The paper examines the variability of vertical humidity profiles over the Indian oceanic region using a set of 1200 radiosonde observations spanning 10 years (1982--1991). The examination is based upon the method of empirical orthogonal function (EOF) analysis. The first EOF explains 61% of the total variance and the first three EOFs together account for 85% of the total variability. The first principal component is almost perfectly correlated with the total precipitable water (TPW) and the second one is well correlated with the ratio of boundary layer moisture and TPW. This fact and an inequality derived from the analysis of the variances of individual terms of the EOF expansion of specific humidity are utilized to establish an algorithm for retrieving humidity profile from satellite microwave measurement of TPW over the region of study. Power of the retrieval technique is demonstrated using 127 independent radiosonde measurements and by plotting the profiles of rms error and bias. The method is found to be distinctly superior compared to a power-law retrieval. A few examples of profile retrieval from satellite measurements of TPW have been checked against colocated radiosonde measurements. Some examples of retrieval show that the method is uniquely able to capture the humidity variability in the boundary layer, particularly the high moisture loading in the monsoon season. ¿ American Geophysical Union 1995