Satellite images of Asia: a buyer’s guide

Choosing imagery for a project in Asia comes down to three numbers and one licence. Match the pixel size to the smallest object you must see, match the revisit time to how fast the ground changes, and check the licence before you download. Free data covers most needs, but national missions fill gaps that Landsat and Sentinel-2 leave.

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

Start from the question, not the sensor. If you are mapping rice fields across a province, 10-metre pixels are enough and free. If you are counting roofs after a typhoon, you need 1 metre or better, and you will pay. If you are tracking monsoon flooding week by week, revisit matters more than sharpness.

Three parameters decide the choice.

Resolution: spatial resolution is the ground size of one pixel. Landsat offers 30 metres in visible bands, Sentinel-2 offers 10, 20 and 60 metres depending on band. Commercial constellations sell 0.3 to 1 metre. A 10-metre pixel cannot show a single house; a 0.5-metre pixel can, at a price.

Revisit: the time between two passes over the same point. Sentinel-2 revisits every five days with two satellites, Landsat every eight days with two. Geostationary weather satellites such as Himawari see the same disc every ten minutes, at coarser resolution. Clouds decide the rest: in monsoon Asia, a five-day revisit can still mean one usable scene a month.

Licence: free does not always mean unrestricted. Landsat and Sentinel data are open, with attribution. Some national products allow research use only, or restrict redistribution. Read the terms before you build a product on top.

Practitioners who write about choosing satellite data in Asia tend to treat the licence as the first filter, not the last, because a scene you cannot legally share is a scene you cannot use in a published map.

Where can I download free satellite images of Asia?

Several archives cover the continent at no cost. Each has its own interface, account rules and download limits.

USGS EarthExplorer carries the full Landsat archive, from 1972 to today, plus radar and elevation products. Registration is free. Downloads come as scene bundles, often a few hundred megabytes each.

Copernicus Data Space, run by the European Union, distributes Sentinel-1 radar, Sentinel-2 optical, Sentinel-3 and others. The browser works well for small areas; the API suits bulk orders. Sentinel-2 coverage of Asia is global and repeated, which makes it the default for time series.

NASA Earthdata hosts MODIS, VIIRS and other moderate-resolution products. MODIS at 250 to 500 metres is coarse but daily, useful for smoke, fires and broad vegetation trends across Southeast Asia.

JAXA distributes ALOS and other Japanese products through its own portals, including global radar mosaics. National platforms in India, South Korea and Thailand publish selected datasets, sometimes with registration limited to domestic institutions.

Google Earth Engine is not an archive but a processing platform. It hosts Landsat, Sentinel, MODIS and some national collections, and lets you run analysis without downloading anything. Access is free for research and non-profit use.

For a first project, Sentinel-2 through Copernicus plus Landsat through EarthExplorer covers optical needs for nearly any Asian country. Add Sentinel-1 when clouds block the view, since radar sees through them.

Which Asian countries fly their own Earth observation satellites?

Several do, and their missions matter for regional work because they add revisit capacity and, in some cases, higher resolution than the open global archives.

Japan runs a long line of missions. ALOS-2 carries L-band radar, useful for ground deformation and forest monitoring. Himawari-8 and Himawari-9 are geostationary weather satellites, imaging the Asia-Pacific every ten minutes. GCOM-W and GCOM-C cover water and climate variables.

India operates the IRS series, one of the oldest civilian programmes in the region, with Cartosat satellites for high-resolution mapping and RISAT for radar. Data access is managed by the national remote sensing centre, with some products free for domestic users.

South Korea flies KOMPSAT satellites, including optical and radar payloads, and sells or shares imagery through national channels. China operates Gaofen, Ziyuan and Fengyun missions, with Fengyun providing geostationary weather coverage. Thailand has THEOS, an optical satellite built with foreign partners. Taiwan runs FORMOSAT, and Vietnam, Indonesia and Malaysia have smaller programmes or hosted payloads.

For a regional study, these missions are not a replacement for Landsat and Sentinel. They are a supplement: a national radar scene can cover a gap in a Sentinel-1 time series, and a geostationary weather satellite gives the hourly context that polar orbiters cannot.

What do resolution, revisit and licence mean in practice?

Take flood mapping in the Mekong Delta. Sentinel-1 radar at 10 metres, revisiting every six to twelve days, sees water under cloud. Sentinel-2 at 10 metres adds optical detail when skies clear. Landsat at 30 metres extends the record back to the 1980s, which is how analysts separate a normal flood from an exceptional one.

Take rice mapping in the Philippines. Sentinel-2 at 10 metres, every five days, tracks transplanting and harvest across a season. Cloud cover is the main obstacle, so analysts combine several dates rather than trusting one.

Take urban heat in Jakarta or Manila. Landsat thermal bands at 100 metres, resampled to 30, give surface temperature. The revisit is coarse, but the archive is long enough to show a trend over decades.

Take deforestation in Borneo. Sentinel-2 and Landsat together give a usable series, and radar adds detection under haze. Licence terms allow publication with attribution, which matters for NGO reports.

In each case the same order applies: define the object, pick the resolution, check the revisit against the rate of change, then confirm the licence.

What does ground truth add to a satellite map?

A classified image is a hypothesis until someone checks it on the ground. Ground truth, the field observations used to train and validate a classification, is what turns pixels into a defensible map.

In practice this means collecting points or plots with known land cover, then comparing them with what the classifier produced. A confusion matrix, the standard table of agreement, tells you where the map fails: often between similar classes such as flooded rice and shallow water, or between two crop types.

Fieldwork in Asia is not always possible. Analysts then use high-resolution reference imagery, local knowledge from partner organisations, or existing maps. Each substitute carries its own error, and stating that error is part of the result.

Free tools lower the barrier. QGIS handles vector and raster work on a laptop. Google Earth Engine runs classification over continental areas without local storage. Both are widely taught in the region, and both accept the open archives described above.

Where does the regional community publish its work?

The Asian Conference on Remote Sensing, held each year, is the main regional meeting for this field. 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, 203 oral and 41 poster.

Proceedings from past conferences are archived online, and they are a practical resource: many papers describe exactly the sensor choices, cloud problems and validation methods that a new project in the same country will face. The Asian Association on Remote Sensing, founded in 1981 and now counting more than 29 member countries, organises the conference and maintains the archive.

For a practitioner starting a project in Asia, the sequence is short. Define the object and the update rate. Check Sentinel-2 and Landsat first, since they are free and global. Add radar if clouds are the problem. Look at national missions if you need finer detail or extra revisit. Confirm the licence. Then plan how you will validate the map on the ground, because that step decides whether anyone else can rely on it.