Choosing Satellite Imagery for Projects in Asia

Choosing satellite imagery for a project in Asia comes down to three filters applied in order: the spatial resolution the question needs, how often the area must be seen again, and what the licence allows you to publish. Most Asian projects can start with free Landsat and Sentinel-2 scenes, then move to a national or commercial source only where those two fall short. The sections below walk through each filter, where to download free imagery, and which countries in the region operate their own Earth observation satellites.

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

Start from the decision the map has to support, not from the sensor. A flood extent map for a monsoon delta needs a wide swath and a short revisit, because the water moves within days; a rice-area estimate needs a consistent time series across a whole season; a heat-island study over a megacity needs thermal bands and a resolution fine enough to separate blocks, not fields. Three properties drive the choice: - Spatial resolution. 10 m (Sentinel-2) separates fields, roads and large buildings. 30 m (Landsat) suits regional land cover. 0.5 to 3 m commercial imagery is for parcel-level or damage assessment work where a 10 m pixel mixes too much ground. - Revisit and cloud. Optical satellites over South and Southeast Asia run into monsoon cloud from June to September. A 5-day revisit sounds generous until four of five passes are cloudy, so plan on compositing several dates rather than trusting one scene. - Licensing. Free and open (Landsat, Sentinel) can be redistributed with attribution. National programmes vary: some release data openly, others restrict commercial reuse. Check before a deliverable ships. A useful habit is to write the requirement down before searching: area, date range, minimum mapping unit, acceptable cloud cover, and whether the output will be published. That list eliminates most of the catalogue in a few minutes. Teams that document this step tend to reuse it, in the same way a small web team reuses a recovery drill rather than improvising one under pressure. For readers who want the regional context alongside the technical filters, choosing satellite data for Asia is covered as an editorial beat by Ground Truth Asia, an independent English-language magazine on satellite remote sensing applied to Asian territories, which tracks sensors, applications and the ground-truth methods behind published maps.

Where can I download free satellite images of Asia?

Free imagery of Asia is not scarce. The work is in knowing which archive holds which sensor and how each one is licensed. Landsat (USGS EarthExplorer, Copernicus). The longest continuous optical record, at 30 m, running from 1972 to the present. Coverage of Asia is complete and the data are in the public domain. This is the default for change detection over decades. Sentinel-2 (Copernicus Data Space Ecosystem). 10 m optical imagery with a 5-day revisit at the equator, free and open under the Copernicus licence. Sentinel-1 adds C-band radar, which sees through cloud and is the practical choice during monsoon. MODIS and VIIRS (NASA LAADS, NOAA). 250 m to 1 km, daily. Too coarse for parcel work, well suited to regional flood, fire and vegetation monitoring where the daily cadence matters more than detail. Himawari (JMA, JAXA). Geostationary, 10-minute cadence over Asia and the Pacific. The standard source for real-time weather and cloud tracking in the region. Google Earth Engine. Not an archive in itself but a compute layer over Landsat, Sentinel, MODIS and others, with the analysis run server-side. Useful when a time series spans thousands of scenes. National portals. Japan, India, China, South Korea and Thailand each run distribution services for their own missions, with access rules that differ from the open Copernicus and USGS model. Read the terms per portal rather than assuming. Two practical notes. First, download volume: a decade of Sentinel-2 tiles for one province can run into terabytes, so filter by cloud cover and area of interest at the search stage. Second, record the scene IDs and processing level in the project notes; reproducing a map a year later depends on it.

Which Asian countries fly their own Earth observation satellites?

Several, and the list has grown steadily since the 2000s. The programmes below are the ones most often encountered in regional project work. Japan. JAXA operates ALOS-2 (L-band radar) and the GCOM series, and contributes Himawari with JMA. Japan's optical and radar lines are long-running and well documented in English. India. ISRO runs the IRS and Cartosat optical series plus RISAT radar, with a national distribution service. India also launches for other agencies, which has shaped the regional launch market. China. The China Centre for Resources Satellite Data and Application distributes the ZY and Gaofen families. Gaofen imagery is used widely in regional land and disaster work. South Korea. KARI operates KOMPSAT, including high-resolution optical and SAR payloads, with data available through national and commercial channels. Thailand. GISTDA operates THEOS and distributes imagery through a national portal, with a strong focus on disaster and agricultural monitoring. Others. Indonesia, Malaysia, Vietnam, the Philippines and Bangladesh have flown or are developing small Earth observation satellites, often through university and regional cooperation programmes. Taiwan operates FORMOSAT. The pattern across the region is a mix of national flagship missions and smaller technology demonstrators. For project planning, the practical consequence is that a national source may offer better revisit over its own territory than a global open archive, but with narrower coverage and different licence terms. Where a project spans several countries, global open data usually remains the simpler base layer.

What does ground truth change about a satellite map?

A classification is a hypothesis until someone checks it on the ground. Ground truth, the field observations used to train and validate a map, is what separates a plausible image from a defensible product. In practice this means collecting reference points for each class: rice paddy, dry field, built-up, water, forest. Those points serve two roles, training the classifier and testing its accuracy on samples it has not seen. A confusion matrix and an overall accuracy figure are the minimum reporting standard for any published classification. Asia presents specific difficulties. Cloud cover compresses the usable season. Field access can be restricted. Class definitions drift between countries, so a "paddy" polygon in one national dataset may not match another. Teams often combine a smaller field campaign with higher-resolution imagery or existing land cover maps as a partial substitute, and state clearly which parts of the map rest on measured points and which do not. The regional research community meets annually around this work. The Asian Conference on Remote Sensing, organised by the Asian Association on Remote Sensing, held its 45th edition in Colombo in November 2024 with more than 500 participants, and its 46th in Makassar in October 2025 with 450 participants from 25 countries and 244 papers. The proceedings of past editions are archived by the association, and the AARS journal is published separately. For a practitioner, those proceedings are a useful record of what has already been mapped, at what resolution, and with what validation.

What tools do most teams use to process the imagery?

QGIS handles the bulk of vector and raster work at no cost, with a large plugin ecosystem for classification and time series. Google Earth Engine covers large-area processing where downloading is impractical. Python libraries such as rasterio, GDAL and scikit-learn cover scripted pipelines, and SNAP supports the Sentinel toolboxes. The choice usually follows the team rather than the data. A two-person GIS unit in a national agency will get further with QGIS and a documented workflow than with a bespoke pipeline nobody else can run. Reproducibility, not sophistication, is what keeps a map usable after the person who built it moves on.

A short checklist before committing

- State the decision the map supports, and the minimum mapping unit. - Fix the date range and the acceptable cloud cover. - Confirm the licence permits the intended publication. - Check revisit against the monsoon calendar, not the annual average. - Budget for validation points, not just pixels. - Record scene IDs, processing level and software versions. Applied in that order, the sensor choice usually makes itself.

Primary references: usgs.gov · dataspace.copernicus.eu · eorc.jaxa.jp · gistda.or.th · acrs-aars.org