After seeing the recent messages about various "Q-flux" methods, I'd like to chip in with a description of how it is handled in the GENESIS model at NCAR. We have had to think about these issues in our applications to Paleozoic geographies, with very different continental distributions from the present. We address both concerns raised by Bob Gallimore and John Kutzbach, namely (i) keeping the heat convergence (W/m2) constant and letting the net heat transport change (PW) for differing continental distributions, and ii) adjusting the heat convergence below sea ice *at each time step* to a low specified value. We first convert the present-day annual-mean net latitudinal heat transport (PW) to a heat convergence by taking its gradient and dividing by the area of ocean at each latitude bin. We "tweak" this raw curve by making it tail to particular values at high latitudes to represent the present-day under-sea-ice convergences (2 W m-2 in NH, 10 W m-2 in SH). This convergence function vs. latitude is applied at each ocean grid point of a new continental map, and a globally constant shift is added or subtracted to restore the global integral to zero. This forms a basic Q-flux input file that is read by the model at the start of a run. These steps correspond to Bob and John's suggestion (A) and Karl's reply (2), and of course the implicit assumption (whether true or not) is that the transport *per unit length of latitude circle* stays constant so the net transport is larger for wider basins. Then at each timestep of the model, the convergence at each grid point is modified by the current predicted fraction of sea ice. The convergence is (1-f)*[basic value] + f*[sea-ice value] where f is the sea-ice fractional cover at the grid point, "basic value" is the convergence from the basic input file, and "sea-ice value" is 2 W m-2 in the NH and 10 W m-2 in the SH. Then a global constant is added at each time step to ensure the global integral is zero. This gets at Bob and John's concern (B). If the sea ice can advance at all, there will be a positive feedback since the advancing ice edge reduces the heat convergence under itself, and does not have to endure open-ocean values (eg, 50 W m-2 at ~51 deg N mentioned by Bob and John). We have not yet done a Pleistocene glacial maximum simulation, so we don't know how far the scheme would allow sea ice to advance. In our paleo-simulations to date (Paleozoic, Mesozoic, Tertiary) the sea-ice distributions have at least seemed well behaved. Dave (pollard@ncar.ucar.edu, 303-497-1344).