User guide

This guide explains how APbase turns irregular field data into validated maps. APbase is available as the Python package apbase and the Rust crate apbase; each example has a tab for each language. Start with Installation and Quickstart to install it and create your first map.

For normal map generation, start with Map or create_map, which run the full pipeline described in Map and grid generation. Use the lower-level components only when you need to inspect or control one stage.

For service, worker, container, and cloud deployment guidance, see Production and cloud deployment.

Reading the parameters in Rust

The parameter lists in this guide use the Python names and types. The Rust crate keeps the same names and defaults, with these conventions:

• Required settings are constructor arguments (MapOptions::new(resolution)); optional ones are fields of a <Name>Options struct that implements Default, so you set only what you change: IDWOptions { power: 3.0, ..Default::default() }.
• Estimators follow new(options)? -> fit(&x, &y, &z)? -> interpolate(&targets)?.
• Fitted values are read through getter methods, not attributes: kriging.model_params()?.range, ck.model_values()?.
• None becomes Option: radius: Some(50.0), seed: Some(42), filter_kwargs: Some(SpatialFilterOptions { .. }).
• String choices become enums, such as FilterLevel::Light or CoKrigingMethod::Icm; they also parse from the Python strings ("icm".parse()?).
• array_like is any slice, Vec, array, or ndarray array of a numeric type; targets is an (n, 2) ndarray array, a slice or Vec of [x, y] pairs, a slice or Vec of point geos::Geometry values, or a fitted Grid.
• Dictionaries of variables are IndexMap<String, (Array1<f64>, Array1<f64>, Array1<f64>)>, and results by method are IndexMaps in a stable order (cv.by_method["icm"].rmse). The cokriging examples need the indexmap crate, and some examples use ndarray types directly.
• Every fallible call returns Result<T, ApbaseError>. e.is(ErrorKind::InterpolationError) is true for that kind and for every more specific kind under it, such as KrigingInvalidRadiusError.

Pages in this guide

• Map and grid generation: the automatic Map pipeline, its result, and manual target grids.
• Spatial filtering: removing invalid rows and spatial outliers.
• Variogram: fitting and inspecting the variogram.
• Inverse distance weighting (IDW): inverse distance weighting.
• Ordinary kriging: local ordinary kriging.
• Cross-validation: the IDW-vs-kriging comparison behind Map.
• Cokriging: estimating with secondary variables (advanced).

For the mathematics behind the automatic IDW-vs-kriging choice, see Methodology. For public signatures, see API Reference.