This is a good example of why you just can't take data at face value and make an assumption that the data you receive has no errors. Without setting up some basic sanity checks, you'd end up computing a 60 degree temperature difference for the entire day just from one bad point.
