Mapa Deserto Do Atacama

How to actually map the Atacama Desert when satellite data fails you

The Atacama Desert is one of the most challenging environments for cartographic work I've encountered. Standard satellite imagery from Sentinel-2 or even Landsat 8 leaves massive gaps in the central hyper-arid zone because the surface reflectance is so extreme that automated classification algorithms just break down. You get clouds of mislabeled pixels that look like noise but are actually valid surface data being rejected by the processing chain. I learned this the hard way while trying to produce a detailed geological map for a mining survey in the Salar de Atacama region back in 2019. The official topographic maps from IGAC Chile cover the area, but they're at 1:50,000 scale and dated to the early 2000s. The terrain has shifted enough in areas with active salt flat extraction and seasonal water flow that relying on them gave us errors in boundary calculations we caught too late. The final product was off by roughly 200 meters in three separate grid cells compared to what we found on the ground using RTK GPS.

Using a mapa deserto do atacama as your base layer

When building your own map from scratch, start with a fusion of sources rather than trusting any single dataset. The approach that actually works combines ASTER GDEM V3 for elevation, Sentinel-2 MSI for spectral data, and the GMTED2010 global dataset as a fallback where newer imagery has persistent cloud cover or sensor artifacts. This combination gives you usable coverage across the entire desert corridor from about 23°S to 27°S latitude and 68°W to 70°W longitude. The real problem isn't collecting the data. It's the radiometric calibration. The Atacama's surface has some of the highest albedo values on Earth in certain zones, particularly around the salar boundaries where evaporite crusts reflect up to 70% of incoming solar radiation. If you don't apply a proper atmospheric correction using something like FLAASH or Dark Object Subtraction, your NDVI and other vegetation indices will be completely meaningless. You'll see green patches where there is literally nothing growing because the algorithm interprets the reflectance spike as biomass. I ended up spending three days just recalibrating the spectral bands for one sector near Valle de la Luna. The default Sen2Cor processor in QGIS was producing nonsense because the aerosol optical depth in that region during the dry season is near zero, and the default atmospheric model assumed a much higher moisture content. Switching to a custom low-aerosol preset fixed it, but you won't find that documented anywhere in the user guides. It came from posting in a European Space Agency processing forum and getting a reply from someone who'd worked on the Sahara similar conditions. That's how these things get learned.

Processing workflow that doesn't waste your time

Here's the sequence I use now. Download Level-2A Sentinel-2 products from Copernicus Open Access Hub for your area of interest, filtered to scenes with less than 5% cloud cover. Run Sen2Cor with the atmos_param option set to arid or custom, then export the bottom-of-atmosphere reflectance bands. Merge the bands in QGIS using GDAL's mosaicking tools with nearest-neighbor resampling to avoid introducing artificial smoothing across the sharp spectral boundaries that define Atacama's geology. For the topographic component, download ASTER GDEM tiles and fill any voids using SRTM 90m data from the CGIAR portal. The ASTER data has some known spikes and dropouts, particularly in steep terrain around the Andean foothills east of the desert, so you'll want to run a median filter with a 3x3 kernel before merging. Don't skip this step. Uncorrected voids in DEM data create catastrophic errors when you're generating contour lines or slope aspects, and catching those errors after you've already produced a final map means starting over. Slope and aspect rasters are critical for understanding water flow patterns in a desert that might go years without rain but experiences intense flash flooding when it does. The canyon systems like Cajón del Azufre channel water unpredictably, and a bare elevation model won't tell you where runoff concentrates. After generating the flow direction and accumulation rasters in Whitebox GAT, I classify high-accumulation cells and overlay them on the composite image. These become my field verification points. If the model says water should pool in a certain area and the satellite imagery shows dry cracked mud or salt crust, I flag it for ground truthing. More often than not, the model is right and the imagery is misleading because of shadowing from nearby ridges.

Common mistakes that will ruin your map

The biggest error I see people make is assuming higher resolution always means better results. WorldView-3 and Pleiades data are available for the Atacama through commercial providers, and the spatial resolution is outstanding. But the revisit time is two to three days under ideal conditions, and the cloud cover in the transitional zones between the desert and the Andean slope can persist for weeks. You end up with a mosaic made from images taken months apart under different sun angles, which creates visible seam lines that are nearly impossible to remove without manual editing. For most applications, a carefully processed Sentinel-2 mosaic at 10m resolution produces a more consistent result than a poorly stitched high-resolution composite. Another issue is coordinate reference systems. Chile uses multiple datums depending on the source. Old government maps are on SAD-69, modern surveys use SIRGAS-Chile, and satellite data defaults to WGS-84. If you're combining datasets without reprojecting everything to a common CRS, your features will be offset by anywhere from 50 to 200 meters depending on the local datum shift. I always reproject to SIRGAS-Chile / UTM zone 19S for anything west of 72°W and zone 18S for the eastern portions near the volcano chains. It takes extra time upfront but prevents the kind of alignment problems that show up only after you've submitted the deliverable.

What this approach can't do

Even with all of this, remote sensing mapping in the Atacama has hard limits. The mineralogical complexity in certain zones simply cannot be resolved at the spatial scales available to public-domain sensors. Areas with mixed exposures of halite, gypsum, sodium nitrate, and lithium brine residue show up as spectrally similar gray tones in optical imagery. You need hyperspectral data from sensors like PRISMA or the EnMAP mission to differentiate those materials, and that data isn't freely available for high-revisit coverage. If your project requires mineral classification at that level, you're looking at paid data access and specialized processing software like ENVI or eCognition, not a QGIS workflow. Ground penetration radar or LiDAR would solve some of these problems, but deploying those in the Atacama is logistically expensive and permits from the Chilean military are required for any airborne survey in the border regions. The desert overlaps with restricted zones near the Bolivian and Argentine borders, and you'll need clearance from Defensa Nacional before any drone or aircraft operations. I've seen projects delayed for six months over this alone. For a practical, no-frustration map that covers the major features, transportation routes, hydrological boundaries, and basic geological zones, the Sentinel-2 plus ASTER fusion method described above will get you there in about two to three weeks of focused work if you're processing everything yourself. Hire someone who's done it before and it's closer to five days. The alternative is spending three weeks fighting with atmospheric correction presets and wondering why your contour lines are generating straight lines across flat salt flats instead of following the subtle gradients that actually exist.