I would look at the wait stats and top running queries over the same period of time for the spike.

It looks like you're using SQL Monitor. You can use the "Rewind Time" feature to get the queries and wait stats for the time period in question. You can also then sort them by either runtime or IO (I'd do both). Identify what was being processed, and that will help you identify what changed (if it's code, structure, or just data volume). Something is changed. You won't just get a random, sustained spike for no reason at all.