# COSMOS: Validation Report (FULL)

Master catalogue used: __master_catalogue_cosmos_20180207.fits__<br>

Number of rows: 2,572,347

<br>

Surveys included:<br>

​

| Survey | Telescope / Instrument  | Filters (detection band in bold)  | Location        |

|--------|-------------------------|:---------------------------------:|-----------------|

| 3D-HST+CANDELS | HST/WFC3        | F140W, F160W, F606W, F814W, F125W |dmu0_CANDELS-3D-HST |

| CFHTLS        | CFHT/Megacam     | u,*g',r',i',z'                    | dmu0_CFHTLS     |

| DECaLS        | DEC              | g,r,z                             | dmu0_DECaLS     |

| HSC-SSP       | Hyper Suprime    | g,r,i,z,y                         | dmu0_HSC        | 

| KIDS          | VLT OmegaCAM     | u,g,r,i                           | dmu0_KIDS       |  

| PanSTARRS-3SS | GPC1             | g,r,i,z,y                         | dmu0_PanSTARRS-3SS |     

| LAS           | UKIRT WFCAM      | Y,J,H,K                           | dmu0_UKIDSS-LAS |     

| CFHT-WIRDS    | CFHT/WIRCAM      | J,H,Ks                            | dmu0_CFHT-WIRDS |

## I. Caveats

### I.a. Magnitude errors 

At faint magnitudes (mag > 24), some surveys have very large errors on the magnitude. These objects may be unreliable for science puposes.<br>

This in includes __KIDS aperture and total__ magnitudes (at mag > 26),  __DECaLS aperture and total__ magnitudes (at m > 27), __PanSTARRS aperture and total__ magnitudes (at mag > 23), __HSC-SSP aperture and total__ magnitudes (at mag > 25; also the magnitude could be as faint as 60 with corresponding error of 10$^{13}$) and __CFHT-WIRDS aperture and total__ magnitudes (at mag > 27).<br>

​

<img src="help_plots/COSMOS_magVSmagerr_KIDS_u_mag_total.png" />

### I.b. Aperture corrections

In most of the case when comparing the aperture magnitudes between surveys, we observed a two peak distribution in the difference between the magnitudes ($\Delta_{mag} = mag_{survey1} - mag_{survey2}$). We have one peak around 0 for point-source objects, with a small spread. And a second peak at higher $\Delta_{mag}$ with a larger spread for extended objects; implying a different aperture correction between surveys for these objects.<br>

That means that galaxies will not have the same aperture magnitude in different surveys. <br>

​

In the griz bands, for bright sources, there is a two peaks distribution when comparing Pan-STARRS, DECaLS, HSC-SSP and KIDS aperture magnitues. Except when comparing KIDS and HSC-SSP magnitudes, $\Delta_{mag}$ is similar for point-sources and extended objects.<br>

​

<img src="help_plots/COSMOS_apcorrIssues_KIDS_r_aperture_-_DECam_r_aperture.png" />

​

## II. Flags

### II.a. Pan-STARRS aperture and total magnitudes

Few Pan-STARRS sources have exactly the same error (of <font color='blue'>0.0010860000038519502</font>) on the __aperture and total__ magnitudes in all the grizy bands. The corresponding aperture magnitude should not be trusted for these objects.<br>

​

<img src="help_plots/COSMOS_gpc1Issues_GPC1_g_mag_aperture.png" />

### II.c. Outliers

By comparing magnitude in the same band between different surveys, we can see that some magnitudes are significanlty different could not be trusted. <br>

The outliers are identified to have a large weighted magnitude difference (equivalent of the $chi^2$).

​

$$chi^2 = \frac{(mag_{1}-mag_{2})^2}{magerr_{1}^2 + magerr_{2}^2}$$ 

<br>

We used the 75th and 25th percentile to flagged the objects 5$\sigma$ away on the large values tail of the $chi^2$ ditribution. (__NB:__ bright sources tend to have their errors underestimated with values as low as $10^{-6}$, which is unrealistic. So to avoid high $chi^2$ due to unrealistic small errors, we clip the error to get a minimum value of 0.1% (i.e. all errors smaller then $10^{-3}$ are set to $10^{-3}$).)

<br><br>

$$outliers == [chi^2 >  (75th \;percentile + 3.2\times (75th \;percentile - 25th \;percentile))]$$

​

<img src="help_plots/COSMOS_outliers_SUPRIME_i_aperture_-_KIDS_i_aperture.png" />