German Centre for Cosmological Lensing - The Buzzard Flock: … · 2021. 1. 19. · Role of...
Transcript of German Centre for Cosmological Lensing - The Buzzard Flock: … · 2021. 1. 19. · Role of...
-
5/6/20 — GCCL Seminar
The Buzzard Flock: Synthetic Sky Catalogs for Precision CosmologyJoe DeRose (UC Santa Cruz & UC Berkeley)
-
Low Redshift Universe Tests of LCDM
Stage III and Stage IV galaxy surveys will put LCDM to the test, but only if we can understand our systematics.
Hildebrandt et al. 2018
-
Role of Cosmological Simulations in Modern Surveys
The Mock as the TestPipeline and algorithm developmentSystematics estimation and tests of marginalization schemesCase study - the Dark Energy Survey
The Mock as the Model (not in this talk, but I also work on this)Accurate predictions for non-linear observablesCovariances
-
The Dark Energy SurveyImaging survey of the southern sky
~5000 sq. degrees4m Blanco Telescope on Cerro Tololo, Chile5 bands: grizy
Done taking 6 years of data, results published for first year (Y1) and working on analyzing first 3 years (Y3)
-
DES Year 1 Cosmology: 3x2-point
galaxies x galaxiesangular clustering
lensing x lensingcosmic shear
galaxies x lensinggalaxy-galaxy lensing
-
Testing 3x2-point with Mocks
Is my pipeline accurate enough for the statistical precision of my data?Robustness to modeling assumptions/observational systematics
galaxy bias, redshift distributions, intrinsic alignments, baryons, shear calibration, etc.
-
Is my pipeline accurate enough for the statistical precision of my data?Robustness to modeling assumptions/observational systematics
galaxy bias, redshift distributions, intrinsic alignments, baryons, shear calibration, etc.
Testing 3x2-point with Mocks
-
Is my pipeline accurate enough for the statistical precision of my data?Robustness to modeling assumptions/observational systematics
galaxy bias, redshift distributions, intrinsic alignments, baryons, shear calibration, etc.
RequirementsHigh enough resolution to model all measurements accurately, e.g. clustering and lensingFlexible enough to model many samples and their cross-correlationsMany times the volume of the survey (must be inexpensive)
Testing 3x2-point with Mocks
-
Our Solution: The Buzzard Flock
JDR et al 2019
Low Resolution Lightcones
1050, 2600, 4000 Mpc/h
Source Galaxy Catalog(Metacalibration)
Error Model
Photo-z
Sample Selection
Lens Galaxy Catalog(redMaGiC)
Cluster Catalog(RedMaPPer)
CALCLENS
ADDGALS
High Resolution Tuning Box
SHAM
p(delta|L,z)
Input Cosmology
LuminosityFunction
f_red(L,z)Color ModelTraining Set
L_cen(M,z)
-
Sub-halo Abundance Matching
Simple 1-to-1 matching between (sub)halos in simulation and galaxies via rank order in cumulative abundanceTraditionally 1 free parameter: scatter in galaxy property at fixed halo property
-
AM Can Model a Diversity of Clustering Statistics
100
200r p
wp(r
p)
0
10
20
s⇠̂ 0
(s) SHAM
SDSS
100 101
r[h�1 Mpc]
�10
0
10
20
s⇠̂ 2
(s)
9.8 < M⇤ < 10.2
100 101
r[h�1 Mpc]
10.2 < M⇤ < 10.6
100 101
r[h�1 Mpc]
10.6 < M⇤ < 11.2
Use different proxy, ! , where ! is a free param yields good
fits to all statistics other than in lowest mass sample.
vα = vvir (vpeakvvir )
α
α
JDR in prep.
-
AM Can Model a Diversity of Clustering Statistics
Can fit lowest stellar mass bin with orphan model, but inconsistent with more massive samples. A persistent problem!
100
200r p
wp(r
p)
0
10
20
s⇠̂ 0
(s)
SHAM
SDSS
100 101
r[h�1 Mpc]
�10
0
10
s⇠̂ 2
(s)
9.8 < log M⇤ < 10.2
100 101
r[h�1 Mpc]
10.2 < log M⇤ < 10.6
100 101
r[h�1 Mpc]
10.6 < log M⇤ < 11.2
JDR in prep.
-
ADDGALS
Adding Density Determined Galaxies to Lightcone Simulations
Run subhalo abundance matching model on high resolution simulation, measure !p(δ |L, z)
Wechsler, JDR in prep.
Mr, z = [�21.8, 0.0] Mr, z = [�21.8, 0.3] Mr, z = [�21.8, 0.7] Mr, z = [�21.8, 1.2]
Mr, z = [�20.8, 0.0]
Mr, z = [�19.9, 0.0]
10�1 100
Mr, z = [�18.9, 0.0]
10�1 100 10�1 100 10�1 100
Addgals
SHAM
R� [h�1 Mpc]
p(R
�)
-
ADDGALS
Wechsler, JDR in prep.
Adding Density Determined Galaxies to Lightcone Simulations
Run subhalo abundance matching model on high resolution simulation, measure !Sample from this distribution to populate low-res light cones
p(δ |L, z)
0
200
400
600
800
r pw
p(r p
)
Mr < �22100
150
200
Mr < �21
SHAM
Addgals T1
Addgals L1
Addgals L2
SDSS
10�1 100 101
rp[h�1 Mpc]
50
100
150
r pw
p(r p
)
Mr < �2010�1 100 101
rp[h�1 Mpc]
50
75
100
125
Mr < �19
-
SEDs & Colors
Wechsler, JDR in prep.
SDSS
kcorrect
Addgals
�22 < Mr < �21
p(c)
�21 < Mr < �20
1 2
u � g0.0 0.5 1.0
g � r0.2 0.4 0.6
r � i
�20 < Mr < �19
0.0 0.2 0.4 0.6
i � z
Paste SED template fits from SDSS as a function of !
Templates not constrained to bluer than rest frame g.Can generate any optical, NIR, NUV photometry by integrating SEDs over relevant bandpasses
Mr, z
-
SEDs & Colors
Paste SED template fits from SDSS as a function of ! :
Not perfect at higher redshifts:Red sequence well modeled, but blue cloud off by redshift dependent shift.Provides something reasonably complex for algorithm testing
Mr, z 0.00.1
COSMOS
Buzzard
Shifted Buzzard
z = 0.20 � 0.43
0.0
0.1z = 0.43 � 0.63
0.0
0.1z = 0.63 � 0.90
0.0 2.5g � r
0.0
0.1
0 1r � i
z = 0.90 � 1.30
0 1i � z
p(c)
JDR et al 2019
-
Wechsler, JDR in prep.
Color-dependent clustering
Impart color dependence to clustering signal by conditional abundance matching SEDs to galaxies
Rank SEDs by g-r color at fixed !Rank simulated galaxies by distance to nearest massive halo at fixed !
Mr
Mr
100
101
102
103
wp(
r p)
�22 < Mr < �21
Addgals
SDSS
�21 < Mr < �20 �20 < Mr < �19
10�1 100 101
1
2
3w
p(r p
)/w
p(r p
)tot
10�1 100 101
rp [h�1Mpc]10�1 100 101
-
Wechsler, JDR in prep.
Color-dependent clustering
Impart color dependence to clustering signal by conditional abundance matching SEDs to galaxies
Successfully matches quenched fraction as function of rResidual issues due to lack of orphans in SHAM
100
101
102
103
wp(
r p)
�22 < Mr < �21
Addgals
SDSS
�21 < Mr < �20 �20 < Mr < �19
10�1 100 101
1
2
3w
p(r p
)/w
p(r p
)tot
10�1 100 101
rp [h�1Mpc]10�1 100 101
-
CALCLENS
Spherical harmonic transform Poisson solver (Becker 2013)Ray-tracing on nside=8192 healpix gridCalculate shear, convergence for all galaxies
�0.4
�0.2
0.0
0.2
0.4
0.2 z < 0.43
⇠+⇠�
10�1 100 101 102�0.4
�0.2
0.0
0.2
0.4
0.63 z < 0.9
DES Y1 ⇠+ scale cut
DES Y1 ⇠� scale cut
0.0 0.2 0.4 0.6 0.8 1.0
✓[arcmin]
0.0
0.2
0.4
0.6
0.8
1.0
�⇠/
⇠hal
ofit
JDR et al 2019
-
Realistic Observables: Lens Galaxies
0.0
2.5DES Y1
Buzzard Mean
0.0
2.5
0.0
2.5
✓w(✓
)0.0
2.5
101 102
✓[arcmin]
0.0
2.5
0.5
1.0
1.5
hci
g � r
DES Y1
Buzzard
r � i i � z
0 10.0
0.2
�(c
)
0 1z�
0 1
Robust red-sequence allows high fidelity redMaGiC sample selection
-
Realistic Observables: Source Galaxies
Metacalibration like sample selected with similar S/N properties as data.BPZ run on fluxes with Y1 like photometric errors.
101 102
✓[arcmin]
�1
0
1
104⇥
✓⇠+(✓
)
101 102
✓[arcmin]
0.0
2.5
101 102
✓[arcmin]
0.0
2.5
101 102
✓[arcmin]
0
5
DES Y1
Buzzard Y1 Mean
0.0 0.50
2
4
6
n(z
)[de
g�2 ]
0.0 0.50
2
4
6
0 1z
0.0
2.5
5.0
n(z
)[de
g�2 ]
DES Y1
Buzzard BPZ
Buzzard True
0 1 2z
0
1
2
-
Buzzard sims used in a 11/14 of “DES Y1 Results”
Gatti, Vielzeuf et al.Hoyle et al.
Chang et al. 2018
MacCrann, JDR et al. 2018
Gruen, Friedrich, Krause, JDR et al.Friedrich, Gruen, JDR, Krause et al.
Density Split Statistics
Redshift Estimation 3x2pt Parameter Inference
Mass Mapping
-
Highlight: Validating the 3x2pt Pipeline
MacCrann, JDR et al. 2018
Constrained biases on inference to
-
Highlight: Validating the 3x2pt Pipeline
MacCrann, JDR et al. 2018
Results corroborated from an independent simulation.
-
Ongoing/Future Work
-
DES Y3
Many new things coming for DES Y3 3x2pt:Source redshift calibration using self-organizing maps with a combination of many band photo-zs and spectroscopyShear-ratio and clustering redshift informationHigher order bias modeling Higher order IA modelingSmall scale GGL non-locality mitigationLens magnification modelingMultiple lens samples
-
DES Y3
Many new things coming for DES Y3 3x2pt:Source redshift calibration using self-organizing maps with a combination of many band photo-zs and spectroscopyShear-ratio and clustering redshift informationHigher order bias models to mitigate scale dependent bias effects Higher order IA modelsSmall scale GGL non-locality mitigationLens magnification modelingMultiple lens samples
Currently being tested using updated suite of 70 Buzzard simulations(DeRose + DES Collab. in prep.)
-
DESI Lensing Mock Challenge
HSC (~630 deg2)
HSC (~90 deg2)
HSC (~680 deg2)
KiDS (~750 deg2) DES (~500 deg2)
Figure from 1611.00036 w/ additions by C. Blake
-
DESI Lensing Mock Challenge
Stage 0: DESI lenses and generic source catalogues (e.g.compare measurement codes, photo-z dilution corrections for \Delta\Sigma)
Stage 1: DESI lenses and source catalogues tailored to specific lensing surveys (HSC, KiDS, DES), for tests of systematics related to sources (e.g. photo-z dilution, “boost” correction, multiplicative shape calibration) (We are here)
Stage 2: Tests of systematics related to lenses (e.g. optimal lens weights, fibre collisions, redshift incompleteness, inhomogeneity in lens catalogues, lens systematic weights)
Stage 3: Test cosmological fitting pipeline (e.g. combined-probe covariance, modelling)
Ongoing work by Chris Blake, JDR, Johannes Lange, Alexie Leauthaud, Suhkdeep Singh, Ji Yao, and more
-
e.g. KiDS-like Mock
Figs. from C. Blake
-
Summary
Cosmological Simulations are essential for testing nearly all facets of current galaxy surveysWe have designed an algorithm that allows us to produce realistic suites of galaxy catalogs including:
Non-linear biasLensingPhotometric errorsPhoto-zs
We have used these in DES to test our Y1 3x2pt analysis and ongoing Y3 analysesThese will also be used for lensing work in DESI
Thanks!