These record types use the START and END fields in the same manner, so they are dealt with together here. The only exception is that the fields in the DESCRIP3 records for the same Level 3 dimension must not overlap, while there can be any overlap in the PROCSPEC, PROCDUP, AUXSPEC, and PAKSPEC record types.
We will consider again the wind vector example from Example 2 in Section 3.2.2. As described, the START and END fields of the DESCRIP3 record extend over all dimensions of the data object. This is the most straightforward use of these fields.
Now suppose that we process the Northern Hemisphere differently from the Southern Hemisphere, and furthermore, we change the southern hemisphere processing from day 20 onwards. These three types of processing would be compactly represented as locally defined integer codes, which we denote as ``codenh,'' ``codesh1,'' and ``codesh2.'' We would need, then, three PROCSPEC records to define these processing codes. Note that the data at the equator, because they may be considered to belong to both hemispheres, will be associated with two processing codes: one for the northern hemisphere and one for the southern. The elements of the START and END arrays will correspond to the wind component, longitude, latitude, pressure, and day-of-year dimensions. The relevant record fields will contain the following (indices start at 1):