Times are reported in Universal Time (UT), which is the same as Greenwich Mean Time (GMT). Drift of instrument clocks
is determined by comparing instrument times with accurate times (from the Global Positioning System (GPS) or clocks synchronized with the National Institute of Standards and Technology in Boulder, CO; (http://tf.nist.gov/)) before and after deployments. Ordinarily, the observed offsets were small compared to the sample intervals, typically a few seconds; therefore most instrument times in this dataset are not adjusted for clock
Time is stored in the EPIC-compliant NetCDF files as two variables named "time" and "time2". This technique avoided round-off error on older computers that did not have sufficient resolution when time is stored as an integer. "Time" is the time in whole Julian days, where midnight on May 23, 1968 = 2,440,000 and "time2" is the time in milliseconds since midnight. Thus, the time in Julian days is computed as
time + (time2 / (1,000 ms/s × 3,600 s/hr × 24 hr/d)) where the product in the denominator is the number of milliseconds in a day. In the CF-1.6 version NetCDF files, time is stored as a single variable “time” in seconds since January 1, 1970 00:00, again using universal time (UT).
Burst and Wave Data
Burst and wave data files contain variables that are shaped differently than evenly spaced time-series data. Most of the files in this database have evenly spaced sample intervals, and the variables are vectors of the length of the time variable. Burst data are sampled rapidly at an even interval but have gaps between bursts; time is represented a matrix. For example, an ADV might be programmed to sample hourly bursts of 17.5 minutes, at 8 Hz, so there are 8,400 points per burst. The time variable(s) are defined so that dimension 1 is the number of samples in the burst, and dimension 2 is the number of bursts, so for this example, the time variable(s) is shaped (8,400 × 2,045). Consequently, the data variables in the time series also have that shape. The burst mean variables presented in the ADV statistics file for this example are vectors of length 2,045.
Wave data files from acoustic current meters contain power spectra and possibly directional spectra; these variables are dimensioned by time (as a vector), frequency, and possibly direction. Some of the pressure-based waves files will contain power spectra dimensioned by time and frequency. The manufacturer’s waves-processing software is used to generate the variables stored, so the contents will vary based on what the code computes. Units and computation techniques may be inconsistent between a ‘pspec’ variable from an ADCP, an AWAC and a seagauge file. Evaluation of using DIWASP (Johnson, 2011) to process directional wave data in a consistent way is in process.
An objective in treating data from a wide variety of irregularly sampled data types (bursts, waves) is to have enough consistency in structure to allow use of one of two viewers to visualize the data. A versatile visualization tool was developed to view ADV, PCADP, Aquadopp, and Vector burst data as well as the data from Seagauge, D|wave, and HOBO wave loggers, and a related program was developed to view files containing wave spectra data.
This section pertains to the EPIC version of the data. There are no models for representing burst and wave spectra data in CF yet, so these types are only served in the EPIC-compliant form. It is likely that the waves and burst data will be served in a CF-compliant form in future, and in that case, the data may be represented as a vector, irregularly spaced in time.
When interpreting the current meter data, processing software must consider two aspects of orientation: (1) whether the instrument's (transducers) face up or down, and (2) the orientation of the instrument's beams relative to the Cartesian system. Each of the ADVs, PCADPs, AQDs, ADPs and ADCPs are equipped with internally mounted flux-gate compasses. These are calibrated prior to deployment according to manufacturer's recommendations. The ADCP velocity data are recorded in raw beam coordinates by instrument firmware and are rotated to geographic coordinates (east, north, up) in post processing using the internal compass and are corrected for local magnetic variation. Data from the ADV, PCADP, and Vector are recorded in instrument coordinates (referred to as x, y, z coordinates here) and later rotated into geographic coordinates and corrected for local magnetic variation.
In cases where several current meters are mounted on the same tripod, possibly facing different directions, all data from the raw beam coordinate system are initially converted to the east-north-up coordinate system. Because the current is expected to be uniform over the spatial scale of the platform, the primary axes of data from similar sampling heights may be aligned and used to determine the actual instrument orientation.