Extended SRC Database and Final Catalog Generation

The 2MASS extended source database (ESDB) contains information derived for sources deemed "extended" by the 2MASS GALWORKS subprocessor. The criteria toward this end (discriminating between point sources and extended sources) are slightly relaxed in order to error on the side of extended source completeness. The ESDB, therefore, contains a mix of real galaxies, real galactic fuzzies (e.g., nebulae, H II regions, etc), stars, double stars, triple stars, and "artifacts" (generally pieces of meteor streaks, but they also comprise pieces of bright stars). It is the job of the final product generator to cull the ESDB of "false" extended sources, leaving a final catalog that is 99% reliable for most of the sky.

For a description of the level-1 (final catalog) specifications for extended sources, see T. Chester's memo: Brief Summary of 2MASS Facts or simply, SURVEY LEVEL 1 REQUIREMENTS .

At this time, both the northern and southern RTB scans (most of) have been fully processed and analysized, so we now have a very good idea what is going into the ESDB. The tables and plots below give a high level summary of what is in the ESDB. We will first concentrate upon the high GLAT fields (low stellar number density), which are relevant to the level-1 specs. The last section of this memo addresses the issue of how we may build a "final product generator" to construct a clean ES catalog. The first scheduled release of 2MASS ES data is in late March of 1999.

TChester has sketched a plan toward building & verifying a final product generator (generalized to both the point and extended src cats) as well as the documentation that goes with it (the mini-supplement). See Final Product Generator Tasks and The Mini-Explanatory Supplement.

Summary of the Extended Source Database C & R

Low Stellar Density Scans

That is to say, the starcount density at Ks <=14 of less than 3.1 in the log. The following table breaks down the classification results for all sources going into the ESDB (RTB fields), covering over 250 sq. degrees and ~14,000 total sources. These are the raw numbers -- we have not applied any limits to cull false sources, except that we have eliminated the duplicate observations (dupes occur at both inscan and cross-scan overlaps). See Appendix for a table that includes dupes.

Classification Results for All Sources Going into the ESDB (no dupes)

J-	J+	nG	nBog	reliab	nS	nD	nA	ntr	nU	nG/mgdg^2	elow	ehigh
5.00	9.00	0	1	.00	1	0	0	0	0	.000	.000	.000
9.00	10.00	0	4	.00	3	1	0	0	0	.000	.000	.000
10.00	11.00	3	29	.09	2	27	0	0	0	.011	.005	.017
11.00	12.00	52	47	.53	0	45	0	2	0	.185	.160	.211
12.00	13.00	227	119	.66	1	109	1	8	0	.809	.756	.863
13.00	13.50	417	135	.76	6	116	5	8	1	2.973	2.828	3.119
13.50	14.00	853	238	.78	15	202	5	16	7	6.082	5.874	6.290
14.00	14.50	1830	418	.81	70	310	12	26	48	13.048	12.743	13.353
14.50	15.00	3388	1028	.77	413	531	44	40	327	24.157	23.742	24.572
15.00	15.50	2571	488	.84	286	169	22	11	459	18.332	17.970	18.693
15.50	16.00	309	34	.90	17	8	9	0	90	2.203	2.078	2.329
-	-	-	-	-	-	-	-	-	-	-	-	-
-	-	-	-	-	-	-	-	-	-	-	-	-
H-	H+	nG	nBog	reliab	nS	nD	nA	ntr	nU	nG/mgdg^2	elow	ehigh
5.00	9.00	0	1	.00	1	0	0	0	0	.000	.000	.000
9.00	10.00	1	12	.08	4	8	0	0	0	.004	.000	.007
10.00	11.00	32	37	.46	0	36	0	1	0	.114	.094	.134
11.00	12.00	174	68	.72	0	64	0	4	0	.620	.573	.667
12.00	13.00	819	205	.80	4	190	0	11	2	2.920	2.818	3.022
13.00	13.50	1204	242	.83	11	213	6	12	23	8.585	8.337	8.832
13.50	14.00	2512	498	.83	110	331	25	32	93	17.911	17.554	18.268
14.00	14.50	3889	1071	.78	501	499	34	37	583	27.729	27.284	28.174
14.50	15.00	975	368	.73	188	151	17	12	233	6.952	6.729	7.175
15.00	15.50	77	61	.56	27	25	9	0	27	.549	.486	.612
15.50	16.00	11	24	.31	10	7	6	1	10	.078	.055	.102
-	-	-	-	-	-	-	-	-	-	-	-	-
-	-	-	-	-	-	-	-	-	-	-	-	-
K-	K+	nG	nBog	reliab	nS	nD	nA	ntr	nU	nG/mgdg^2	elow	ehigh
5.00	9.00	0	1	.00	1	0	0	0	0	.000	.000	.000
9.00	10.00	5	13	.28	4	9	0	0	0	.018	.010	.026
10.00	11.00	55	39	.59	0	38	0	1	0	.196	.170	.223
11.00	12.00	278	82	.77	0	78	0	4	0	.991	.932	1.051
12.00	13.00	1551	255	.86	8	220	10	17	12	5.529	5.389	5.670
13.00	13.50	2370	357	.87	46	269	24	18	73	16.898	16.551	17.246
13.50	14.00	4256	864	.83	365	429	36	34	539	30.346	29.881	30.811
14.00	14.50	1079	791	.58	388	361	15	27	314	7.693	7.459	7.928
14.50	15.00	95	225	.30	124	85	10	6	43	.677	.608	.747
15.00	15.50	7	74	.09	34	34	4	2	8	.050	.031	.069
15.50	16.00	5	25	.17	11	8	5	1	3	.036	.020	.052

Table Key

nG == number of galaxies
nBog == total number of false detections (non galaxies)
reliab == raw reliability
nS == number of single stars
nD == number of double stars
nA == number of artifacts
nU == number of unknown objects
NG/mag deg^2 == galaxy number counts (per mag per deg^2)

For the level-1 spec, we want to generate a catalog that is 90% complete and 99% reliable. The spec refers to galaxies brighter than Ks = 13.5, H < 14.3 and J < 15.0, for which the "sh" value is larger than 0.5 (see the small print in the lev-1 spec doc; link given above). The "sh" limit is designed to limit consideration to only sources that are extended beyond the PSF (that is, to the ability of measuring a significant difference between the radial extent of a galaxy and the PSF). A raw "sh value" of 0.5 corresponds to a "sh score" of ~10.0. The latter parameter is what is written to the ESDB. The following table shows the classification breakdown (see table 1, above) with a "sh" score limit of 10.0.

Classification Results "sh" score >= 10.0

J-	J+	nG	nBog	reliab	nS	nD	nA	ntr	nU	nG/(mgdg^2	elow	ehigh
5.00	9.00	0	0	.00	0	0	0	0	0	.000	.000	.000
9.00	10.00	0	0	.00	0	0	0	0	0	.000	.000	.000
10.00	11.00	3	9	.25	0	9	0	0	0	.011	.005	.017
11.00	12.00	51	16	.76	0	15	0	1	0	.182	.156	.207
12.00	13.00	226	47	.83	1	41	1	4	0	.806	.752	.859
13.00	13.50	416	51	.89	1	41	5	4	0	2.966	2.821	3.112
13.50	14.00	838	86	.91	1	69	3	13	2	5.975	5.769	6.181
14.00	14.50	1775	145	.92	4	114	7	20	26	12.656	12.356	12.956
14.50	15.00	3101	356	.90	48	242	36	30	211	22.111	21.713	22.508
15.00	15.50	2139	157	.93	53	76	21	7	268	15.251	14.922	15.581
15.50	16.00	163	13	.93	2	3	8	0	39	1.162	1.071	1.253
-	-	-	-	-	-	-	-	-	-	-	-	-
-	-	-	-	-	-	-	-	-	-	-	-	-
H-	H+	nG	nBog	reliab	nS	nD	nA	ntr	nU	nG/(mgdg^2	elow	ehigh
5.00	9.00	0	0	.00	0	0	0	0	0	.000	.000	.000
9.00	10.00	1	1	.50	0	1	0	0	0	.004	.000	.007
10.00	11.00	32	11	.74	0	11	0	0	0	.114	.094	.134
11.00	12.00	173	21	.89	0	19	0	2	0	.617	.570	.664
12.00	13.00	807	76	.91	2	71	0	3	1	2.877	2.776	2.978
13.00	13.50	1156	85	.93	1	71	5	8	12	8.242	8.000	8.485
13.50	14.00	2274	173	.93	19	114	19	21	56	16.214	15.874	16.554
14.00	14.50	3237	380	.89	104	219	32	25	321	23.080	22.675	23.486
14.50	15.00	597	87	.87	19	48	14	6	98	4.257	4.082	4.431
15.00	15.50	15	14	.52	1	6	7	0	8	.107	.079	.135
15.50	16.00	1	0	1.00	0	0	0	0	1	.007	.000	.014
-	-	-	-	-	-	-	-	-	-	-	-	-
-	-	-	-	-	-	-	-	-	-	-	-	-
K-	K+	nG	nBog	reliab	nS	nD	nA	ntr	nU	nG/(mgdg^2	elow	ehigh
5.00	9.00	0	0	.00	0	0	0	0	0	.000	.000	.000
9.00	10.00	5	1	.83	0	1	0	0	0	.018	.010	.026
10.00	11.00	55	9	.86	0	9	0	0	0	.196	.170	.223
11.00	12.00	274	25	.92	0	22	0	3	0	.977	.918	1.036
12.00	13.00	1368	71	.95	2	52	10	7	6	4.877	4.745	5.009
13.00	13.50	1841	125	.94	6	87	23	9	33	13.127	12.821	13.432
13.50	14.00	2811	272	.91	68	152	31	21	245	20.043	19.665	20.421
14.00	14.50	490	195	.72	63	106	12	14	97	3.494	3.336	3.652
14.50	15.00	20	22	.48	5	14	3	0	5	.143	.111	.174
15.00	15.50	0	4	.00	0	2	2	0	0	.000	.000	.000
15.50	16.00	0	0	.00	0	0	0	0	0	.000	.000	.000

The reliablity now hovers between 75 and 95%. Just looking at the last half mag bin relevant to the spec (e.g., 13.0 < Ks < 13.5), the reliablity is around 90%. The true completeness is difficult to measure (one needs a "truth" set independently measured and verified), but based on repeatibility tests, the completeness is likely to be well above 95%. See Coma Repeatibility from 10 scans of 970521n and GALWORKS Performance on the Spiral-Rich Cluster Hercules and 2MASS Validation Fields .

We therefore require more stringent limits upon the star-galaxy discrimination parameters in the ESDB to cull out additional false galaxies to bring the relability up to >98%.

Methods Toward Producing a Reliable Extended Source Catalog

In order to cull the ESDB of unwanted double and triple stars, we need to look more carefully at the star-galaxy discrimination parameters. The following link shows the results for all sources going into the ESDB sans dupes.

Star - Galaxy Discrimination

It can be seen that the galaxies and doubles star are most easily separated with the "wsh" and with J-K color (or alternatively, JHK color score). The following tables summarize the C&R with a color-color criterion applied and with a combination of color-color and "wsh".

Classification Results "sh" score >= 10.0 and "color" score >= -0.25

J-	J+	nG	CJ	nBog	reliab	nS	nD	nA	ntr	nU	nG/(mgdg^2	elow	ehigh
5.00	9.00	0	.00	0	.00	0	0	0	0	0	.000	.000	.000
9.00	10.00	0	.00	0	.00	0	0	0	0	0	.000	.000	.000
10.00	11.00	3	1.00	2	.60	0	2	0	0	0	.011	.005	.017
11.00	12.00	51	1.00	6	.89	0	5	0	1	0	.182	.156	.207
12.00	13.00	226	1.00	18	.93	0	16	0	2	0	.806	.752	.859
13.00	13.50	416	1.00	22	.95	1	19	0	2	0	2.966	2.821	3.112
13.50	14.00	838	1.00	35	.96	0	27	1	7	2	5.975	5.769	6.181
14.00	14.50	1770	1.00	87	.95	3	70	5	9	24	12.620	12.320	12.920
14.50	15.00	3048	.98	196	.94	29	125	25	17	196	21.733	21.339	22.126
15.00	15.50	2090	.98	93	.96	28	45	14	6	243	14.902	14.576	15.228
15.50	16.00	158	.97	12	.93	2	3	7	0	39	1.127	1.037	1.216
-	-	-	-	-	-	-	-	-	-	-	-	-	-
H-	H+	nG	CH	nBog	reliab	nS	nD	nA	ntr	nU	nG/(mgdg^2	elow	ehigh
5.00	9.00	0	.00	0	.00	0	0	0	0	0	.000	.000	.000
9.00	10.00	1	1.00	0	1.00	0	0	0	0	0	.004	.000	.007
10.00	11.00	32	1.00	5	.86	0	5	0	0	0	.114	.094	.134
11.00	12.00	172	.99	7	.96	0	5	0	2	0	.613	.566	.660
12.00	13.00	807	1.00	34	.96	1	31	0	2	1	2.877	2.776	2.978
13.00	13.50	1155	1.00	45	.96	0	36	4	5	12	8.235	7.993	8.478
13.50	14.00	2265	1.00	104	.96	14	63	16	11	54	16.150	15.810	16.489
14.00	14.50	3175	.98	227	.93	60	128	25	14	302	22.638	22.236	23.040
14.50	15.00	567	.95	41	.93	8	21	9	3	85	4.043	3.873	4.213
15.00	15.50	12	.80	3	.80	0	0	3	0	8	.086	.061	.110
15.50	16.00	1	1.00	0	1.00	0	0	0	0	1	.007	.000	.014
-	-	-	-	-	-	-	-	-	-	-	-	-	-
K-	K+	nG	CK	nBog	reliab	nS	nD	nA	ntr	nU	nG/(mgdg^2	elow	ehigh
5.00	9.00	0	.00	0	.00	0	0	0	0	0	.000	.000	.000
9.00	10.00	5	1.00	0	1.00	0	0	0	0	0	.018	.010	.026
10.00	11.00	55	1.00	3	.95	0	3	0	0	0	.196	.170	.223
11.00	12.00	274	1.00	12	.96	0	10	0	2	0	.977	.918	1.036
12.00	13.00	1368	1.00	44	.97	1	28	10	5	6	4.877	4.745	5.009
13.00	13.50	1841	1.00	94	.95	6	59	22	7	33	13.127	12.821	13.432
13.50	14.00	2808	1.00	241	.92	68	127	29	17	242	20.021	19.644	20.399
14.00	14.50	458	.93	148	.76	60	70	10	8	97	3.266	3.113	3.418
14.50	15.00	15	.75	7	.68	2	5	0	0	2	.107	.079	.135
15.00	15.50	0	.00	1	.00	0	1	0	0	0	.000	.000	.000
15.50	16.00	0	.00	0	.00	0	0	0	0	0	.000	.000	.000

Notes:
CJ, CH, CK are the "completeness" measures for J,H and Ks, respectively. The completeness is by computing the ratio of galaxies per mag bin for "sh" score >=10 only versus "sh" score >=10 + color "score" >= -0.25.

"color" score = (J-Ks - 0.9) + ( (H-Ks)>0.3 ) * ( (H-Ks)-0.3 ),

which adds the color distance from the blue line, for all sources, to the distance from the point sources with the reddest H-Ks colors, only for those sources with redder H-Ks colors. For sources with H-Ks > 0.3, the color score reduces to:
color score = J-Ks + H-Ks - 1.2 = J + H - 2*Ks - 1.2

Classification Results for "sh" score >= 10.0 and "color" score + "wsh" >= 0.25

J-	J+	nG	CJ	nBog	reliab	nS	nD	nA	ntr	nU	nG/(mgdg^2	elow	ehigh
5.00	9.00	0	.00	0	.00	0	0	0	0	0	.000	.000	.000
9.00	10.00	0	.00	0	.00	0	0	0	0	0	.000	.000	.000
10.00	11.00	3	1.00	0	1.00	0	0	0	0	0	.011	.005	.017
11.00	12.00	51	1.00	2	.96	0	1	0	1	0	.182	.156	.207
12.00	13.00	226	1.00	1	1.00	0	1	0	0	0	.806	.752	.859
13.00	13.50	416	1.00	10	.98	1	5	2	2	0	2.966	2.821	3.112
13.50	14.00	838	1.00	19	.98	0	14	1	4	2	5.975	5.769	6.181
14.00	14.50	1775	1.00	85	.95	4	61	7	13	24	12.656	12.356	12.956
14.50	15.00	3099	1.00	250	.93	36	154	35	25	206	22.096	21.699	22.493
15.00	15.50	2135	1.00	119	.95	37	57	18	7	264	15.223	14.893	15.552
15.50	16.00	163	1.00	12	.93	2	3	7	0	39	1.162	1.071	1.253
-	-	-	-	-	-	-	-	-	-	-	-	-	-
H-	H+	nG	CH	nBog	reliab	nS	nD	nA	ntr	nU	nG/(mgdg^2	elow	ehigh
5.00	9.00	0	.00	0	.00	0	0	0	0	0	.000	.000	.000
9.00	10.00	1	1.00	0	1.00	0	0	0	0	0	.004	.000	.007
10.00	11.00	32	1.00	0	1.00	0	0	0	0	0	.114	.094	.134
11.00	12.00	173	1.00	0	1.00	0	0	0	0	0	.617	.570	.664
12.00	13.00	807	1.00	14	.98	1	11	0	2	1	2.877	2.776	2.978
13.00	13.50	1155	1.00	29	.98	0	22	4	3	12	8.235	7.993	8.478
13.50	14.00	2274	1.00	107	.96	17	57	19	14	55	16.214	15.874	16.554
14.00	14.50	3234	1.00	278	.92	81	146	31	20	318	23.059	22.653	23.464
14.50	15.00	596	1.00	61	.91	9	35	13	4	96	4.250	4.075	4.424
15.00	15.50	14	.93	10	.58	0	6	4	0	8	.100	.073	.127
15.50	16.00	1	1.00	0	1.00	0	0	0	0	1	.007	.000	.014
-	-	-	-	-	-	-	-	-	-	-	-	-	-
K-	K+	nG	CK	nBog	reliab	nS	nD	nA	ntr	nU	nG/(mgdg^2	elow	ehigh
5.00	9.00	0	.00	0	.00	0	0	0	0	0	.000	.000	.000
9.00	10.00	5	1.00	0	1.00	0	0	0	0	0	.018	.010	.026
10.00	11.00	55	1.00	0	1.00	0	0	0	0	0	.196	.170	.223
11.00	12.00	274	1.00	2	.99	0	2	0	0	0	.977	.918	1.036
12.00	13.00	1368	1.00	25	.98	1	10	10	4	6	4.877	4.745	5.009
13.00	13.50	1841	1.00	82	.96	6	49	22	5	33	13.127	12.821	13.432
13.50	14.00	2811	1.00	240	.92	68	122	31	19	244	20.043	19.665	20.421
14.00	14.50	489	1.00	163	.75	60	80	11	12	97	3.487	3.329	3.644
14.50	15.00	20	1.00	20	.50	5	12	3	0	5	.143	.111	.174
15.00	15.50	0	.00	4	.00	0	2	2	0	0	.000	.000	.000
15.50	16.00	0	.00	0	.00	0	0	0	0	0	.000	.000	.000

Thus a simple color-color cutoff plus "wsh" score results in a vast improvement to the reliablity without significantly affecting the internal completeness. The level-1 spec of 99% still remains elusive, however. The last couple % on R will require more sophisticated methods, either using the other star-galaxy parameters in some clever combination, or employing an OBDT method that uses all information in a complicated (non-intuitive) fashion.

Tasks to Produce the 2MASS Extended Source Catalog

• Apply the OBDT method to the RTB classification data and find the optimum combination of the star-galaxy discrimination parameters (including color)

• Develop algorithm to identify artifacts in the ESDB by searching coord position space and p.a. space.

• Decide how to eliminate and/or combine in-scan and cross-scan dupes

• Choose "shortform" fields for catalog (i.e., decide which ESDB fields should be highlighted the user)

• What to do with Algor 2 sources? (they do not apply to the lev-1 specs)

• What to do with Algor 3 sources? (they do not apply to the lev-1 specs)

• Cross-ID 2MASS galaxies with NED ?

The following table shows the classification breakdown for all sources in the ESDB (located along sites of low stellar number density), including duplicates (there should be between 10 and 15% dupes due cross and inscan coadd overlap).

Classification Results for All Sources Going into the ESDB (w/ dupes)