapeake Bay Environmental Observatory. The kriging methods generally outperform inverse distance weighting for all parameters and depths. Incorporating output from the water quality model through universal kriging appears to improve some of the interpolations by specifically accounting for some physical and biogeochemical features of the estuary. Such integration of process-based information with statistical interpolation warrants further study.