How every number is measured
DYAAN — Data You Analyse & Act oN
Dyaan exists to answer a simple question honestly: what does the satellite record actually say about this place? Every figure you see is computed on demand from a public Earth-observation archive — counted from real pixels, or clearly labelled as a model sample where no per-pixel measurement exists. Nothing is invented, and nothing is filled in to look tidier than the data. This page is the receipt.
The one rule
A headline number is either counted or it is sampled, and we always say which.
- Counted means we read the raster pixels inside your area and add them up — land cover class areas, water gained and lost, elevation along a line, what a viewshed can see. The count is reproducible: the same area returns the same number.
- Sampled means the underlying product is a model on a coarse grid, not a per-pixel image — air quality and weather. We take the model's value at the area and label it a sample, never a pixel count.
When a read fails, Dyaan says the read failed. It does not substitute a guess, a cached value dressed up as fresh, or a zero that looks like a measurement.
Wildfire — NASA FIRMS VIIRS
Counted. Active-fire detections from the VIIRS 375 m instrument (Collection 2, near-real-time) aboard Suomi NPP and NOAA-20, read from NASA FIRMS' public 24-hour CSV feeds. Dyaan keeps every detection whose coordinates fall inside your area and reports the count, the total and peak fire radiative power, and the day/night split. NRT detections are provisional and can be revised by NASA; a detection is a thermal anomaly, not a guarantee of a wildfire.
Surface water — JRC Global Surface Water v1.4
Counted. Water transitions over 1984–2021 at 30 m resolution (Pekel et al., Nature 2016), read by HTTP byte-range directly from the public Global Surface Water archive. Dyaan counts each 30 m pixel by transition class — permanent, new, lost, seasonal — inside your area, and reports change as a share of the pre-existing 1984 water extent, so the denominator is stated rather than implied. Requests over a hard pixel ceiling are refused rather than approximated.
Land cover — ESA WorldCover 2021
Counted. The ESA WorldCover 2021 v200 global map at 10 m, read by byte-range from the public Cloud-Optimized GeoTIFFs. Dyaan tallies the pixels of each of the eleven land-cover classes inside your area and reports each class's hectares and share. The colours in the map, legend, chart and table all come from ESA's official class palette, so what you see is what was counted.
Vegetation — Sentinel-2 NDVI
Counted. Surface reflectance from Sentinel-2 L2A, bands B04 (red) and B08 (near-infrared), located through the public STAC catalogue and read by byte-range from the scene COGs. Dyaan computes NDVI per pixel — (NIR − red) / (NIR + red) — and, for change, differences two periods you choose. Contains modified Copernicus Sentinel data; cloud and haze affect individual scenes.
Terrain & elevation — Copernicus DEM GLO-30
Counted. The Copernicus GLO-30 digital elevation model at 30 m, read by byte-range from the public COGs. Terrain reports elevation, slope and aspect from the counted grid; the elevation profile samples the DEM along the exact line you draw. Over open water the DEM is flat sea-level data, and Dyaan says so rather than reporting a confident landform where there is none.
Viewshed — line of sight over the DEM
Counted. From an observer point and eye height, Dyaan walks the line of sight across the Copernicus DEM grid and reports the share of the area that is visible. Visibility rises with eye height exactly as real terrain would dictate; a mostly no-data or all-sea-level area is refused rather than returned as a confident sightline.
Air quality — Copernicus CAMS
Sampled. Copernicus CAMS air-quality analysis (roughly 11 km) served via Open-Meteo. This is a modelled product on a coarse grid, not a per-pixel image, so Dyaan takes point samples across the area, averages them, and labels the result a model sample — never a counted pixel. Figures are compared against WHO guideline values for context.
Weather — MET Norway
Sampled. A numerical weather forecast from MET Norway at the area's representative point. It is a forecast, not a measurement of the ground, and it is labelled as such: today's range and conditions come straight from the model's own fields, with no invented "feels like" or fabricated min–max.
Provenance you can audit
For the raster analyses, Dyaan shows where the pixels came from: how many were counted, how many byte-ranges were downloaded now and their size, and how many were served from this server's cache. A cached re-run is never presented as a fresh download, and a partial read is flagged rather than smoothed over. Every outbound read carries a timeout, so one slow archive can't hang a request forever.
Place boundaries
Search and named-feature boundaries (a lake, a park, a district) come from OpenStreetMap via Nominatim, © OpenStreetMap contributors under the Open Database License (ODbL). You can also draw an area, trace a shape, or import your own GeoJSON or KML boundary — every analysis then clips to that exact shape.
Keyless by design
Every source above is public and read without an API key. That keeps a free tier genuinely free, keeps the data provenance transparent, and means the numbers you get are the numbers the archives publish — not a reseller's re-interpretation.
Honest limits
- Near-real-time products (fire) are provisional and can change.
- Optical products (NDVI, land cover) are affected by cloud, and reflect the date of the underlying scene, not this instant.
- Model products (air, weather) are samples of a coarse grid, not per-pixel truth.
- Very large areas are refused rather than approximated, to keep counts real.
Dyaan is an independent product and is not affiliated with, endorsed by or sponsored by NASA, the European Commission, the Joint Research Centre, MET Norway or the OpenStreetMap Foundation. For citations and licences, see the data attribution page.