Author
 |
LESCH, S - UC RIVERSIDE, CA |
|
Submitted to: Proceedings of the International Salinity Forum
Publication Type: Proceedings Publication Acceptance Date: 4/11/2005 Publication Date: 4/25/2005 Citation: Lesch, S.M. 2005. Statistical techniques for calibrating regional (multi-field) EM survey data. In: Proceedings of the International Salinity Forum, Managing Saline Soils and Water: Science, Technology, and Soil Issues. April 25-27, 2005. Riverside, CA pp:289-292. Interpretive Summary: In this study we examine the strengths and weaknesses of two commonly used spatial prediction techniques, cokriging (CoK) and hierarchical spatial regression (HSR), specifically with respect to predicting various soil properties from calibrating regional EM survey data. We demonstate that the HSR modeling approach possesses a number of advantages over CoK, and suggest that it is better suited for modeling regional EM survey data. Some specific examples of these advantages will be discussed (in the oral presentation) using the 1991 Broadview Water District regional EM / soil salinity survey. Technical Abstract: The purpose of this research is to review and compare two commonly used spatial prediction techniques, cokriging (CoK) and hierarchical spatial regression (HSR), specifically with respect to predicting various soil properties from calibrating regional EM survey data. A brief review of the mathematical modeling assumptions behind each statistical technique is presented, and then the relative merits of each approach are discussed. The four advantages of the HSR approach are then highlighted; i.e., (i) model flexibility, (ii) estimation efficiency, (iii) inference techniques for parameter testing, and (iv) succinct model specification. |
