MapEvaluator¶

class gammapy.cube.MapEvaluator(model=None, exposure=None, background=None, psf=None, edisp=None, background_model=None)[source]¶

Bases: object

Sky model evaluation on maps.

This is a first attempt to compute flux as well as predicted counts maps.

The basic idea is that this evaluator is created once at the start of the analysis, and pre-computes some things. It it then evaluated many times during likelihood fit when model parameters change, re-using pre-computed quantities each time. At the moment it does some things, e.g. cache and re-use energy and coordinate grids, but overall it is not an efficient implementation yet.

For now, we only make it work for 3D WCS maps with an energy axis. No HPX, no other axes, those can be added later here or via new separate model evaluator classes.

We should discuss how to organise the model and IRF evaluation code, and things like integrations and convolutions in a good way.

Parameters:	model : `SkyModel` Sky model exposure : `Map` Exposure map background : `Map` Background map psf : `PSFKernel` PSF kernel edisp : `EnergyDispersion` Energy dispersion background_model: `~gammapy.cube.models.BackgroundModel` Background model to use for the evaluation. Can be specified instead of `background`.

Attributes Summary

`bin_volume`	Map pixel bin volume (solid angle times energy bin width).
`energy_bin_width`	Energy axis bin widths (`astropy.units.Quantity`)
`energy_center`	True energy axis bin centers (`Quantity`)
`energy_edges`	Energy axis bin edges (`Quantity`)
`geom`	This will give the energy axes in e_true
`geom_image`
`lat`
`lon`
`lon_lat`	Spatial coordinate pixel centers.
`solid_angle`	Solid angle per pixel

Methods Summary

`apply_edisp`(npred)	Convolve map data with energy dispersion.
`apply_exposure`(flux)	Compute npred cube
`apply_psf`(npred)	Convolve npred cube with PSF
`compute_dnde`()	Compute model differential flux at map pixel centers.
`compute_flux`()	Compute model integral flux over map pixel volumes.
`compute_npred`()	Evaluate model predicted counts.

Attributes Documentation

bin_volume¶: Map pixel bin volume (solid angle times energy bin width).

energy_bin_width¶: Energy axis bin widths (astropy.units.Quantity)

energy_center¶: True energy axis bin centers (Quantity)

energy_edges¶: Energy axis bin edges (Quantity)

geom¶: This will give the energy axes in e_true

geom_image¶

lat¶

lon¶

lon_lat¶

Spatial coordinate pixel centers.

Returns lon, lat tuple of Quantity.

solid_angle¶: Solid angle per pixel

Methods Documentation

apply_edisp(npred)[source]¶

Convolve map data with energy dispersion.

Parameters:	npred : `Map` Predicted counts in true energy bins
Returns:	npred_reco : `Map` Predicted counts in reco energy bins

apply_exposure(flux)[source]¶

Compute npred cube

For now just divide flux cube by exposure

apply_psf(npred)[source]¶: Convolve npred cube with PSF

compute_dnde()[source]¶

Compute model differential flux at map pixel centers.

Returns:	model_map : `Map` Sky cube with data filled with evaluated model values. Units: `cm-2 s-1 TeV-1 deg-2`

compute_flux()[source]¶

Compute model integral flux over map pixel volumes.

For now, we simply multiply dnde with bin volume.

compute_npred()[source]¶

Evaluate model predicted counts.

Returns:	npred.data : ~numpy.ndarray array of the predicted counts in each bin (in reco energy)