Satellite Imagery Sources

Every satellite sensor is described by four kinds of resolution, and confusing them causes most beginner mistakes.

Spatial resolution is the ground size of one pixel - 30 m for Landsat, 10 m for Sentinel-2. It sets the smallest thing you can see. Spectral resolution is how many, and how narrow, the colour bands are - not just red, green and blue but near-infrared and shortwave-infrared, which is how you measure vegetation and water. Temporal resolution (revisit) is how often the satellite returns to the same place - days for the free workhorses, so you can track change. Radiometric resolution is how finely each pixel records brightness. For built-environment work the first three decide almost everything: how small, in what colours, how often.

The Landsat programme (NASA and USGS) has imaged the Earth since 1972, which makes it the go-to record for long-term change. Current Landsat 8 and 9 deliver multispectral imagery at about 30 m (with a 15 m panchromatic band for sharpening), and you download it free from USGS EarthExplorer, in the public domain.

Thirty metres sounds coarse, and for a single plot it is. But its value is the archive: because you can pull comparable scenes from the 1980s to today, Landsat is unmatched for tracking how a city has sprawled, how a wetland has contracted, or how green cover has changed over decades. When the question is how has this changed over a long time, Landsat is usually the honest first answer.

The European Union's Copernicus programme runs Sentinel-2, a pair of satellites delivering optical imagery at 10, 20 and 60 m depending on the band, with a revisit of only a few days. You download it free and openly from the Copernicus Data Space Ecosystem. For most built-environment analysis on a budget, Sentinel-2 is the default: 10 m is fine enough to read neighbourhoods, land use, water bodies and vegetation, and the frequent revisit means you can almost always find a recent cloud-free scene.

The partner Sentinel-1 carries radar (SAR) rather than an optical camera, so it sees through cloud and at night - invaluable in the monsoon, when optical sensors are blinded for weeks. Together the Sentinels are the free imagery backbone of serious Indian GIS work.

India flies its own Earth-observation fleet through ISRO, and you reach it through the National Remote Sensing Centre (NRSC) - the Bhoonidhi ordering portal, with many layers also viewable on Bhuvan. Two families matter here. Resourcesat carries the LISS-III sensor at about 23.5 m and AWiFS at about 56 m - medium-resolution multispectral imagery for land and vegetation, some of it free. Cartosat is the high-resolution optical family, from sub-metre to a few metres, used for detailed mapping.

Here is the honest India caveat this course insists on: the finest ISRO imagery is not free to the public. High-resolution Cartosat data is priced under the NRSC/ISRO data policy, and the exact per-scene cost varies by sensor and licensee (we do not quote a figure because it is not publicly fixed). So the practical Indian stack is: free Sentinel-2 and Landsat for most work, free Resourcesat/AWiFS where it helps, and priced Cartosat only when a task genuinely needs sub-metre detail and the budget exists.

Put the pieces together as a decision, not a habit. Long-term change over a large area - Landsat, for the archive. Current land use, vegetation and water at neighbourhood scale, on any budget - Sentinel-2, for the 10 m and the frequent revisit. Cloud-blind monsoon monitoring - Sentinel-1 radar. India-specific medium-resolution products, or an official Indian source - Resourcesat via Bhoonidhi. Genuine building-scale detail - priced Cartosat, or commercial very-high-resolution imagery, budget permitting.

And always record the scene's date and sensor on your map. Imagery is a measurement taken at an instant; a land-cover map from a February scene and one from a July scene can disagree completely, and the reader deserves to know which day you are showing them.