A geospatial methods concept defining how Earth-related information is measured, represented, and analyzed in space. It governs coordinate reference, data quality, and analytical operations used to derive patterns, change, and connectivity. It does not guarantee validity without scale awareness, validation, and uncertainty handling for inputs and outputs. It supports decision-making by producing repeatable spatial indicators and maps suitable for review. The concept is generally stable, though sensors, standards, and computation evolve over time.

Principle

Model spatial dependence with a variogram or covariance function, then compute weights that minimize estimation variance under the unbiasedness constraint (ordinary, simple, universal kriging variants reflect differing assumptions about the mean and trend).

Demonstration

Use ordinary kriging with an empirically fitted variogram to estimate soil contaminant concentrations across a remediation site, yielding prediction maps and kriging variance that guide additional sampling and risk assessment.

Misapplication

Fitting variograms to insufficient or clustered data without testing stationarity or anisotropy, or applying kriging when the underlying assumptions (second‑order stationarity or intrinsic stationarity) are violated, leading to biased predictions and overconfident variance estimates.

Consequence

Provides statistically principled predictions with quantified uncertainty, enabling more rigorous risk assessment, sampling optimization, and resource estimation when variogram modeling and diagnostics are performed correctly.

Reversal

Replacing kriging with deterministic interpolators (e.g., IDW, spline) removes the formal error model and the optimality under the chosen covariance, trading quantified uncertainty for simpler, often ad hoc estimates.

Boundary

Requires enough spatial data to estimate a reliable variogram and assumptions about stationarity and trend to be examined; not appropriate where covariance structure cannot be estimated or where non‑stationary, process‑driven variability dominates without transformation or detrending.

Semantic Tension

Contrasted with deterministic methods: kriging emphasizes model‑based uncertainty quantification and optimal linear prediction under a covariance model, while deterministic methods prioritize simplicity or smoothness without formal prediction variances.

Synthesis

Kriging is a geostatistical framework that uses empirical covariance modeling to produce minimum‑variance linear predictions and explicit uncertainty measures; it is most powerful when data support variogram estimation and diagnostic checks confirm modeling assumptions.