Data quality of TRITON meteorological sensors

The meteorological sensors on TRITON buoys except for the rain gauge were investigated.   The anemometer, thermometer and hygrometer, and short wave radiometer were developed and manufactured at Woods Hole Oceanographic Institute (WHOI), and the barometer was manufactured at Paroscientific, Inc (Table 1).   All sensors were calibrated at manufacturers before delivery, and then calibrated again at JAMSTEC.   The reference sensors of JAMSTEC are all calibrated once per year by standard instruments to be traceable in Japan, except the sensor for temperature calibrated by Sea-Bird Electronics.   The calibration in JAMSTEC is basically to measure residuals of the sensor output to the standard instruments in measurement units.   If the residuals become larger than a certain level for each sensor, the coefficients in sensor will be rewritten.   The RMS (root mean square) residual errors for each sensor are almost within 0.1 degree-C for air temperature (with the average RMS of 0.06 degree-C and the largest residuals of 0.2 degree-C), almost within 2%RH for relative humidity (with the average RMS of 1.3%RH ) and almost within 0.2hPa (with the average RMS of 0.09hPa) for atmospheric pressure.   The residuals of wind direction sensor are 2 degrees bias of differences in average (with the standard deviation of 6 degrees and the largest differences of 10 degrees).   The averaged torque of wind speed sensors is 1.6 gram*cm with the standard deviation of 0.9 gram*cm, which is within the manufacture's recommendation (2.6 gram*cm) and guaranteed to be within the manufacturer's specification of 0.3m/s.   For the short wave radiation sensors, the residuals of all sensors are within ?6% to +4% in comparison with the reference radiometer at JAMSTEC (Table 2).   As mentioned above, the calibrations of meteorological sensors are just to measure residuals from the reference sensors, and we seldom change the coefficients in sensors. 

The drifts of meteorological sensors are all small compared to the RMS errors of each calibration. For hygrometers, it is not significant, but small drift of about ?1.2%RH per year in average was found.

The effects of uncertainty of sensor calibration on estimating the surface heat flux, particularly the latent heat flux, were calculated using the TOGA-COARE bulk formula 2.6.   Using all real time meteorological data from the TRITON buoys since 1998, the errors of latent heat flux originated in the uncertainties of meteorological data based on laboratory calibration were estimated.   For the averaged latent heat flux of 120 Watt/m^2, the errors from the uncertainties of wind speed (0.3m/s from the manufacturer's specification), SST (0.01degree-C), air temperature (0.1degree-C), relative humidity (2%RH) were 6.8 Watt/m^2, 0.3Watt/m^2, 2.4Watt/m^2 and 8.9Watt/m^2 respectively in average for all data of available periods.   The total of these errors was 18.4 Watt/m^2, and counts 15.4% for the total averaged latent heat flux.   The result indicates that the calibration of relative humidity should be improved for better heat flux estimation.  

Table 1  List of sensors installed on TRITON buoy

Sensor	Type	Manufacture	Installed location	Height or Depth
Anemometer	ASIMET, 7050-A (with R.M. Young, 05103)	WHOI (Woods Hole Oceanographic Institution	Top of buoy's tower	3.5 m
Thermometer and hygrometer	ASIMET, 7030-A (with Rotronic, MP-101A)	WHOI	Lower part of buoy's tower	2.2 m
Barometer	DIGIQUARTZ, MET 1-485	Paroscientific, Inc.	Lower part of buoy's tower	2.2 m
Rain gauge	ORG-115 (mini-ORG)	Scientific Technology, Inc.	Upper part of buoy's tower	2.2 m
Short wave radiometer	ASIMET, 7070-A (with Epply Precision Spectral Pyranometer)	WHOI	Upper part of buoy's tower	2.2 m
Thermometer	MicroCAT, SBE37-IM	Sea Bird Electronics	Under sea	Under water 1.5m

Table 2  Calibration results of TRITON meteorological sensors

Parameter	Calibration system	Stability of calibration system	RMS errors in calibrations	Drift of sensors	Conservatively estimated error for real-time data
Air Temp.	SBE3 temperature reference sensor for CT sensor calibration. Accuracy of 2 mK (0.002 C)	within 0.001 C	Almost within 0.1 C in RMS. Average of RMS is 0.06 C. Between +0.2 C to ?0.2 C for maximum residuals.	Not apparent, conservatively within 0.1 C per year.	0.2 C
Relative Humidity	Humidity generator (EM25) calibrated by HYGRO M4-RH/D-2-XR (Accuracy within 2.3 %)	within 0.05 %RH	Almost within 2 %RH in RMS. Average of RMS is 1.3 %RH. In +2 to ?6 %RH for maximum residuals.	Not apparent, but tend to drift ?1.2%RH per year.	4 %RH
Atmospheric Pressure	Digital pressure controller calibrated by RUSKA (Accuracy of 0.008 %F.S.)	within 0.05 %RH	Within 0.2 hPa in RMS. Average of RMS is 0.09hPa at 1000hPa.	Not apparent, conservatively, within 0.2hPa per year.	0.2 hPa
Wind Speed	Pulse counter and Torque meter	N/A	2 % higher wind speed in pulse counting check. 1.6 gm*cm in average with the S.D. of 0.6 gm*cm in torque check (Within 2.6 gm*cm is manufacturer's recommendation)	N/A	N/A
Wind Direction	Calibration stand set in outdoor	N/A	2 degrees in average with S.D of 6 degrees. In +10 to ?10 degrees for maximum residuals.	Not apparent.	0.2 hPa
Short wave Radiation	Referred to short wave radiometer manufactured by Ishikawa Sangyo, and calibrated by JMA (Accuracy of   2%). Comparison using artificial sun.	within 0.4 %	Within 5 % in most sensors. In +4 % to ?6 % for maximum residuals.	Not apparent, conservatively, within 4-6 % per year.	6 %

Wind speed sensors (colored by gray) are not calibrated, but generation of right pulse number and having the torque in the certain level (2.6 gram*cm) are checked.  

The calibration stand for wind direction sensors has set in October 1999, so the statistics in this table does not include the calibration data before October 1999.

RMS is calculated in each calibration for each sensor (cf. each type of sensor).

All calibration data used in this calculation is based on the calibrations from January 1998 to April 2001 except for wind direction sensors.