I just noticed something strange resulting from the use of the Bulk Flux formulation.
Basically if I do a realistic ROMS model run (with tidal, river, atmospheric, open boundary forcing, etc.) and then look at the water temperatures recorded at various stations, there are "hot" spikes in them at and close to the ocean surface. These spikes occur from time to time and appear to be limited to the first 3-5 grid points in the vertical (measured down from the surface). Especially at these spikes, the temperature difference between that at the top level and one level below it is appreciable - ie. about 3-6 degrees C. The effects of the spikes last 2-8 hours in time depending on the application.
We have examined this phenomena in multiple applications and it only occurs when we use the Bulk Flux formulation either with the meteorological inputs as horizontal fields OR even as 1-grid point time-series (ie. it occurs in both situations). If however, we bypass the Bulk Fluxes by specifying some wind stresses and net heat flux fields directly, the spikes disappear and we end up with smooth vertical temperature profiles.
Has anybody else noticed this? The easiest way to check this would be to open a stations.nc file from a ROMS run, read the whole matrix of temperature for a particular station (which will be Nz (number of vertical levels) x Nt (time dimension of the stations.nc file) in size) and then plot the whole matrix (say using pcolor(x,y,T) in Matlab) at once as in a Hovmuller plot. It should then be possible to see any spikes as thin, long slivers of water much warmer than their surroundings and extending 3-5 grid points in the vertical for 2-8 hours.
Alternatively, if at a particular station, the temperature time-histories corresponding to levels Nz and Nz-1 are plotted together, appreciable differences between them (even though they may be a few mm apart in reality) can be seen to occur as spikes from time to time.
These spike also manifest themselves as highly localized structures in the horizontal plots of surface temperature (ie. temperaure at level Nz) but only if the fields have been saved in to history files sufficiently frequently (eg. every 1/2 hour or every 1-hour).
Potential problem/bug in Bulk Fluxes
Re: Potential problem/bug in Bulk Fluxes
We found this heating problem occurs due to the diffusion coefficient associated with the temperature variable (ie. AKt) becoming very small (locally and from time to time - irregularly) near the ocean surface; then, ROMS applies all of the net heat flux to the top sigma level and it heats up dramatically. This problem could be alleviated by using a larger background value for the AKt variable in toms.in but this affects the vertical temperature stratification in the water column. Hence, a possible solution would be to use the SOLAR_SOURCE CPP option in ROMS to allow deeper peneration of the heat in to the water column. However, we then have to specify physically (and ecologically) reasonable Jerlov water types and this itself has two complications : (i) it is not easy even from the literature to figure out the Jerlov water type for a particular Bay/Estuary under consideration and (ii) in ROMS the Jerlov water type is a scalar and it is not allowed to vary in space as is the expected case in reality (for a Bay/Eastuary).