Skip to content

DHI/modelskill

Repository files navigation

ModelSkill: Flexible Model skill evaluation.

Python version Python package PyPI version

ModelSkill is a python package for scoring MIKE models (other models can be evaluated as well).

Contribute with new ideas in the discussion, report an issue or browse the documentation. Access observational data (e.g. altimetry data) from the sister library WatObs.

Use cases

ModelSkill would like to be your companion during the different phases of a MIKE modelling workflow.

  • Model setup - exploratory phase
  • Model calibration
  • Model validation and reporting - communicate your final results

Installation

From pypi:

> pip install modelskill

Or the development version:

> pip install https://github.com/DHI/modelskill/archive/main.zip

Example notebooks

Workflow

  1. Define ModelResults
  2. Define Observations
  3. Match Observations and ModelResults
  4. Do plotting, statistics, reporting using the Comparer

Read more about the workflow in the getting started guide.

Example of use

Start by defining model results and observations:

>>> import modelskill as ms
>>> mr = ms.DfsuModelResult("HKZN_local_2017_DutchCoast.dfsu", name="HKZN_local", item=0)
>>> HKNA = ms.PointObservation("HKNA_Hm0.dfs0", item=0, x=4.2420, y=52.6887, name="HKNA")
>>> EPL = ms.PointObservation("eur_Hm0.dfs0", item=0, x=3.2760, y=51.9990, name="EPL")
>>> c2 = ms.TrackObservation("Alti_c2_Dutch.dfs0", item=3, name="c2")

Then, connect observations and model results, and extract data at observation points:

>>> cc = ms.match([HKNA, EPL, c2], mr)

With the comparer object, cc, all sorts of skill assessments and plots can be made:

>>> cc.skill().round(2)
               n  bias  rmse  urmse   mae    cc    si    r2
observation                                                
HKNA         385 -0.20  0.35   0.29  0.25  0.97  0.09  0.99
EPL           66 -0.08  0.22   0.20  0.18  0.97  0.07  0.99
c2           113 -0.00  0.35   0.35  0.29  0.97  0.12  0.99

Overview of observation locations

ms.plotting.spatial_overview([HKNA, EPL, c2], mr, figsize=(7,7))

map

Scatter plot

cc.plot.scatter()

scatter

Timeseries plot

Timeseries plots can either be static and report-friendly (matplotlib) or interactive with zoom functionality (plotly).

cc["HKNA"].plot.timeseries(width=1000, backend="plotly")

timeseries