Dear Mike and Calendarists:
I have prepared an Excel spreadsheet showing the San Fransisco average temperature 1995 to end of 2008, based on U Dayton records.
It is a ZIP file, so I can't email it to anybody because most email servers reject ZIP files, but you can download it from our file transfer server:
The file will be deleted automatically from the server in seven days.
Try to use a version of Excel that is older than the Windows 2007 version, because that latest version is extremely sluggish when working with this file, and even more profoundly so is the Macintosh 2008 version (Apple Numbers spreadsheet -- totally hopeless, spinning beach ball mouse cursor for hours, I complained to Apple.) I suspect that the newer versions have "new" but very inefficient calendrical calculations in them, probably made highly inefficient because of extra overhead for handling all of the global date format variants.
Within the Excel file are a few worksheets:
The "Landscape" chart shows just 2 years, you can change the x-axis to choose other ranges.
The "Long" chart shows all of the data as one long chart in portrait orientation, but don't try to print it -- impossible! This is the worksheet that contain my sine-wave fits. You can tweak the sine-wave fit coefficients over at the top right, sorry if their meaning may be obscure, this was not intended to be "user friendly", and I haven't worked out how to automatically calculate those coefficients. It is impossible to obtain a perfect fit, because the same sine wave can't fit all of these years, and the peaks in September that Victor pointed out often occur are obviously going to be beyond the curve there, but anyhow overall the data is again consistent with a 35-day lag.
The "Data" worksheet contains the raw data. Degrees Celcius, so there. The SF average annual temperature is about 13.9°C, and the average half-span, well moderated by the nearby Pacific, is only ±4°C.
The "EqSolst" worksheet lists all of the equinox and solstices from 1995 to 2009. The numbers are Windows date serial numbers (day #1 = December 31, 1899, don't believe it when Microsoft tells you otherwise, because Windows thinks that Feb 29 came after Feb 28, 1900), and the fraction is the fraction of day elapsed from midnight in terms of San Fransisco standard time. You can format these to display date and time without adversely affecting the charts, but I didn't bother because normally there should be little interest in looking at this worksheet. The Low and High cells are just to set the axis limits for the drawn lines that indicate the equinoxes and solstices.
-- Irv Bromberg, Toronto, Canada