Notes to accompany the 1 December 1999 version of metdata.txt:

This is a tab separated file.  It's pretty long at over 30000 lines.

I am quite sure that times are accurate to within a minute.  They 
are probably good to within a few seconds if the clock rate drift 
is constant.

These times are not directly from the datalogger.  I took the datalogger
times, corrected for clock drift, and then interpolated back to minute
boundaries.  This means there is a slowly varying 1 to 2 point smoothing 
on the data.  If people want the actual times rather than the resampled 
ones, I can supply that.

Values represent averages over the period _ending_ at the time indicated 
(and beginning at the time of the previous line).  This is a potential 
source of confusion.  I can switch it around if that would help.  I could
also resample the data so the time periods center around the minutes.

The date and time formats can be changed easily.

The flag for invalid data is 'NaN', which means 'Not a Number' in Matlab.
If -9999 or something such would be easier to handle, I can do so (by 
running sed 's/NaN/-9999/g' on the file).

Temperatures, wind speed and wind direction data are (probably) final.  
RH is as yet uncorrected.  Radiation is uncalibrated.  The rain gauge 
works, but tends to miss the light drizzles so common at the site, since
the funnel has to be wetted down before water reaches the tipping bucket.

Units:
Temperature, degrees C
Relative humidity, 0-100%
Wind speed, meters/second
Wind direction, degrees clockwise from true north
Rain, mm
Radiation, uncalibrated