This page describes a simple ModelSkill workflow when model results and observations are already matched. See workflow page for a more elaborate workflow.
uv is an extremely fast Python package and project manager that is 10-100x faster than pip, and also makes it easy to install Python and manage projects. With uv, creating a virtual environment is as easy as uv venv.
To install ModelSkill, run this command in a terminal:
pip install modelskilluv pip install modelskillThe simplest use-case for skill assessment is when you have a dataset of matched model results and observations in tabular format.
import pandas as pd
import modelskill as ms
df = pd.read_csv("../data/Vistula/sim1/6158100.csv", parse_dates=True, index_col="Date")
df.head()| Qobs | Qsim | Prec | |
|---|---|---|---|
| Date | |||
| 2000-01-02 | 5.2 | 4.641 | 0.11 |
| 2000-01-03 | 5.2 | 4.666 | 0.05 |
| 2000-01-04 | 5.2 | 4.556 | 0.72 |
| 2000-01-05 | 5.2 | 4.470 | 0.30 |
| 2000-01-06 | 5.2 | 4.391 | 1.38 |
cmp = ms.from_matched(df, obs_item="Qobs", mod_items="Qsim", quantity=ms.Quantity("Discharge", "m3/s"))
cmp<Comparer>
Quantity: Discharge [m3/s]
Observation: Qobs, n_points=3653
Model(s):
0: QsimA time series plot is a common way to visualize the comparison.
cmp.plot.timeseries()
Another more quantitative way to analyze the compared data is to use a scatter plot, which optionally includes a skill table (Definition of the metrics).
cmp.plot.scatter(skill_table=True)
The skill table can also be produced in tabular format, including specifing other metrics.
cmp.skill(metrics=["bias", "mae", "rmse", "kge", "si"])| n | bias | mae | rmse | kge | si | |
|---|---|---|---|---|---|---|
| observation | ||||||
| Qobs | 3653 | -5.303471 | 7.473344 | 14.903921 | 0.360125 | 0.866969 |