Report for Simulation #1

compiled by Heidi Newberg

Information on the simulation

The simulation covers 9 sq. deg in the direction of Northern Galactic Pole down to a limiting magnitude i'<19.5. Xiaohui Fan produced the simulation of four classes of objects : 1. quasars, from a simulation by Don Schneider (212 in total); 2. normal stars, using a population synthesis+galactic struture code developed by Y.Ng (Padova), including the young, old, thick disk, halo and bulge populations. (13116 in total); 3. white dwarf stars, assuming a galatic distribution as an exponential disk with 250 pc scale height, and the colors calculated from white dwarf atomsphere models of Bergeron (16 in total); 4. narrow emission line galaxies, using model colors and statistics based on the spectroscopic observation of the Schneider-Schmidt-Gunn sample (58 in total). The magnitudes of the objects are then reddened according to Burstein-Heiles extinction map, and proper photometric errors added accoring to the Black Book numbers. The proper motion of each galactic object is calculated assuming a simple galactic kinematic model, and then random errors are added for all objects (the proper motions are not currently used in target selection, however). More detailed information can be found in two documents written by Fan. The first one describes the star, white dwarf and quasar simulations (quite out of date, the quasar simulation is slightly different from the Schneider's simulation used here, but gives very consistent results); the second one describes the emission-line galaxy simulation. Figures are missing in both documents.

The Tunable Input Parameters

Xiaohui's simulations were converted to survey formats and run through the pipeline by Heidi Newberg. The only parameters in the quasar input parameters file that are currently used by the program are: 
CAPILIMIT	= 20.0
SKIRTILIMIT	= 18.0
and the locus files as specified by: 
LOCUSFILEDIR /data/dp1.17/data/qso
CAPLOCUSMAGS -30 15 16 17 18 19 30
SKIRTLOCUSMAGS -30 30
The *LOCUSMAGS lists give the breakpoints for using different locus fits. In this case, there is only one locus for the skirt (it is not used since the target area doesn't go into the skirt). There are 6 locus fits for the cap, corresponding to the i' magnitude ranges: -30 < i' < 15, 15 < i' < 16, 16 < i' < 17, 17 < i' < 18, 18 < i' < 19, and 19 < i' < 30. The blue (and I guess red) endpoints and the number of sigma are set separately in each of the locus files. In this case I used 4 sigma for all magnitude ranges. The locus parameter files also contain the entire description of each locus fit.

All of the locus fits were automatically generated with r_start = {0.8, 0.1, -0.15}; r_end = {2.75, 1.3, 0.6}; d = {0.2, 0.2, 0.2}; Niter = 50; and N_sigma_spacing = 3. See Newberg and Yanny (1997) for definitions of the input parameters. The entire simulation was used to fit the loci.

The other parameters may eventually be used for FIRST target selection, high redshift target selection, point source discrimination, and identification of regions in which target selection failed.

Results of the Simulation

The quasar target selection pipeline selected 334 out of the 13402 objects in the simulation. The figure of merit was: 

Targets selected (by type):     1 (young disk) 	13/991
                                2 (thick disk) 	35/9801
                                3 (bulge/halo) 	5/2324
                                4       	0/0
                                5       	0/0
                                6       	0/0
                                7       	0/0
                                8 NELG      	58/58
                                9 WD     	14/16
                                10 QSO     	209/212
                                        	-----------------
                                        	334/13402

QSO completeness (z):           -10 - 1 85/85
                                1 - 1.5 48/48
                                1.5 - 2 32/32
                                2 - 2.5 24/24
                                2.5 - 3 15/16
                                3 - 3.5 5/6
                                3.5 - 4 0/1
                                4 - 6   0/0
                                6 - 10  0/0
                                        -----------------
                                        209/212

QSO completeness (m):           -10 - 15        0/0
                                15 - 16 1/1
                                16 - 17 1/1
                                17 - 18 20/20
                                18 - 19 91/92
                                19 - 30 96/98
                                        -----------------
                                        209/212
The three QSOs which were not selected are: 
u	g	r	i	z	redshift

22.387 	20.285 	19.483 	19.346 	19.136 	3.617
19.974 	18.879 	18.694 	18.579 	18.478 	2.775
22.418 	19.989 	19.346 	19.164 	18.963 	3.491
It is possible to vary the input parameters to increase the completeness at the expense of efficiency. The most obvious parameter to vary is the number of sigma away from the center of the stellar locus to exclude from target selection. The results of varying this parameter are as follows: 
# sigma		efficiency		completeness

5		208/321 = 0.65		208/212 = 0.981
4		209/334 = 0.63		209/212 = 0.986
3		210/492 = 0.43		210/212 = 0.991
2.5		211/938 = 0.22		211/212 = 0.995
2		212/2123 = 0.10		212/212 = 1.000

While it is true that the completeness does not change much as a function
of the width of the locus, note that the QSOs missed are (relatively)
high redshift, and thus worth some effort to dig out.

I have generated the following color plots of the output.  Objects which
were selected by the target by the current algorithm are circled.  The
locus fit for each magnitude range is shown as a black line.