A worldwide investigation of continental erosion is carried out by the study of large drainage basins, on the basis of hydrological data, environmental factors, and basin relief distribution. Inside each basin, mean geochemical and mechanical denudation rates are defined. A multicorrelation analysis shows that the mechanical denudation rates Ds are uncorrelated with environmental factors and correlated with mean basin elevation H, while chemical denudation rates Dd are insensitive to relief but correlated with mean annual precipitation. Furthermore, two linear relationships between H and •6Hare detected: (1) Ds (m/10 •yr) = 419 x 1 (m) -0.245, with V (explained variance) = 95.1%; this law concerns basins related to orogenies younger than 250 Ma. The negative intercept is interpreted as a continental sedimentation rate of 245 m/m.y. An alternative model in which one invokes a critical elevation, separating erosion from sedimentation, is equally successful and leads to lower sedimentation rates (60-110 m/m.y.). For both models, one derives from the slope of the adjustments, erosion time constants Copyright 1988 by the American Geophysical Union Paper number 7T0908. 0278-7407/88/007T-0908510. O0 on the •der of 2.5 m.• (2) Ds (m/103 yr) = 61 x 10 VH (m), with = 86.5%

Summary The first part of this article gives new travel‐time data for the phase PKiKP at subcritical distances (Δ < 110°) obtained at the Warramunga array in Northern Australia. A compilation of the PKiKP subcritical travel times obtained from previous studies has also been carried out. In the second part, differential PKiKP‐PcP travel times at short distances (Δ<45°) are used to determine the inner core ellipticity and mean radius, subject to hypotheses on the equilibrium figure of the liquid core and the topography of the core‐mantle boundary (CMB). The coherence of the PKiKP‐PcP travel‐time residuals is increased when Morelli & Dziewonski's CMB topography is taken into account. According to various hypotheses, the values obtained for the ellipticity range from (1.9±2.2)×10−3 to (5.6±1.7)×10−3. This denotes a small inner core flattening corresponding to a decrease of the polar radius from (1.6±1.8) to (5.0±1.5) km with respect to a spherical inner core. The preferred solutions which involve the CMB topography, are consistent with the ellipticity 2.43times10−3 deduced from hydrostatic equilibrium of a rotating inner core. The last part of the paper concerns the amplitude variations of PKiKP. The amplitude ratio PKiKP/PcP at short distances (Δ<45°) is used to constrain the density jump δρ at the inner core boundary. δρ values of 1.35–1.66 g cm−3 are obtained for a quality factor in the liquid core ranging from 10 000 to infinity. These values, as well as a possible stratification in the inner core, are in favour of an inner core composition which is not pure iron.

Fully two‐dimensional analytic boundary layer solutions are used to model the thermomechanical structure of the oceanic upper mantle when a shallow horizontal return flow helps balance the lithospheric transport of mass from ridge to trench. The following are all incorporated in the solutions: horizontal and vertical advection of heat, vertical heat conduction, viscous dissipation, adiabatic heating and cooling, buoyancy, and the pressure‐ and temperature‐dependent nonlinear rheology of olivine. Depth profiles of horizontal and vertical velocities, temperature, and shear stress are calculated for several ages of the ocean floor. Such solutions are used to construct accurate isotherm and streamline patterns within the rigid lithosphere and high‐shear, return flow asthenosphere of the oceanic upper mantle boundary layer. Ocean floor topography is inferred from the thermal contraction of the cooling lithosphere and asthenosphere and from the adverse horizontal pressure gradient required by the dynamics to drive the shallow return flow. The flattening of the topography of the old ocean floor can be attributed to the retardation of boundary layer cooling by shear heating and/or to the adverse pressure gradient of a shallow return flow. The latter effect would be dominant if a shallow return flow did occur in the earth's upper mantle. Solutions which provide adequate fits to the ocean floor bathymetry data for the Pacific plate can be found if the activation volume for the creep of olivine is small, i.e., about 11 cm3/mol, and if the deep mantle temperature T is high, i.e., if T is about 1500°C at depths of 100–200 km (depending on age) for dry olivine or about 1400°C at similar depths for wet olivine. The upper mantle temperatures required for a shallow return flow to be compatible with bathymetry data imply partial melting in the lower lithosphere and upper asthenosphere from the ridge axis to ages of about 10 m.y. Shear stresses at the base of the rigid lithosphere generally range from about 1 to several tens of bars, and horizontal adverse pressure gradients at great depth vary from about 10 to 100 mbar/km for wide variations in the extent of return flow and deep mantle temperature. For a relatively cold deep mantle, shear stresses in the lower lithosphere and asthenosphere decrease almost linearly with depth, implying horizontal pressure gradients essentially constant with depth. For a hot deep mantle, shear stress versus depth profiles show curvature in the lower lithosphere and upper asthenosphere and a linearly decreasing pattern at greater depths. Under such circumstances, horizontal pressure gradients exhibit considerable depth variation including the possibility of a reversal in sign between the top and the bottom of the asthenosphere.

A large set of published great circle Rayleigh wave phase velocities in the period range 125-350s is used to compare three recent tectonic models (Okal, UvEque, Jordan). Prior to any regionalization, the symmetry property of the great circle integrals is used to obtain a lower limit of the signal/noise ratio in the data. It turns out that the signal is responsible for at least 30 per cent of the data variance in the period range 175-300 s.A standard regression method is applied for computing the 'pure path' velocities and the model efficiency is derived from a variance analysis. It is shown that, even at great depth, none of the three models explains more than 60 per cent of the energy due to the long-wavelength lateral heterogeneities (A 2 6500 km).The three models have nearly the same efficiency for explaining the shortperiod data ( T -125s). Between 200 and 300s, the higher performance of Okal's model indicates that it is important to separate the subduction zones from the other orogenic zones. By perturbing the lateral extension of the subduction zones, it comes out that they constitute on both sides of the subducting slabs wider anomalies than often assumed, suggesting large downgoing flows. On the contrary, the effect of surface features such as marginal seas are restricted to a close region in front of the trenches. Finally, the anomalous ellipticity values deduced directly from great circle data are partly explained by a coupling between tectonics and ellipticity.

In situ stress measurements in thick limestone layers of the Parisian Basin and in neighboring regions yield a remarkably homogeneous stress field. The flat jack method has been used. Comparison with over‐coring gave good agreement for the azimuth of the principal axes. In the Jurassic of Burgundy and Poitou as well as in a Carboniferous outcrop near the English Channel, maximum compression is found along the NNW‐SSE direction.