Choosing Satellite Imagery for a Project in Asia

Start with the question your map has to answer, then work backwards to the sensor. A flood outline needs a short revisit time and radar; a rice-area estimate needs a consistent optical archive over several seasons; a city heat map needs thermal bands and a resolution fine enough to separate blocks, not fields. Once you know the question, the free archives and the licence terms narrow the field quickly.

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

Write down four things before you download anything: the area, the period, the smallest object you must see, and how often you need to see it. The area decides the scene size and the projection. The period decides whether an archive exists at all. The smallest object sets the resolution floor. The frequency sets the revisit requirement.

Then match those four numbers to a sensor. If your smallest object is a single rice paddy, you need 10 to 30 metres per pixel, which points to Landsat or Sentinel-2. If it is a building footprint, you need under 5 metres, which usually means a commercial task or a national archive with restricted access. If it is a monsoon flood, you need a sensor that sees through cloud, which points to radar rather than optics.

A common mistake is to pick the highest resolution available and then discover the archive is too short or the licence too narrow. Resolution is only one axis. A 10-metre image every five days for ten years will answer more territorial questions than a 1-metre image of one afternoon. For a wider view of how practitioners in the region frame these trade-offs, the editorial work at choosing satellite data asia treats sensors, licences and ground validation as one subject rather than three.

Budget your time for the boring part: cloud masking, atmospheric correction, and checking that two dates are actually comparable. In tropical Asia, a wet-season composite can lose half its pixels to cloud, and a dry-season composite can hide the very change you are looking for.

Where can I download free satellite images of Asia?

Three archives cover most of the continent at no cost.

Landsat, run by the US Geological Survey and NASA, goes back to 1972 and gives 30-metre optical bands with a 16-day revisit. The long record is the point: it is the only free archive deep enough for multi-decade change analysis over Asian coastlines, deltas and cities.

Sentinel-2, from the European Union's Copernicus programme, gives 10-metre optical bands with a five-day revisit when you combine both satellites. It is the workhorse for rice mapping, urban growth and vegetation indices. Sentinel-1 adds radar, which works at night and through cloud, and is the standard choice for monsoon flood mapping.

National portals fill the gaps. Japan distributes ALOS and Himawari data through JAXA; India distributes Resourcesat and Cartosat products through ISRO's Bhuvan platform; China, South Korea and Thailand run their own distribution services. These portals often carry higher resolution than the global free archives, but access rules vary by country and by product, and some require registration or a research justification.

For weather in near real time, Himawari-8 and Himawari-9 deliver a full-disk image every ten minutes. That cadence is useful for tracking convection over the maritime continent, but the resolution is coarse, around 500 metres to 2 kilometres depending on the band, so it complements rather than replaces land imaging.

How do I read resolution, revisit time and licences?

Resolution comes in several kinds, and they are not interchangeable. Spatial resolution is the ground size of one pixel. Spectral resolution is how many wavelength bands the sensor records and how narrow they are. Radiometric resolution is how many brightness levels each pixel can hold. Temporal resolution is how often the sensor returns. A project can fail on any one of them.

Revisit time is the interval between two passes over the same point. It is not the same as the time between two usable images. Cloud, off-nadir angle and acquisition planning all cut into it. Sentinel-2's nominal five days can become three weeks in the Mekong basin during the wet season.

Licences decide what you may publish. Landsat and Sentinel data are open, with attribution, and can be redistributed. National products often carry restrictions: no commercial use, no redistribution, or a requirement to acknowledge the agency. Read the terms before you build a workflow around a dataset, because switching archives late means reprocessing everything.

Two practical habits help. Keep a small table of every scene you download, with its identifier, date, cloud cover and licence. And record the processing steps in a text file alongside the data, so a colleague can reproduce the map without asking you.

Which Asian countries fly their own Earth observation satellites?

Several do, and the list has grown steadily since the 1990s.

Japan operates a broad programme through JAXA, including the ALOS radar and optical series, the GCOM-W and GCOM-C climate missions, and the Himawari geostationary weather satellites.

India runs one of the largest national constellations, with the IRS series, Resourcesat, Cartosat and the Oceansat and Scatsat missions, all managed by ISRO.

China operates the Gaofen and Fengyun families, among others, with both optical and radar instruments and a geostationary weather series.

South Korea flies the KOMPSAT series, including high-resolution optical and radar payloads, through KARI.

Thailand operates THEOS, an optical satellite built with European cooperation, and has continued with follow-on smallsat work.

Others in the region have launched smaller or shared missions, and several countries participate through regional programmes rather than owning a full satellite. The practical consequence for a project is that national data often exists, but the access route is a national agency rather than a global open portal.

What about ground validation and the wider community?

A satellite image is a measurement, not a map. Turning it into a map means checking it against something on the ground: field plots, gauge readings, drone surveys, or official statistics. This is what practitioners call ground truth, and it is the step that decides whether a classification is defensible.

Free tools lower the entry cost. QGIS handles vector and raster work on a laptop. Google Earth Engine runs classification and time-series analysis on server-side archives, which matters when your study area spans a decade and a whole country.

The regional community meets through the Asian Conference on Remote Sensing, held annually since the early 1980s, with proceedings archived by the Asian Association on Remote Sensing. For students and early-career analysts, those proceedings are a useful record of what has already been tried on Asian landscapes, and often include the validation details that journal papers compress.

A short checklist before you commit

• State the question, the area, the period, the smallest object and the needed frequency.
• Pick the archive that satisfies all five, not the one with the finest pixels.
• Check the licence for redistribution and commercial use.
• Plan for cloud and for gaps in the record.
• Budget time for validation, not just for download.

None of this requires a large budget. It requires deciding what counts as evidence before you start looking at images, and then letting the sensor specifications, the licence terms and the ground checks settle the rest. Readings taken by eye are covered in naming clouds without instruments.