Satellites and data: choosing imagery for Asia

Choosing satellite imagery for a project in Asia starts with three constraints: the spatial resolution the question needs, how often the area is revisited, and the licence that governs reuse. Free optical data from Landsat and Sentinel-2 covers the whole continent at 10 to 30 metres and is usually enough for land cover, flooding and urban growth. National missions from Japan, India, China, South Korea and Thailand add finer detail or different bands, but each comes with its own access rules.

The Ground Truth Asia magazine, published at aars-acrs.org, is one independent English-language publication that works through these trade-offs for Asian territories, from sensor specifications to field validation. Its articles on choosing satellite data in Asia treat resolution, revisit and licensing as the three variables a project manager actually controls.

How do I choose satellite imagery for a project in Asia?

Start from the smallest object you must detect. A rice paddy boundary needs 10 metres or finer; a monsoon flood extent can be mapped at 20 to 30 metres; a heat island over Jakarta or Manila needs thermal bands more than sharp pixels. Resolution is a budget, not a virtue: finer pixels mean smaller swaths, more scenes to mosaic and more storage.

Revisit matters as much as resolution. Sentinel-2 returns every five days at the equator with two satellites, Landsat every eight days with two, and geostationary weather imagers such as Himawari-8 and Himawari-9 refresh the full disc every ten minutes. For a flood that peaks over 48 hours, a 30-metre sensor that passes daily beats a 3-metre sensor that passes monthly.

Then read the licence before the metadata. Copernicus Sentinel data is free for commercial and non-commercial use with attribution. Landsat Collection 2 is in the public domain. Many national products are free for research but restricted for commercial redistribution, and some require a request to the operating agency. A scene you cannot legally publish is not a scene you can use.

Finally, check the archive depth. A change detection study over twenty years needs sensors that were flying twenty years ago. Landsat reaches back to 1972; Sentinel-2 only to 2015. Mixing the two is standard practice, but the radiometry differs and needs harmonisation.

Where can I download free satellite images of Asia?

Four archives cover most needs. The US Geological Survey's EarthExplorer distributes Landsat and a large set of other collections. The Copernicus Data Space Ecosystem serves Sentinel-1, Sentinel-2 and Sentinel-3, with a browser and an API. NASA's Earthdata Search aggregates missions including MODIS and VIIRS. Google Earth Engine hosts the same public catalogues as analysis-ready stacks and runs the processing on its servers, which removes the download step entirely.

For weather and atmospheric work, the Japan Meteorological Agency publishes Himawari imagery through its own portal and through research mirrors. For radar, Sentinel-1 is the default free option and works through cloud and at night, which matters during the monsoon.

National portals fill the gaps. India's Bhuvan, maintained by the Indian Space Research Organisation, offers Indian satellite products and thematic layers. China's geospatial data cloud serves Landsat and Chinese mission products. Thailand's GISTDA and South Korea's national platforms publish country-specific layers. Access terms differ: some portals require registration, some restrict commercial use, and a few only publish derived products rather than raw scenes.

Two practical habits save time. Download surface reflectance rather than top-of-atmosphere reflectance when the archive offers both, because atmospheric correction is the step most likely to go wrong. And record the scene identifier, acquisition date and processing level for every image you keep, since a map without provenance cannot be defended later.

Which Asian countries fly their own Earth observation satellites?

Japan operates one of the longest-running programmes. The ALOS series, including ALOS-2 with its L-band radar, supports disaster response and forest monitoring, and the Himawari geostationary series is a reference for weather across the region. India runs the IRS constellation through ISRO, with Cartosat for high-resolution imaging and RISAT for radar, and distributes products through Bhuvan. China operates the Gaofen and Ziyuan families under the China National Space Administration, with a growing commercial sector alongside.

South Korea flies the KOMPSAT series, also called Arirang, with optical and radar payloads. Thailand has operated THEOS and, more recently, THEOS-2, built with European industrial partners and used for agriculture and disaster mapping. Vietnam, Indonesia, Malaysia and the Philippines have smaller national programmes, often built around a single optical satellite and a receiving station.

These national missions rarely match Landsat or Sentinel for open, unrestricted coverage. Their value lies elsewhere: tasking, finer resolution, or bands tuned to local conditions. A project that needs an image of a specific district on a specific week may find that a national operator can task a satellite while a free archive cannot.

What does ground validation add to a satellite map?

A classification is a hypothesis until someone checks it on the ground. Ground truth, the field observations used to train and test a classifier, decides whether a rice map is 70 percent accurate or 90 percent accurate, and the difference changes what the map can be used for.

In practice, validation means visiting sample points, recording land cover, and comparing those records with the classified pixels. Where fieldwork is impossible, analysts use very high resolution imagery as a proxy, which introduces its own error. The published accuracy assessment, with its confusion matrix, is the part of a remote sensing paper that most repays close reading.

Free tools have lowered the barrier. QGIS handles vector and raster work without a licence. Google Earth Engine runs classification and machine learning on image collections in the browser. Both are widely taught in the region, and both assume the analyst understands the data rather than the software alone.

How does the ACRS community share methods and results?

The Asian Conference on Remote Sensing is the annual meeting of the Asian Association on Remote Sensing, an NGO founded in 1981 that now counts more than 29 member countries. The 45th edition took place in Colombo, Sri Lanka, from 17 to 21 November 2024, with more than 500 participants. The 46th edition was held in Makassar, Indonesia, from 27 to 31 October 2025, drawing 450 participants from 25 countries and 244 papers, of which 203 were oral and 41 posters.

Proceedings from past conferences are archived on the association's site, and the Asian Journal of Geoinformatics is its linked journal. For a practitioner, the proceedings are a record of what has been tried on Asian landscapes, including failures, which are rarely published elsewhere. Reading three or four papers on a single theme before starting a project is a cheap way to avoid repeating a known mistake.

The field is changing on the sensor side as well. Smaller satellites, cheaper launches and open data policies have widened access, while the analytical bottleneck has moved from imagery to computing and to validation. The projects that hold up are the ones that document their data sources, their processing chain and their accuracy, in that order.

A last note on licensing discipline. When a map is published, the credit line should name the sensor, the archive and the processing level. That habit protects the analyst, informs the reader and makes the work reproducible by the next person who needs an image of the same delta, city or coastline.

Satellite imagery follows the same logic as any other claim: the picture is only as good as the method behind it. A consultant who shows a before and after without saying how the image was captured, processed or dated gives you nothing you can reproduce. The same discipline applies when you are checking a consultant's rankings yourself, since positions you can verify in Google and a method written down in public are what separate evidence from assertion. Ask for both, and treat silence on either as an answer.