Geostatistical estimation solvers for the GeoStats.jl framework.
This solver provides a high-performance implementation of the inverse distance weighting scheme introduced in the very early days of geostatistics (see Shepard 1968). It is perhaps the simplest first attempt in the literature to perform estimation based on the notion of proximity to data locations.
This implementation makes use of k-d trees from the NearestNeighbors.jl package, which leads to a fast estimation method for large or high-resolution spatial domains. Although this method is recommended for fast assessment of a new field, it has poor statistical properties (lacks covariance model) and should mainly be used for qualitative purposes.
This solver provides an implementation of locally weighted regression (a.k.a. LOESS) introduced by Cleveland 1979. It is the most natural generalization of inverse distance weighting in which one is allowed to use a custom weight function instead of distance-based weights.
Like in the inverse distance weighting solver, this solver makes use of k-d trees from the NearestNeighbors.jl package for fast data lookup. Locally weighted regression (LWR or LOESS) is a popular non-parametric method, however it still has poor statistical properties compared to other estimation methods such as Kriging that explicitly model spatial correlation.
In the current implementation, the estimation variance is computed assuming Gaussian residuals.
This polyalgorithm solver provides an implementation of various forms of Kriging introduced by Matheron 1971. Kriging is a popular method in various industries due to its good statistical properties and flexibility. Unlike the previous solvers, Kriging relies on the specification of a variogram model, which can be fit to geospatial data.
Get the latest stable release with Julia's package manager:
] add GeoEstimation
This package is part of the GeoStats.jl framework.
For a simple example of usage, please check the main documentation.
Asking for help
If you have any questions, please contact our community on the gitter channel.