U.S. Dept. of Commerce / NOAA / OAR / PMEL / Publications
The depths at which moored velocity and temperature measurements have been made at 0°, 110°W and 0°, 140°W have changed periodically due to shifting scientific priorities in EPOCS. Also, although VACM/VMCM temperature and velocity data return has been high (>90%), occasional instrument failures and the loss of a mooring at 0°, 110°W between April and October 1982 have led to data gaps. We have filled these data gaps where appropriate and gridded the time series to a set of standard levels (Plate 2) to facilitate the computation of mean seasonal cycles (Appendix B), and the comparison of the 1982-1983 and 1986-1987 ENSO events (section 6). For example, at 0°, 110°W we substituted SST as well as temperature and velocity data at 15 m, 50 m, and 100 m from from a backup mooring at 0°, 108°W during April-October 1982; and we substituted velocity at 80 m from the backup mooring during July-November 1986. We also used linear interpolation between vertically adjacent instruments and, where this was not possible or advisable, linear extrapolation and/or interpolation by least squares regression. Some specific examples of these procedures and how they affect subsequent analyses are presented in this appendix. In each case the examples are in terms of daily or weekly averaged data; results (quoted in terms of cross correlations and rms differences) improve with further temporal averaging. Additional examples are given by McPhaden et al. [1988]; similarly, Halpern [1987a] discusses the representativeness of 108°W data as a substitute for 110°W data for variations with periods longer than 1 week.
At 45 m at 110°W we used a regression fill for zonal velocity during October 1987 to December 1988, because mean zonal velocity curvature precludes linear interpolation (Figure 4). The regression equation was based on 3 years (November 1983 to November 1986) of simultaneous measurements at 25 m and 45 m. The correlation between weekly averaged data and simulated data derived from this regression fill was 0.95 with a mean (rms) difference of 0 cm s (15 cm s). We likewise used regression fills for 10-m data based on 25-m or 45-m data at 140°W for March-June 1986, March-May 1987, and May-November 1988. The correlation of simulated daily 10-m data based on the regression formula with actual daily data at 10 m for a 4½ year period (April 1983 to October 1987) was 0.94 with a mean (rms) difference of 0 cm s (14 cm s). Zonal velocity data prior to November 1983 at 110°W were linearly extrapolated based on the underlying vertical shear from 15 m or 20 m depth to 10 m depth for consistency with the more recent 10-m time series.
The SST time series at 140°W was very gappy, so we have substituted 10-m temperatures in Figures 2b and B1. The correlation between 576 contemporaneous daily averaged SST and 10-m temperatures spanning April 1984 to June 1986 was 0.99 with a mean difference of 0.02°C (10 m colder) and an rms difference of 0.14°C. Thus, for our purposes, 10-m temperatures are equivalent to SST at 140°W. At 110°W the correspondence between SST and 10- to 20-m temperatures is less perfect because of greater near-surface stratification. Thus SST data gaps were filled with regression formulae at 110°W using the closest near-surface temperature record. For instance, several months of missing SST data between July 1981 and April 1983 were filled with a regression formula based on 15-m temperatures. The correlation between 233 days of regression-derived SST data and actual SST data during this period was 0.99 with a mean (rms) difference of 0.00°C (0.30°C).
Figure B1. Monthly mean climatologies of sea surface temperature and zonal winds at 0°, 110°W and 0°, 140°W (solid lines). Superimposed are monthly means from the Reynolds [1988] sea surface temperature climatology and Wyrtki and Meyers [1975] wind climatology (dotted lines).
Temperature time series data were sometimes missing below 200 m (e.g., June-November 1988 at 140°W and February-October 1988 at 110°W). For dynamic height calculations relative to 250 dbar at these times, we appended the mean temperature gradient below 200 m to the 200-m temperature record to create an artificial time series at 250 m. This method, discussed by Kessler et al. [1985] and McPhaden et al. [1990a], assumes that temperature changes at depth are due to vertical displacements of the mean thermal structure. Sensitivity experiments in which we substituted a mean gradient estimate for existing moored temperature data at 250 m indicate that the method leads to daily averaged 0-/250-dbar dynamic height errors of O(0.1 dyn. cm).
Go back to previous section or forward to next section