"Hold this," says Chester Gillmore, the bow-tied director of manufacturing operations for Planet Labs, handing WIRED something that vaguely resembles a toner cartridge. "This is a spaceship. You are holding a spaceship."
If the object does not look like anything one might take for an extraterrestrial vehicle, then where WIRED is standing – in a cluttered warren of offices in San Francisco’s SoMa – hardly seems the typical skunkworks of the space-industrial complex. A row of fixies hangs on a wall and the bins are full of energy-bar wrappers. "Mission control" seems to be a lone guy at a computer wearing a SpaceX T-shirt. Aerospace engineers stopped wearing skinny ties long ago but, apart from Gillmore (whose neckwear is wonderfully out of place), office attire is startup casual. Indeed, were it not for the Export Control security check WIRED successfully cleared at reception – because satellites, which Planet Labs makes, are considered a controlled technology – this could be a visit to any number of tours of fledgeling tech concerns.
When the word "launch" is uttered around here, it is meant literally: on a Russian Dnepr rocket or an Antares headed to the International Space Station (ISS), to name just two of the company’s distribution channels. Planet Labs’s product is low-cost imaging satellites, known as Doves. Since the first were launched in 2013 – joining the more than 1,100 active satellites in orbit – more than 70 others have followed, designed, engineered and deployed at breakneck pace. Once the company gets to its estimated magic number of 150 satellites – the largest armada in orbit – it can begin to deliver on its founding proposition, as briskly -encapsulated by Will Marshall, the company’s hyperkinetic CEO and cofounder: “To image the entire planet, every day.”
Marshall, English-born, Oxford-educated and Nasa-trained, is perched on the edge of a chair like a svelte, excitable bird. He sketches a vivid picture of how "a line scanner moving across planet Earth" can aid humanity. It’s easy, as in Mike Judge’s Silicon Valley, to parody the streak of often fatuous altruism in the tech sector, where self-serving entrepreneurs describe every obscure app as a world-saver. But even without Marshall’s infectious optimism it is not hard to imagine how being able to see the world better – getting a daily status update, if you will – might help make it a better place. Even the name Doves speaks to better-world imaging, unlike the Kestrels and Talons that pervade the military-heavy satellite industry.
These similarities, in atmospherics, messaging and workflow, between Planet Labs and San Francisco’s startup community, are hardly accidental. "We model ourselves on a Silicon Valley hardware startup," Marshall says. Employees are as likely to have come from Facebook as the aerospace industry. The company, says Gillmore, “thinks about hardware like software”, focusing on fast iteration, rolling release schedules and updates in the field. What is as remarkable as the promise of Planet Labs’s technology – a data set for the planet – is the way it tackled the problem: off-the-shelf components, in-house production and testing and disaggregating resilient networks. Marshall calls it "agile aerospace". He started the company with the idea that space, as they like to say, is broken. They believe they have a fix.
A few years ago, when Marshall was working at the Nasa Ames Research Center, the one closest to Silicon Valley in geography and spirit (it once housed the world’s largest supercomputer, the ILLIAC IV), he saw a senior engineer hold up a smartphone in a conference room. The engineer had a simple question about satellites: why were they so complicated? Most of what you need, he pointed out, was right there, in that phone. "It’s true," Marshall agrees. "If you take the component list for a satellite and a smartphone, there’s a lot of overlap: GPS, radios, a hard drive, a CPU, accelerometers."
And so Marshall, along with former Nasa colleagues Robbie Schingler and Chris Boshuizen, opened the "Small Spacecraft Office" at Ames. They had helped form a nonprofit called the Space Generation Advisory Council, a young professionals group hoping to inject new thinking into aerospace research, so were accustomed to working together. "We added the couple of things it didn’t have that you’d need for a satellite, and shot it into space," Marshall says. "It” was called the PhoneSat and, even if the "phone" was the size of a brick, the trio had helped to make a point. "Consumer electronics, and the rest of the tech industry had developed a lot of technology that is useful for satellites," argues Marshall, "[but] a lot of the space sector is largely ignoring this on the basis that they already know how to do it."
According to Marshall, part of the problem is the aerospace industry’s aversion to risk. “We’re spending a billion dollars [£645m] on this satellite -- obviously we need it to work. How do we know it’s going to work? Well, we look at the previous satellites we’ve put in space and know have worked, and we use only those, which reinforces antiquated technology.” And when you are talking about a behemoth such as Nasa’s truck-sized, nearly-billion-dollar Landsat 8, which was launched, bristling with thousands of sensors and powerful optics, in February 2013, that approach seems rational.
But Marshall and his colleagues had another idea. What if, instead of big, expensive, super-capable satellites, you put up simpler, cheaper, lower-flying (and thus shorter-lived) craft? The CubeSat, devised by professors at California Polytechnic State University and Stanford University, had proved the viability of that concept. And what if they were launched in such numbers that instead of individual redundancy, you had system redundancy? “If a few fail, it’s not the end of the world,” Marshall says. Sure, the images might not be as good but, to use a terrestrial metaphor, when you’ve left the big DSLR at home, the smartphone camera is good enough. So the three left Ames to start Planet Labs (PhoneSat soldiers on -- it recently became part of a Nasa mission to study the effect of the Sun on the Earth’s magnetic field), joining what seems an increasingly crowded industry. Besides imaging competitors such as Skybox, Elon Musk in January opened an office in Seattle where his company, SpaceX, will develop cadres of small, low-cost satellites to distribute lower-cost internet. OneWeb, backed by Richard Branson, plans to do much the same.
As Robbie Schingler describes it, traditional aerospace engineers design for the particular "phenomenology" that observers want to research. "After they design that piece of machinery, they build the satellite around it," he says. This is all done with what he calls a "requirements-driven waterfall approach": teams of engineers spending years working on mitigating risk on very specific subsystems, typically with chronic cost overruns. "That satellite is very likely to work," Marshall says. "But it’s how you end up with a 2MP camera on the mission to Mars." In other words, by the time you are fully optimised for one technology, on launch day it has been succeeded several times over by better, faster stuff. Planet Labs’s approach is bottom-up, Schingler says, rather than top-down. "Our engineers are incentivised to think about the entire system," he says, "rather than get all the way down a really tight set of engineering requirements that they have to optimise for. It’s a very different approach."
The workstations, some shrouded by server racks, cling film and HEPA filters – home-made "clean rooms" – look less like a high-tech assembly floor and more like an enterprising maker’s workshop. Gillmore shows WIRED a large, round cardboard tabletop divided into small compartments like a roulette wheel. "This is actually a highly sophisticated manufacturing-assisting robot," he says. Click a button and it spins to the next "jewellery box" holding the part the assembly worker needs. The emphasis is on speed. "If at any point you detect anomalies," he says, "you pull it and start on another" – which is how the company has more satellites in space than it does employees.
In the same way that geneticists use short-lived fruit flies to study -generations of evolutionary change in a matter of weeks, the company iterates at hyper speed by burning through actual satellites, rather than by complex modelling of prototypes. "We’ve learned lessons by breaking things rather than by simulating and overdesigning," says Ben Haldeman, a spacecraft architect at the company. Unlike the phrase used during Apollo 13’s launch, failure is definitely an option. "We treat the space environment as part of our extended lab," Marshall says. With 12 builds in three years, there is little in the way of legacy systems. There have been times when the team redesigned systems in the launch month. “This is pretty unheard of in aerospace,” notes Haldeman.
Not everything is within their control, however. Just after WIRED’s visit, the company lost 26 Doves when an Orbital Sciences Antares rocket and Cygnus cargo spacecraft on an ISS mission exploded six seconds after launching from Nasa’s Wallops Flight Facility in Virginia. As Chris Boshuizen, the company’s CTO and third cofounder, puts it, the incident essentially confirmed the company’s "strength in numbers" model of lower-cost satellites. "You can’t claim to have disposable satellites and then cry when you lose them," he says.
Because the Doves are made mostly of hardware drawn from the consumer-electronics industry, they are able to ride Moore’s law all the way to outer space. "Our solid-state drive has 480GB," Gillmore says. "Just a couple of months ago, it was 240GB. We tested it, spun it and it flew. Now we’re already testing a terabyte." A warhorse such as Landsat 5, meanwhile, for all its usefulness and unexpectedly long life, had technology from circa 1984 – the year of the first Mac – when it finally deorbited in 2013.
Although Planet Labs’s hardware may share some components with a smartphone, it is not simply a matter of throwing a handset into space. A lot of intricate work needs to be crammed into a 3U form factor (the roughly shoebox-sized standard used on vessels such as the ISS) before the Dove is handed over to a launch-integration-services company such as NanoRacks, which preps it for launch. It needs some radiation protection, although much less than higher-flying birds (the radiation hazard begins to increase exponentially the higher you fly). It needs to withstand the shock and vibration of the launch as well as the -40°C to +80°C temperature fluctuations encountered in low Earth orbit.
If something does go wrong with the satellite, there are some ways to correct its trajectory or install system upgrades. Richard Walker, the company’s director of operations, says that whereas most fleets of spacecraft are in constant contact with the ground, "here it could be 16 hours between contact. If something breaks, we don’t know when or exactly how. All we know is that when we talk to it again, it’s not happy. Now we have to take another 16 hours to look at our data and figure out what to ask it the next time it comes around".
The array of consumer-grade technology displayed on the benches at Planet Labs suggests there has been a striking change in the way practitioners think about space. For decades, the space race gave society the trickle-down of technology transfer -- everything from smoke detectors to Teflon-coated -fibreglass to, indeed, the CMOS sensors that enable your phone to take pictures. Now, it is the fruits of the -consumer-electronics wars -- the battle to cram ever more functionality and power into the same foot-print -- paying dividends in space.
This is not to say that the research and exploration colossus that is Nasa does not matter; after all, it is where the founders of Planet Labs and rival satellite company Skybox -originated. Yet the kind of media attention that was once lavished on rocket launches is now directed towards iPhone launches. The Wall Street Journal even speculated that one reason Google thought it wise to acquire Skybox was so that it could track movements at Foxconn factories to tell when new iPhones were going to ship.
In May 1906, in the aftermath of the catastrophic San Francisco earthquake, a Chicago photographer called George R Lawrence used a series of massive kites he called “captive airships” to lift a 636kg camera -- capable of taking a negative some 2.4 metres long -- into the sky and snapped a soon-to-be-famous photograph showing the devastation wrought in dramatic detail. Lawrence would go on to commercialise the production of aerial images of landscapes in the American west; in an echo of today’s satellite-guided “precision agriculture”, his images could, for instance, reveal a property’s water prospects. In his work lay a tantalising possibility: this new view would give us new ways of seeing, thinking about – or improving – our life on Earth.
The first civilian-oriented land-imaging satellite did not arrive until 1972, when the US launched Landsat. It gathered images; it even discovered a small island off the coast of Labrador before fizzling out six years later. Since then, there has been a steady march towards higher-resolution images: whereas Landsat 1 images came in a 30-metre resolution, Planet Labs’s Doves boast a resolution of between three and five metres.
Even this is no longer a particularly impressive number: Digital Globe’s recently launched WorldView-3 shoots images with resolutions of up to 31 centimetres (below the 50-centimetre US security restriction, which itself is gradually being eased), good enough to reveal not only the make and model of a car travelling on a motorway but the surface condition of the road.
But resolution is not the only game in satellite imagery. Cost and coverage matter too. Digital Globe’s new satellite, for example, is said to cover an area the size of Texas every day. But that means the company is not photographing most of the world, and to get the satellite in a position to take that great a photo takes operational time -- meaning that the cost for that imagery is also, generally, Texas-sized. "Right now you can go to Landsat and get a snapshot of any place on Earth about every two weeks," says Erika Reinhardt, a Planet Labs software engineer. The cost is basically free. But if there’s cloud cover, that two weeks is likely to take longer. "A lot of things can change at a much higher cadence than that." Your crops could dry up; a wildfire could have long since spun out of control. Weather satellites could give you daily imagery, but the resolution is low. What Planet Labs is trying to hit, Reinhardt says, is that "sweet spot" of high-frequency, high (enough) resolution and lower cost.
Fully imaging Earth is no small task. Just getting into space means you are dependent on weather and the other whims of launches. Then you need to fly multiple orbits. The Sun-synchronous orbit, delivered by Dnepr, is optimal because, says Walker, "It’s always the same local time directly underneath the satellite. No matter what picture the satellite takes, the Sun angle is all the same." That reduces processing costs. The ISS orbit is cheaper to launch but higher altitude, thus lower resolution. Then you need to get the images back to Earth (the lower altitude helps with data-transmission speeds). For that you need ground stations, and lots of them (Schingler says a lesson of Flock 1 – launched in January 2014 – was that they needed more). Many of those ground stations, Walker says, are not optimised or sufficiently automated to communicate with the number of satellites Planet Labs is working with. So the company has built its own, in places like New Zealand. Places where you can’t, as Walker says, "ring and ask, ‘Hey, for $100 a pass will you give this to us?’"
Once you get images, they need to be processed. Software developer Frank Warmerdam, who came to Planet Labs from Google (he has also created several major open-source geospatial apps), lists any number of potential issues, from "terrain effects" to vignetting. Once those are cleaned up, the images still have to be rectified – working out where the image is on Earth – because, he says, "there’s a limit to how precisely you can know where they are just from the pointing info on the satellite".
"When you’ve got the whole Earth to pan through, it’s hard to know where to look," says Boshuizen, although change-detection algorithms can narrow down the search base of where you need to look. Then you have to consider what you have, and how it differs from what you had yesterday. A lot of this is automated, but, as Warmerdam says, "What you can see in a picture is a valuable information product too."
The amount of analysis that will be done by Planet Labs and the amount that will be done by its clients is something the company doesn’t want to discuss. Boshuizen says the current clients are Fortune 100 companies; from there, the company wants to move down the chain of how much data complexity users are equipped to handle, from government agencies and universities and non-profits down, eventually, to consumer level. "Our ultimate goal is to give this data to everybody," he says. But the market is well established. Companies already use satellites to help them make decisions; for example, an aerospace analyst used Digital Globe images to look at a decade’s worth of activity at the Gulfstream plant in Savannah, Georgia, to estimate how soon the company would unveil a new jet. Companies such as Remote Sensing Metrics will sell you images of car parks at big-box retailers (taken "from approximately 11am to 1.30pm when satellites pass overhead"), which companies can link to actual in-store purchases to get a rough estimate of how many people visit the store versus actually buying something.
But there are places where the large range, high resolution and lower cost of Planet Labs’s images could make a difference. Disaster response, for example. Just like Lawrence photographing post-earthquake San Francisco, satellites have been imaging events such as the Christchurch earthquake in February 2011 and Typhoon Haiyan in November 2013 – getting faster and more detailed every time.
But sometimes disasters unfold more slowly. Take deforestation. "If you wait until the end of the year to take a picture – which is the typical rate – at the end of the year you have a big hole in the Amazon," Marshall says. "If you take it every day, you can catch people in the act who are logging in the wrong place and send the co-ordinates to the response team who go and stop them."
In a study published last year in Environmental Research Letters, a team of researchers found that deforestation rates in the Democratic Republic of the Congo were higher than had been estimated from looking at the pictures transmitted by Landsat. The problem? Image resolution. The clearing activity in the Congo was largely the result of subsistence farmers and therefore in multiple, but small, patches. The mapping resolution that works in a country such as Brazil, with its large-scale agriculture, would "omit the majority of change" in the DRC, according to the researchers. Humans are changing the Earth at a granularity lower than older satellites can pick up. These almost invisible statistical glitches in the imagery – "cryptic disturbances" – have been estimated to add as much as half to deforestation estimates.
It is not just scale that’s important, but time. Looking at the Dove images, Boshuizen was struck by the transformation of agriculture in parts of Asia and Africa. Looking on Google Earth, the view was of traditional agricultural patterns, such as terraced smallholdings, but in the new images much of that was replaced by the squares and circles of Western-style crop planting with pivot irrigation systems.
Marshall says there’s a "pent-up demand" for this global snapshot, even though no one has worked out the full range of its possibilities. Rather than focus on a few applications of the imagery, he says the goal is to "democratise" access. "Everyone can come in and run their algorithms," he says. "We want to enable an ecosystem of all those companies using our dataset." He does not, he says, want to be too "prescriptive" about what the images are used for. After all, the tech world is filled with businesses now trading on what were originally ancillary features. For example, hotel reviews were a peripheral aspect of early TripAdvisor, and the telcos hardly envisioned the promise of text messaging when they first included it as a diagnostic tool. "The most common use case is out there," Marshall says. "And it won’t come from us."
Three other satellite startups:
SpaceX
The outer-space arm of Elon Musk’s empire is in the process of developing "micro-satellites operating in large formations", according to a Musk tweet. SpaceX will build and deploy these satellites with the aim of providing cheap, unfettered internet access, but that is all Musk has given away so far.
Satellogic
Emiliano Kargieman’s Argentine startup has the same basic vision as Planet Labs: launch hundreds of cheap, disposable satellites that communicate with each other while monitoring Earth constantly. Although its goal of 300 satellites is double Planet Labs’, to date it has only three in space (Planet Labs has 28).
Skybox Imaging
Acquired by Google in 2014, Skybox Imaging makes the biggest of the microsatellites. Their size means better resolution and they can also film in HD from space. Google has said they will help keep Google Maps accurate, and hopes the tech will help it improve internet access and disaster relief.
This article was originally published in April 2015.
This article was originally published by WIRED UK
