Taking the cloud to the edge of space
In this guest post, Chris Roberts, head of datacentre and cloud at Goonhilly Earth Station, wonders if enterprises are ready to tap into the business opportunities the explosion in data generated by Earth observation satellites is on course to bring?
During the summer, satellite images of fires burning in Brazil’s Amazon basin shocked the world. Using Copernicus Sentinel-3 data, as part of the Sentinel-3 World Fires Atlas, almost 4,000 fires were detected in August 2019, compared with just 1,110 fires at the same time the previous year, according to the European Space Agency.
But the satellite image data revealed a more nuanced story. Most of the fires were burning on agricultural land that was cleared of trees some time ago, but the most intense fires were burning where forest had recently been cleared.
The world is waking up to the insight and detail satellite data offers any organisation prepared to invest in it. Morgan Stanley estimates that the global space industry could generate revenue of more than $1 trillion by 2040, up from $350 billion, currently. Much of the increase comes from the booming satellite sector.
At the same time, the industry is miniaturising. The emergence of small satellites, including cubesats with shorter development cycles and smaller development teams, and consequently, is cutting launch costs.
Demand for data drives Earth observation satellite interest
Driving demand for investment in the satellite industry is a thirst for data. Companies know that, used in the right way, satellite data can help decipher global business, economic, social and environmental trends.
For example, in 2017 US retail giant JC Penney announced the closure of 130 stores after five years of struggling to attain full-year profitability.
But Orbital Insight, a satellite data company, knew something was afoot well before the announcement. It found the number of vehicles in JC Penney car parks had fallen by 5% year-over-year the last quarter of 2016 and was down 10% year-on-year for Q1 of 2017.
The start-up tracks 250,000 parking lots for 96 retail chains across the US. It found JC Penny’s parking lot figures track its stock price closely.
Research from UC Berkeley added to evidence of the value of satellite data. It found that if, during the weeks before a retailer reported quarterly earnings, investors had bought shares when parking traffic increased abnormally and sold its shares when it declined, they would have earned a return that was 4.7% higher than the typical benchmark return.
They also discovered that stock prices did not adjust as sophisticated investors used the satellite data to profit from trading shares. Instead, during the period before earnings reports, the information stayed within the closed loop of those who had paid for it.
According to The Atlantic, hedge-fund managers use machine-learning algorithms that incorporate car counts as well as other types of alternative data emanating from the geolocation capabilities of mobile phones to monitor consumer behaviour, consumer transactions and retail footfall.
This is just one of the many applications of satellite data. Companies use it to price competitors’ assets, economists use it to predict GDP, and political analysts use it to understand global conflict. But this is just the beginning.
Open satellite data is being combined with an increasing range of other data sources to provide insight to any business with an interest in understanding the present or predicting the future, sometimes in unexpected ways.
For example, Google Street View has been used by the Rochester Institute of Technology to find invasive plants, using images from its cars to do something completely unrelated to map building.
In the future, autonomous cars will constantly be uploading and downloading data, and talking to the cities they drive through. Delivery drones and robots will add to the vast pool of insight.
Within the home environment, sensors, cameras and smart speakers are becoming commonplace. Companies such as Audio Analytic in Cambridge, UK, for example, can use speaker systems’ audio data to recognise smoke alarms, breaking windows, babies crying, and dogs barking.
Counting the cloud cost of Earth observation satellite data
Working out how to manage the sheer volume of data being collected while minimising cloud storage costs is becoming a priority. Machine learning and AI will become part of an industrialised effort to sift through and analyse satellite data and combine it with other data sources to provide businesses with new valuable insight.
While the advent of cloud computing has given the illusion of infinite computing power anywhere, the truth is a different story. According to infrastructure provider NEF, the price to transfer 30 TB per month is $2,500. The key is only to move the data you need.
This is one of the reasons that edge computing is on the rise. The concept is that, with an exponential increase in IoT data, organisations will analyse it with machine learning models close to the location where it is produced, rather than shifting volumes of raw data across their infrastructure.
Satellite data enters the ecosystem at Earth stations. Co-locating these Earth stations with high-performance computing is a natural extension, allowing data scientists to build machine-learning models to analyse data close to the point of delivery. This can help businesses and academia to be more responsive and exploit these burgeoning information resources more efficiently. Then they need only move their most valuable, insightful data elsewhere. This is the next frontier in edge computing: computing at the edge of space.