eWaterCycle-DA

Code to run Data Assimilation with hydrological models on the eWaterCycle platform.

Installation

Install this package alongside your eWaterCycle installation

pip install ewatercycle-DA

Then HBV becomes available as one of the eWaterCycle models

from ewatercycle_DA import DA

docs

Documentation can be found here

Changelog

Changelog can be found in CHANGELOG.md on GitHub.

Quick Usage overiew

(maybe migrate this to docs?)

Can be used with or without assimilating, this will run 10 versions of the same model. By varying the setup_kwargs you can vary the model run itself.

Without assimilating

HBVForcing = ...

ensemble = DA.Ensemble(N=10)
ensemble.setup()

ensemble.initialize(model_name="HBV",
                   forcing=HBVForcing
                   setup_kwargs={'parameters':'7.6,0.5,460,3.8,0.19,1.3,0.082,0.0061',
                                 'initial_storage':'0,100,0,5'}
                    )

ref_model = ensemble.ensemble_list[0].model

lst_Q = []
while ref_model.time < ref_model.end_time:
    ensemble.update(assimilate=False)
    lst_Q.append(ensemble.get_value("Q"))

For running HBV see seperate docs

With assimilating

#... same as above just add two more definitions
def H(Z):
    """returns discharge which is the last value on the state vector for HBV"""
    return Z[-1] 


ensemble.initialize_da_method(ensemble_method_name = "PF", 
                              hyper_parameters = {
                                               'like_sigma_weights' : 0.05,
                                               'like_sigma_state_vector' : 0.01,
                                                 },
                              state_vector_variables = "all", 
                              # the next three are keyword arguments but are needed:
                              observation_path = ds_obs_dir,
                              observed_variable_name = "Q",
                              measurement_operator = H, 
                           
                            )
lst_Q = []
while ref_model.time < ref_model.end_time:
    ensemble.update(assimilate=True)
    lst_Q.append(ensemble.get_value("Q"))