Homework

Exercise 1

  1. Download module4_model.zip into a new project folder and extract its content.
  2. Review the folder structure and files of the model. It should be as follows:
module4_model
├── data
│   ├── flow_meter_A_2020_09.dfs0
│   ├── flow_meter_B_2020_09.dfs0
│   └── rainfall_events_2020-09.dfs0
├── Dyrup_uncalibrated.mupp
├── Dyrup_uncalibrated.sqlite
├── LTS
│   └── rainfall_events_LTS_2020_09.MJL
└── results
    ├── rainfallBaseDefault_Network_HD.res1d
    └── rainfallBaseDefault_Surface_runoff.res1d
  1. Open Dyrup_uncalibrated.mupp in MIKE+ and familiarize yourself with the model.
    • Review the boundary conditions. Can you see where the rainfall is applied?
    • Find the system’s only node outlet: G60F360.
    • Locate the two flow meters (A and B). Which reach IDs are they associated with?
    • Review the simulation setup.
  2. Make a new LTS job list for the simulation. How does it compare with LTS/rainfall_events_LTS_2020_09.MJL?
  3. Run a simulation. Do you get the same results as those in the results folder?

Exercise 2

  1. Create a new Jupyter Notebook and save it in your project folder.
  2. Create a PointObservation object for “Flow Meter A” using data/flow_meter_A_2020_09.dfs0.
  3. Create a PointObservation object for “Flow Meter B” using data/flow_meter_B_2020_09.dfs0.
  4. For both obs_A and obs_B:
    • Print their objects and inspect the data.
    • Plot their time series.

Exercise 3

  1. Use the model network results from Exercise 1 (or the provided reference results).
  2. For “Flow Meter A” (model reach G60F380_G60F360_l1), extract its ‘Discharge’ quantity to a dfs0 file.
  3. Create a PointModelResult object from the dfs0 file created above.
  4. For “Flow Meter B” (model reach G62F070_G62F060_l1), extract its ‘Discharge’ quantity to a dfs0 file.
  5. Create a PointModelResult object from the dfs0 file created above.
  6. For both mod_A and mod_B:
    • Print the object summary.
    • Plot the model result time series.

Exercise 4

  1. Match obs_A with mod_A using ms.match(). Store in comparer_A.
  2. Match obs_B with mod_B. Store in comparer_B.
  3. For comparer_A:
    • Generate time series plot: comparer_A.plot.timeseries().
    • Generate scatter plot: comparer_A.plot.scatter().
    • Calculate and display skill table: comparer_A.skill().
  4. Repeat step 3 for comparer_B.
  5. Examine the skill tables. Qualitatively, which metrics suggest better or worse performance for each location?

Exercise 5

  1. Create a ComparerCollection named cc from comparer_A and comparer_B.
  2. Display cc. Check cc.obs_names and cc.mod_names.
  3. Make a temporal coverage plot: cc.plot.temporal_coverage().
  4. Make a scatter plot: cc.plot.scatter().
  5. Make a residual histogram: cc.plot.residual_hist().
  6. Calculate and display the skill table for the collection: cc.skill().
    • Modify to show only ‘rmse’, ‘bias’, ‘nse’, and ‘kge’ metrics.
  7. Calculate and display the mean skill table: cc.mean_skill().
    • Modify to show only ‘rmse’, ‘bias’, ‘nse’, and ‘kge’.
  8. Using cc.score(), retrieve and print the mean ‘kge’ score for the model.

Exercise 6

  1. Repeat Exercise 4 and 5, but using the max_model_gap parameter of ms.match(). Use a value of 600 seconds.
  2. How do the number of observation points in each Comparer change?
  3. How does the temporal coverage plot change?
  4. How does this approach impact the skill assessment?
  5. Review ModelSkill’s documentation on ‘nse’. How do both flow meters, and the overall model perform on this metric?

Download the exercise solutions: