How to use

Use the layer control (top-right) to compare surface heat patterns with land cover and urban structure. Toggle a few layers at a time for clarity — each active layer's color scale appears at the bottom. Click the ? next to a layer to open its description; click ? again, the ×, press Esc, or click the map to close. Long text scrolls inside the panel.

LST tiles: Landsat land surface temperature in °C — not the same as 2 m air temperature.
Optional ECOSTRESS LST composite (~70 m) — companion to Landsat LST; irregular ISS coverage. Details in Layer guide / layer ? (surface ≠ air).
District summaries: zonal means (ΔT, mean LST, mean NDVI), not per-pixel temperature.
Municipality outlines: official polygons from the City of Skopje GIS portal (ArcGIS MapServer Rock/municipalities_skopje, layer Opstini).
Census 2021 (Skopje) — T-01en population / household / dwelling counts with Wikipedia municipality age shares (0–14, 15–64, 65+); heat exposure indices = min–max(zonal mean LST) × min–max(census count). Polygons use transparent fills and popups (no centroid circles). Land Surface Temperature (Tiles) is the only overlay on by default; turn census layers on from the layer control.
Each layer’s ? tooltip lists its data sources (Landsat, Google Earth Engine, OpenStreetMap, baked XYZ tiles).
Open Layer guide for interpretation notes (surface vs air, baked tiles vs optional live LST).

SmartUp

Changelog (Skopje map)

Changelog — Skopje summer surface heat map

What changed in the Skopje interactive map you are viewing. Open this panel from the header Changelog button; this text is copied into the HTML when the file is exported from the repository.

Format follows Keep a Changelog (see keepachangelog.com). Dates refer to May 2026 work batches.

[Unreleased]

Census on municipalities (2021): Population, households, dwellings, and age bands (0–14, 15–64, 65+) as opština choropleths and heat-exposure style indices. In the layer control, open Census 2021 — Age (Skopje) for the age layers first, then Census 2021 — Skopje (opštini) for the rest. Popups list the fields baked into the export.
2019 heat-island study raster: The city GIS portal “С” map image ships next to this file under toplotni-ostrovi/ and appears as Skopje GIS — Heat islands raster (C) — 2019 Study. It is meant to sit above the default Landsat LST tiles when both are on; use Overlay opacity and draw order in the legend to fine-tune stacking or transparency.
Sharing this map: The page head includes Open Graph and Twitter Card tags for link previews. Before you publish on a subdomain, set the canonical site URL and preview image in map.py so social crawlers get correct og:url and og:image values.
Layer control: The layers button stays in the upper-right area of the screen so it is easy to reach on phones and laptops.

[0.2.1] — 2026-05-14

Changed

How census reads on the map: Age structure is shown on opština polygons (shares and estimated counts where applicable), with tooltips and popups — not as centroid bubble maps.
Colours: Population exposure uses a clear blue ramp; age-band layers use ramps so differences are visible at city scale, including 15–64 as both share and estimated population.
Defaults: Land Surface Temperature (Tiles) turns on automatically when tiles/lst/ sits next to this HTML. ECOSTRESS, Census 2021, municipality outlines, and Toplotni ostrovi layers start off — turn them on in the layer control when you need them.
Legend overlay sliders: When an overlay is active, the legend can show opacity and draw order so overlapping rasters and polygons stay readable.

Notes

Publishing: Ship this HTML together with folders tiles/ (LST, and ECOSTRESS if you built it) and toplotni-ostrovi/ beside it, with the same relative paths as in outputs/ after a local build. Missing folders mean missing layers in the browser.
Refreshing this text: Edit CHANGELOG.md in the repo, then rebuild the map (python app.py --city Skopje --map-only, optionally with --no-street-ranking for speed). The Changelog button only updates after a new export.

[0.2.0] — 2026-05-13

Added

Changelog in the map: This file is converted to simple HTML and shown from the header so colleagues see release notes without leaving the page.
Layer picker offline: Grouped layer control assets are inlined so the map works on static hosting without a Leaflet CDN.
Repo hygiene: Indexing ignores for large local data/ trees so editors stay responsive.

Changed

Layer list UX: The expanded layer panel scrolls more comfortably when many overlays are listed.
LST tiles pipeline: GDAL XYZ generation for bundled tiles/lst/ was tightened for smoother static deploys.
Dependencies: requirements.txt updated for the pipeline and map stack.

Experimental (off by default)

Kernel-style heat surface (developer builds only): a soft surface from sampled points for inspection — not a second official census layer.

[0.1.3] — 2026-05-11

Added

Optional reference overlays: Extra municipal context layers (transit, amenities, water, cycling, etc.) where configured, fetched from public FeatureServer endpoints and wired through the shared map code.

[0.1.2] — 2026-05-11

Changed

Documentation: ECOSTRESS / AppEEARS offline workflow (inventory, masking, composite, XYZ) and when optional composite layers appear on the map.

[0.1.1] — 2026-05-11

Changed

Documentation: CLI entrypoints, outputs layout, and static tile workflow for running the pipeline locally.

[0.1.0] — 2026-05-11

First consolidated pass: summer UHI sampling, static LST tiling, ECOSTRESS tooling, and a richer Folium layer stack for integrated map exports.

Added

Optional deep-dive mode in the codebase: extra indices, summaries, and artifacts under outputs/ when enabled via CLI flags (config.py, app.py).
ECOSTRESS scripts under scripts/: inventory, scene quality, selection, GDAL XYZ tiles, optional map layer when metadata.json exists.
Landsat LST XYZ script (generate_lst_xyz_tiles.py): bake rasters for static hosting (large files stay out of git).
Boundary helpers in scripts/: download official polygons via ArcGIS REST when configured.
clustering.py: k-means zones used in plots and the map.

Changed

Core modules (gee.py, sampling.py, spatial.py, pipeline.py, map.py, utils.py, eda.py, ml.py, app.py, config.py): sampling, Getis-Ord Gi*, optional street pilot hooks, ML + SHAP, configuration, legends, and layer guide plumbing.
requirements.txt, .gitignore: tooling and local artifact paths.

Project scope (this map)

Skopje: Summer LST (Landsat), optional ECOSTRESS composite tiles, City of Skopje GIS municipality outlines, archived Toplotni ostrovi 2019 raster when packaged, and Census 2021 choropleths on opština polygons (skopje_census_layers.py).

---

Older remote history (2025) is mostly template and infrastructure merges; entries from May 2026 onward describe the Skopje map you are shipping today.

Layer guide — for colleagues

This map combines several independent products. Before interpreting colours, check which layer is active and whether the legend encodes temperature, land cover, statistical hot spots, or something else.

Tip: In the layer control (top-right), click the ? next to a layer name for its official description (purpose, what it shows, and data sources). Use this page for interpretation pitfalls and research-oriented advice.

Land surface temperature (LST)

Land Surface Temperature (Tiles) — Landsat thermal satellite surface temperature (PNG XYZ next to the HTML; legend from tiles/lst/metadata.json). Surface temperature is not 2 m air temperature.
ECOSTRESS LST composite (~70 m) — offline median/mean masked ECOSTRESS composite as XYZ tiles (default off). Pair with Landsat for context; see layer guide for masking notes (surface ≠ air).
NDVI / NDBI / elevation / land cover — optional Landsat or reference rasters for context in the same grouped layer control.
Skopje GIS — heat islands (2019 study) — the city’s published raster (layer “С”) from the archived OUT/ToplotniOstrovi service, shown as Skopje GIS — Heat islands raster (C) — 2019 Study under City Reference. Municipality boundaries appear once as Municipalities (City of Skopje GIS). Not the Landsat LST product (different legend and lineage).

The default public layer list focuses on district-scale summaries plus satellite context. Rebuild with --debug-layers for full diagnostic overlays.

Gi* hot / cold spots

Per-pixel Gi* and district Gi* overlays are not listed in the default public Skopje picker. Rebuild with --debug-layers for QA. Use Gi* language for local clustering — not interchangeably with raw LST colour.

Neighbourhood polygons (ΔT, mean LST, …)

These aggregate sampled pixels into administrative or grid units. They are appropriate for zonal summaries and policy maps; point layers show within-zone variability that polygons hide.

How to use this in research (practical)

State the layer you cite (name as in the layer control) and the date range in the map title.
Do not infer statistics from dot density alone — sampled or clustered point layers (e.g. Gi* pixels) are for context; use exported data for inference.
For publication, run analyses on exported tabular or raster outputs from the pipeline, not screenshots alone.
Compare like with like — same season, same LST source (Landsat vs MODIS), same baseline for ΔT when comparing cities or years.

Skopje Summer Surface Heat Map

Landsat-based land surface temperature · 2025-06-01 → 2025-08-31

Surface ≠ air temperature — compare relative patterns, not personal exposure.

Skopje Census 2021 (how to read it)

LST — PNG XYZ tiles above; polygon fills are census or index layers, not pixel LST.
Skopje 2021 — Elderly heat exposure index (LST × 65+) — min–max(mean LST) × min–max(estimated 65+); unitless 0–1 (toggle on under Census 2021 — Skopje (opštini)).
Age (0–14 / 15–64 / 65+) — open Census 2021 — Age (Skopje) in the layer control first; remaining census choropleths sit under Census 2021 — Skopje (opštini). Use popups for exact numbers.

Land Surface Temperature (°C)

Min Mid Max

Overlay opacity & draw order

One slider row per active overlay (layer control, top-right). Order: 1 = bottom.

Active color scales