# Licensed under a 3-clause BSD style license - see LICENSE.rst
"""Cube models (axes: lon, lat, energy)."""
import copy
import numpy as np
import astropy.units as u
from gammapy.maps import Map, MapAxis, RegionGeom, WcsGeom
from gammapy.modeling import Covariance, Parameters
from gammapy.modeling.parameter import _get_parameters_str
from gammapy.utils.fits import LazyFitsData
from gammapy.utils.scripts import make_name, make_path
from .core import Model, Models
from .spatial import ConstantSpatialModel, SpatialModel
from .spectral import PowerLawNormSpectralModel, SpectralModel, TemplateSpectralModel
from .temporal import TemporalModel
__all__ = [
"SkyModel",
"FoVBackgroundModel",
"BackgroundModel",
"create_fermi_isotropic_diffuse_model",
]
[docs]class SkyModel(Model):
"""Sky model component.
This model represents a factorised sky model.
It has `~gammapy.modeling.Parameters`
combining the spatial and spectral parameters.
Parameters
----------
spectral_model : `~gammapy.modeling.models.SpectralModel`
Spectral model
spatial_model : `~gammapy.modeling.models.SpatialModel`
Spatial model (must be normalised to integrate to 1)
temporal_model : `~gammapy.modeling.models.temporalModel`
Temporal model
name : str
Model identifier
apply_irf : dict
Dictionary declaring which IRFs should be applied to this model. Options
are {"exposure": True, "psf": True, "edisp": True}
datasets_names : list of str
Which datasets this model is applied to.
"""
tag = ["SkyModel", "sky-model"]
_apply_irf_default = {"exposure": True, "psf": True, "edisp": True}
def __init__(
self,
spectral_model,
spatial_model=None,
temporal_model=None,
name=None,
apply_irf=None,
datasets_names=None,
):
self.spatial_model = spatial_model
self.spectral_model = spectral_model
self.temporal_model = temporal_model
self._name = make_name(name)
if apply_irf is None:
apply_irf = self._apply_irf_default.copy()
self.apply_irf = apply_irf
self.datasets_names = datasets_names
self._check_unit()
super().__init__()
@property
def _models(self):
models = self.spectral_model, self.spatial_model, self.temporal_model
return [model for model in models if model is not None]
def _check_covariance(self):
if not self.parameters == self._covariance.parameters:
self._covariance = Covariance.from_stack(
[model.covariance for model in self._models],
)
def _check_unit(self):
from gammapy.data.gti import GTI
# evaluate over a test geom to check output unit
# TODO simpler way to test this ?
axis = MapAxis.from_energy_bounds(
"0.1 TeV", "10 TeV", nbin=1, name="energy_true"
)
geom = WcsGeom.create(skydir=self.position, npix=(2, 2), axes=[axis])
gti = GTI.create(1 * u.day, 2 * u.day)
value = self.evaluate_geom(geom, gti)
if self.apply_irf["exposure"]:
ref_unit = u.Unit("cm-2 s-1 MeV-1 sr-1")
else:
ref_unit = u.Unit("sr-1")
if self.spatial_model is None:
ref_unit = ref_unit / u.Unit("sr-1")
if not value.unit.is_equivalent(ref_unit):
raise ValueError(
f"SkyModel unit {value.unit} is not equivalent to {ref_unit}"
)
@property
def covariance(self):
self._check_covariance()
for model in self._models:
self._covariance.set_subcovariance(model.covariance)
return self._covariance
@covariance.setter
def covariance(self, covariance):
self._check_covariance()
self._covariance.data = covariance
for model in self._models:
subcovar = self._covariance.get_subcovariance(model.covariance.parameters)
model.covariance = subcovar
@property
def name(self):
return self._name
@property
def parameters(self):
parameters = []
parameters.append(self.spectral_model.parameters)
if self.spatial_model is not None:
parameters.append(self.spatial_model.parameters)
if self.temporal_model is not None:
parameters.append(self.temporal_model.parameters)
return Parameters.from_stack(parameters)
@property
def spatial_model(self):
"""`~gammapy.modeling.models.SpatialModel`"""
return self._spatial_model
@spatial_model.setter
def spatial_model(self, model):
if not (model is None or isinstance(model, SpatialModel)):
raise TypeError(f"Invalid type: {model!r}")
self._spatial_model = model
@property
def spectral_model(self):
"""`~gammapy.modeling.models.SpectralModel`"""
return self._spectral_model
@spectral_model.setter
def spectral_model(self, model):
if not (model is None or isinstance(model, SpectralModel)):
raise TypeError(f"Invalid type: {model!r}")
self._spectral_model = model
@property
def temporal_model(self):
"""`~gammapy.modeling.models.TemporalModel`"""
return self._temporal_model
@temporal_model.setter
def temporal_model(self, model):
if not (model is None or isinstance(model, TemporalModel)):
raise TypeError(f"Invalid type: {model!r}")
self._temporal_model = model
@property
def position(self):
"""`~astropy.coordinates.SkyCoord`"""
return getattr(self.spatial_model, "position", None)
@property
def evaluation_radius(self):
"""`~astropy.coordinates.Angle`"""
return self.spatial_model.evaluation_radius
@property
def frame(self):
return self.spatial_model.frame
def __add__(self, other):
if isinstance(other, (Models, list)):
return Models([self, *other])
elif isinstance(other, (SkyModel, BackgroundModel)):
return Models([self, other])
else:
raise TypeError(f"Invalid type: {other!r}")
def __radd__(self, model):
return self.__add__(model)
[docs] def __call__(self, lon, lat, energy, time=None):
return self.evaluate(lon, lat, energy, time)
def __repr__(self):
return (
f"{self.__class__.__name__}("
f"spatial_model={self.spatial_model!r}, "
f"spectral_model={self.spectral_model!r})"
f"temporal_model={self.temporal_model!r})"
)
[docs] def evaluate(self, lon, lat, energy, time=None):
"""Evaluate the model at given points.
The model evaluation follows numpy broadcasting rules.
Return differential surface brightness cube.
At the moment in units: ``cm-2 s-1 TeV-1 deg-2``
Parameters
----------
lon, lat : `~astropy.units.Quantity`
Spatial coordinates
energy : `~astropy.units.Quantity`
Energy coordinate
time: `~astropy.time.Time`
Time co-ordinate
Returns
-------
value : `~astropy.units.Quantity`
Model value at the given point.
"""
value = self.spectral_model(energy) # pylint:disable=not-callable
# TODO: case if self.temporal_model is not None, introduce time in arguments ?
if self.spatial_model is not None:
if self.spatial_model.is_energy_dependent:
spatial = self.spatial_model(lon, lat, energy)
else:
spatial = self.spatial_model(lon, lat)
value = value * spatial # pylint:disable=not-callable
if (self.temporal_model is not None) and (time is not None):
value = value * self.temporal_model(time)
return value
[docs] def evaluate_geom(self, geom, gti=None):
"""Evaluate model on `~gammapy.maps.Geom`."""
energy = geom.axes["energy_true"].center[:, np.newaxis, np.newaxis]
value = self.spectral_model(energy)
if self.spatial_model:
value = value * self.spatial_model.evaluate_geom(geom)
if self.temporal_model:
integral = self.temporal_model.integral(gti.time_start, gti.time_stop)
value = value * np.sum(integral)
return value
[docs] def integrate_geom(self, geom, gti=None):
"""Integrate model on `~gammapy.maps.Geom`.
Parameters
----------
geom : `Geom`
Map geometry
gti : `GTI`
GIT table
Returns
-------
flux : `Map`
Predicted flux map
"""
energy = geom.axes["energy_true"].edges
value = self.spectral_model.integral(energy[:-1], energy[1:],).reshape(
(-1, 1, 1)
)
if self.spatial_model and not isinstance(geom, RegionGeom):
# TODO: integrate spatial model over region to correct for
# containment
value = value * self.spatial_model.integrate_geom(geom).quantity
if self.temporal_model:
integral = self.temporal_model.integral(gti.time_start, gti.time_stop)
value = value * np.sum(integral)
return Map.from_geom(geom=geom, data=value.value, unit=value.unit)
[docs] def copy(self, name=None, **kwargs):
"""Copy SkyModel"""
if self.spatial_model is not None:
spatial_model = self.spatial_model.copy()
else:
spatial_model = None
if self.temporal_model is not None:
temporal_model = self.temporal_model.copy()
else:
temporal_model = None
kwargs.setdefault("name", make_name(name))
kwargs.setdefault("spectral_model", self.spectral_model.copy())
kwargs.setdefault("spatial_model", spatial_model)
kwargs.setdefault("temporal_model", temporal_model)
kwargs.setdefault("apply_irf", self.apply_irf.copy())
kwargs.setdefault("datasets_names", self.datasets_names)
return self.__class__(**kwargs)
[docs] def to_dict(self, full_output=False):
"""Create dict for YAML serilisation"""
data = {}
data["name"] = self.name
data["type"] = self.tag[0]
data["spectral"] = self.spectral_model.to_dict(full_output)
if self.spatial_model is not None:
data["spatial"] = self.spatial_model.to_dict(full_output)
if self.temporal_model is not None:
data["temporal"] = self.temporal_model.to_dict(full_output)
if self.apply_irf != self._apply_irf_default:
data["apply_irf"] = self.apply_irf
if self.datasets_names is not None:
data["datasets_names"] = self.datasets_names
return data
[docs] @classmethod
def from_dict(cls, data):
"""Create SkyModel from dict"""
from gammapy.modeling.models import (
SPATIAL_MODEL_REGISTRY,
SPECTRAL_MODEL_REGISTRY,
TEMPORAL_MODEL_REGISTRY,
)
model_class = SPECTRAL_MODEL_REGISTRY.get_cls(data["spectral"]["type"])
spectral_model = model_class.from_dict(data["spectral"])
spatial_data = data.get("spatial")
if spatial_data is not None:
model_class = SPATIAL_MODEL_REGISTRY.get_cls(spatial_data["type"])
spatial_model = model_class.from_dict(spatial_data)
else:
spatial_model = None
temporal_data = data.get("temporal")
if temporal_data is not None:
model_class = TEMPORAL_MODEL_REGISTRY.get_cls(temporal_data["type"])
temporal_model = model_class.from_dict(temporal_data)
else:
temporal_model = None
return cls(
name=data["name"],
spatial_model=spatial_model,
spectral_model=spectral_model,
temporal_model=temporal_model,
apply_irf=data.get("apply_irf", cls._apply_irf_default),
datasets_names=data.get("datasets_names"),
)
def __str__(self):
str_ = f"{self.__class__.__name__}\n\n"
str_ += "\t{:26}: {}\n".format("Name", self.name)
str_ += "\t{:26}: {}\n".format("Datasets names", self.datasets_names)
str_ += "\t{:26}: {}\n".format(
"Spectral model type", self.spectral_model.__class__.__name__
)
if self.spatial_model is not None:
spatial_type = self.spatial_model.__class__.__name__
else:
spatial_type = ""
str_ += "\t{:26}: {}\n".format("Spatial model type", spatial_type)
if self.temporal_model is not None:
temporal_type = self.temporal_model.__class__.__name__
else:
temporal_type = ""
str_ += "\t{:26}: {}\n".format("Temporal model type", temporal_type)
str_ += "\tParameters:\n"
info = _get_parameters_str(self.parameters)
lines = info.split("\n")
str_ += "\t" + "\n\t".join(lines[:-1])
str_ += "\n\n"
return str_.expandtabs(tabsize=2)
[docs] @classmethod
def create(cls, spectral_model, spatial_model=None, temporal_model=None, **kwargs):
"""Create a model instance.
Parameters
----------
spectral_model : str
Tag to create spectral model
spatial_model : str
Tag to create spatial model
temporal_model : str
Tag to create temporal model
**kwargs : dict
Keyword arguments passed to `SkyModel`
Returns
-------
model : SkyModel
Sky model
"""
spectral_model = Model.create(spectral_model, model_type="spectral")
if spatial_model:
spatial_model = Model.create(spatial_model, model_type="spatial")
if temporal_model:
temporal_model = Model.create(temporal_model, model_type="temporal")
return cls(
spectral_model=spectral_model,
spatial_model=spatial_model,
temporal_model=temporal_model,
**kwargs,
)
class FoVBackgroundModel(Model):
"""Field of view background model
The background model holds the correction parameters applied to
the instrumental background attached to a `MapDataset` or
`SpectrumDataset`.
Parameters
----------
spectral_model : `~gammapy.modeling.models.SpectralModel`
Normalized spectral model.
dataset_name : str
Dataset name
"""
tag = ["FoVBackgroundModel", "fov-bkg"]
def __init__(self, spectral_model=None, dataset_name=None):
if dataset_name is None:
raise ValueError("Dataset name a is required argument")
self.datasets_names = [dataset_name]
if spectral_model is None:
spectral_model = PowerLawNormSpectralModel()
if not spectral_model.is_norm_spectral_model:
raise ValueError("A norm spectral model is required.")
self._spectral_model = spectral_model
super().__init__()
@property
def spectral_model(self):
"""Spectral norm model"""
return self._spectral_model
@property
def name(self):
"""Model name"""
return self.datasets_names[0] + "-bkg"
@property
def parameters(self):
"""Model parameters"""
parameters = []
parameters.append(self.spectral_model.parameters)
return Parameters.from_stack(parameters)
def __str__(self):
str_ = f"{self.__class__.__name__}\n\n"
str_ += "\t{:26}: {}\n".format("Name", self.name)
str_ += "\t{:26}: {}\n".format("Datasets names", self.datasets_names)
str_ += "\t{:26}: {}\n".format(
"Spectral model type", self.spectral_model.__class__.__name__
)
str_ += "\tParameters:\n"
info = _get_parameters_str(self.parameters)
lines = info.split("\n")
str_ += "\t" + "\n\t".join(lines[:-1])
str_ += "\n\n"
return str_.expandtabs(tabsize=2)
def evaluate_geom(self, geom):
"""Evaluate map"""
energy = geom.axes["energy"].center[:, np.newaxis, np.newaxis]
return self.evaluate(energy=energy)
def evaluate(self, energy):
"""Evaluate model"""
return self.spectral_model(energy)
def to_dict(self, full_output=False):
data = {}
data["type"] = self.tag[0]
data["datasets_names"] = self.datasets_names
data["spectral"] = self.spectral_model.to_dict(full_output=full_output)
return data
@classmethod
def from_dict(cls, data):
"""Create model from dict
Parameters
----------
data : dict
Data dictionary
"""
from gammapy.modeling.models import SPECTRAL_MODEL_REGISTRY
spectral_data = data.get("spectral")
if spectral_data is not None:
model_class = SPECTRAL_MODEL_REGISTRY.get_cls(spectral_data["type"])
spectral_model = model_class.from_dict(spectral_data)
else:
spectral_model = None
datasets_names = data.get("datasets_names")
if datasets_names is None:
raise ValueError("FoVBackgroundModel must define a dataset name")
if len(datasets_names) > 1:
raise ValueError("FoVBackgroundModel can only be assigned to one dataset")
return cls(spectral_model=spectral_model, dataset_name=datasets_names[0],)
def reset_to_default(self):
"""Reset parameter values to default"""
values = self.spectral_model.default_parameters.values
self.spectral_model.parameters.values = values
def copy(self, **kwargs):
"""Copy SkyModel"""
return self.__class__(**kwargs)
class BackgroundModel(Model):
"""Background model.
Create a new map by a tilt and normalization on the available map
Parameters
----------
map : `~gammapy.maps.Map`
Background model map
spectral_model : `~gammapy.modeling.models.SpectralModel`
Normalized spectral model,
default is `~gammapy.modeling.models.PowerLawNormSpectralModel`
"""
tag = "BackgroundModel"
map = LazyFitsData(cache=True)
def __init__(
self, map, spectral_model=None, name=None, filename=None, datasets_names=None,
):
if isinstance(map, Map):
axis = map.geom.axes["energy"]
if axis.node_type != "edges":
raise ValueError(
'Need an integrated map, energy axis node_type="edges"'
)
self.map = map
self._name = make_name(name)
self.filename = filename
if spectral_model is None:
spectral_model = PowerLawNormSpectralModel()
spectral_model.tilt.frozen = True
self.spectral_model = spectral_model
if isinstance(datasets_names, list):
if len(datasets_names) != 1:
raise ValueError(
"Currently background models can only be assigned to one dataset."
)
self.datasets_names = datasets_names
super().__init__()
@property
def name(self):
return self._name
@property
def energy_center(self):
"""True energy axis bin centers (`~astropy.units.Quantity`)"""
energy_axis = self.map.geom.axes["energy"]
energy = energy_axis.center
return energy[:, np.newaxis, np.newaxis]
@property
def spectral_model(self):
"""`~gammapy.modeling.models.SpectralModel`"""
return self._spectral_model
@spectral_model.setter
def spectral_model(self, model):
if not (model is None or isinstance(model, SpectralModel)):
raise TypeError(f"Invalid type: {model!r}")
self._spectral_model = model
@property
def parameters(self):
parameters = []
parameters.append(self.spectral_model.parameters)
return Parameters.from_stack(parameters)
def evaluate(self):
"""Evaluate background model.
Returns
-------
background_map : `~gammapy.maps.Map`
Background evaluated on the Map
"""
value = self.spectral_model(self.energy_center).value
back_values = self.map.data * value
return self.map.copy(data=back_values)
def to_dict(self, full_output=False):
data = {}
data["name"] = self.name
data["type"] = self.tag
data["spectral"] = self.spectral_model.to_dict(full_output)
if self.filename is not None:
data["filename"] = self.filename
if self.datasets_names is not None:
data["datasets_names"] = self.datasets_names
return data
@classmethod
def from_dict(cls, data):
from gammapy.modeling.models import SPECTRAL_MODEL_REGISTRY
spectral_data = data.get("spectral")
if spectral_data is not None:
model_class = SPECTRAL_MODEL_REGISTRY.get_cls(spectral_data["type"])
spectral_model = model_class.from_dict(spectral_data)
else:
spectral_model = None
if "filename" in data:
bkg_map = Map.read(data["filename"])
elif "map" in data:
bkg_map = data["map"]
else:
# TODO: for now create a fake map for serialization,
# uptdated in MapDataset.from_dict()
axis = MapAxis.from_edges(np.logspace(-1, 1, 2), unit=u.TeV, name="energy")
geom = WcsGeom.create(
skydir=(0, 0), npix=(1, 1), frame="galactic", axes=[axis]
)
bkg_map = Map.from_geom(geom)
return cls(
map=bkg_map,
spectral_model=spectral_model,
name=data["name"],
datasets_names=data.get("datasets_names"),
filename=data.get("filename"),
)
def copy(self, name=None):
"""A deep copy."""
new = copy.deepcopy(self)
new._name = make_name(name)
return new
def cutout(self, position, width, mode="trim", name=None):
"""Cutout background model.
Parameters
----------
position : `~astropy.coordinates.SkyCoord`
Center position of the cutout region.
width : tuple of `~astropy.coordinates.Angle`
Angular sizes of the region in (lon, lat) in that specific order.
If only one value is passed, a square region is extracted.
mode : {'trim', 'partial', 'strict'}
Mode option for Cutout2D, for details see `~astropy.nddata.utils.Cutout2D`.
name : str
Name of the returned background model.
Returns
-------
cutout : `BackgroundModel`
Cutout background model.
"""
cutout_kwargs = {"position": position, "width": width, "mode": mode}
bkg_map = self.map.cutout(**cutout_kwargs)
spectral_model = self.spectral_model.copy()
return self.__class__(bkg_map, spectral_model=spectral_model, name=name)
def stack(self, other, weights=None):
"""Stack background model in place.
Stacking the background model resets the current parameters values.
Parameters
----------
other : `BackgroundModel`
Other background model.
"""
bkg = self.evaluate()
other_bkg = other.evaluate()
bkg.stack(other_bkg, weights=weights)
self.map = bkg
# reset parameter values
self.spectral_model.norm.value = 1
self.spectral_model.tilt.value = 0
def __str__(self):
str_ = self.__class__.__name__ + "\n\n"
str_ += "\t{:26}: {}\n".format("Name", self.name)
str_ += "\t{:26}: {}\n".format("Datasets names", self.datasets_names)
str_ += "\tParameters:\n"
info = _get_parameters_str(self.parameters)
lines = info.split("\n")
str_ += "\t" + "\n\t".join(lines[:-1])
str_ += "\n\n"
return str_.expandtabs(tabsize=2)
@property
def position(self):
"""`~astropy.coordinates.SkyCoord`"""
return self.map.geom.center_skydir
@property
def evaluation_radius(self):
"""`~astropy.coordinates.Angle`"""
return np.max(self.map.geom.width) / 2.0
def create_fermi_isotropic_diffuse_model(filename, **kwargs):
"""Read Fermi isotropic diffuse model.
See `LAT Background models <https://fermi.gsfc.nasa.gov/ssc/data/access/lat/BackgroundModels.html>`_
Parameters
----------
filename : str
filename
kwargs : dict
Keyword arguments forwarded to `TemplateSpectralModel`
Returns
-------
diffuse_model : `SkyModel`
Fermi isotropic diffuse sky model.
"""
vals = np.loadtxt(make_path(filename))
energy = u.Quantity(vals[:, 0], "MeV", copy=False)
values = u.Quantity(vals[:, 1], "MeV-1 s-1 cm-2", copy=False)
kwargs.setdefault("interp_kwargs", {"fill_value": None})
spatial_model = ConstantSpatialModel()
spectral_model = (
TemplateSpectralModel(energy=energy, values=values, **kwargs)
* PowerLawNormSpectralModel()
)
return SkyModel(
spatial_model=spatial_model,
spectral_model=spectral_model,
name="fermi-diffuse-iso",
apply_irf={"psf": False, "exposure": True, "edisp": True},
)