What happened at 2:22pm on 6/23/06?
What went wrong from 5/30/05 to 6/12/05?
Did something change on 11/5/02?
How should I specify dates for historical data access?
How far back do you have data?
Images not loading / broken images?
Graphs load, but are blank.
Rain gauge?
How accurately do readings here reflect actual conditions?
How's the time accuracy on the historical data?
Why'd you mention Community Boating and MIT specifically?
Can I have access to the raw data?
What software is doing all this?
What's with all the different names?

What happened at 2:22pm on 6/23/06?

Lightning!

I don't think it was a direct hit, because nothing seems burnt, but a lightning strike must have been very very near by, because at 2:22pm on friday several electronic things in my apartment went haywire. The weatherstation itself started giving bogus temperature readings, and at the same time some lines in the anemometer shorted out, killing the wind readings too. The lightning also caused one of my computers to reboot and killed my stereo and caused a random wierd pattern to appear on my vacuum fluorescent two line text screen in the front room.

On saturday [6/24] I swapped wires so wind speed is back online, but not direction. Now that it's spinning again I think the anemometer seems to have more friction than it used to. It seems to take a stronger breeze to start it turning than it used to.

On monday [6/26] I ordered a new weatherstation with 2 day delivery so it should be back up on Wednesday night. Better than ever... with rain guage and humidity sensor, even!

On [7/16] I finally got the weatherstation up using a new database logging scheme which should simplify integration of more weatherstations.

What went wrong from 5/30/05 to 6/12/05?

Mea culpa. Due to sloppy work on my part, the junction box on my roof was exposed to moisture [full of water] and the connections to the anemometer corroded. Direction data for this period was lost since as a quick hack I started using the direction wire for speed. What surprised me is that the direction wire's connection was flaky and [perhaps when the T rumbled by] would pulse on and off causing the weater station to think that the anemometer was spining quickly.

Looking at the raw data, it's trivial to automatically identify and remove [almost] all these false spikes. Real gusts have a characteristic "decay" profile, as the anemometer cups gradually slow down, whereas these recent false gusts spike up for a few samples, then fall right back down to a more normal level. If I get any requests I'll write something to go through the data for these last couple weeks and filter out the noise. The junctions in the junction box are now waterproof connectors designed for telephone line splices, and the junction box itself is somewhat sheltered from water now.

Did something change on 11/5/02?

Yes; they're reroofing my building's roof. I temporarily moved the anemometer from being bolted to the wood roofdeck to being bolted to the sewer vent pipe a couple feet away. It was about two feet south and about six inches higher than it had been for the last three years. Today, I had to move it entirely off my roof; it's temporarily on my neighbor's roof, and it's unfortunately about 3 feet lower and is substantially wind shadowed to the West. It's also only a couple feet from the roof edge, and it might be subject to updrafts / wind funneled by the building. I'll move it back asap!

Images not loading / broken images?

If you're looking up historical data, see the info below. If the images are downloaded, but show up as broken, your browser may not be able to handle PNG images.

Graphs load, but have no data.

If you're not looking back at old data this might be a real problem, send me an email or give me a call if you know my number. If the graph loads, with the title, and axis labeled, but no data line then that's definitely something for me to fix asap. If you're looking back at old data maybe you're looking back too far in time for that particular database / weatherstation. See the later section on how far back I have data.

How should I specify dates for historical data access?

Anything that the gnu date utility will accept. "1pm yesterday", "2pm last saturday", "04/17/2002 20:00", etc. If your images aren't loading perhaps the date spec wasn't understood, try clicking on an image to download it individually. It might give you a more descriptive error.

How far back do you have data?

Data are on line back to 9/04/2000; I have data back to 6/19/1999, but are not in an easily useable format. To get data between 9/2000 and 7/2006 you have to select the "flatfile" option, eg db=flatfile in the url. The new-improved west cedar roof data, which includes rain, pressure, humidity go back to 7/2006 choose the postgres database and fields either std, dew or full. CBI Dock data goes back to October 2006 and to get this choose postgres database and select the cbi fields, eg db=postgres&fields=cbi.

How accurately do readings here reflect actual conditions?

Pretty well. I have pictures from my roof of the directions that my station is in a wind shadow, I need to make time to put them up. Shifty days show up with a big spread in the direction dots. Gusty days show peak speeds triple or more average speeds, consistent days show peak speeds only double average speed. You tell me! [esw at alum dot mit dot edu]

How's the time accuracy on the historical data?

Well, don't go setting your watch by it. As of 2/13/2002 the server's been using ntp to keep time correctly, so you can trust it now. Up until then it has been fairly poor at keeping time, perhaps up to a half hour off. By looking for discontinuities I might be able to reconstruct the true time for any time in the past... but not easily.

Rain gauge?

It's been up and working nicely since July 2006.

Why'd you mention Community Boating and MIT specifically?

I work for a lab of MIT and I've been sailing and windsurfing at Community Boating for the several years. Check out http://wile.org/ for more information about me, including my use of GPS tracklog files for windsurfing and rollerblade performace analysis.

Can I have access to the raw data?

Sure! Send me an email asking for particular old days. If you want current [near real time] data, suggest a method for me to provide it that minimizes hassle/webserver processing as well as traffic to my site. [ie, not one that downloads the same data multiple times over the course of a day.]

What software is doing all this?

A bunch of Perl scripts. Some to take the data from the Peet Brothers Ultimeter 100, Ultimeter 2100 and the CBI Davis weatherstation and write it to a timestamped log file. Another script loads this data in to a postgres database. A materialized database view provides 1 minute, 5 minute and 1 hour averages of this data. The data for the time period of interest is extracted and plotted using gnuplot. Just ask if you want the source. If there's sufficient [eg, any!] interest I'll put up a tarball on my site and link to it here.

What's with all the different names?

Well, I first started calling this "Wind at West Cedar." I still like that name, but it's kind of long and nobody knows where West Cedar street is, even if they walk by it everyday. Then I started calling it "wwwind" which has a nice ring to it, it's easy to remember... but wwwind.org was taken, and while wwwind.net was available, I didn't think that name was very appropriate. So, I settled on cbiwind.org because it was also fairly catchy and memorable plus .org .net and .com were available. I only registered .org. That's the coolest one anyway.

-erik wile esw at alum dot mit dot edu