A Comparison of Linear and Nonlinear Random Field Estimators
Name
10422941.pdf
Size
7.77 MB
Format
Adobe PDF
Checksum (MD5)
2721603267d46e91406a275312956f24
Author(s)
Angulo, Carlos Enrique Puente
Bras, Rafael L.
Date Issued
November 1982
Publisher
Cambridge, Mass. : Ralph M. Parsons Laboratory, Hydrology and Water Resource Systems, Massachusetts Institute of Technology, Dept. of Civil Engineering
Series/Report no.
R (Massachusetts Institute of Technology. Department of Civil Engineering) ; 82-51.
Report (Ralph M. Parsons Laboratory for Water Resources and Hydrodynamics) ; 287.
Abstract
The estimation of random fields from limited samples is an important issue in most fields of geophysics, such as Hydrology and Meteorology. Work by Matheron and others at the Paris School of Mines has popularized Kriging techniques to estimate random fields at specified locations or to get areal averages. This work presents the theoretical and practical aspects of both Linear and Nonlinear (Disjunctive) Kriging estimators, and provides a comparison of their performance in estimating point and areal values of generated fields. The experiments performed were designed to closely resemble actual and practical situations. The results show that small sample based inconsistencies lead to a Disjunctive Kriging solution which does not give more accurate estimates than the theoretically less precise Linear Kriging estimator. The results also suggest the use of a multi-realization approach when using these techniques in network design problems.
Description
Scanning notes: Disclaimer inserted for illegible graphs and text.
Prepared with support of National Science Foundation and Office of Surface Mining through Grants CME-7919836 G5105071
Persistent DSpace Link
