Satellites Must Now Pass the Landsat Test

Earth observation has plenty of pictures. What it lacks is a way to tell which ones can be trusted.

EarthDaily announced on October 26, 2026 that NASA had awarded it a $3 million task order to supply constellation data and technical support to NASA and the U.S. Geological Survey. The money is small. The scope is not. The two agencies will test the data against Landsat standards, including calibration, stability, radiometric and geometric performance, harmonizability and fitness for purpose. In effect, a commercial constellation is sitting an entrance exam.

That exam matters now because supply is arriving faster than anyone can vet it. A mining team deciding where to drill, or whether a pit wall is moving, needs a measurement that holds up against last year's. A fresh image does not give it that. "to understand change, you first need measurements you can trust and compare over time." said EarthDaily CEO Don Osborne.

Landsat as the yardstick

NASA's framing is deliberately cautious. Its Sustainable Land Imaging Commercial Transition Initiative is a phased, product-based approach meant to preserve the scientific quality, consistency and continuity standards of the Landsat record. NASA says it and USGS are partnering with industry to integrate commercial data or products into that record, "accelerating the EO economy into the 2030s." EarthDaily, for its part, says it is committed to approaches that preserve Landsat's continuous record of more than 50 years, which is freely and openly available worldwide.

The mechanism is practical. NASA says the first USGS "Quick Win" products include bathymetry, surface water extent, burned area and burn severity. Commercial offerings must augment Landsat-derived equivalents by showing enhanced spectral and spatial resolution, improved temporal frequency, or both. NASA also says the partnership will identify quality and consistency gaps and share open-source, government-supplied processing workflows so providers can meet Landsat-quality requirements. The evaluation will also test whether commercial observations can sharpen revisit frequency and spatial resolution when combined with government systems.

The queue at the door

The candidates are numerous. EarthDaily says eight of its 10 satellites are in orbit and cover 100% of Earth's landmass every two days. At full operations it expects to image about 92% of landmass daily, excluding Antarctica. It collects 22 spectral bands, with thermal-infrared native resolution of 120 m and no sharpening applied.

Planet Labs said in a press release that it launched 20 satellites on SpaceX's Transporter-18 rideshare, among them the hyperspectral Tanager-2 and 18 SuperDoves that will supplement its PlanetScope daily-monitoring mission. Tanager captures 426 contiguous bands at 30 m resolution, a configuration Planet says suits rare-earth identification, water quality and agricultural analysis. It plans at least three more Tanagers with Carbon Mapper.

Radar is multiplying too. Tech Times reports that iQPS has put a second satellite into one of its four orbital planes for the first time. Placing it about 180 degrees ahead of its sibling halves the minimum revisit interval for that plane. The target is 36 satellites, nine in each plane, covering every point on Earth with a 10-minute minimum revisit. As of August 2026 the company had about 10 operational satellites. Its roadmap calls for 24 by late 2028 and the full network by 2030. Each spacecraft weighs about 100 kg, and Tech Times says the platform is roughly one-twentieth the mass and one-hundredth the cost of conventional SAR spacecraft. Its spotlight mode reaches 46 cm azimuth resolution, at night and through cloud.

Cheap radar and spectral satellites are only useful if their outputs line up with each other. That is the gap the NASA test is built to close.

What miners actually get

The government's own spectral tools are fading, which raises the stakes. New Space Economy reports that ASTER's thermal-infrared subsystem was permanently shut down on January 16, 2026 because of power constraints on Terra. The shortwave-infrared subsystem has been unavailable for many years. Private operators are filling in unevenly. Pixxel's Firefly documentation describes 135 bands from roughly 472 to 890 nanometers at about 5.4-meter sampling, with no shortwave infrared. Its Honeybee generation is scheduled for 2027. Esper Satellites says its 2026 Four Leaf Clover mission will bring a major leap in spectral resolution and calibration stability.

PhotoSat is selling the idea directly. Canadian Mining Journal reported on January 6, 2026 that the company launched a regional hyperspectral targeting product, R-HET. PhotoSat says it lets teams identify a larger set of alteration minerals than were previously detectable through multispectral data. A case study over the Cerro Casale and Caspiche region mapped alteration minerals that multispectral data could not confidently detect. A property-scale version is due in 2026.

The physics sets the limits. New Space Economy notes that a 30-meter pixel covers 900 square meters, so a narrow vein can vanish inside a mixed pixel. Ordinary optical and hyperspectral satellites cannot see through thick soil, vegetation, overburden or hundreds of meters of rock, so they will not replace drilling. The publication puts the value of the data as "better target selection rather than remote reserve estimation." It cites EUSPA's March 2026 account that field work confirmed several targets Norwegian explorer Kuniko had flagged with TerraEye. That matters in a year when, per the International Energy Agency as cited by New Space Economy, global mineral exploration spending fell by more than 10% in 2025. Cheaper targeting is attractive when budgets are shrinking.

Millimeters, with a caveat

Ground movement is the second problem. A study published November 4, 2026 in the journal Remote Sensing fused Sentinel-1 InSAR with GNSS to track subsidence at the Banji Coal Mine in Anhui Province. InSAR offers high spatial resolution but sparse sampling in time. GNSS offers the reverse. The study used 32 Sentinel-1 scenes with a revisit of about 12 days. It reported root mean square errors of 6.37 mm for strict spatial leave-one-out validation and 2.84 mm for retrospective reconstruction with missing InSAR phases. Its authors say the statistical stability of differences among models needs further verification with additional independent data. New Space Economy also reports that an EUSPA-supported project involving Eramet used Sentinel-1 radar to pick out millimeter-scale movement at mining areas.

The Banji authors' caveat applies to the whole sector. More satellites make for more candidates, not more answers. The constellations that matter to a drill program or a tailings dam will be the ones whose numbers survive comparison with a long-term archive, and a camera that cannot say what it saw last year is still just a camera.