The ultimate aim in the simulation of weld pools is to depict the final state of the solidified metal. Solidification history determines the metallurgical state, and can be partially derived from macroscale simulation. This requires realistic initial conditions—weld pool temperatures and flows at power off time—in order to produce accurate solidification histories for the pool points. Numerical simulation of the heat and mass tranfer is carried out here with a finite volume finite difference scheme on a flat surfaced pool (appropriate for currents of present interest). Suface tension, buoyancy and Lorentz forces are included in the flow model. When the arc heating and Lorentz forces are shut off, continuation of the calculation allows examination of the solidification thermal environment. Welds on stainless steel are simulated with fluid flow driven by an “effective” surface tension coefficient of ∂γ/∂T = −0.01 dyne/cm K. Solidification events consist of an initial phase which smooths the fusion zone boundary and removes the superheat from the pool, followed by a quasisteady stage, and end with a terminal boundary layer with time dependence similar to the singular spherical Stefan solution. Introduction of “numerical macrographs” allows convenient comparison of simulated conditions with actual weld macrographs.