Working with satellite data for archaeological surveys
I spent about three years doing systematic remote sensing work across parts of Mesopotamia before handing it off to other people. The basic workflow is straightforward once you understand what each dataset can actually give you, but there are enough landmines in there that most beginners end up wasting weeks on data they can't use. Satellite Remote Sensing For Archaeology is really just about picking the right sensor for the problem you're trying to solve. Most people jump straight into Sentinel-2 because it's free and 10-meter resolution sounds good on paper. Then they try to find subtle crop marks or subsurface wall features and realize they need something better than a blurry thumbnail of the landscape.
Getting started with the actual data
Download links aren't something I can provide directly, but the data sources are all public. Copernicus opens access data comes from the USGS Earth Explorer or the Copernicus Data Space portal. If you need higher resolution, Planet has a free tier for research that gives you roughly 3-meter data with daily revisits. Landsat 9 is another solid option at 30 meters, though the temporal coverage makes it better for seasonal analysis than fine detail detection. The workflow I used for most of my projects went something like this. First you grab a cloud-free composite for your area of interest, usually from the dry season if you're hunting for crop marks. Then you run band ratio calculations in QGIS or ENVI. The ndwi index or simple red-to-near infrared ratios often reveal vegetation anomalies that correspond to buried structures. From there you might shift to hillshade models derived from SRTM or ALOS World 3D topography if you're looking for earthworks and trench systems. I ran into a real problem once in southern Iraq where the satellite imagery showed what looked like an extensive canal network. My initial interpretation was Bronze Age irrigation features based on the geometry. The ground truthing revealed something completely different. The "canals" were actually modern agricultural drainage ditches from the 1990s that had been partially backfilled. The satellite signature was nearly identical because both features produce the same kind of vegetation stress pattern. I ended up cross-referencing historical aerial photography from the 1960s and some declassified Corona frames to distinguish the ancient features from the modern ones. That took another two weeks I hadn't budgeted for.
What beginners get wrong about resolution
Resolution matters more than most people think, but not in the obvious way. People assume higher resolution is always better. It isn't. Very high resolution data like Maxar's 30-centimeter imagery can actually hurt you when you're looking for regional settlement patterns because the detail becomes overwhelming and the footprints are so small that covering a meaningful study area gets expensive fast. The sweet spot for most archaeological applications sits somewhere between 1 and 10 meters. Sentinel-2 at 10 meters works well for broad settlement distribution studies and landscape context. Planet at 3 meters is useful for site-level feature identification. When I need sub-meter detail for excavation planning, I usually pull from WorldView or use drone photogrammetry instead of paying for satellite prices per square kilometer. Another thing nobody tells you about multispectral data: the bands matter more than the resolution. A 10-meter Sentinel-2 image with proper near-infrared and shortwave infrared bands will often outperform a 1-meter panchromatic image with no spectral information. For detecting subsurface features through vegetation, the near-infrared band around 840 nanometers is usually the most diagnostic. Shortwave infrared around 2200 nanometers can help identify altered soils and certain types of building materials.
Get the Full Details
Processing steps that actually matter
You have to do atmospheric correction before anything else. Raw digital numbers from satellites include atmospheric scattering and absorption effects that will mess up your spectral analysis. Sen2Cor handles Sentinel data for free. LaSRC works for Landsat. If you skip this step your band ratios will be off and you'll waste time chasing artifacts. Once corrected, I typically pan sharpen the 10-meter bands to 5 meters using the panchromatic band when available. The improvement in visual interpretability is noticeable without adding much processing time. Then I run the ratio analysis. Red/NIR ratios between 0.6 and 0.8 often highlight vegetation stress patterns associated with buried features. Values outside that range are usually soil or moisture variation rather than archaeology. Digital elevation model processing deserves its own attention. A single hillshade at one sun angle is almost never enough. I generate hillshades at four different sun elevation angles and four azimuth angles. That gives you 16 different perspective views of the terrain. Put them in a virtual raster stack and you can toggle through them to pick out subtle earthworks that would be invisible in a single render. This process takes about 20 minutes on a decent machine and usually reveals something you missed the first three times around.
When satellite data just won't work
Let me be clear about the limitations. Dense forested areas are essentially a loss for optical satellite archaeology. The canopy blocks everything below it and the vegetation response gets muddied by natural forest variability. LiDAR is the only real solution there and it's expensive. If your project area is under heavy forest cover you're better off spending your budget on airborne LiDAR flights or ground survey rather than buying satellite imagery you can't interpret. Satellite remote sensing also struggles with recent sites. Features from the last few decades tend to have been disturbed by plowing, construction, or vegetation management that erases the subtle signals you'd be looking for. The method works best for pre-modern sites in areas with minimal recent ground disturbance. Agricultural intensification has destroyed the crop mark potential in many regions that were productive targets thirty years ago. There's also the issue of temporal resolution. Sentinel-2 revisits every five days under ideal conditions, but cloud cover in many archaeological regions makes that unrealistic. You might go weeks between usable images in monsoon climates or mountainous terrain. Planning your field season around satellite overpasses is usually worth the logistical headache. Knowing you have a clear image window coming up in ten days changes how you prioritize which sites to target.
The bottom line is that satellite remote sensing is a screening and prioritization tool first. It tells you where to look and what looks promising. It rarely replaces ground survey or excavation. The best projects I've been part of used satellite data to narrow down a 500 square kilometer area to about 40 candidate sites, then spent the remaining budget on targeted ground walking and test trenching at those locations.
