An Alternative to Cokriging for Situations with Small Sample Sizes

Abstract

Lack of large datasets in soil protection studies and environmental engineering applications may deprive these fields of achieving accurate spatial estimates as derived with geostatistical techniques. A new estimation procedure, with the acronym Co_Est, is presented for situations involving primary and secondary datasets of sizes generally considered too small for geostatistical applications. For these situations, we suggest the transformation of the secondary dataset into the primary one using pedotransfer functions. The transformation will generate a larger set of the primary data which subsequently can be used in geostatistical analyses. The Co_Est procedure has provisions for handling measurement errors in the primary data, estimation errors in the converted secondary data, and uncertainty in the geostatistical parameters. Two different examples were used to demonstrate the applicability of Co_Est. The first example involves estimation of hydraulic conductivity random fields using 42 measured data and 258 values estimated from borehole profile descriptions. The second example consists of estimating chromium concentrations from 50 measured chromium data and 150 values estimated from a relationship between chromium and copper concentrations. The examples indicate that in situations where the size of the primary data is small, Co_Est can produce estimates which are comparable to cokriging estimates.