In their 2006 paper, Monestiez and coauthors returned to statistical “first-principles” to
develop the Poisson kriging model [42]. The thought process was as follows: if properties
of the conditional and marginal distributions of the Poisson kriging model were properly
exploited, then the theoretical semivariogram of the latent distribution could be represented
as a function of the observed counts. After finding the theoretical form of the semivariogram,
a natural estimator could be derived using only the observed Poisson counts. Monestiez et.
al were able to use simple properties of the variances and expected values to derive their
final