Introduction to Spatial Data Analysis (COURSE FULL)

Session 5 – 01:15:00 – Spatial relationships & joins
We formalise spatial predicates (intersects, within, contains, touches, within-distance, nearest) and use them to drive reproducible spatial joins. Boundary policies, CRS/units, and performance considerations are covered to ensure accurate results at scale.

Session 6 – 01:15:00 – Network & accessibility basics
Lines are translated into graphs to compute realistic accessibility metrics. You’ll learn how to assign edge weights (distance/time), evaluate shortest paths, and outline simple service areas, contrasting Euclidean with network-based measures and noting practical tooling.

Session 7 – 01:15:00 – Raster fundamentals
We define extent, resolution, origin, alignment, and CRS for rasters, then align disparate layers to a reference grid. You’ll choose appropriate resampling methods for categorical vs continuous data and manage NoData and datatypes for reliable downstream analysis.

Session 8 – 01:15:00 – Raster algebra & terrain
This session introduces map algebra (local/focal), reclassification of categorical rasters, and terrain derivatives (slope, aspect, roughness) from DEMs, highlighting parameter choices and sanity checks for defensible outputs.

Session 9 – 01:15:00 – Areal neighbours & weights
You’ll construct contiguity (rook/queen), distance, and kNN neighbours and build spatial weights matrices with row standardisation. We handle islands, check graph connectivity, and visualise neighbour structures to justify modelling assumptions.

Session 10 – 01:15:00 – Global & local autocorrelation (Moran’s I, LISA)
We explain why autocorrelation matters for inference, compute global Moran’s I with permutation tests, and produce Local Moran cluster maps with appropriate significance interpretation and cautions about multiple testing.

Session 11 – 01:15:00 – Point pattern analysis
We frame point processes and CSR, apply quadrats and kernel intensity estimation, and interpret Ripley’s K/L with edge corrections. The session also introduces inhomogeneous patterns and the role of covariates when intensity varies across space.

Session 12 – 01:15:00 – From points to surfaces: density & grids
You’ll create kernel density estimation (KDE) surfaces and hex/square grids, compare bandwidth/bin-size choices, and normalise counts appropriately to produce clear, honest maps that support decision-making.

Session 13 – 01:15:00 – Variogram theory & practice
We define semivariance and interpret nugget, sill, and range. Directional variograms are used to detect anisotropy, and valid models (Spherical/Exponential/Gaussian) are fitted and assessed for stability and residual structure.

Session 14 – 01:15:00 – Interpolation I: IDW vs Ordinary Kriging
This session contrasts IDW’s deterministic weighting with model-based ordinary kriging. You’ll learn core assumptions, parameter choices, prediction vs uncertainty outputs, and how to compare approaches using principled validation.

Session 15 – 01:15:00 – Interpolation II: Co-Kriging & trend surfaces
We incorporate external drift and covariates via trend surfaces or kriging with external drift (KED), and outline when co-kriging is appropriate. The emphasis is on alignment, scaling, and parsimonious modelling to avoid common pitfalls.

Session 16 – 01:15:00 – Zonal statistics & raster summaries
You’ll compute zonal means and percent cover with exact area weighting and assemble multi-layer summaries for modelling, ensuring consistent CRS/units and defensible choices for polygon handling and NoData.

Session 17 – 01:15:00 – Areal regression: SAR/SEM/SAC & diagnostics
We position SLX against SAR/SEM/SAC models, fit and compare them, interpret SAR impacts (direct/indirect/total), and re-test residual spatial structure to ensure valid inference under your chosen weights matrix.

Session 18 – 01:15:00 – Spatial GLMs & smooths
Poisson/Binomial outcomes are modelled with spatial structure via ICAR/BYM (areal), Matérn Gaussian fields (point-referenced), or Gaussian-process-style smooths. We also include clear cautions about geographically weighted regression (GWR).

Session 19 – 01:15:00 – Spatial cross-validation & leakage
You’ll design blocked or leave-location-out cross-validation, tune models with spatial folds, and avoid feature/target leakage. The outcome is a realistic estimate of generalisation performance.

Session 20 – 01:15:00 – Reproducible pipelines, reporting & capstone presentations
We wrap analyses in targets + renv, organise data and metadata, apply a map-design checklist, and communicate uncertainty. The course concludes with short capstone presentations.