Statistical analysis of errors in estimating wet deposition using five surface estimation algorithms

Jeffrey W. Grimm, James A. Lynch

Research output: Contribution to journalArticlepeer-review

13 Scopus citations

Abstract

Wet deposition measurements of H+, SO42- and NO3- from 29 monitoring sites located in (16) and around (13) Pennsylvania, U.S., were analyzed to quantify errors associated with extrapolating point estimates of deposition using five surface-fitting algorithms. The influence of site density on estimation errors associated with each surfacing algorithm was also investigated. The five surfacing differed little in their abilities to predict the concentration or deposition of individual ions found in precipitation in Pennsylvania. However, the size of estimation errors for all parameters, even those based on the densest network, were quite high relative to the variation observed among monitoring sites in Pennsylvania. All monitoring site observations were within 22.8, 17.6 and 23.1 per cent of the median concentration and 33.9, 35.3 and 36.7 per cent of the median deposition for H+, NO3- and SO42-, respectively. Maximum per cent errors indicate that estimation errors may severely obscure actual surface features in at least some portions of the estimated concentration and deposition grids in Pennsylvania. Deposition and concentration estimates based on higher density networks were generally more accurate; however, the improvements afforded by the additional sites were quite modest. Based on the magnitude of estimation errors, kriging produced the most accurate estimates, although no single algorithm consistently yielded the most accurate estimates for all parameters.

Original languageEnglish (US)
Pages (from-to)317-327
Number of pages11
JournalAtmospheric Environment Part A, General Topics
Volume25
Issue number2
DOIs
StatePublished - 1991

All Science Journal Classification (ASJC) codes

  • Pollution

Fingerprint Dive into the research topics of 'Statistical analysis of errors in estimating wet deposition using five surface estimation algorithms'. Together they form a unique fingerprint.

Cite this