Jahangard Mohammadi,
Volume 3, Issue 1 (4-1999)
Abstract
The analysis of the EC data set indicated that the spatial distribution of EC data of different depths are closely related to one another. It means that they are spatially cross correlated on one another and can be considered to be co-regionalized. It also implies that EC values at a particular depth contain useful information about the other depths which can be used to improve their estimation. In this research, we aimed to investigate the effects of using relevant ancillary information in the estimation procedure. To do this, cokriging was used. To evaluate this algorithm as a potential tool for mapping EC, its performance on the independent test data was evaluated and compared with the results obtained from studies using kriging. The results of the co-regionalization of EC at different depths indicated that cokriging the salinity data, although more rigorous from theoretical point of view, displayed no advantage over independent ordinary kriging at each depth. The results confirmed that cokriging improves little over ordinary kriging if the primary and auxiliary variables are almost equally sampled and all the variograms are identical. Also, ordinary kriging showed to be quite self-consistent since the predicted average salinity profile over the three depths was almost identical to the one predicted by cokriging. Considering the complexity of the cokriging and the LMC modeling, it is clear that there is no gain in using co-regionalization.
