Abstract
Abstract A new algorithm for combining separately digitized fast and slow temperature signals from the Neil Brown Mark III CTD is presented. Ideally, one would like to exploit the stability of the platinum resistance thermometer (PRT) while retaining the high resolution of the fast response thermistor (FRT). Practical methods for combining the two in the past have not solved the problem of low frequency drift in the FRT. This drift can introduce errors as large as 0.2°C over vertical scales comparable to the CTD cast depth. A running mean filter which is applied in the time domain to both temperature time series is introduced in this paper. The estimated temperature is then found as T c (t) = T FRT + 〈T PRT 〉 − 〈T FRT 〉 , wher 〈〉 indicate filtered values and t is time. The length of the filter is important in the elimination of salinity spikes. A minimum filter length of 1500 scans, or about 60 m, of data is required to calculate acceptable salinities. Because the FRT itself has a longer time constant than that of the conductivity cell, the estimated temperature T c still lags conductivity by about 80 ms. For applications in which the data are further averaged, this small lag can generally be ignored. Alternatively, the lag can be corrected with a lag shift of the temperature time series relative to conductivity, or by more sophisticated techniques such as that introduced by Horne and Toole (1980, Journal of Physical Oceanography, 10, 1122–1130).
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call