Light curves for flares#

Compute the light curve of a PKS 2155-304 flare on 10 minutes time intervals.


  • Understanding of how the light curve estimator works, please refer to the light curve notebook </tutorials/analysis-time/light_curve.


Frequently, especially when studying flares of bright sources, it is necessary to explore the time behaviour of a source on short time scales, in particular on time scales shorter than observing runs.

A typical example is given by the flare of PKS 2155-304 during the night from July 29 to 30 2006. See the following article.

Objective: Compute the light curve of a PKS 2155-304 flare on 5 minutes time intervals, i.e. smaller than the duration of individual observations.

Proposed approach#

We have seen in the general presentation of the light curve estimator, see the light curve notebook, Gammapy produces datasets in a given time interval, by default that of the parent observation. To be able to produce datasets on smaller time steps, it is necessary to split the observations into the required time intervals.

This is easily performed with the select_time method of Observations. If you pass it a list of time intervals it will produce a list of time filtered observations in a new Observations object. Data reduction can then be performed and will result in datasets defined on the required time intervals and light curve estimation can proceed directly.

In summary, we have to:

  • Select relevant Observations from the DataStore

  • Apply the time selection in our predefined time intervals to obtain a new Observations

  • Perform the data reduction (in 1D or 3D)

  • Define the source model

  • Extract the light curve from the reduced dataset

Here, we will use the PKS 2155-304 observations from the H.E.S.S. first public test data release. We will use time intervals of 5 minutes duration. The tutorial is implemented with the intermediate level API.


As usual, we’ll start with some general imports…

import logging
import numpy as np
import astropy.units as u
from astropy.coordinates import Angle, SkyCoord
from astropy.time import Time
from regions import CircleSkyRegion

# %matplotlib inline
import matplotlib.pyplot as plt

log = logging.getLogger(__name__)

from import DataStore
from gammapy.datasets import Datasets, SpectrumDataset
from gammapy.estimators import LightCurveEstimator
from gammapy.makers import (
from gammapy.maps import MapAxis, RegionGeom
from gammapy.modeling.models import PowerLawSpectralModel, SkyModel

Check setup#

from gammapy.utils.check import check_tutorials_setup


        python_executable      : /home/runner/work/gammapy-docs/gammapy-docs/gammapy/.tox/build_docs/bin/python
        python_version         : 3.9.15
        machine                : x86_64
        system                 : Linux

Gammapy package:

        version                : 1.0
        path                   : /home/runner/work/gammapy-docs/gammapy-docs/gammapy/.tox/build_docs/lib/python3.9/site-packages/gammapy

Other packages:

        numpy                  : 1.23.4
        scipy                  : 1.9.3
        astropy                : 5.1.1
        regions                : 0.7
        click                  : 8.1.3
        yaml                   : 6.0
        IPython                : 8.6.0
        jupyterlab             : not installed
        matplotlib             : 3.6.2
        pandas                 : not installed
        healpy                 : 1.16.1
        iminuit                : 2.17.0
        sherpa                 : 4.15.0
        naima                  : 0.10.0
        emcee                  : 3.1.3
        corner                 : 2.2.1

Gammapy environment variables:

        GAMMAPY_DATA           : /home/runner/work/gammapy-docs/gammapy-docs/gammapy-datasets/1.0

Select the data#

We first set the datastore.

data_store = DataStore.from_dir("$GAMMAPY_DATA/hess-dl3-dr1/")

Now we select observations within 2 degrees of PKS 2155-304.

target_position = SkyCoord(329.71693826 * u.deg, -30.2255890 * u.deg, frame="icrs")
selection = dict(
    radius=2 * u.deg,
obs_ids = data_store.obs_table.select_observations(selection)["OBS_ID"]
observations = data_store.get_observations(obs_ids)
print(f"Number of selected observations : {len(observations)}")
Number of selected observations : 21

Define time intervals#

We create the list of time intervals. Each time interval is an astropy.time.Time object, containing a start and stop time.

t0 = Time("2006-07-29T20:30")
duration = 10 * u.min
n_time_bins = 35
times = t0 + np.arange(n_time_bins) * duration
time_intervals = [Time([tstart, tstop]) for tstart, tstop in zip(times[:-1], times[1:])]
[53945.85416667 53945.86111111]

Filter the observations list in time intervals#

Here we apply the list of time intervals to the observations with select_time().

This will return a new list of Observations filtered by time_intervals. For each time interval, a new observation is created that converts the intersection of the GTIs and time interval.

short_observations = observations.select_time(time_intervals)
# check that observations have been filtered
print(f"Number of observations after time filtering: {len(short_observations)}\n")
Number of observations after time filtering: 44

GTI info:
- Number of GTIs: 1
- Duration: 600.0 s
- Start: 175898401.00000018 s MET
- Start: 2006-07-29T20:41:05.184 (time standard: TT)
- Stop: 175899001.00000018 s MET
- Stop: 2006-07-29T20:51:05.184 (time standard: TT)

As we can see, we have now observations of duration equal to the chosen time step.

Now data reduction and light curve extraction can proceed exactly as before.

Building 1D datasets from the new observations#

Here we will perform the data reduction in 1D with reflected regions.

Beware, with small time intervals the background normalization with OFF regions might become problematic.

Defining the geometry#

We define the energy axes. As usual, the true energy axis has to cover a wider range to ensure a good coverage of the measured energy range chosen.

We need to define the ON extraction region. Its size follows typical spectral extraction regions for HESS analyses.

# Target definition
energy_axis = MapAxis.from_energy_bounds("0.4 TeV", "20 TeV", nbin=10)
energy_axis_true = MapAxis.from_energy_bounds(
    "0.1 TeV", "40 TeV", nbin=20, name="energy_true"

on_region_radius = Angle("0.11 deg")
on_region = CircleSkyRegion(center=target_position, radius=on_region_radius)

geom = RegionGeom.create(region=on_region, axes=[energy_axis])

Creation of the data reduction makers#

We now create the dataset and background makers for the selected geometry.

dataset_maker = SpectrumDatasetMaker(
    containment_correction=True, selection=["counts", "exposure", "edisp"]
bkg_maker = ReflectedRegionsBackgroundMaker()
safe_mask_masker = SafeMaskMaker(methods=["aeff-max"], aeff_percent=10)

Creation of the datasets#

Now we perform the actual data reduction in the time_intervals.

Define the Model#

The actual flux will depend on the spectral shape assumed. For simplicity, we use the power law spectral model of index 3.4 used in the reference paper.

Here we use only a spectral model in the SkyModel object.

spectral_model = PowerLawSpectralModel(
    index=3.4, amplitude=2e-11 * u.Unit("1 / (cm2 s TeV)"), reference=1 * u.TeV
spectral_model.parameters["index"].frozen = False

sky_model = SkyModel(spatial_model=None, spectral_model=spectral_model, name="pks2155")

Assign to model to all datasets#

We assign each dataset its spectral model

Extract the light curve#

We first create the LightCurveEstimator for the list of datasets we just produced. We give the estimator the name of the source component to be fitted.

lc_maker_1d = LightCurveEstimator(
    energy_edges=[0.7, 20] * u.TeV,

We can now perform the light curve extraction itself. To compare with the reference paper, we select the 0.7-20 TeV range.

Finally we plot the result for the 1D lightcurve:

plt.figure(figsize=(8, 6))
light curve flare

Total running time of the script: ( 1 minutes 13.227 seconds)

Gallery generated by Sphinx-Gallery