| ... | @@ -23,21 +23,15 @@ With the 3D model a tracer simulation with an inert tracer ($`r`$ = 0) was done. |
... | @@ -23,21 +23,15 @@ With the 3D model a tracer simulation with an inert tracer ($`r`$ = 0) was done. |
|
|
|
|
|
|
|

|
|

|
|
|
|
|
|
|
|
# Artificial data
|
|
|
|
|
|
|
|
|
|
The results of the tracer simulation with the 3D hydrodynamic model were used to generate artificial data. For each DEMO model compartment, the depth-weighted average fraction North Sea water was calculated for each time step (hour).
|
|
|
|
|
|
|
|
|
|
[Top](Hydrodynamics)
|
|
|
|
|
|
|
|
|
|
# Calibration exchange rates
|
|
# Calibration exchange rates
|
|
|
|
|
|
|
|
The exchange rates between the neighboring compartments are estimated by fitting the output of the DEMO tracer model with an inert tracer ($`r`$ = 0) to the artificial data generated by the 3D GetM/GotM model. Initial conditions for each compartment were calculated from the artificial data as the average fraction of North Sea water in each compartment during the first four tidal cycles. At the North Sea boundary the tracer concentration was 1. The models are fitted using the modFit function in R (Soetaert and Petzold, 2010)[<sup>2</sup>](#fn2-7947). The 0D model and the models with the 4 and 28 compartments are fitted separately. The parameters were estimated using the Levenberg-Marquardt algorithm where the objective was to minimize the sum of squared residuals (SSR) between model and "observations".
|
|
The exchange rates between the neighboring compartments are estimated by fitting the output of the DEMO tracer model with an inert tracer ($`r`$ = 0) to "artificial" data generated by the 3D GetM/GotM model. These artificial data were the result of the tracer simulation with the 3D hydrodynamic model from Jiang et al. (2019) [<sup>1</sup>](#fn1-7947). For each DEMO model compartment, the depth-weighted average fraction North Sea water was calculated for each time step (hour). Initial conditions for each compartment were calculated from the artificial data as the average fraction of North Sea water in each compartment during the first four tidal cycles. At the North Sea boundary the tracer concentration was 1. The models are fitted using the modFit function in R (Soetaert and Petzold, 2010)[<sup>2</sup>](#fn2-7947). The 0D model and the models with the 4 and 28 compartments are fitted separately. The parameters were estimated using the Levenberg-Marquardt algorithm where the objective was to minimize the sum of squared residuals (SSR) between model and "observations".
|
|
|
|
|
|
|
|
[Top](Hydrodynamics)
|
|
[Top](Hydrodynamics)
|
|
|
|
|
|
|
|
## 0D model
|
|
## 0D model
|
|
|
|
|
|
|
|
For the 0D model only 1 parameter (exchange rate between the North Sea and the Oosterschelde) had to be fitted. In the figure below, the fitted model is compared with the artificial data. The fitted model is used to calculate the average residence time ($`RT`$, d) and turnover time ($`TT`$, d) of the water within the compartment.
|
|
For the 0D model only 1 parameter (exchange rate between the North Sea and the Oosterschelde) had to be fitted. In the figure below, the fitted model is compared with the artificial data. The fitted model is used to calculate the average residence time ($`RT`$`, d) and turnover time (`$`TT`$, d) of the water within the compartment.
|
|
|
|
|
|
|
|

|
|

|
|
|
|
|
|
| ... | | ... | |