Evaluating weather forecast accuracy: an interview with Eric Floehr — LessWrong