The following was reproduced from an original hard copy report and does not contain the complete text or figures.
METEOROLOGICAL AND OCEANOGRAPHIC MEASUREMENTS FROM CANADIAN WEATHER BUOYSFinal Report
A Review of Sensors, Data Reduction, Transmission, Quality Control and Archival Methods
Sidney, B.C. V8L 3S8
Tel: (604) 656-0881
Fax: (604) 656-4789
Table of contents
- 1.0 INTRODUCTION
- 2.0 SENSORS
- 2.1 Temperature - Air and Water
- 2.1.1 Location of Sensors
- 2.1.2 Data Collection and Processing
- 2.1.3 Temperature and Sensor Drift in the ZENO
- 2.2 Pressure
- 2.3 Wind Speed and Direction
- 2.4 Wave Height and Period
- 2.1 Temperature - Air and Water
The meteorological and oceanographic data that are reported from the Environment Canada Oceanographic Data Acquisition Systems (ODAS) weather buoys are used daily by meteorologists and forecasters for real time input into forecasting procedures. The data are then archived in a number of locations. The archived data are used for many purposes including verification of modeling activities, establishment of engineering design criteria and evaluation of climate change.
The purpose of this document is to clarify how ODAS buoy data are sampled, processed, reported and archived.
The Canadian Network of weather buoys includes a variety of hull types. The main ones are the 6m NOMAD, and the 3m Discus. There are also two 12 m Discus buoys in Lake Ontario, and a Hexoid and a Toga buoy in Lake Winnipeg. A buoy status report (Table 11 gives details of all the buoys in the network with the exception of the 12 m Discus buoys. Figures 1, 2 and 3 show drawings of the NOMAD, 3 m Discus and 12 m Discus buoys respectively).
At the time of writing, the Canadian Network was beginning to change over from ZENO payloads to WATCHMAN 100 payloads. While this is basically seamless from the user's point of view, any significant differences in how the data are sampled or handled are outlined in this report. This change underscores the fact that data networks are always changing and that care should be taken when using the information from this document to ensure it is still valid.
The first draft of this report was discussed at a workshop in Victoria, B.C., on March 4th and 5th, 1996. The recommendations from the workshop are included in Section 5.0. The series of action items agreed to at the workshop are included in Section 5.3. The workshop agenda and attendees are given in Appendix E.
The temperature reported in the G0ES message is the arithmetic mean of 150 four second samples taken through out the 10 minute meteorological sampling period.
Over the years the ZENO payloads have shown signs of temperature drift. Each time a ZENO is checked for deployment, the coefficients are adjusted to produce the correct temperature. The following discussion identifies the sources of the drift and examines the possible ranges. Table A-1 in Appendix A shows the drift values for a particular ZENO payload between calibration dates for the West Coast buoys. The table also gives the values of air and water temperature ground truth data that were taken at the time of deployment. The ground truth data were recorded from the deployment vessel for three hours after the buoy was deployed. The data were measured using standard WMO procedures. Table A-2 In Appendix A gives the details of when the ZENO payloads were changed at each WMO location, for the West Coast buoys.
The temperature is measured by a Yellow Springs 44018 Thermistor composite. This element is thermally connected to the hull below the water line for the water temperature measurement, and contained in a probe within a mast-mounted radiation shield for air temperature (see Figures 1 and 2 for the locations of these sensors for the NOMAD and 3m Discus buoys). The thermistor composite is made up of two different precision thermistors which, together with two precision external fixed resistors, are specified to produce a resistance ratio which is linear over a temperature range from 35eC to +35aC. The sensors are specified to give interchangeable accuracy to within 0.15ºC.
The thermistor sensor is excited by a 1.00 Volt reference generated by an Analog Devices AD537KH voltage to frequency converter. The sensor divides this voltage according to its temperature. The resulting temperature variable voltage controls the frequency deviation of the AD53KH. The basic frequency of the AD537KH chip is set by the product of the values of two components, a resistor and a capacitor, as well as the characteristics of the chip itself. The frequency output of this device, which corresponds to sensor temperature, is fed to a four-digit decimal counter. This counter is read and re-set by the ZENO controller every four seconds. The counter reading is divided by four and multiplied by the quadratic values from the calibration conversion table. These resulting temperature readings per four-second period are averaged for the ten-minute meteorological sample period to give a final temperature reading.
Due to a number of variables, the frequency of the voltage to frequency converter (and hence temperature) is not known accurately when assembled. The primary contributions are the tolerances of the timing resistor, the capacitor, and the AD537KH itself. This initial uncertainty is corrected through software calibration coefficients. The software calibration coefficients are determined by connecting standard precision calibration resistors to the ZENO top plate connector and producing the desired temperature readout.
To calibrate the system to within 0.1 ºC readout resolution, the ratio of the two external calibration resistors must be correct to within 0.07% (one part in 1,400). While this accuracy is difficult to purchase in standard values, it is simple enough to measure and adjust and check with accurate electronic meters.
The timing resistor used is a metal film type of 2% tolerance. Due to our sources of supply, there is no way of knowing which manufacturer made them, or which is installed where. Resistors of this general type are specified as having a long-term drift of "less than 1 %". The way this is specified on the data sheets gives the impression that they are much better than that, however, there is no other indication.
The capacitor used is a polystyrene type, chosen for stability. It appears that at least a couple of different manufacturers' capacitors have been used in various production runs. Again, specific data are not obtainable. Typical specifications for capacitors of this material indicate a long-term drift (two years) of "less than 0.5%".
Given the above, the combined component drift of the resistor and capacitor over a year could be in the order of 1%. The sensor change due to temperature (the parameter that is actually being measured is about 0.7%/ºC). Thus, the possible effect of a 1% frequency drift due to components could be as large as 1 .5ºC. This is not far off that noted by observation.
The contribution of the A0537KH chip to long-term drift is not stated. Contributions due to temperature and power supply are specified in the 0.01 % range, so it is assumed likely that the chip contribution is this size, or smaller. This would be an order of magnitude less than the contribution of the timing resistor and capacitor.
A degree Celsius represents about 40-50 counts per second in the counter, depending
upon the initial values of the components. No drift is expected in the counter design, and
the drift of the processor clock is negligible in this measurement.
There should be no "drift" change as a result of the software time taken to remove the data and reset the counter, nor in the computations performed to adjust the output to temperature.
Recent changes in the specification of the resistor and capacitor in the temperature circuit should greatly reduce the drift characteristics. The new timing resistors are 0.05% with an absolute temperature coefficient of 10 ppm/°C. The new capacitors have a published drift characteristic of ±(0.05% +0.1pF). The temperature drift over the next two-year period will be closely monitored to determine if the drift problem has been eliminated.
There are three types of barometers in use throughout the Network. These are the Atmospheric Instrumentation Research (AIR DB 2A), the new AIR-SB-2A and the Paroscientific Series 1000. The pressure sensor is mounted within the ZENO housing and vented to the hull compartment via a desiccant box. Its power is only switched on when it is in use. For ocean buoys, the sensor can be assumed to be located at sea level. For buoys in inland lakes, the correction to mean sea level must be taken into account.
The Atmospheric Instrumentation Research Inc. AIR SB 2A Intellilsensor II is a dual diaphragm pressure sensor. It has a total accuracy of ±0.5 mb. This is the current sensor of choice because of technological specifications, reliability and price.
The Atmospheric Instrumentation Research Inc. AIR-DB-2A Intellisensor Air DB is a dual diaphragm pressure sensor. The quoted accuracy over 12 months is ±0.5 mb.
The paroscientific Series 1000 digiquartz pressure transducer is temperature compensated over the range of -25ºC to +85ºC. It has an accuracy of 0.15 Mb and an aging (drift) factor of 0.10 mb per year, which will cause the pressure transducer to indicate a lower-than-actual pressure.
Barometric pressure readings are taken each second for 10 seconds before and 10 seconds after the ten-minute meteorological sample. The twenty readings are then averaged.
The wind speed and direction are collected from two anemometers (see Figures 1 and 2). Both anemometer values are reported in the raw data bulletins (SXCN52 KWAL), but only one in the SMVD17/SIVD17/SNVD17 CWVR and CWHX bulletins. However, this latter bulletin does not indicate which anemometer reading has been included/rejected.
The wind speed and direction are averaged over a period of 10 minutes. The sample frequency is 1 Hz. Thus, 6oo samples are taken in the 10-minute period.
The wind speed is reported in the bulletin both as a vector average in Section 1 Meteorological Data and as a scalar average in Section 4 Comments (see Section 3.1). The wind direction used to calculate the vector average is as measured each second throughout the 1O-minute sample length.
The gust wind speed is the highest 8-second moving scalar average of the wind speed within the 10-minute sample length.
The compass headings that are reported in the comments section (A5 group) of the GOES message are the compass headings from the last reading in the 10-minute sample length. That is, they are not averaged over the 10-minute period.
The four-figure time group within the GOES message (see Section 3) is the time at the beginning of the meteorological sample (i.e., the time of the start of the 10-minute wind sample). Note that for NDBC buoys and C-MAN Stations this time is the end of the acquisition time. This is the same practice as airport weather observations in the U.S. (SAs, soon to be METARs).
There are two separate sensors that are used to measure waves; the fully suspended (gimbaled) heave sensor manufactured by Datawell, and the single-axis strapdown accelerometer manufactured by Columbia Research or Schaevitz. Because of the space limitations in the 3m Discus buoys, in Canada these buoys have always been fitted with the strapdown accelerometer. In the NOMADS, the Datawell gimbaled heave sensor has always been used on the West Coast (WMO positions 46184, 46004, and 46036). Only one Datawell sensor is used on the East Coast. The following summary (Table 2 indicates the historical usage of the East Coast Datawell heave sensor). At all other times, stations 44143 and 44139 have used a strapdown accelerometer. The other NOMAD stations on the East Coast, namely 44137, 44136, 44140, 44141 and 44142, have used a strapdown accelerometer since they were first deployed.
The single axis strapdown accelerometer that is used as a heave sensor is either a Columbia Research SA-107B or a Schaevitz. It is not gimbaled, but is strapped down inside the ZENO payload. This unit is installed on all Canadian 3m Discus buoys and on five of the six operational NOMAD buoys on the East Coast. See above for their deployment history. See Figure 1 for the location of the strapdown accelerometer.Potential for Error
The strap down accelerometer measures the component of gravity perpendicular to the deck of the buoy. If the buoy is level and heaving, the measured component of gravity will have the vertical acceleration due to the waves added to and subtracted from it. This alternating signal is seen and used to calculate the magnitude of the waves. However if the buoy is rolling and/or pitching and not heaving, the component of gravity perpendicular to the deck is still changing. and is interpreted as being waves. If the rolling and/or pitching of the buoy is small (say less than ±15º) than the effect is very small (less than 4%) and by this assumption it ignored (Gilhousen, 1986). However if the buoy is rolling and/or pitching more, say 45º. then the error is larger. If the buoy is heaving at well as pitching and/or rolling. it has not been calculated how they combine together, and what the size of the errors might be. The results from the SWS-1 program, which compares both a strapdown accelerometer and a gimbaled Datawell heave sensor, indicate that the strapdown accelerometer consistently undervalues the wave heights by about 10% through a wide range of sea states.
The Datawell heave sensor works on the principle of the deflection of the tip of a clamped cantilever. This deflection it a measure of the vertical acceleration since the accelerometer it mounted on a gravity stabilized platform. It is housed in a cylinder, 0.4 m diameter, 0.5 m high, and weighs 33 kg. This sensor is installed on the West Coast NOMAD buoys and on one East Coast NOMAD buoy (see Table 2). For the Canadian buoys the maximum range is ±15m or 0.667 V/m. See Figure 1 for the location of the Datawell heave sensor in a NOMAD buoy.
The heave sensor is turned on for two minutes before measurements begin, to allow it to stabilize. The heave is measured by the 12 bit A/D converter card. The minimum theoretical resolution is thus less than 4 mm. A collection of 256 heave samples is made at one-second intervals. and an FFT analysis used to break up the data into spectral bands, This analysis takes 24 seconds. This is repeated eight times, and the results averaged. This whole process takes 8*(256 + 24) seconds, which is 37 minutes and 20 seconds, and occurs just before the meteorological sample starts. The sensor signal is low-pass filtered with a single corner at 10 Hz to reduce aliasing due to noise, etc.
Mean of the valid raw data is removed.
A 10% cosine taper is applied to each 256-second block (i.e., the first and last 10% of the block is tapered by a cosine function).
The FFT is calculated for block
A spectrum is calculated for each block. The spectra are ensemble-averaged into a single 128-band wave spectrum.
The spectrum is corrected for the cosine taper.
The spectrum between 2 and 30 seconds is corrected for the transfer function of the heave sensor. This is:
|1/[1-1.414ia-a²)(1-ib)]||where a = T/30.8|
|b = T/170|
|T = Period Time|
Band averaging is done as indicated in Table 3 to reduce the number of spectral bands Ha and Tp are calculated for frequencies between 2 and 30 seconds (0.5 and 0.033 Hz).Band averaging
The first band (Band 0) includes all energy up to 0.002 Hz;
Band 1 is centered at 0.003953 Hz;
Each band from Band 1 to Band 27 has a band width of 0.003906 Hz (i.e., 1/256 seconds); and,
Beyond Band 27, the averaging is as per Table 2.
|Table 2 - Band Averaging|
Band 1 is centered at 0.003953
|Frequency Range Hz||Band Centre|
The significant wave is calculated as 4 (total variance)½. The total variance, which is also called the total wave energy, is obtained by integrating the wave spectrum over all the frequencies.
The reported maximum wave (Hmax) is a calculated value. It is calculated to be twice the maximum positive excursion of the wave above the mean level computed from the sample. The maximum positive excursion is a measured value and is in fact the maximum crest height. If the reported maximum wave is divided by two, this will correspond to the maximum crest elevation of the wave above the mean level. The reported data give no indication of the depth of the wave trough, although the maximum negative excursion is reported in the spectral wave message. The mean level is calculated as the average height of all the wave measurements in the record.
The maximum dynamic range of the wave sensors is presently set at ±15 m or 0.667 Vm. Recently, a number of storm events have occurred that indicate a larger range should be considered (see Recommendation Number 3, Section 6-1).
The peak period is taken as the period at the peak of the spectrum. In a bi-modal sea, the peak period is the period corresponding to the highest energy peak at the time of calculation.
The data for each spectral band indicated in Table 3 are transmitted from the buoy.
........ The remainder of this document may be obtained from Axys or MSC.