While the advancements in computation and learning algorithms have been of value, there remains uncertainty when modeling spatially-explicit data. Specifically, attention needs to be given to spatial autocorrelation and the modifiable areal unit problem that are interlinked on some level. Spatial auto-correlation is a correlation between values of a single variable at nearby (?) locations that lead to deviation of the independence assumption in classical statistics (CITE). It could be described as a varying map pattern,  redundant information in the dataset, or an underlying spatial process mechanism. Interpreting auto-correlation as an outcome of aerial unit demarcation relates it to the modifiable areal unit problem (MAUP). By simply changing the surface partitioning used to demarcate areal units, results of statistical analyses of spatial data can be varied. Similarly, by changing the resolution to satellite imagery to coarser resolution will decrease the level of auto-correlation and could lead to negative auto-correlation. 
Previous work applied an effective sample size estimation (ESS) to account for spatial autocorrelation (SA) in the data. The application of ESS in statistical inference is based on interpretation of SA as redundancy in the data. In other words, there is an equivalent amount of information present in uncorrelated samples as the correlated observations(\citet{Cressie_1993,Schabenberger2005,Griffith2003,Vallejosa}). Effective sample size is a function of both spacing between and intensity of locations at which measurements are taken. For areal units, it refers to resolution. One measure that quantifies interpoint spacing is the first nearest neighbor (nn) statistic,whose index is given by