Speed is the most likely reason. Think about it from the flip side of the argument. Their argument rests on the assumption that if a known terrorist is in contact with a US person and there is reason to believe that US person is in the US for nefarious purposes, then they may already be moving to attack somewhere and the NSA needs his call records ahead of time. They want to know what other foreigners is he in contact with, are they known terrorists, are the people in the US that he's talking to also talking with each other (which could indicate a terrorist cell), etc. Answer these questions, package up the information and shoot it over to the FBI for follow-up investigation.
If they didn't have the records ahead of time, the NSA or the FBI would have to get a warrant, submit the warrant to all N of the major phone providers, wait for however long it takes them to process it, get responses from N-1 providers that the phone number doesn't belong to them. They then have to analyze the data, then repeat the process for the second layer, etc. What would be done in a few minutes if they already had the data now might take a few weeks. If you believe that an attack is likely, you may not have a week to identify the initial contact, two weeks to identify the cell he's working in, etc.
All of this rests on the assumption that analyzing phone metadata is good way to get leads for an investigation on international terrorism. Are there better ways to get leads? Can the information still be useful if gathered through a traditional warrant process? If not, is it worth the privacy tradeoff to foil X number of cases over the course of a decade, and are there good ways to handle the information in order to mitigate privacy concerns?
If they didn't have the records ahead of time, the NSA or the FBI would have to get a warrant, submit the warrant to all N of the major phone providers, wait for however long it takes them to process it, get responses from N-1 providers that the phone number doesn't belong to them. They then have to analyze the data, then repeat the process for the second layer, etc. What would be done in a few minutes if they already had the data now might take a few weeks. If you believe that an attack is likely, you may not have a week to identify the initial contact, two weeks to identify the cell he's working in, etc.
All of this rests on the assumption that analyzing phone metadata is good way to get leads for an investigation on international terrorism. Are there better ways to get leads? Can the information still be useful if gathered through a traditional warrant process? If not, is it worth the privacy tradeoff to foil X number of cases over the course of a decade, and are there good ways to handle the information in order to mitigate privacy concerns?