We combined measurements from laser ranging to the moon, laser ranging to the LAGEOS satellite, and very long baseline interferometric observations of extragalactic radio sources, to produce a series of universal time (UT) and length-of-day (LOD) variations for 1981--1983. The UT1 series was obtained by convolving the ensemble of measurements from all three techniques, suitably weighted, with a Gaussian function whose full width at half-maximum (FWHM) amplitude was 5 days; values of LOD at 3-day invertals were obtained by differencing. These values of lOD have 80% of their one standard deviation (sigma) errors ranging from 0.03 to 0.06 ms. We compared our UT1-derived series of LOD values with tow series of LOD derived from variations in atmospheric angular momentum (AAM). One series was developed by Rosen and Salstein from wind analyses performed by the U.S. National Meteorological Center, the other by Barnes et al. from analyses performed at the European Centre for Medium Range Weather Forecasts (ECMWF).

Our attempt to estimate the one-sigma errors of both yielded bounds between 0.02 and 0.05 ms (80% limits). After removing a model for tidal effects from the UT1-derived LOD series, we find that the major harmonic components of the UT1-derived and the ECMWF AAM-derived series agree within the root-sum-square uncertainties, except for amplitude differences (UT1-AAM) of 0.05¿0.03 ms. -0.02¿&0.01 ms, and 0.02¿0.01 ms, for periods of 183. 122 days, and 27.5 days, respectively. The first two of these amplitude differences may be accounted for by some combination of high-altitude winds, ocean currents, atmospheric pressure and groundwater effects, and errors in the LOD series. Half of the difference in the 27.5-day component may be accounted for by errors in the tidal model, due to uncertainties in the effects of the oceans and the fluid core, as described by Yoder et al. The root-mean-weighted-square (temporal) differences between UT1-derived and AAM-derived LOD series, after high-pass filtering (Gaussian FWHM=45 days), are 0.066 ms if the ECMWF series is used and 0.07 ms if the National Meteorological Center series is used. Including the effect of pressure, where available in the AAM computations reduces by about 10% the short-period differences bettween the observed (UT1) and inferred (AAM) changes in LOD. For the characteristic 30- and 70-Day fluctuations there is no significant evidence (¿0.5 day) of a temporal lag between atmospheric excitation and the response of the solid earth.

Interpretation of the differences in AAM-derived and UT1-derived LOD is limited by our lack of knowledge of the uncertainties in the AAM observations. To within these limits it appears that atmospheric and tidal contributions are sufficient to account for the observed variations in length of day for periods of 1 year or less. Thus there is no evidence for significant contributions from any other geophysical effects to such variations in length of day.