Skip to contentScholarGate
LibraryBookshelfDeskReview StudioAssistant
Sign in
On this page
IntuitionHow it worksWhen to use itStrengths & limitationsCommon pitfallsApplicationsFrequently asked🔒 Read the full methodSourcesRelated methods
Cite this pageSpotted an issue on this page? Report or suggest a fix →
Home›Bayesian›Spatial Kalman Filter
Bayesian methodsBayesian / computational

Spatial Kalman Filter

Spatial Kalman Filter for Spatio-Temporal State-Space Models · Also known as: spatial state-space filter, spatio-temporal Kalman filter, SKF, spatial dynamic linear model

The spatial Kalman filter applies classical Kalman filtering to spatio-temporal state-space models, treating a spatially distributed latent field as the hidden state that evolves over time. At each time step, the filter recursively predicts the spatial field forward and then updates the prediction with new spatial observations, producing optimal linear estimates of the field and its uncertainty across all locations.

ScholarGate
  1. Bayesian methods
  2. v1
  3. 2 Sources
  4. PUBLISHED
Cite this page →
Tools & resources
Download slides
Learn & explore

Read the full method

Members only

Sign in with a free account to read this section.

Sign in

Method map

The neighbourhood of related methods — select a node to explore.

Spatial Kalman Filter
Dynamic Bayesian Inferen…Kalman FilterParticle FilterSequential Monte CarloSpatial Bayesian Inferen…Spatial MCMC

When to use it

Use the spatial Kalman filter when you have repeated measurements from a spatial network over time and want to (1) reconstruct the full spatial field at all locations, (2) propagate and quantify spatial uncertainty, and (3) update estimates sequentially as data arrive. It is particularly suited to environmental monitoring, remote sensing gap-filling, and geophysical data assimilation. Do not use it when the spatial field is truly static (use kriging instead), when observations are not sequential in time, when non-Gaussian or nonlinear dynamics dominate without a particle filter extension, or when the spatial dimension is so large that even reduced-rank approximations are impractical.

Strengths & limitations

Strengths
  • Provides optimal linear estimates of the spatially distributed field and its full uncertainty covariance at each time step.
  • Sequential updating is computationally efficient: new data are assimilated without reprocessing all past observations.
  • Naturally handles missing observations at any time step — the predict step still runs, and the update step simply uses whatever sensors are available.
  • Reduced-rank and basis-function variants scale to continental or global spatial domains.
  • Produces spatially smooth, physically coherent reconstructions because the covariance structure encodes spatial continuity.
Limitations
  • Assumes linear state and observation equations and Gaussian noise; nonlinear or non-Gaussian settings require extended variants (EKF, UKF, or particle filter).
  • The state covariance matrix grows quadratically with the number of spatial locations, making exact computation infeasible for large grids without approximations.
  • Results depend on the correct specification of the spatial covariance structure (Q_t, H_t, P_0); misspecification leads to over- or under-smoothed fields.
  • Estimating unknown model parameters (transition matrices, covariance hyperparameters) requires an outer optimisation loop (maximum likelihood or MCMC), which can be expensive.

Frequently asked

How does the spatial Kalman filter differ from ordinary kriging?

Kriging produces a single optimal spatial interpolation from a static dataset, with no temporal dynamics. The spatial Kalman filter extends this to a time-series setting: the spatial field evolves according to a dynamic model and estimates are updated each time step as new observations arrive. Kriging is a special case of the spatial Kalman filter's update step when the field is treated as static.

What if the spatial domain is very large (thousands of grid cells)?

The exact Kalman filter is computationally infeasible for large grids because the state covariance matrix scales as O(n^2). Common solutions include reduced-rank representations using a small number of spatial basis functions (e.g., empirical orthogonal functions or fixed-rank kriging), sparse precision matrices exploiting the SPDE/Matern connection, or Vecchia approximations that condition on only nearby neighbours.

Can the spatial Kalman filter handle nonlinear dynamics?

Not directly. The standard filter assumes linear Gaussian dynamics. For nonlinear state equations you can linearise locally using the Extended Kalman Filter (EKF) or use the Unscented Kalman Filter (UKF) for smoother approximations. For strongly non-Gaussian or highly nonlinear problems, ensemble Kalman filters or particle filters are preferred.

What is the difference between filtering and smoothing?

Filtering provides the best estimate of the spatial field at time t using only data up to t — this is the online, real-time estimate. Smoothing uses all data (past and future) to produce retrospective estimates at each time step, which are more accurate but require storing and reprocessing the full time series in a backward pass (e.g., the Kalman smoother / RTS smoother).

How do I estimate the covariance parameters (Q, H, P0)?

These hyperparameters are typically estimated by maximising the marginal likelihood (the log-likelihood computed via the innovations from the prediction step), which can be done with gradient-based optimisation. Alternatively, a fully Bayesian approach places priors on the hyperparameters and samples them via MCMC alongside the latent states.

Sources

  1. Cressie, N. & Wikle, C. K. (2011). Statistics for Spatio-Temporal Data. Wiley. ISBN: 978-0-471-69274-4
  2. Kalman, R. E. (1960). A new approach to linear filtering and prediction problems. Journal of Basic Engineering, 82(1), 35-45. DOI: 10.1115/1.3662552 ↗

How to cite this page

ScholarGate. (2026, June 3). Spatial Kalman Filter for Spatio-Temporal State-Space Models. ScholarGate. https://scholargate.app/en/bayesian/spatial-kalman-filter

Related methods

Dynamic Bayesian InferenceKalman FilterParticle FilterSequential Monte CarloSpatial Bayesian InferenceSpatial MCMC

Which method?

Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.

  • Dynamic Bayesian InferenceBayesian↔ compare
  • Kalman FilterBayesian↔ compare
  • Particle FilterBayesian↔ compare
  • Sequential Monte CarloBayesian↔ compare
  • Spatial Bayesian InferenceBayesian↔ compare
  • Spatial MCMCBayesian↔ compare
Compare side by side →

Similar methods

Time Series Kalman FilterKalman Filter with Measurement ErrorKalman Filter with Missing DataSpace-Time KrigingKalman FilterState Space ModelSpace-Time Ordinary KrigingHierarchical Kalman Filter

Related reference concepts

Data AssimilationGaussian Process ModelsHidden Markov ModelsSpatial Point ProcessesTime-Series Models • Dynamic Quantile Regressions • Dynamic Treatment Effect Models • Diffusion Processes • State Space ModelsEM Algorithm

Spotted an issue on this page? Report or suggest a fix →

ScholarGate — Spatial Kalman Filter (Spatial Kalman Filter for Spatio-Temporal State-Space Models). Retrieved 2026-07-21 from https://scholargate.app/en/bayesian/spatial-kalman-filter · Dataset: https://doi.org/10.5281/zenodo.20539026
Quick facts
Originator
R. E. Kalman (base filter, 1960); extended to spatial settings by Cressie, Wikle and colleagues
Year
1960 (base); spatial extensions 1990s–2000s
Type
Bayesian state-space model
DataType
spatio-temporal observations (gridded or point-referenced)
Subfamily
Bayesian / computational
Related methods
Dynamic Bayesian InferenceKalman FilterParticle FilterSequential Monte CarloSpatial Bayesian InferenceSpatial MCMC
ScholarGate

A content-first reference library for research methods — what each one is, how it works, and where it comes from.

Open data (CC-BY)

Explore

  • Library
  • Search the library…
  • Browse by field
  • Fields
  • Journey
  • Compare
  • Which method?

Reference

  • Subjects
  • Atlas
  • Glossary
  • Methodology
  • Philosophy

Your tools

  • Bookshelf
  • Desk
  • Chat

Company

  • About
  • Pricing
  • Contact
  • Suggest a method

Entries are compiled from published sources for reference. Verifying the accuracy and suitability of any information for your own use remains your responsibility.

© 2026 ScholarGate · A research-method reference library
  • Privacy
  • Cookies
  • Terms
  • Delete account