Construction of the Model object¶
EMC^2 uses the emc2.core.Model object to know what fields to load from a given
file in order to obtain the required information. In order to define your
emc2.core.Model object, it is highly recommended to use class inheritance. For example:
$ class ModelE(Model):
$ def __init__(self, file_path):
$ """
$ This loads a ModelE simulation with all of the necessary
$ parameters for EMC^2 to run.
$
$ Parameters
$ ----------
$ file_path: str
$ Path to a ModelE simulation.
$ """
In particular, EMC^2 will require information about the mixing ratio and the number concentration of four different species: cloud liquid (cl), cloud ice (cl), precipitating liquid (pl), and precipitating ice (pi). These precipitation classes are commonly used in many models in order to represent the cloud microphysical properties. EMC^2 derives the radar and lidar parameters for all 4 of these species. First, in order to specify the fields that EMC^2 needs to look for, certain entries whose keys correspond to the names of these four precipitation species must be specified:
$ # Names of mixing ratios of species
$ self.q_names = {'cl': 'qcl', 'ci': 'qci', 'pl': 'qpl', 'pi': 'qpi'}
$ # Number concentration of each species
$ self.N_field = {'cl': 'ncl', 'ci': 'nci', 'pl': 'npl', 'pi': 'npi'}
$ # Convective fraction
$ self.conv_frac_names = {'cl': 'cldmccl', 'ci': 'cldmcci', 'pl': 'cldmcpl', 'pi': 'cldmcpi'}
$ # Stratiform fraction
$ self.strat_frac_names = {'cl': 'cldsscl', 'ci': 'cldssci', 'pl': 'cldsspl', 'pi': 'cldsspi'}
$ # Effective radius
$ self.re_fields = {'cl': 're_mccl', 'ci': 're_mcci', 'pi': 're_mcpi', 'pl': 're_mcpl'}
$ # Convective mixing ratio
$ self.q_names_convective = {'cl': 'QCLmc', 'ci': 'QCImc', 'pl': 'QPLmc', 'pi': 'QPImc'}
$ # Stratiform mixing ratio
$ self.q_names_stratiform = {'cl': 'qcl', 'ci': 'qci', 'pl': 'qpl', 'pi': 'qpi'}
In addition, other fields must also be specified:
$ # Water vapor mixing ratio
$ self.q_field = "q"
$ # Pressure
$ self.p_field = "p_3d"
$ # Height
$ self.z_field = "z"
$ # Temperature
$ self.T_field = "t"
$ # Name of height dimension
$ self.height_dim = "plm"
$ # Name of time dimension
$ self.time_dim = "time"
What if your model does not produce output for all 4 species, as is common in many GCMs? Simply place in zero arrays that are the same shape as your model fields!
Finally, we need to load in the model dataset. In order to do this, the model must be loaded into a format that is compatible with xarray. If you are loading a netCDF file, this is quite easy to do as xarray has native support for netCDF files. For example, ModelE’s files are in netCDF format, so we can simply do:
$ self.ds = xr.open_dataset(file_path)
One thing to be aware of is that you must ensure that all of your fields that you load are 64-bit double precision floating point numbers in order to conform with the assumed data types in ECM^2. Otherwise, underflow errors are likely for many of the calculations. To ensure that this is the case, one can simply loop over the variables in the file like this:
$ super().prepare_variables()
Finally, there are many assumptions that go into the calculation of the forward modelled radar moments. For example, there are various fall-speed relationships of the form \(V = aD^b\) for different types of particles. Therefore, if you are looking at a case where you think a specific kind of ice species may be dominant, it is important to adjust these \(a\) and \(b\) constants. There are numerous papers on this subject that are included in the references below. It is best to match these coefficients with what is used in your model for the best comparison. In order to adjust the constants that are used in the various routines in EMC^2, you would have to fill in these dictionaries for each hydrometeor species:
$ # Bulk density
$ self.Rho_hyd = {'cl': 1000. * ureg.kg / (ureg.m**3),
$ 'ci': 500. * ureg.kg / (ureg.m**3),
$ 'pl': 1000. * ureg.kg / (ureg.m**3),
$ 'pi': 250. * ureg.kg / (ureg.m**3)}
$ # Lidar ratio
$ self.lidar_ratio = {'cl': 18. * ureg.dimensionless,
$ 'ci': 24. * ureg.dimensionless,
$ 'pl': 5.5 * ureg.dimensionless,
$ 'pi': 24.0 * ureg.dimensionless}
$ # Lidar LDR per hydrometeor mass content
$ self.LDR_per_hyd = {'cl': 0.03 * 1 / (ureg.kg / (ureg.m**3)),
$ 'ci': 0.35 * 1 / (ureg.kg / (ureg.m**3)),
$ 'pl': 0.1 * 1 / (ureg.kg / (ureg.m**3)),
$ 'pi': 0.40 * 1 / (ureg.kg / (ureg.m**3))}
$ # a, b in V = aD^b
$ self.vel_param_a = {'cl': 3e-7, 'ci': 700., 'pl': 841.997, 'pi': 11.72}
$ self.vel_param_b = {'cl': 2. * ureg.dimensionless,
$ 'ci': 1. * ureg.dimensionless,
$ 'pl': 0.8 * ureg.dimensionless,
$ 'pi': 0.41 * ureg.dimensionless}
$ super()._add_vel_units()
References¶
Locatelli, J. D., and Hobbs, P. V. (1974), Fall speeds and masses of solid precipitation particles, J. Geophys. Res., 79( 15), 2185– 2197, doi:10.1029/JC079i015p02185.
Brown, P.R. and P.N. Francis, 1995: Improved Measurements of the Ice Water Content in Cirrus Using a Total-Water Probe. J. Atmos. Oceanic Technol., 12, 410–414, https://doi.org/10.1175/1520-0426(1995)012<0410:IMOTIW>2.0.CO;2
Heymsfield, A.J., G. van Zadelhoff, D.P. Donovan, F. Fabry, R.J. Hogan, and A.J. Illingworth, 2007: Refinements to Ice Particle Mass Dimensional and Terminal Velocity Relationships for Ice Clouds. Part II: Evaluation and Parameterizations of Ensemble Ice Particle Sedimentation Velocities. J. Atmos. Sci., 64, 1068–1088, https://doi.org/10.1175/JAS3900.1