A three-dimensional picture of the horizontal velocity divergence is constructed at seven times, at twelve-hour intervals, with use of observed winds as a basis for divergence computations. No level of non-divergence is found; however, a surface of least divergence and a level of least divergence are located at each of the seven times, and their behavior is noted. A defense of the method includes comparison of vertical velocities computed by another method, and by the same method but with application of various types and degrees of smoothing. Principle conclusions are (1) that a surface of least divergence can be located but is not a physically significant surface, (2) that vertical velocities computed from the integrated low-level divergence correlate very well with the precipitation pattern, (3) that various kinds and degrees of smoothing do not appreciably change the order of magnitude of the divergence, and (4) that interesting diurnal variations in divergence intensity occur when values at 0300 GCT and 1500 GCT are compared.