Font size

Construct the Simulation Model

Subtask Description:
Construct the simulation model
Conduct a hindcast simulation with policy or other change.

Action points of the implementation:
The model should now be running in a logical manner; however it now needs to be validated using hindcasting. This involves undertaking a simulation run on the model using historical driving data as a scenario and comparing the outputs from the model to what occurred in the real world. If these simulations do not match with the observed data it may be necessary to re-examine the model to discover where errors lie.

Result: A calibrated and validated simulation model of the system.

Area:
Himmerfjärden , Sweden

Policy Issue:
Eutrophication status and reduction.

Human Activities:
Urban sewage discharge, agriculture and industrial activity, tourism.

General Information:
Nutrient loading has caused increased turbidity, loss of biodiversity, including submerged aquatic vegetation, deep water oxygen def iciency, phytoplankton blooms and biodiversity loss. The main stakeholder concerns are connected with tourism, recreational activities and nature enjoyment, and the sustainable implementation of WFD that poses economic challenges for several activities in the area.

Example of Implementation:
Ecological hindcast.
The objective of the ecological model is to simulate the response in terms of water transparency (Secchi depth) of the coastal ecosystem of the Himmerfjärden area to changes in nutrient loads in terms of nitrogen concentrations. The study area consists of three sub-basins (HA, NA and HI), differing in size, sill depths, and dominant nutrient sources.

A four year period (1997-2000) was used for hindcasting and calibration of the ecological model. This included a year with relatively high load from the STP (1997) and a major fresh-water inflow event at the end of the period (2000). In the first step of the calibration process, the estuarine water flow was calculated for the inner model basin Hallsfjärden by calibrating with salinity data. A large variety of parameter settings was tested to achieve the best fit according to minimum sums of quadratic difference between model and data, for average basin salinity and for the individual depth layers. After the inner basin behaved satisfactorily, the process was repeated for the outer basins, which depend on modelled surface water flow from the inner basins. However, real salinity data was used for the inputs to each basin during calibration. Except during the last year, the calculated Secchi depth showed fairly good agreement for model basin HI (see table below). For two of the years, tot-N was overestimated by up to 13% (571 µg/l), and Secchi depth underestimated for the model basin NA. In model basin HA, total N was overestimated by up to 14% and the empirical relationship also seemed to give too low a Secchi depth.

Table 1. Modelled and observed total nitrogen and Secchi depth of the hindcast years 1997-2000. Model run 1 was done for each basin with inputs from adjacent basins according to measured concentrations. In Model run 2 input concentrations are generated by the model. Secchi depth is calculated from modelled tot-N, and, for comparison, for measured tot-N. SD= standard deviation.

For the ecological model, the data used in the hindcast runs are given in the table of inputs and marked as forcing function (ff) or calibration data (c). An additional table presents values of various constants or parameters, the latter modified during the calibration process.

Simulated (blue) and interpolated measurements (pink) of total nitrogen concentration in the hindcast run over four years (1997-2000) for HI basin

Modeled (lines) and measured (crosses) salinity for the full volume in HI basin

Comments:
It is highly recommended that the forcing data and data used for model comparison have been measured during a major policy change in order to test that the model responds correctly to this change.

Contact: Jakob Walve jakob.walve@ecology.su.se