SUMMARYWe present a new global thermochemical model of the lithosphere and underlying upper mantle constrained by state of the art seismic waveform inversion, satellite gravity (geoid and gravity anomalies and gradiometric measurements from ESA's GOCE mission), surface elevation and heat flow data: WINTERC-G. The model is based upon an integrated geophysical–petrological approach where seismic velocities and density in the mantle are computed within a thermodynamically self-consistent framework, allowing for a direct parametrization in terms of the temperature and composition variables. The complementary sensitivities of the data sets allow us to constrain the geometry of the lithosphere–asthenosphere boundary, to separate thermal and compositional anomalies in the mantle, and to obtain a proxy for dynamic surface topography. At long spatial wavelengths, our model is generally consistent with previous seismic (or seismically derived) global models and earlier integrated studies incorporating surface wave data at lower lateral resolution. At finer scales, the temperature, composition and density distributions in WINTERC-G offer a new state of the art image at a high resolution globally (225 km average interknot spacing). Our model shows that the deepest lithosphere–asthenosphere boundary is associated with cratons and, also, some tectonically active areas (Andes, Persian Gulf). Among cratons we identify considerable differences in temperature and composition. The North American and Siberian Cratons are thick (>260 km) and compositionally refractory, whereas the Sino-Korean, Aldan and Tanzanian Cratons have a thinner, fertile lithosphere, similar to younger continental lithosphere elsewhere. WINTERC-G shows progressive thickening of oceanic lithosphere with age, but with significant regional differences: the lithospheric mantle beneath the Atlantic and Indian Oceans is, on average, colder, more fertile and denser than that beneath the Pacific Ocean. Our results suggest that the composition, temperature and density of the oceanic mantle lithosphere are related to the spreading rate for the rates up to 50–60 mm yr–1: the lower spreading rate, the higher the mantle fertility and density, and the lower the temperature. At greater spreading rates, the relationship disappears. The 1-D radial average of WINTERC-G displays a mantle geothermal gradient of 0.55–0.6 K km–1 and a potential temperature of 1300–1320 °C for depths >200 km. At the top of the mantle transition zone the amplitude of the maximum lateral temperature variations (cratons versus hotspots) is about 120 K. The isostatic residual topography values, a proxy for dynamic topography, are large (>1 km) mostly in active subduction settings. The residual isostatic bathymetry from WINTERC-G is remarkably similar to the pattern independently determined based on oceanic crustal data compilations. The amplitude of the continental residual topography is relatively large and positive (>600 m) in the East European Craton, Greenland, and the Andes and Himalayas. By contrast, central Asia, most of Antarctica, southern South America and, to a lesser extent, central Africa are characterized by negative residual topography values (>–400 m). Our results show that a substantial part of the topography signal previously identified as residual (or dynamic) is accounted for, isostatically, by lithospheric density variations.