I wonder whether someone knows a clear definition/description of

1) Robust diagnostic models Forcing with data from initial state aimed at future predictions?

2) Prognostic models Tie model results back to observations using a relaxation factor?

'Googling' didn't help me much unfortunately.

  • $\begingroup$ Welcome to Earth Science! Are you talking about models in general, or within a specific domain (atmosphere, ocean, ...)? The precise usage of terminology can differ between domains. $\endgroup$ – gerrit May 14 '14 at 15:02
  • $\begingroup$ I will apply it to the ocean models, but I think it is the same for both, as described in the answer below. $\endgroup$ – user2003479 May 15 '14 at 10:04

As far as I know in the ocean modeling community the terms diagnostic and prognostic have quite different meanings from what you are stating.

A prognostic simulation is used to predict future state of the system (forecast) using the model equations. The model is initialized with the initial conditions and it is used to predict using the Navier-Stokes equations the future ocean state under specified boundary conditions. There is no constraint by observations because the model numerically follows the specified equations. In the case of models with data assimilation, the prognostic solution is statistically modified (e.g., Kalman filter, adjoint method) to optimally adjust to the observations, but the new solution can't be considered prognostic as there has been a departure from the primitive forward (Navier-Stokes) equations.

On the other hand, a diagnostic simulation does not provide a prediction of the fluid state. There is no time-evolution based on the model equations. An example of such simulation is the case in which the velocity field is calculated based on a fixed density (temperature and salinity) field. While these simulations can be useful for short-term ocean state estimation, they can't be used for ocean state evolution as the dynamics are restricted.

As far as I understand it, this is the standard in the ocean modeling community, but I don't know if it can be expanded to other fields.

| improve this answer | |
  • 2
    $\begingroup$ The atmospheric science community has the same view of these terms as far as I'm aware. Diagnostic to diagnose (tell us more about a state of the atmosphere based on what information we know) and prognostic to forecast (integrate a (sometimes sparsely) known state forward in time). $\endgroup$ – casey May 14 '14 at 20:56
  • 1
    $\begingroup$ @BHF I think you mean prognostic in the very end of your comment (I can fix that for you if so). Solving pressure diagnostically to allow prognosis is a good example of the intertwining of the two. $\endgroup$ – casey May 15 '14 at 0:20
  • $\begingroup$ Numerical models of the atmosphere usually have both kinds of equations. Prognostic equations are integrated (forwards) in time; diagnostic equations do not include a time derivative, but may need to be solved every time step to support the prognostic equations. $\endgroup$ – BHF May 15 '14 at 5:00

In terrestrial biogeochemistry models, or TBMs, diagnostic models assimilate remote sensing observations while prognostic models are driven by forcing data [source]. This is because diagnostic models are used to diagnose the representation of processes in models (i.e., validation) while prognostic models are used to predict future states (i.e., forecasting). Perhaps the most well-known diagnostic model is CASA [link]. This distinction no longer exists in most current models, as they may perform diagnosis or prognosis in different modes.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.