General parameters

General parameter definition include the selection of data assimilation simu­lation mode as well as definition of a few general, mode-specific parameters.

Mode selection

Three data assimilation modes are available:

·         Updating with weighting function: this mode performs a state updating of model parameters applying user-defined weighting functions to esti­mate the updates applied to model simulation result during simulation up to the Time of Forecast.
Additionally, an ‘Error forecast’ correction option may be applied after Time of Forecast, if selected.

·         Updating with Kalman filter: this mode updates the model with the ensemble Kalman filter based on Monte Carlo simulation techniques.

·         Uncertainty prediction: uses Monte Carlo simulation to propagate uncertainties in the boundary conditions, and assess uncertainties in the simulated model outputs.

Basic parameters

First updating time step. The time step number at which the model starts to be updated. Parameter is only available for updating using either the weight­ing function or the Kalman filter.

A value of 0 corresponds to the first time step of the simulation and e.g. a value of 100 corresponds to the updating only starting at simulation time step number 100 and forward. Starting the update later than time step 0 may ensure that the stochastic process is sufficiently evolved to give good esti­mates of the uncertainty before the first update, thereby getting rid of poten­tial instabilities occurring at the beginning of the simulation.

Ensemble size. Number of simultaneous runs that are to be carried out to evaluate the statistical properties needed for the uncertainty assessment out­put and also for determining the updating parameters applied in the Kalman filter. Parameter is hence only available for Kalman filter or uncertainty predic­tion simulation mode.

The quality of the statistical estimates are strongly dependent on the ensem­ble size: the larger the ensemble size the higher the confidence in the results. On the other hand, the ensemble size has a linear effect on the run time, that is, when the ensemble size is doubled the run-time is also doubled. Recom­mended values are 50-200.

If only reliable estimates of standard deviations are of importance, a smaller value may be chosen. When producing confidence intervals an estimate of the full uncertainty distribution is needed, thus an increase in the ensemble size is recommended for such cases.

Forecast

Time of forecast. Date and time at which the model switches to forecasting mode. Parameter is available when performing updating either with the Weighting function or the Kalman filter. Model state updating is only per­formed till the Time of Forecast.

Forecast mode. This option is only available for the Kalman filter option and gives a choice of applying deterministic or stochastic forecasting. At time of forecast, ‘Deterministic’ forecasting runs a single ensemble member based on the ensemble mean state at time of forecast. The ‘Stochastic’ type contin­ues evaluating all ensembles after time of forecast, applying boundary pertur­bations similarly to an uncertainty prediction run.