Twinning water quality modelling in Latvia

1
Kickoff meeting Twinning on development of
modelling capacity to support water quality
monitoring in Latvia Modelling rationales and
approaches
Photo Lake Övre hammardammen, Fredrik Ejhed
2
Monitoring vs. Modelling

• Monitoring at strategic sites
• Regionalised data WFD typology
• Modelling gives overview
• Modelling gives physical processes
• Models needed to plan measures
• No model without monitoring data

Monitoring sites in Sweden used for WFD and
HELCOM PLC-4
3
Modelling - for what purpose ?

• The purpose needs to be defined before
  application
• Source apportionment
• Programmes of measures
• Modelling investigation of failure to meet good
  quality

4
Resolution temporal and spatial

• Daily drive data but annual results
• Dynamic results needed in lakes and coastal
  processes
• Spatial resolution time and cost proportional
• WFD demand both high spatial resolution and
  overview
• Solution provide overview model results and
  high resolution model results for sensitive water
  bodies
• Monitoring data dependence

5
Model concepts
Rewritten from EUROHARP documentation

• Combination of models above

6
Model concepts Limitations and advantages

• Advantages
• Process descriptions
• Scenarios possibilities
• Low data requirements
• Simple models

• Limitations
• Expert user
• High data requirements
• Timeconsuming
• Few scenario possibilities
• Valid only for model range

7
Calibration of models

• The conceptual and processbased model is
  calibrated using monitoring data
• Empirical models e.g. regression analysis are
  only valid within the data range used

Fig. Total nitrogen concentration before and
after calibration of soil retention, Skivarpsån
Sweden, model for HELCOM PLC4 and WFD
8
Validation and assessment analysis

• Validation of model in time split or spatial
  split of data
• Validation of target expectations
• Validation of e.g. flow, concentration, load
• Using simple statistical tools
• In EUROHARP annual timestep results validated
  using mean deviation, mean absolute deviation and
  standard deviation

9
Results weakest link rule

• A good model need good driving data
• Ex. problems with faulty point source coordinates
  in an inlet watercourse to lake Vättern in
  Sweden, model for HELCOM PLC-4 and WFD. Total
  nitrogen concentrations vs. time.

10
Model choice

• Well tested models for the region
• Experience of the models
• Keep an open flexible structure in model systems
• developments in model system may be hindered by a
  single choice of model.
• continuous developments necessary
• recalculations of old results for assessment of
  developments towards the environmental targets