Selected PS1 data for USNO (2016 August 22)

Selected regions from current PSPS database that might be usable for testing in astrometry processing by USNO. These objects were selected by searching the Detection table for objects that have more than 30 entries. There are two spots in the sky that have such data. Then those lists were expanded to include all objects in the immediate area.

NOTE: I recommend that you remove single-epoch objects (ObjectThin.nDetections=1) before working with this data. Those objects are mainly spurious. I have included them here for completeness but we generally will recommend that they not be used except in unusual cases. For objects this bright, there should be very few real objects with small numbers for nDetections. Removal of objects with nDetections=2 may also be helpful; you should experiment with using nDetections as a filter to reduce contamination by spurious objects.

Data

Here are the files:

• tabc-objid.csv (2.7 MB) is a simple text file with a list of the object IDs. This has 153,881 rows.
• tabc_ObjectThin_rwhite.fit (38 MB) is the ObjectThin table, which has information on the positions, number of detections, detections in each filter, etc. Basically this has all the summary information on the objects except for magnitudes. Also has 153,881 rows.
• tabc_MeanObject_rwhite.fit (65 MB) is the MeanObject table, which has information on mean magnitudes in the various filters and using various algorithms (PSF fitting, aperture photometry, Kron photometry, etc.) Also has 153,881 rows.
• tabc_Detection_rwhite.fit (684 MB) has the information on detections at for each epoch. It has multiple entries for each object; the table has 1,810,873 rows, for an average of almost 12 detections for each object. If the Detection table were complete, this would have as many detections for an object as the nDetections field in ObjectThin. But for the current database, on average only about 25% of the detections appear in the table.

Here are PDF versions of the database schema descriptions for these tables:

• schema-objectthin.pdf
• schema-meanobject.pdf
• schema-detection.pdf

How the sample was defined

To define the sample, I started with all the objects in the Detection table as of 2016 August 18 that have at least one magnitude in the range 15<mag_AB<18. Note that I do not require all the magnitudes to be in that range, and I also do not discriminate based on filter (grizy). There are 1.89M unique objects in that magnitude range in the Detection table, and the table has 12.0M measurements of those objects (an average of 6.3 measurements per object). The RA and Dec positions for those objects are shown in Figure 1.

Figure 1: Positions of detections in the magnitude range.

If the Detection table was complete for the objects that are currently in it, it would have nDetections entries for each object, where nDetections comes from the ObjectThin table that summarizes the properties of the object. But Figure 2 shows that there are many fewer entries in the Detection table than expected. The black histogram is the distribution of nDetections for these objects: it peaks at about 70 detections/object. The red histogram is the distribution of the actual number of rows in Detection for these objects: it has much smaller values. The current Detection table has only about 12% of all the expected detections for these objects. So not only is the table limited to a small part of the sky (Figure 1), but it also has only a small subset of the detections in those regions.

Figure 2: Comparison of ObjectThin nDetections distribution with Detection table count.

However, there are some regions that have better sampling of the detections. Notice the hump on the red histogram in Figure 2 that extends from about 20 to 40 entries. Those come from two regions of the sky that have a larger fraction of the detections. Looking back at Figure 1, the red points are objects that have more than 30 detections. They are concentrated at 2 positions (RA=115, Dec=+41 and RA=248, Dec=−7). Figure 3 shows the objects in the vicinity of those positions, with the points color-coded by the number of entries in the Detection table (green to red = many detections, blue to cyan = few detections). The core of high detections is visible in this plot in yellow and red.

Figure 3: Regions with more than 30 Detection table entries.

To create the sample of objects included in the data linked at the top of this page, I did this:

1. Selected the objects with more than 30 entries in the Detection table (11,402 objects).
2. Expanded the region by selecting all objects with more than 15 entries and within 2 degrees of the >30 entry sample (57,514 objects).
3. Expanded the sample further by selecting all objects within 0.2 degrees of the >15 entry sample, regardless of the number of detections (153,881 objects)
4. Extracted from the PS database the data for all 153,881 objects along with all the Detection entries for those objects (1,810,873 measurements).

Figure 4: Objects included in the sample dataset.

With these cuts, we are left with the objects shown in Figure 4. The halo of objects with few entries in Detection (the dark blue region) has mostly been removed. Over these small areas the sample retains all the objects currently in the Detection tables. On average each object has 11.8 entries in the table; and on average 26.6% of the expected detections are present. This should be a more useful sample for experimentation although it is certainly not complete.

The last plot, Figure 5, shows the distribution of the ObjectThin nDetection values (black histogram) with the actual number of entries in Detection (red histogram). The red histogram definitely shows a larger number of entries in the Detection table, as desired.

Figure 5: Comparison of ObjectThin nDetections distribution with Detection table count for the final sample.

---

Richard L. White, rlw@stsci.edu
2016 August 23