Maurizo's point about clouds will work, but the cloud data will not be good enough to do this. Anders couldn't sort out the Swedish cloud data - even in the 20th century. I don't think the cloud data - even if digitized - would be a help. There are so many issues with observation times, and the fact that observers appeared to sometimes base the observation on the time they looked and sometimes over the whole day. The Swedish categories also reduced the further back in time. I don't want to slow down the paper, so this is only a suggestion, but I think the time series of rainday counts may be a better homogeneous proxy for cloudiness. Cloud observations will be a waste of time!
Madness at #AGU2018 – probability based tornado warnings
42 minutes ago