Premium
Cokriging Limited Transmissivity Data Using Widely Sampled Specific Capacity from Pump Tests in an Alluvial Aquifer
Author(s) -
Hughson Lance,
Huntley David,
Razack M.
Publication year - 1996
Publication title -
groundwater
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.84
H-Index - 94
eISSN - 1745-6584
pISSN - 0017-467X
DOI - 10.1111/j.1745-6584.1996.tb01859.x
Subject(s) - kriging , mathematics , statistics , regression , linear regression , data set , regression analysis , aquifer properties , soil science , aquifer , environmental science , geology , groundwater , geotechnical engineering , groundwater recharge
Use of the specific capacity of a pumping well to predict aquifer transmissivity is desirable due to the cost of pumping tests and the availability of specific capacity measurements. The geostatistical technique of cokriging is a method of incorporating the spatial variability of a correlated variable (e.g., specific capacity) in estimating a related undersampled variable (e.g., transmissivity). This study examines the reliability of cokriging transmissivity estimates using a data set of 215 pairs of transmissivity and specific capacity. Subsets of pairs of transmissivity and specific capacity were selected and cokriged to estimate transmissivity at the remaining well locations. The estimates of transmissivity were then compared to actual measurements of transmissivity. The same subsets of pairs were used to estimate transmissivity with loglinear regression of transmissivity on specific capacity and ordinary kriging of transmissivity alone. Comparison of these three methods indicates the number of wells with both transmissivity and specific capacity data necessary to obtain improvement in transmissivity estimates with cokriging over the simpler regression and kriging methods. The results show that significant improvement in the transmissivity estimate is obtained by cokriging with 50 or more pairs of transmissivity and specific capacity, and that loglinear regression is superior when less than 30 pairs are available. With between 30 and 50 pairs of available data measurements, cokriging does not reliably improve the estimate over loglinear regression.