Constraining properties of mantle circulation models using disparate observations

Properties of the mantle are difficult to constrain and critical for controlling mantle evolution and dynamics. We attempt to constrain these properties by comparing the outputs from mantle circulation models (MCMs) to 9 disparate observations. Over 250 MCMs driven at the surface by 1 Ga of plate motion history are considered. A metric is developed to quantify the fit/misfit between each observation and MCM prediction. The observations include, global seismic tomography, SOLA seismic inference of the Pacific upper mantle, global surface wave phase velocity data set, gradients of seismic velocity in the deep mantle, dynamic topography, geoid, geomagnetic reversals, temperature difference between MORB and OIB source regions, and the difference in amount of recycled oceanic crust in MORB versus OIB source regions. The comparisons are done with (i) heatmaps of each metric for each MCM, (ii) correlation between the metrics and input parameters, (iii) analyses of sub-sets where only a single MCM parameter is changed, (iv) random forest analysis where the importance and partial dependence plot of MCM parameters are produced for each metric. From this analysis we find that parameters can be constrained, including for example the temperature at the core mantle boundary, the preferred equation of state, the preferred plate motion history model, the presence of a basal layer, the buoyancy number of the recycled basalt, viscosity profile. For example the MCMs prefer a cooler core-mantle boundary, a mantle reference frame-based plate motion history, a Murnaghan EoS and a basalt buoyancy number in the lower mantle of around 0.4-0.5. Methods, analyses and further results will be presented.