ATTO is a 325-meter tower in the Uatumã Sustainable Development Reserve, about 150 km northeast of Manaus. It is equipped to continuously measure greenhouse gases, reactive gases, aerosols, and energy and water fluxes at various heights (ATTO Scientific Brief). The tower has operated since 2015 and cost €8.4 million to build, split between the German government and Max Planck Society on one side and Brazil’s MCTIC via FINEP on the other (Agência FAPESP).
The question worth asking is not whether ATTO has a carbon cost. Any station with diesel generators, road and river logistics, climate-controlled labs, and a satellite link does. The question is why that cost is not published as a ledger alongside the measurements it enables.
What the station’s own documentation records
The ATTO Scientific Brief is unusually candid about the physical plant. Power is continuously supplied by diesel generators. Supplies and people move between Manaus and the station by trucks and boats, and additional heavy vehicles at the site maintain the road and transport fuel. Sensitive instruments sit in 12 climate-controlled laboratories built from modified shipping containers. The field camp lodges up to 45 people during campaigns. In 2023, a new satellite connection was installed for improved internet and data transfer.
These are the components of a material budget: fuel, freight, cooling load, and data transmission. What the brief does not provide is any of the quantities. No annual diesel volume. No kilowatt-hour figure for the station or the container labs. No freight tonnage between Manaus and the reserve. No data volume or energy estimate for the satellite link. No water use. No decommissioning or e-waste pathway for the more than 100 pieces of equipment that an INPA technician monitors on daily climbs during roughly ten-day monthly visits (Agência FAPESP).
That absence is the story. It is not a scandal. It is a reporting convention: research infrastructure is documented by its scientific capability, not by its material throughput. The same convention applies to most observatories, and it is increasingly out of step with how other digital infrastructure is being counted.
How Brazil is learning to count digital load
Brazil’s electricity planning agency, EPE, has begun collecting data-center load directly from distributors. In a December 2025 presentation, EPE reported that 20 distributors said they host data centers and that those distributors account for about 65% of the electricity market (EPE, Coleta Data Center na Distribuição 2025). The same presentation projects data-center load rising from hundreds of megawatts in 2025 toward several gigawatts by 2030, with the caveat that the collected data are not yet reflected in official planning.
EPE’s September 2025 presentation cites IEA figures of 415 TWh of data-center electricity consumption globally in 2023, about 1.5% of world electricity, and notes that IT equipment accounts for roughly 30–50% of a data center’s energy use and cooling for another 30–50% (EPE, A Expansão da Carga de Data Centers). Those ranges are for commercial facilities, not research stations, but the accounting categories are instructive: compute, cooling, and everything else.
ATTO’s container labs are cooling-dominated infrastructure in the same sense, though at a radically smaller scale. The difference is that a data center’s load is now being made legible to the electricity system, while a research station’s load remains legible mainly to its own maintenance staff.
Why the ledger matters more than the number
It would be possible to estimate ATTO’s diesel consumption from generator capacity and typical load factors, or its freight emissions from distance and tonnage. Those estimates would be speculative, and this article does not make them. The point is structural: a station that produces high-precision measurements of carbon exchange should be able to state its own material inputs with comparable precision. Not because the two numbers are commensurable — they are not — but because the credibility of a measurement institution depends on its willingness to measure itself.
The comparison with data centers is not about scale. ATTO’s footprint is almost certainly trivial next to a hyperscale facility. The comparison is about method. EPE is building a reporting habit for digital infrastructure: ask the operator, collect the load, publish the range. A similar habit for research stations would require, at minimum, annual figures for diesel or other fuel consumed on site, electricity generated and used, cooling load for instrument containers, freight and personnel transport between the nearest city and the station, data volume transmitted and the energy intensity of the link, water use where relevant, and a decommissioning or e-waste pathway for retired instruments.
None of these fields is exotic. Most are already tracked for operational reasons — fuel must be purchased, generators maintained, freight scheduled. The gap is publication, not measurement.
What ATTO measures, and what that justifies
The case for ATTO does not rest on its being low-impact. It rests on what long-term records at a single undisturbed Amazon site can show that no satellite or short campaign can: the continuous exchange of carbon, water, aerosols, and reactive gases between the forest and the atmosphere, at heights that integrate hundreds of square kilometers of upwind forest. The station is run jointly by German and Brazilian scientists and is designed to record meteorological, chemical, and biological data continuously (ATTO Project). That is a public-good argument, and it is strong.
But public-good arguments are strengthened, not weakened, by transparency about costs. A station that publishes its diesel liters and its satellite data volume alongside its carbon flux data gives critics one less thing to speculate about and gives funders a clearer basis for comparing infrastructure options. The alternative — leaving the material budget undocumented — invites exactly the kind of unanchored estimate that rigorous science should avoid.
Frequently asked questions
Does ATTO publish its energy consumption?
Not in the documents reviewed for this article. The ATTO Scientific Brief documents diesel generators, transport, container labs, and a satellite link but gives no kWh, fuel, or data-volume figures.
Is ATTO’s carbon footprint large compared with a data center?
Almost certainly not, but no published figure allows a direct comparison. The relevant contrast is methodological: commercial digital infrastructure is now being counted through distributor reporting, while research infrastructure is not.
Would publishing a material ledger undermine ATTO’s scientific case?
No. The scientific case rests on the uniqueness and continuity of the measurements. A material ledger adds a dimension of accountability without changing what the instruments record.
What would a station-level ledger need to include?
At minimum: on-site fuel consumption, electricity generated and used, cooling load, freight and personnel transport, data volume and link energy intensity, water use, and an e-waste or decommissioning pathway. Most of these are already tracked operationally; the gap is publication.