Atmospheric simulations

Federico Di Pierro (INFN Torino), [email protected]

for the ccf-atmosim group (partial, sorry!): Bianca Maria Dinelli, Michele Doro,
Markus Gaug, Tarek Hassan, Raquel de Los Reyes, Enzo Papandrea, Michael Prouza,
Stefan Shssler, Stanislav Stefanik, Piero Vallania, Carlo Vigorito, Michal Vrastil,...

CTA CCF f2f, Barcelona, June 22 nd, 2016 1

 Outline
Goals of the simulations

Studied atmospheric profiles

atmospheric transmission from MODTRAN

MC atmosim production

Event reconstruction

Results
ideal case: perfect knowledge
of the atmosphere
extreme case: only an
average profile

CTA CCF f2f, Barcelona, June 22 nd, 2016 2

 Goals
The Atmosphere, as part of our detector, has a strong influence on the data.
Two main aspects are the molecular absorption profile and the presence of
aerosols. In CTA the atmosphere will be monitored with several Atmospheric
Calibration instruments and using external data (GDAS).

We want to estimate, by means of detailed MC:

the effect of different atmospheric profiles on CTA performances

effective area (energy threshold, flux)
energy bias and resolution
angular resolution
the effect of aerosols (dust, clouds) under different conditions.
the effect of atmospheric calibration uncertainties on recontructed
energy and flux uncertainties
their compliance with CTA performance requirements (for instance:
A-PERF-0240, Energy resolution)
the possible strategy for generating observation-wise MC simulations
inputs for CORSIKA: density, thickness, refraction index profiles
inputs for sim_telarray: atmospheric transmission (vs and photon
production heigth)
CTA CCF f2f, Barcelona, June 22 nd, 2016 3
 Atmospheric profiles
We focused on the La Palma site
very well known atmosphere, lots of measurements already available
profiles from GDAS* (< 25 km) and NRLMSISE-00 (25 - 100 km) and
exponential extrapolation up to 120 km (technically needed by corsika)

Name Description
Thanks to
Average Winter Averaged over all winter profiles. Markus Gaug!
Average Summer Averaged over all summer profiles.
Extreme14.0_low Air density has the minimum value at 14 km.
Extreme16.0_high Air density has the maximum value at 16 km.
Extreme7.0_low Air density has the minimum value at 7 km.
Extreme5.0_high Air density has the maximum value at 5 km.
the different "extreme" altitudes have been chosen to maximize the
effects on the HE or on the LE events.
we used also the atmospheric profiles used in CTA-N MC Production 3

* GDAS (Global Data Assimilation System) data of NCEP (National Centers for
Environmental Prediction, USA)

CTA CCF f2f, Barcelona, June 22 nd, 2016 4

 Atmospheric profiles
We focused on the La Palma site
very well known atmosphere, lots of measurements already available
profiles from GDAS (< 25 km) and NRLMSISE-00 (25 - 100 km) and
exponential extrapolation up to 120 km (technically needed by corsika)

CTA CCF f2f, Barcelona, June 22 nd, 2016 5

 Atmospheric profiles
We focused on the La Palma site
very well known atmosphere, lots of measurements already available
profiles from GDAS (< 25 km) and NRLMSISE-00 (25 - 100 km) and
exponential extrapolation up to 120 km (technically needed by corsika)

CTA CCF f2f, Barcelona, June 22 nd, 2016 6

 Atmospheric profiles
We focused on the La Palma site
very well known atmosphere, lots of measurements already available
profiles from GDAS (< 25 km) and NRLMSISE-00 (25 - 100 km) and
exponential extrapolation up to 120 km (technically needed by corsika)
thickness [g/cm2]

CTA CCF f2f, Barcelona, June 22 nd, 2016 7

 Atmospheric profiles
We focused on the La Palma site
very well known atmosphere, lots of measurements already available
profiles from GDAS (< 25 km) and NRLMSISE-00 (25 - 100 km) and
exponential extrapolation up to 120 km (technically needed by corsika)
refractive index

CTA CCF f2f, Barcelona, June 22 nd, 2016 8

 Atmospheric transmission
Obtained using MODTRAN, with previous profiles as input Thanks to Enzo
Only molecular absorption (Haze = 0, no aerosol attenuation) Papandrea!
In Prod3 aerosols* were included (atm_trans_2147_1_3_0_0_0.dat)

*The atmospheric transmission component due to the aerosols will be measured

by the LIDARs.
CTA CCF f2f, Barcelona, June 22 nd, 2016 9
 Atmospheric transmission
Obtained using MODTRAN, with previous profiles as input
Only molecular absorption (Haze = 0, no aerosol attenuation)
In Prod3 aerosols* were included (atm_trans_2147_1_3_0_0_0.dat)

*The atmospheric transmission component due to the aerosols will be measured

by the LIDARs.
CTA CCF f2f, Barcelona, June 22 nd, 2016 10
 Production

3AL4M15-1-F
CORSIKA
Primary: gamma, point-like
Zenith angle: 20 deg
Azimuth: South
Energy range: 3 GeV - 100 TeV
Spectral index: -2
Sampling area (radius): 1000 m,
after optimization (< 0.1% lost
events at any energy)
Number of showers: 5000/50000
core re-scattering: 10
3000/300 jobs = ~ 150 M events

SIM_TELARRAY
from Production 3: 4 LSTs and 15
MSTs (Flash-cam)

CTA CCF f2f, Barcelona, June 22 nd, 2016 11

 Production
run on the GRID
scripts adapted from Prod3, several issues solved (new DIRAC version,
security update on the CE, multi CE improvement,...)

Thanks to Luisa Arrabito!

CTA CCF f2f, Barcelona, June 22 nd, 2016 12
 Production
run on the GRID
scripts adapted from Prod3, several issues solved (new DIRAC version,
security update on the CE, multi CE improvement,...)
small production (only gammas!), 1 model = ~300 cores, ~5 days, ~3 TB.

Atmospheric Corsika Sim_telarray Evndisp Evndisp

model (stero rec) (LUT)
Prod3: with Done Done Done Done
aerosols 1.6 TB 1.8 TB 3.2 GB 3.5 GB
Average winter Done Done Done Done
Average summer Done Done Done Done
Extreme_14.0_low Done Done Done Done
Extreme_16.0_high Done Done Done Done
Extreme_5.0_high Done Done Done Done
Extreme7.0_low Done Done Done Done

Thanks to Michal Vrastil!

CTA CCF f2f, Barcelona, June 22 nd, 2016 13
 Reconstruction
Eventdisplay package (version 2016.01.11)
converter
stereo reconstruction
LUT (energy, mscw/l)

Reconstruction options and cuts: standard prod3 analysis,

job.setCalibrationFile( 'ped.20151106.evndisp.root' )
job.setNNcleaninginputcard( 'EVNDISP.NNcleaning.dat' )
job.setReconstructionParameter( 'EVNDISP.prod3.reconstruction.runpar
ameter.NN' )
signals extraction:

FADCANALYSIS 2 (sliding window)

FADCSUMMATIONWINDOW 6 (LST) and 4 (MST-F)
image cleaning: "TIMENEXTNEIGHBOUR" (arXiv:1307.4939v1)

image edge fit: if Loss >= 0.1 and ntubes >= 5 (MST-F)

telescope-wise pre-cuts: ntubes 4, max loss 0.1/0.2, no Size cut

stereo-wise: minangle 10 deg (LST), 15 deg (MST-F)

Simple reconstruction quality cuts (only to avoid too poorly reconstructed

events), next performances are pseudo-performances.
number of telescopes (>= 2), linear reconstructed energy differences

(dES < 1.5), Mean Scaled Width (-2 < MSCW < 0.5),...
CTA CCF f2f, Barcelona, June 22 nd, 2016 14
 Analysis: the selected events
Example: Prod 3 atmospheric model.

Minimum reconstructed energy (~ 20GeV)

Reconstructed energy shows features < 60 GeV
CTA CCF f2f, Barcelona, June 22 nd, 2016 15
 Cross-check with Prod3
From Gernot's analysis:
3A4L15M-3-F, February (red)
3A4L15M-1-F, April (black)
My production, same atmosphere (with aerosols), 3A4L15M-1-F, February (green)

CTA CCF f2f, Barcelona, June 22 nd, 2016 16

 Cross-check with Prod3
From Gernot's analysis:
3A4L15M-3-F, February (red)
3A4L15M-1-F, April (black)
My production, same atmosphere (with aerosols), 3A4L15M-1-F, February (green)

CTA CCF f2f, Barcelona, June 22 nd, 2016 17

 Cross-check with Prod3
From Gernot's analysis:
3A4L15M-3-F, February (red)
3A4L15M-1-F, April (black)
My production, same atmosphere (with aerosols), 3A4L15M-1-F, February (green)

The difference could be due to improved analysis...(to be investigated)

The atmospheric models comparisons are done with the same analysis.
CTA CCF f2f, Barcelona, June 22 nd, 2016 18
 Effective area
Effective area after trigger and reconstruciton
spectra rescaled to spectral index -2.5
uncertainty on the effective area reflects on the flux uncertainty

All following results should be intended as pseudo-effective area,

pseudo-Energy resolution, etc...because it was not applied the full analysis.
CTA CCF f2f, Barcelona, June 22 nd, 2016 19
 Effective area

Maximum difference for the atmosphere including aerosols

At high energy the atmospheres denser in their higher part (~14 km) produce
larger effective areas, at LE the atmospheres denser in their lower part (<7km)
produce larger effective areas. All models are consistent.
The largest effects is at the threshold, however always of the order 5-10%.
CTA CCF f2f, Barcelona, June 22 nd, 2016 20
 Energy bias
mean of E rec E true
E true
not expected a large effect, we are using the same atmosphere for filling the
LUTs and for the data production

CTA CCF f2f, Barcelona, June 22 nd, 2016 21

 Energy bias

The difference of the energy biases w.r.t. Average Winter's one

The large effects < 20 GeV are due to a selection of the events which suffered
large fluctuations
Above 20 GeV, 2 % effect.

CTA CCF f2f, Barcelona, June 22 nd, 2016 22

 Energy resolution
E interval including 68 % of the events (w.r.t. to mean(Erec - Etrue))

structure below 100 GeV, transition from LST to MST?

compliant with requirements

CTA CCF f2f, Barcelona, June 22 nd, 2016 23

 Energy resolution

as expected, very small effect 2 % effect.

CTA CCF f2f, Barcelona, June 22 nd, 2016 24

 Angular resolution
68 % containment radius

We expect a second order effect on the angular resolution due to the different
atmospheric profiles
CTA CCF f2f, Barcelona, June 22 nd, 2016 25
 Angular resolution
68 % containment radius

We expect a second order effect on the angular resolution due to the different
atmospheric profiles
CTA CCF f2f, Barcelona, June 22 nd, 2016 26
 Extreme case: no atmo calibrations
Using the Average Winter LUTs to reconstruct data
produced with all other atmospheric profiles
we expect a large effect on the Energy bias
same Size corresponds to different energy

only minor effects on the Energy resolution

the spread of the reconstructed energies should not

increase

Very recent results: Preliminary!

CTA CCF f2f, Barcelona, June 22 nd, 2016 27

 Extreme case: energy bias

Preliminary!

CTA CCF f2f, Barcelona, June 22 nd, 2016 28

 Extreme case: energy bias

Preliminary!

The difference with the atmosphere with aerosols is the largest (rather constant
8% energy underestimation)
The purely molecular profiles show an Energy dependency ( 4 %)
dangerous for energy spectra!
largest effects for the extreme atmospheric profiles occuring at high altitude
CTA CCF f2f, Barcelona, June 22 nd, 2016 29
 E.c.: Energy and Angular resolutions
Energy resolution Angular resolution
Preliminary! Preliminary!

The effects are rahter small as expected

CTA CCF f2f, Barcelona, June 22 nd, 2016 30

 Next steps

To study with MODTRAN the dependencies of the Optical

Depth on the variations of the atmospheric profiles (density,
thickness, refractive index, RH, Water Vapor fraction...)

To introduce the aerosols in the atmospheric transmission

different kind of aerosols
different altitude and thickness
George has already provided measured atmospheric
transmission profiles at the HESS site.

Same studies at different zenith angle

To study the effects of the uncertainties of the currently

foreseen CTA atmospheric measurements
how much and how fast do the molecular and aerosols
profiles change?
how precisely will they be measured?
CTA CCF f2f, Barcelona, June 22 nd, 2016 31
 Backup

CTA CCF f2f, Barcelona, June 22 nd, 2016 32