Challenges in solid Earth data assimilation

 

Alik Ismail-Zadeh

 

Geophysical Institute, Karlsruhe University, Hertzstr. 16, Karlsruhe 76187, Germany; and

International Institute of Earthquake Prediction Theory and Mathematical Geophysics, Russian Academy of Sciences, Profsoyuznaya str. 84/32, Moscow 117997, Russia;

e-mail: Alik.Ismail-Zadeh@gpi.uka.de

 

Modern seismic tomography of the Earth's interior facilitates the inference of the complex trajectories of contemporary mantle movements. To understand the dynamics of the crust and mantle in the geological past a quantitative tool is required to solve the inverse problem of thermal convection. Data assimilation techniques can be used to constrain the initial conditions for the mantle velocity and temperature from their present observations. Data assimilation in this case is defined as the incorporation of present (observations) and past data (initial conditions) in an explicit dynamic model to provide time continuity and coupling among the physical fields. The basic principle of data assimilation is to consider the initial condition as a control variable and to optimise the initial condition in order to minimize the discrepancy between the observations and the solution of the model. A practical implementation of data assimilation techniques in modelling of geodynamic processes depends strongly on data accuracy (e.g., uncertainties in the present temperature distribution in the Earth's interior). Since there are no direct measurements of mantle temperatures, the temperatures can be estimated indirectly from either seismic wave (and their anomalies), geochemical analysis or through the extrapolation of surface heat flow observations. Many models of mantle temperature are based on the conversion of seismic tomography data into temperature; and seismic tomography data incorporate their own errors. If the present mantle temperature models are significantly biased, information on temperature can be improperly propagated to the geological past, and the resulting reconstruction will have nothing with true dynamics of the region in the past.