The Oceans Their Physics, Chemistry, and General Biology "d2_3

TREATMENT AND ANALYSIS OF SERIAL OBSERVATIONS

After a number of temperature observations, salinity determinations, or other analyses have been made, it is necessary to eliminate those values that are in error and to put the data into a convenient form to make them readily comparable with other material of a similar character. The methods by which observations obtained with reversing thermometers are corrected to yield the temperatures in situ and by which the depths of sampling can be computed from the readings of unprotected thermometers have been described on preceding pages. The determination of salinity by chlorinity titration and other methods is discussed in chapter III. In addition, numerous chemical tests and analyses may be made on the water samples by methods listed in chapter VI. After the temperature in situ and the thermometric depths have been calculated, the observational data are generally listed on a summary sheet which shows the following information about each water sample: wire depth of each sampling instrument (the amount of wire rope between the surface and the sampling device), wire angle for each cast, temperature in situ (the corrected protected thermometer reading, or, if two thermometers were attached to the bottle, the average), depth of sampling as computed from the unprotected thermometer (these are usually not attached to each bottle), chlorinity, salinity, and other analyses that have been made.

The next step in the analysis of serial oceanographic observations is to establish the depths of sampling. A considerable amount of personal judgment enters, and this can be improved only by general experience, knowledge of the area in which observations were made, and familiarity with the behavior of individual instruments. The general procedure that must be followed is somewhat as follows: (1) if there was any error in the meter wheel, the necessary corrections must be applied to the meter-wheel readings; (2) if the wire angle was 5 degrees or less, the depths of sampling can be taken as equal to the wire depths as

― 358 ―

obtained from the corrected readings of the meter wheel; (3) if the wire angle was greater than 5 degrees, the ratio of the wire depth to the thermometric depth at those levels where unprotected thermometers were used is calculated. If the wire angle was small and the cast shallow, these ratios will be virtually constant, and the average of the ratios obtained from any single cast may then be used to arrive at the depths of sampling for all instruments on that cast. If the wire angle was large or the cast deep, the ratio will usually be found to increase with increasing depth. This increase indicates that the wire rope approaches the vertical at greater depths, and under such circumstances a ratio that changes with wire depth must be used. In extreme cases the “wire curve” may be plotted. The wire curve, which is the actual form assumed by the wire in the water, is constructed from the thermometric depths, the wire angle, and the amount of wire between the sampling bottles. In all cases the accepted depths arrived at in the above manner should be examined to see that the distances between sampling bottles are compatible with the wire distance between them. The above method of treatment presupposes that the observations are themselves valid. If, for any reason, the thermometers did not function properly or if the reversing bottles closed when lowering, the thermometric depths may not be correct and the results must be discarded.

After the accepted depths have been decided upon, the next step in the analysis of the observations is to plot vertical distribution curves. In one diagram, one of the observed properties may be plotted as a function of depth at a number of stations, or all of the observed properties at a single station may be plotted. From such curves it is often possible to detect those samples that are incorrect because of faulty functioning of the thermometers or because the water-sampling device closed prematurely or leaked when being hauled up. From the vertical distribution curves the following data are taken: (1) the interpolated values at standard depths, and (2) the depths of selected values of a given property at a number of stations. These data are necessary in order to construct vertical sections of the distribution of any property.

Besides the vertical distribution curves, it is common practice to prepare other plots of various kinds which serve either to detect errors or to bring out characteristic features of the data. One of the most common of these is the temperature-salinity curve (T-S curve), in which the corresponding values of temperature and salinity from a single station are plotted in a graph, with temperature and salinity as the coordinates, and are then joined by a curve in order of increasing depth. In any given area the shape of the T-S curve has a fairly definite form, and hence errors in observation may sometimes be detected from such a graph. Many other uses of graphs of this nature are discussed in chapter V. Similar constructions can be used for the other serial data whenever

― 359 ―

any two kinds of observations may be plotted against each other. After interpolated values have been read from the vertical distribution curves, they too should be entered upon such diagrams, and, if the points corresponding to any interpolated depth do not fit the previously constructed curve, certain adjustments in the construction of the vertical distribution curve may be necessary.

The interpolated values of temperature and salinity at standard depths are obtained before the density, specific volume anomaly, and other calculations based upon these data are made (see chapter III). As a check upon the correctness of the data, it is often advisable to plot the vertical distribution of density (σ_ι) and the specific volume anomalies for each station as functions of depth.

Various types of distribution diagrams may be prepared from the serial observations, many of which are shown elsewhere in this volume, particularly in chapter XV.

The interpretation of serial oceanographic observations depends so much on the nature of the data, the distribution of observations in space and time, and the characteristic features of the region under investigation that it is impossible to attempt to formulate any “system” of analysis. Only from a knowledge of oceanographic problems and of the significance of the various kinds of data can the most fruitful method of attack be decided upon in any given investigation. The development of new theories and new viewpoints may invalidate earlier conclusions, but the observational data remain valid if they have met the required standards of accuracy.