Density, temperature, and composition of the North American lithosphere - New insights from a joint analysis of seismic, gravity, and mineral physics data: 2. Thermal and compositional model of the upper mantle

Temperature and compositional variations of the North American (NA) lithospheric mantle are estimated using a new inversion technique introduced in Part 1, which allows us to jointly interpret seismic tomography and gravity data, taking into account depletion of the lithospheric mantle beneath the cratonic regions. The technique is tested using two tomography models (NA07 and SL2013sv) and different lithospheric density models. The first density model (Model I) reproduces the typical compositionally stratified lithospheric mantle, which is consistent with xenolith samples from the central Slave craton, while the second one (Model II) is based on the direct inversion of the residual gravity and residual topography. The results obtained, both in terms of temperature and composition, are more strongly influenced by the input models derived from seismic tomography, rather than by the choice of lithospheric density Model I versus Model II. The final temperatures estimated in the Archean lithospheric root are up to 150 degrees C higher than in the initial thermal models obtained using a laterally and vertically uniform fertile compositional model and are in agreement with temperatures derived from xenolith data. Therefore, the effect of the compositional variations cannot be neglected when temperatures of the cratonic lithospheric mantle are estimated. Strong negative compositional density anomalies (<-0.03 g/cm(3)), corresponding to Mg # (100 x Mg/(Mg+Fe)) >92, characterize the lithospheric mantle of the northwestern part of the Superior craton and the central part of the Slave and Churchill craton, according to both tomographic models. The largest discrepancies between the results based on different tomography models are observed in the Proterozoic regions, such as the Trans Hudson Orogen (THO), Rocky Mountains, and Colorado Plateau, which appear weakly depleted (>-0.025 g/cm(3) corresponding to Mg # approximate to 91) when model NA07 is used, or locally characterized by high-density bodies when model SL2013sv is used. The former results are in agreement with those based on the interpretation of xenolith data. The high-density bodies might be interpreted as fragments of subducted slabs or of the advection of the lithospheric mantle induced from the eastward-directed flat slab subduction. The selection of a seismic tomography model plays a significant role when estimating lithospheric density, temperature, and compositional heterogeneity. The consideration of the results of more than one model gives a more complete picture of the possible compositional variations within the NA lithospheric mantle.