Relationships between grid sampling spacing and kriging variance in some grasslands of Ireland and its implications on soil sampling

Sampling plays an important role in acquiring precise soil information which is pursued for modern agricultural production world-wide, because it determines both the cost and quality of final products. It had been proposed... [ view full abstract ]

Sampling plays an important role in acquiring precise soil information which is pursued for modern agricultural production world-wide, because it determines both the cost and quality of final products. It had been proposed that the relationships between kriging variance and grid spacing from an area with existing information could be applied to estimate the grid spacing for sampling neighboring and similar areas in order to achieve desired soil prediction accuracy. However, this approach had never considered the issues of location-dependent soil variation and the parameter uncertainty. In order to investigate these issues in sampling design, this study established and compared the relationships between kriging variance and grid spacing in four grasslands of Ireland, taking into account of parameter uncertainty using the Monte Carlo Markov Chain (MCMC) approach. Results showed that relationships between kriging variance and grid spacing were not monotonic, but quite complicatedly related to study area and sample repeats. Actually, kriging variance was strongly correlated with sample variances. Moderate parameter uncertainty did not have much influences on the relationship between kriging variance and grid spacing, since the total kriging variance, i.e., mean kriging variance (MKV) plus prediction variance due to parameter uncertainty, varied with grid spacing in the same way to the MKV generated using a parameter set having maximum likelihoods. It was concluded that it is still challenging to apply the relationships between kriging variance and grid spacing built on one area to another area to attain desired kriging variance.