We present deep 450 \mu m and 850 \mu m observations of a large , uniformly covered 394 arcmin ^ { 2 } area in the COSMOS field obtained with the Scuba-2 instrument on the James Clerk Maxwell Telescope ( JCMT ) .
We achieve root-mean-square noise values of \sigma _ { 450 } =4.13 mJy and \sigma _ { 850 } =0.80 mJy .
The differential and cumulative number counts are presented and compared to similar previous works .
Individual point sources are identified at > 3.6 \sigma significance , a threshold corresponding to a 3–5 % sample contamination rate .
We identify 78 sources at 450 \mu m and 99 at 850 \mu m , with flux densities S _ { 450 } = 13–37 mJy and S _ { 850 } = 2–16 mJy .
Only 62–76 % of 450 \mu m sources are 850 \mu m detected and 61–81 % of 850 \mu m sources are 450 \mu m detected .
The positional uncertainties at 450 \mu m are small ( 1–2.5″ ) and therefore allow a precise identification of multiwavelength counterparts without reliance on detection at 24 \mu m or radio wavelengths ; we find that only 44 % of 450 \mu m-sources and 60 % of 850 \mu m-sources have 24 \mu m or radio counterparts .
450 \mu m-selected galaxies peak at \langle z \rangle = 1.95 \pm 0.19 and 850 \mu m-selected galaxies peak at \langle z \rangle = 2.16 \pm 0.11 .
The two samples occupy similar parameter space in redshift and luminosity , while their median SED peak wavelengths differ by \sim 20–50 \mu m ( translating to \Delta T _ { dust } = 8 - 12 K , where 450 \mu m-selected galaxies are warmer ) .
The similarities of the 450 \mu m and 850 \mu m populations , yet lack of direct overlap between them , suggests that submillimeter surveys conducted at any single far-infrared wavelength will be significantly incomplete ( \buildrel > \over { \sim } 30 % ) at censusing infrared-luminous star formation at high- z .