Mostar from orbit
The Neretva valley, one cloud-free summer composite per year from 1984 to 2026. Every frame is the same 31 × 40 km of ground on the same grid, so the only things that move are the things that actually moved.
Drag the slider, or use the arrow keys. Space plays.
What you are looking at
Mostar sits where the Neretva cuts through the karst, and the shape of the city is dictated by that: it runs north–south along the river because the limestone gives it nowhere else to go. The pale strip southeast of the centre is the airport. The dark mass to the north is the reservoir behind the dam.
At 30 m per pixel — 20 m for the recent Sentinel-2 years — the Old Bridge is about one pixel, so this is not the view for that. What this resolution is good for is everything the eye misses from the ground: which hillsides lost their vegetation and which grew it back, how far the built-up edge crept up the valley sides, and how the land either side of the river changed at different rates and different times.
How it was built
Sensors, compositing, and the corrections that mattered
Each year is a median composite of the least-cloudy Jun–Sep scenes, taken from every satellite footprint that overlaps the area. Compositing rather than picking one scene fills gaps, suppresses haze the cloud mask misses, and — for 2012 — repairs the wedge-shaped scan gaps that Landsat 7 has produced ever since its scan line corrector failed in 2003.
| Years | Instrument | Why |
|---|---|---|
| 1984–2011 | Landsat 5 TM | the long backbone of the series |
| 2012 | Landsat 7 ETM+ | the only satellite over Mostar that year — Landsat 5 retired, Landsat 8 not yet launched |
| 2013–2014 | Landsat 8 OLI | — |
| 2015–2026 | Sentinel-2 MSI | 10 m, resampled to the common 20 m grid |
Everything is converted to surface reflectance before compositing, so a value means the same physical thing in 1984 as in 2026. Every frame then gets the same fixed stretch. It would be easy to histogram-match each year to a reference and get a perfectly smooth animation, and it would be wrong: matching the distributions guarantees that any real change in them has been erased, and change in those distributions is the entire subject.
What is corrected is brightness only, anchored on pixels that genuinely should not change — bare karst, selected where it looks like rock in at least 85% of the forty-three years. The city is never used to normalise the city.
Two corrections turned out to matter more than expected. Sentinel-2 gained a radiometric offset in processing baseline 04.00 on 25 January 2022; without it every frame from 2022 on comes out too bright, which reads convincingly as four hazy summers rather than as a units error. And scene selection has to draw from every footprint the area touches — Mostar straddles a boundary in both the Landsat and Sentinel tiling grids, and choosing simply the least-cloudy scenes leaves whole corners of the valley unobserved.
Built from Microsoft Planetary Computer's copies of USGS/NASA Landsat Collection 2 Level-2 and ESA Copernicus Sentinel-2 Level-2A. Contains modified Copernicus Sentinel data. Landsat courtesy of the U.S. Geological Survey. 43 annual composites, 1984–2026.