Selected measurements performed with Multi-Filter Rotating Shadow-band Radiometers under cloudless conditions at an urban tropical site, Sao Paulo, Brazil (23.65 o S, 46.62 o W, 800 m above the mean sea level) are analyzed. Water vapor content (U) values are estimated from the analysis of measurements performed near 940 nm with two bandwidths (10 and 35 nm). Narrow- and wide-band measurements performed near 940 nm are analyzed separately with the help of line-by-line radiative model computed transmittances, allowing independent U N and U W estimates. The ratio of these measurements is also analyzed, allowing a third water vapor content estimate (U N W ). Selected approaches for computing water vapor absorption are compared, including two line parameter databases and two line-by-line models. After pooling, all results obtained from measurements performed during 16 cloudless mornings and representing the water vapor absorption through HITRAN-2000 line parameters and CKD-2.4 continuum model, estimates U N , U W , and U N W were on average 1.8, 0.5, and 4.0 kg m - 2 lower than respective values provided by balloon-borne soundings, respectively; and U N estimates were 3% to 14% smaller than respective U W values. Assuming that the independent values provided by soundings are accurate, the ad hoc reduction of continuum absorption coefficients to 50% of values provided by CKD-2.4 model improved the reliability of narrow-band water vapor content estimates in this study. Lastly, the influence of aerosol extinction on water vapor content was minimized from the analysis of measurements performed near 870 and 1035 nm. The corresponding turbidity exponent α assumed values between 0.7 and 1.1, which is somewhat weaker than those usually reported for urban conditions.