Wednesday, August 12, 2009

regarding the accuracy of the station's barometer

[The following are excerpts of an email conversation between myself and Rex Hervey of NDBC.  When CHAMP first started feeding data from this station to NDBC, some questions were raised about the accuracy of its barometric pressure readings.  It so happened we had some similar questions about the WXT barometric pressures at LCIY2 at the same time.  In due course of time we determined that the pressure readings were all correct within instrument tolerances, and later (March 2010) we would add a redundant analog barometer to PVGF1 just to be sure.  Still, the model analyses shared by Rex are interesting enough to include on this blog.]

On June 25th, 2009, Rex wrote:
We are releasing the data from Port Everglades with the exception of pressure.  It looks like it is about one mb low.  We will continue to watch it.  If you can apply an offset, I can give you one after we have more time to look at it.  Picture is not up yet, but here is the web page: http://www.ndbc.noaa.gov/station_page.php?station=PVGF1
My reply on July 5th, 2009 was:
Can you tell me more about how you know the station's air pressure is off?  I assume you're comparing it with reports from surrounding stations but I'd like to know which ones.  The WXT520 weather transmitter is newly-calibrated and I'm not aware that we've had any problems with them in the past.

I could also put a standalone barometer out there for comparison purposes, although that will have to wait until we can coordinate another visit.
Rex followed up on July 6th, 2009 with:
It is compared to surrounding stations and numerical model analyses. Now that we have had a week or so to look at it, the bias is only about .3 to .4 mb low compared to model analyses.  That is within limits, so we will release it.
On August 12th I asked for another look at the situation, and Rex replied the same day with the following text and images:
I sent some plots comparing the baro reading with the NWS numerical model analyses.  Green Xs are from the model. One is over the last 2 months; the other is over the last month.  It appears as though it may be drifting closer to the model results over time.  Overall it is -.26 mb off, and more recently only -.15 mb off.  My initial evaluation after start up was only over a few observations.  I clearly should have given it more time.

One month.

Two months.

Monday, August 3, 2009

station goes offline, and is revived

On July 20th, 2009, I received an email from Rex Hervey (NDBC) about issues relating to data feeds (current and planned) from our St. Croix and Little Cayman CREWS stations.  As an afterthought Rex said:  "Also, we haven't received data from the Port Everglades station for a while now."

He wasn't the only one who'd noticed.  Tom Carsey on July 21st, 2009, sent his own trouble report:  "Evidently the Pt. Everglades site has not transmitted since July 12."

Following up on Rex's report (I was in Little Cayman helping install the new CREWS station at the time, and hadn't see Tom's message yet), I commented in email:
Rex at NDBC mentioned that our Port Everglades data feed went offline.  Checking the archives, it seems like it's been offline for about nine days.  It's difficult for me to investigate further from the field, but I thought I'd mention in case others wanted to look into it further.

The last transmission (that I see) was at 13:42 UTC on day 193 (Sunday morning, the 12th??).
Lew Gramer, who was the acting CHAMP sysadmin in my absence, followed up on July 21st, 2009:
Yes, the last transmission from the Port Everglades station that I see in the archives on our server is still the one on Sunday 12 July at 13:24 GMT. I checked the "usual" diagnostic data from the last day's transmissions (station and datalogger voltages, transmitter forward vs. reflected power, data and error counts, etc.), all seemed normal. Jack is out of town all week as well, but Shoe located that key and will travel to check the site out tomorrow... 
As Lew indicated, Mike Shoemaker (a/k/a "Shoe") traveled up to the Port Everglades site on the morning of July 22nd and was able to bring the station back online.  Lew sent out the following report about Mike's intervention later that afternoon:
Mike Shoemaker visited the Port Everglades monitoring station this morning, reset everything, opened the station door for a while and let it air out / cool off. The station has begun transmitting again: the 14:24 and 15:24 UT transmissions have been received, and were both well-formated. (Shoe, the last hourly transmission has reasonable looking salinity and sea temperature also.)

Shoe believes part of the reason for the sudden failure in early July may be that the enclosure holds too much heat. We checked, and the hourly mean "panel temperature" climbed above 36oC (97oF) or 50 of the hours when we got regular transmissions in late June and early July. Max PTemp was 38.8oC. (Shoe, went down from 32.7 last hour to 31.9 just now, so at least isn't getting worse.) Jack and Mike J., when you both get back, Shoe has ideas for how to fix this.
 Following Jack's next data-recovery visit on July 31st, I reviewed the patched data archives and was able to produce the following email summary of the incident on August 3rd, 2009:
There was NO interruption in datalogger activity in this period.  As far as I can tell from a brief look at the logger/transmitter diagnostics, the problem was specific to the transmitter.  Either the transmitter stopped working, or it merely stopped communicating with the datalogger.  Either way, Shoe's power-reset appears to have fixed the problem, although there's no guarantee that it won't happen again.