Geostatistics is a branch of statistics that focuses on data sets of variables in space, known as regionalized variables, in which each value is associated with a particular position in space.[1]​[2]​ These are not standard statistical techniques applied to solving problems specific to geology, but rather a set of particular methods that aim to model this particular type of variables. In any phenomenon with spatial expression in which the values ​​of the variable(s) of interest are influenced by the proximity between the points studied, a basic assumption of conventional statistical methods is not met, which is the independence between the observations.[1]​ For example, the characteristics of an aquifer at two nearby points will be associated with the regional geological aspects so that, for example, the depth variable will have more similar values ​​when said points are closer. That is why geostatistical methods were developed, which incorporate this dependence between values ​​into the models.

Geostatistics began its development in relation to the prediction of probability distributions of ore grades for gold mining operations in South Africa, in the Witwatersrand region, also known as the Rand, promoted by mining engineer Danie G. Krige.[2][3] Subsequently, French engineer Georges Matheron with his group at the School of Mines in Paris, France, developed these ideas rigorously.[4]

It is currently applied in various disciplines such as: petroleum geology, hydrogeology, hydrology, meteorology, oceanography, geochemistry, geometallurgy, geography, forestry engineering, forestry, environmental control, landscape ecology, pedology, agriculture (especially precision agriculture), climatology, among others. Geostatistics is applied in various branches of geography, such as the study of disease spread (epidemiology), trade and military planning (logistics), and the development of efficient spatial networks. Geostatistical algorithms have been incorporated in many places, including geographic information systems (GIS) and the R statistical environment (R programming language).

Geostatistics develops various estimation and simulation procedures, which are used to study spatially distributed variables. It is carried out from a number of samples taken in locations within the domain, in which a phenomenon to be studied is manifested and considered representative of its reality, which is generally always unknown. Therefore, its main objective is to estimate unknown values ​​from the known ones, seeking to minimize the variable in the estimation error.[5]​.

• - Multivariable interpolation.

• - Spatial analysis.

• - Kriging or Kriging.

• - Theory of regionalized variables.