Open maximerischard opened 7 years ago
@maximerischard My apologies on not responding to this sooner -- my Harvard email expired, so I didn't get the notification. Now updated to my UCAR email. I'll get back to you this week.
Great! There's no rush, this isn't a bottleneck. I've written a bit more since I posted this, but the summary statistics stuff is in section 1 (and the relevant plots have ended up on pages 2 and 3; I'll take care of things like figure references and captions at the end).
OK, just took a look! As an aside, I very much like your writing style so far -- it's quite clear and fun to read.
In terms of metrics, I agree that mean Tx and Tn for the year is a good metric.
I also agree that the second metric should focus on variability. Since climate people tend to think more about the width of distributions than the day-to-day temperature change (although this is also important/interesting, and there's some cool work on it), I wonder if we should instead use the 95% range (or similar) of the annual distributions, i.e. 97.5th minus 2.5th percentile. It would also be interesting to break that metric down by season to see the effect of the seasonal cycle on this bias.
Following our chat last week, I explored some possibilities for summary statistics that would demonstrate the effects of the measurement time bias. After a few attempts, I settled on two:
I've written up that section of the manuscript, which is on github. The rest of the paper is for now just an outline. No need for detailed feedback on the writing yet, as this is an early draft, but please let me know if you think this is a good way to explain the issue that we're trying to address.