Construct an integrator

In this example, we showcase how to construct a pyshqg.core.integration.RungeKuttaModelIntegrator instance.

[1]:

from pyshqg.backend.numpy_backend import NumpyBackend as Backend
from pyshqg.preprocessing.reference_data import load_test_data
from pyshqg.core.integration import RungeKuttaModelIntegrator
from pyshqg.core.constructors import construct_model, construct_integrator

Construct the model

The first step is to construct the model. In the present example, we use the test model at spectral resolution T21 and grid resolution T31, implemented with the numpy-based backend.

See the notebook ‘Construct a model’ for more details about how to construct the model.c

[2]:

ds_test = load_test_data(
    internal_truncature=21,
    grid_truncature=31,
)

backend = Backend('float64')
model = construct_model(backend, ds_test.config)

Manual construction

To construct the integrator, we need:

model: the model to integrate;
dt: the integration time step;
method: the name of the integration method.

Currently, four integration methods are implemented:

method='ee': explicit Euler;
method='abm': second-order Adams-Bashforth;
method='rk2': second-order Runge–Kutta;
method='rk4': fourth-order Runge–Kutta.

[3]:

integrator = RungeKuttaModelIntegrator(
    model=model,
    dt=3600,
    method='rk4',
)

Construction using a configuration dictionary

Following what has been done for the model, it is possible to use the pyshqg.core.constructors.construct_integrator() function to construct the integrator using a configuration object stored as a dictionary.

For example, the test data contains a dictionary that will produce the integrators used in the unit tests.

[4]:

integrator = construct_integrator(ds_test.config['rk4_integration'], model)