Space-Time Spatial Autocorrelation
Space-Time Spatial Autocorrelation Analysis · Also known as: STSA, spatiotemporal autocorrelation, space-time Moran's I, temporal spatial dependence
Space-Time Spatial Autocorrelation extends classic spatial autocorrelation measures — most notably Moran's I — to data that vary across both geographic units and time periods. It detects whether nearby locations that are also temporally close tend to share similar attribute values, revealing clusters, trends, or anomalies that purely spatial or purely temporal analyses would miss.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
+3 more
When to use it
Use space-time spatial autocorrelation when you have georeferenced observations recorded at multiple time points and you suspect that the spatial clustering structure changes over time or that temporal propagation (e.g., disease spread, crime diffusion, economic contagion) is present. It is especially valuable for public health surveillance, environmental monitoring, and regional economics. Do not apply it when observations are truly independent across time, when the time series is very short (fewer than three periods), or when spatial units change their boundaries across periods, as these conditions invalidate the weight matrix assumption.
Strengths & limitations
- Simultaneously captures spatial dependence and temporal dynamics in a single coherent framework.
- Detects emerging clusters or diffusion processes that pure cross-sectional or pure time-series methods overlook.
- Local LISTA statistics pinpoint specific location-period combinations driving the global signal.
- Non-parametric permutation inference makes it robust to non-normal attribute distributions.
- Widely supported in mature software (GeoDa, PySAL, R spdep) with established visualisation tools.
- Requires balanced panel data — the same set of spatial units observed at each time period.
- The choice of spatial weight matrix (contiguity vs. distance, bandwidth) and temporal lag window strongly influences results, and there is no universally optimal selection rule.
- Computational cost scales with n×T, making large space-time grids (e.g., satellite pixels × daily observations) expensive.
- The statistic detects linear association only; nonlinear or threshold-based space-time dependence may be missed.
Frequently asked
How is space-time Moran's I different from running separate Moran's I for each time period?
Running Moran's I period-by-period treats each cross-section independently and cannot detect whether clusters persist, migrate, or emerge across time. Space-time Moran's I uses a joint weight matrix that links nearby observations in both dimensions simultaneously, so it captures the continuity and propagation of spatial dependence over time.
What software can compute space-time spatial autocorrelation?
GeoDa (free, GUI-based) implements space-time local autocorrelation statistics. The Python library PySAL (esda module) and the R package spdep both provide space-time extensions of Moran's I with permutation inference.
How do I choose the temporal lag?
The lag should reflect the substantive process: one week for disease spread, one year for economic cycles. Where theory is unclear, compute the statistic at multiple lags and inspect the space-time correlogram, but apply a Bonferroni correction or FDR adjustment to guard against spurious significance.
Can I use this method with irregular time intervals?
Standard implementations assume regular (equally spaced) time intervals because the temporal weight matrix uses fixed lag windows. Irregular intervals require custom weight specifications or continuous-time spatial models such as geostatistical space-time kriging.
What if my spatial units have changed boundaries over time (e.g., census tract redrawing)?
Changing boundaries invalidate the assumption of a fixed spatial weight matrix across periods. You must either interpolate data to a consistent geography using areal interpolation, or restrict the analysis to periods with stable boundaries.
Sources
- Clifford, P., Richardson, S., & Hemon, D. (1989). Assessing the significance of the correlation between two spatial processes. Biometrics, 45(1), 123–134. DOI: 10.2307/2532039 ↗
- Anselin, L., & Getis, A. (1992). Spatial statistical analysis and geographic information systems. The Annals of Regional Science, 26(1), 19–33. DOI: 10.1007/BF01581478 ↗
How to cite this page
ScholarGate. (2026, June 3). Space-Time Spatial Autocorrelation Analysis. ScholarGate. https://scholargate.app/en/spatial-analysis/space-time-spatial-autocorrelation
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.
- Geographically Weighted RegressionSpatial analysis↔ compare
- Global Moran's ISpatial analysis↔ compare
- Local Moran's ISpatial analysis↔ compare
- Spatial Lag ModelSpatial analysis↔ compare
- Spatial Panel ModelSpatial analysis↔ compare