Orion Weather station
The observatory has a orion weather station
(manufactured by Columbia weather systems). It is located on
the north east corner of the platform (next to the old radar
A single board computer (on the platform) controls the
sensors. It is configured to read the sensors every 15 seconds.
Once a minute it will append the new data (ASCII csv format) to a
file on the ao network via ethernet.
The weather station is located on the north east
corner of the platform next to the faa radar blanker. A single
board computer (192.168.100.216) reads the weather station and
writes this data to disc via ftp. The datafile is stored on maestro.
- Data stored on maestro systemdisk: It is exported as
- In OrionWeatherStationData/oriondata.txt
- See ~phil/doc/windowsMount on how to mount it.
- Once a month, the monthly data is extracted from the large
file and moved to a monthly idl binary save file. Plots
are then made summarizing the data for this month.
- plotting script in /share/megs/phil/x101/wst/plotmon.pro
The data for a month is extracted from the idl
archives and plotted. The input data has been sampled at 15 second
intervals. The plots (vs date) are fist averaged to 15 minute
intervals before plotting. If you see any data dropouts in the
plots, then data was not acquired during that time (most likely
trouble with the weather station).
The plots contain:
- Page 1: wind velocity:
- top: average (red) and peak velocity is plotted vs date. The
averaging is over 15 minutes
- middle: wind velocity vs hour of day for the month.
The data is averaged to 15 minutes
- bottom: wind velocity vs direction for the month. Data is
average over 1 degree bins.
- data with wind velocity less than 1mph is excluded (small
amplitudes don't have well defined directions).
- the blue graph is a scaled version of the histogram of the
wind direction. It counts how many times the wind pointed in
- note (there may be some directional shielding from
equipment adjacent to the weather station).
- Page 2: temp,rainfall,
humidity, and pressure:
- top: temperature vs date (averaged to 15 minutes).
- 2nd: cumulative rainfall for the month.
- if the value goes from non-zero to zero, the weather
station was probably rebooted.
- 3rd: relative humidity
- 4th: barometric pressure. This has been adjusted for the
- 0906- cumulative rainfall for month got reset 25jun09. lost
- 0908- cumulative rainfall got reset 18aug09,. probably can
- 1005- cumulative rainfall for month got reset around 13,18
may10. lost data.
- 1004- cumulative rainfall for month got reset around 20apr10.
- 1012- lost some of the cumulative rainfail.. got reset.
- 1108- lots of missing data.
- 1303- monthly rainfail reset around 20mar13.. lost data.
- 1307 -lost cumulative rainfail 29-31jul.
- 1311- monthly rainfall reset on 23nov13
- 1312- lots of missing data.
- 1401- no timestamps 1.5 to 7.5 hours. cumulative rainfall
- 1402 - no timestamps 0 to 7.5 hours. cumulative rainfall reset
- 1403 - no timestamps 0 to 8 hours. cumulative rainfall is
- 1404 - no timestamps 0 to 8 hours. cumulative rainfall is
- 1405 - no timestamps 0 to 8 hours. cumulative rainfall is
- 1406 - no timestamps 0 to 8 hours. cumulative rainfall is
- 1407 - no timestamps 0 to 12.5 hours. cumulative rainfall is
- 1709 - hurricane maria
- 1801 - lost about
1/2 the month. rain totals are not correct.
- 1 -jan14 -> Jul14.
missing hours 0am thru 8am. Lots of dropped data. On
01aug14 a loose sensor cable at the wst was found and repaired.
things started to work better after this.
- temp info updated every 15 seconds.
- Rain records are sent to micro server from sensor whenever .01
inches of rain has been accumulated.
- After sending rain record, the counters are reset to 0.
- Rainintensity (same as rain rate?) this is a 5minute rolling
- The problem is when it stops raining, there are no records
- The 15 second outputs from the micro server keeps the old
intensity value. Even though there is no rain.
- Trouble is:
- rain rate=A Inches/hour
- on Next temp read still A inches/hour.
- Can't tell difference between this and no rain on second
read unless you look at the accumulated rain value. probably
just as easy to use the rain day, value and differentiate
- Nice to have rain year. This allows you to differentiate
adjacent values and see if matches with accumulated day.
station user manual (.pdf)
users manual (.pdf)
Where is the data:
- raw data:
- looks like the data is written to disc about once every 15
secs via ftp from weather station.
- before 07oct14:
- windows: maestro:
- weather station wrote to maestro systemDisc
- every couple of months jorge would move to the maestro
public disc, same directory. with names that contained
last write date.
- i've copied all of jorge's raw files to
- names: oriondata_yymmdd_yymmdd.txt start,end dates
for data in file.
- 07oct14 and after
- weather station is now writing (ftp via pat) to
/share/orion/oriondata.txt (on pserverk)
- a crontab script:
/share/megs/phil/x101/wst/orioncron.sc moves this file
to wst_yymmdd.txt every morning at 1 minute after midnite
(yymmdd is the date for the previous day .. prior to
- monthly idl save files:
- contains subset of raw data.
- wst_yyyymm.sav (about 20mb/month).
- created during monthly processing of data
- before 07oct14:
- /share/megs/phil/x101/wst/procrawfiles.pro .. to get rid
of bad dates
- n=wstsavearchive(bar) to merge with idl archive files
- after 07oct14
- need to write an idl routine to read in files for a
month,then call wstsavearchive.
- idl lib:
- @phil, @wstinit (svn/lib/wst/xxx.pro)
- routines to read raw data, and idl save file data.
defines the struct for idl monthly save file
- orioncron.sc (daily crontab to move raw data files)
- this was the routine used to read the old raw files that
had many months of data in one file. Needed to be run
manually. tried to check the dates.. probably no longer
- user passes in arrary (bar) of wst structs. will merge
- unused routine to read raw data from large file and output
to archive. Not used since it didn't do a good enough job of
automatically checking the timestamps of raw records.
- make monthly plot for .html page.
- calls wstplotmon(). data has to already have been put in
the idl save file.
weekly, monthly, yearly rain accumulations reset
The weather station reports weather info every 15
seconds. Rain reports occur when a .01 inches change in the
accumulated rain occurs. The micro server keeps daily, weekly,
monthly, and yearly accumulations of rain.
On 04nov10 the weather station reset the counts
of weekly,monthly, and yearly rain accumulations. This should not
have occurred. The circumstances were:
The plots shows data from 04nov10
when the counters were reset (.ps) (.pdf):
- power dip around 4:48 am. this lasted for about 35 minutes.
- After the power dip, time stamps for data records was no
longer spaced by a fixed 15 secs.
- time between records went from about 0 secs to about 3
- the jumping around lasted till 12:48. There was a 3 minute
dropout and then the time interval between records went back
to every 15 seconds.
- the cumulative counters for rain were reset at 12:49 when the
records began to be regularly spaced again.
- Page 1:
- top: time between records.
- The jump at 4:48 lasted for about 35 minutes
- after this the time jumped around from 0 secs to 3
- 2nd: rain total for today. This was zero all day since there
was no rain.
- bottom: monthly accumulation of rain.
- There was about .78 inches this month.
- At 12:48 it jumped back to zero.
- You can see the 35 minutes dropout at 4:48 during the
- Page 2: blowup around 12:48
- top: time between records.
- Records should have been spaced by 15 seconds. They were
being timestamped as low as 2 secs apart and as high as 180
- middle: rain today
- bottom: monthly rain accumulation.
- This goes to zero a 12:48.
What could be going on:
- time between records became irregular after the power fail.
- the timestamps are put on the record at the micro server so
the problem must have been at the sensors or micro server.
- the accumulation totals were reset when the microserver
started recording data regularly .
- The timestamp is put on the data read from the sensors
by the microserver.
- If the microserver was busy waiting on a network write, you
might think that you could get a time longer than the requested
- The time should never be < 15 seconds since that is the
interval that was requested for recording the data.
- I wonder if the micro server rebooted itself at 12:48?
- 25aug14: moved data may08 thru aug14 to idl monthly save files
- 31Jul14: found loose input cable upstairs. reconnected
- a bunch of date jumps
- not recording data midnight to about 8am
<- page up