It is an interesting idea but I think quite a bit of work to get the calendar accurate.
I guess one of the problems that I see with this is at what point is the forecast considered the actual forecast for comparison reasons?
For example I am using WXSIM to generate a forecast for me every hour. At what point do you compare the forecast against actual?
3 hours prior to the "compare" time, and what is the compare time? 9am, 3pm, both?
My forecast generally does not change a lot the same day but small changes do occur during the day.
Do you break down the day into 2 halves? or maybe night and day? or Morning, Afternoon, and Evening?
Do you have any thoughts on how you want the data displayed? Would you have a temperature graph displayed? the list is almost endless.
I know most people do not have very accurate data for cloud as it is hard to estimate clouds without additional hardware.
I have a hardware cloud sensor (quite cheap) and I am trying to get more accurate cloud readings for my site as I use its data to calculate my current conditions which changes every minute.
It has a sensor temperature and a sky temperature. But on a really hot clear day the sensor is very high like 40-60 degrees C and the sky can be 10-15C. But on a mild day with no clouds the sensor can be 25 degrees and the sky in the negatives.
I am currently trying a matrix of cells 19 wide by about 30 long ranging from -35C to 55C along the x axis (sensor temperature) and -38C to 60C along the y axis (sky temperature). I then specify the cloud value based on XY coordinate. It is a bit hit and miss but it may work out better than my previous approach just comparing the values.
So even determining that it was "Cloudy" all morning is quite a bit of work - some of this raw data can bee see at
http://weather.crowe.co.nz/SkyClarity/Rain is another issue, at what point do you say it "rained" today 0.2mm, 1mm? More than xx minutes?
Interesting idea though for a calendar.
Chris