There is something both reassuring and deeply depressing about scientists building an ark for data (see here: https://e360.yale.edu/features/germany-us-climate-data). That is more or less what is happening in Germany at the moment. Apparently the German Climate Computing Center (DKRZ) in Hamburg is preparing additional storage capacity for copies of important climate data from the United States. The project even has an appropriate name: CMIP7-ark. CMIP is a project of the World Climate Research Programme (WCRP).
The idea is relatively simple. Climate research produces enormous amounts of data, much of which is stored at research institutions and government organizations. Scientists around the world depend on that information. Historical measurements, climate simulations and long-running datasets are especially valuable because you cannot simply recreate decades of observations if they disappear.
Normally, you would probably assume that data stored by major American scientific institutions is reasonably safe. But that assumption is apparently no longer considered sufficient.
According to an interesting article in Yale Environment 360, German researchers and research organizations are increasingly concerned about the long-term availability of American climate data. The German Research Foundation has therefore established a program to create safe copies of publicly available U.S. scientific datasets. The program has a budget of $35 million.
In Hamburg alone, DKRZ plans to add around five petabytes of storage capacity for backups of climate simulations. For comparison: DKRZ already stores around 254 petabytes of data. It looks like CMIP7-ark is not the only project. Other initiatives with names such as SHIELD and SEND — Secure Endangered Earth Data — apparently are doing similar things. The objective is essentially to make sure that important scientific information remains available, regardless of what happens to the original repositories.
There is actually a broader lesson here that goes well beyond climate science. We have become accustomed to thinking of digital information as something almost abstract. Data is “online”, therefore it exists. But of course that isn’t how it works. Somewhere there is a server. Or a storage system. Or magnetic tape. Someone maintains it. Someone pays for it. Someone decides whether it remains accessible. And sometimes there is only one really important copy.
That suddenly makes scientific data surprisingly vulnerable. Not just to political decisions, but also to cyberattacks, war, natural disasters, technical failures or simply a lack of money and maintenance. The Yale article mentions examples ranging from datasets in the United States to scientific information threatened by the war in Ukraine.
So perhaps this German initiative points towards a much more resilient model for scientific information in general. Instead of assuming that one institution or one country will preserve important datasets forever, make multiple copies and distribute them across organizations and countries.
That is basically the same principle we use for important personal files. One copy is not a backup. And one backup sitting next to the original isn’t a particularly good backup either. For information that may be important to researchers for decades — perhaps centuries — geographically distributed copies suddenly sound remarkably sensible.
So yes, I think this is a good development. Important scientific data should probably never depend on a single institution, storage facility or even country. If projects such as CMIP7-ark eventually lead to a more distributed international infrastructure for preserving scientific data, that would be valuable. But at the same time, it is of course rather sad that we apparently need projects like this in the first place.