Despite whimsical advertisements about computing “in the cloud,” the internet lives on the ground. Data centers are built on land, and most of the physical elements of the internet—such as the cables that connect households to internet services and the fiber-optic strands carrying data from one city to another— are buried in plastic conduits under the dirt. That system has worked quite well for many years, but there may be less than a decade within which to adapt it to the changing global climate.
Most of the current internet infrastructure in the United States was built in the 1990s and 2000s to serve major population centers on the coasts. As new connections were needed, companies installed them alongside roads and railways, and such rights-of-way often hug coastlines. (Installations were done along these routes for reasons of both access and security, and because anyone doing construction or repairs there would already be cautious about buried lines.)
A public map of the physical infrastructure of the internet was first published in 2013 by computer scientists Paul Barford of the University of Wisconsin–Madison and Ram Durairajan of the University of Oregon and their colleagues. The map was the result of four years of painstaking work to assemble data mostly from public records of cable-laying permits. On the map, Barford and Durairajan identified the number of key network locations in coastal areas, and showed exactly how close they are to the shore. Building on that work, I joined them to study the risk to the internet from rising oceans.
The basic approach was simple: Take the map of internet hardware and line it up with a map of projected sea-level rise to see where network infrastructure may be under water in the coming years. To fuse the two datasets, we reconciled the different geographic projections and spatial resolutions, placing the data in a common spatial grid. Then we were able to visualize and analyze the overlap of projected average sea level with current internet infrastructure.
Understanding the Threats
Where the internet is not underground, much of it is actually under water already: A physical web of undersea cables, some 885,000 kilometers of them in about 300 cable systems, carries massive amounts of data between continents in milliseconds. Those cables are encased with tough steel housings and rubber cladding to protect them from the ocean. However, they connect to the land network, which was not designed with water in mind. Although the standard buried fiber conduits are resistant to water and weather, most of the deployed conduits are not meant to be under water permanently. If these plastic pipes carrying wires underground were to flood, the water could freeze and thaw, damaging or even breaking wires. It could also corrode electronics and interrupt fiber-optic signals. The fact that a great deal of conduit infrastructure was deployed over the past 20 years and is aging means that all seals and cladding are likely to be more vulnerable to damage, especially if they are under water.
To identify what was now dry but will one day likely get wet, we had to sort through a wide range of potential scenarios, based mainly on varying estimates of how human-generated greenhouse gas emissions will change over time. We used Sea Level Rise Inundation (SLRI) data from the Digital Coast project of the National Oceanic and Atmospheric Administration (NOAA). This diverse data set includes a collection of geo-based sea-level rise projections for the United States over the next century. We settled on one that NOAA created in 2012 and recommended for analysis of situations involving expensive long-term investments, such as for infrastructure projects.
Based on the assumption that global greenhouse gas emission trends will continue in their current relationship to human population and economic activity, that SLRI model expects global average sea levels to rise 0.3 meters by 2030, and a further 1.5 meters by 2100.
Although this rate of increase may sound improbably high, a 2017 report by NOAA also includes an even more “extreme” scenario, which takes into account the mounting evidence of more rapid melting of the Greenland and Antarctic glaciers, such as a 2018 study by geophysicist R. Steven Nerem of the University of Colorado–Boulder and colleagues, which uses historical satellite altimeter data to show that global sea-level rise is accelerating.
In addition, our analysis is conservative because it does not consider the threat of severe storms that would cause temporary sea-level incursions beyond the predicted average.
The Effects of Rising Waters
What we found was not particularly surprising, but it was alarming: The internet is very vulnerable to damage from sea-level rise between now and 2030. Thousands of kilometers of cables now safely on dry land will be under water. Dozens of ocean-cable landing stations will be too, along with hundreds of data centers and network-interconnection locations called points of presence.
There will be further damage by 2100, though the vast majority of the danger is between now and 2030. In some metropolitan areas, between one-fifth and a quarter of local internet links are at risk, and nearly one-third of intercity cables.
Our study also found that risks to internet infrastructure are not the same everywhere. Our results show that communication infrastructure in New York, Miami, and Seattle, respectively, are at highest risk. New York City and New Jersey are especially vulnerable, in part because they are home to many ocean landing sites and data centers, as well as lots of metro and long-haul cables. In addition, studies by geologist Timothy Dixon of the University of South Florida and his colleagues have shown that the mid-Atlantic U.S. coast is sinking up to 2.5 centimeters per decade. The Atlantic coast is also relatively close to the Greenland ice cap, which Nerem and his colleagues have also shown to have regional effects on sea level.
Questions for the Future
It’s important to note that these risks do not necessarily mean that U.S. internet service will get worse or be disconnected by 2030. For one thing, the companies that operate these cables and facilities may choose to relocate them to safer ground—but the costs of doing so may be passed on to customers. Mitigation measures, such as seawalls and hardened enclosures for submarine cable landing points, may buy some time, but will likely not be effective over the long term. In addition, structures that are buried will be much more difficult to access and secure.
In some metropolitan areas, between one-fifth and one quarter of local internet links are at risk, and nearly one-third of intercity cables.
Satellites may become an additional route to pick up some of the traffic. Right now, even though most of our end connections to the internet are wireless, satellites only account for about 1 percent of human interactions with the internet. Because most telecommunications satellites are in geosynchronous orbit, about 35,000 kilometers above Earth, radio waves take about 230 milliseconds to get to a satellite and back, which is about nine times slower than, say, a signal traveling between New York and London on a fiber-optic cable. That kind of latency is too slow for real-time communications, such as teleconferencing. And satellite services have also been more expensive than fiber-optic cable. Some companies, such as SpaceX, are looking to grow satellite internet infrastructure at lower orbits (allowing faster transmission speeds), but these programs are still in the planning stages.
But even if companies don’t move their equipment, the internet already has many redundant pathways for data. Even a single email message is broken into small pieces that may follow separate paths to the recipient’s computer. The systems that manage this routing could potentially handle the additional traffic around wet areas— but that may affect service quality.
Future deployments of internet infrastructure will need to consider the impact of climate change. These plans must include consideration of such issues as acquiring and paying for new rights-of-way for laying cables, and costs and projections relating to how populations will move. Other aspects of risk-aware deployment include developing new methods for hardening fiber cables, conduits, and other infrastructure to be more resistant to the severe weather that will be a consequence of climate change.
We’re planning to study the potential effects to the network and its users in future research. We also plan to expand our study beyond the United States. For now, though, it’s safe to say that internet service in several U.S. coastal cities will need to adapt to sea-level rise soon, and someone will need to pay for it.
- Durairajan, R., C. Barford, and P. Barford. 2018. Lights out: Climate change risk to internet infrastructure. In Proceedings of the Applied Networking Research Workshop 2018, in Montreal, Canada, July 16, pp. 9–15. doi:10.1145/3232755.3232775
- Durairajan, R., S. Ghosh, X. Tang, P. Barford, and B. Eriksson. 2013. Internet atlas: A geographic database of the internet. In Proceedings of ACM SIGCOMM HotPlanet in Hong Kong, China, August. doi:10.1145/2491159.2491170
- Karegar, M. A., T. H. Dixon, and S. E. Engelhart. 2016. Subsidence along the Atlantic Coast of North America: Insights from GPS and late Holocene relative sea level data. Geophysical Research Letters 43:3126–3133.
- Leuliette, E. W., and R. S. Nerem. 2016. Contributions of Greenland and Antarctica to global and regional sea level change. Oceanography 29(4):154159.
- Nerem, R. S., B. D. Beckley, J. T. Fasullo, B. D. Hamlington, D. Masters, and G. T. Mitchum. 2018. Climate-change–driven accelerated sea-level rise. Proceedings of the National Academy of Sciences of the U.S.A. 115:2022–2025. doi:10.1073/pnas.1717312115
- Sweet, W. V., et al. 2017. Global and Regional Sea Level Rise Scenarios for the United States. NOAA Technical Report NOS CO-OPS 083. https://tidesandcurrents.noaa.gov/publications/techrpt83_Global_and_Regional_SLR_Scenarios_for_the_US_final.pdf