12CO (3–2) High-Resolution Survey (COHRS) of the Galactic Plane: Complete Data Release

We present the full data release of the 12CO (3–2) High-Resolution Survey (COHRS), which has mapped the inner Galactic plane over the range of 9.°5 ≤ l ≤ 62.°3 and ∣b∣ ≤ 0.°5. COHRS has been carried out using the Heterodyne Array Receiver Program on the 15 m James Clerk Maxwell Telescope in Hawaii. The released data are smoothed to have a spatial resolution of 16.″6 and a velocity resolution of 0.635 km s−1, achieving a mean rms of ∼0.6 K on TA* . The COHRS data are useful for investigating detailed three-dimensional structures of individual molecular clouds and large-scale structures such as spiral arms in the Galactic plane. Furthermore, data from other available public surveys of different CO isotopologues and transitions with similar angular resolutions to this survey, such as FUGIN, SEDIGISM, and CHIMPS/CHIMPS2, allow studies of the physical properties of molecular clouds and comparison of their states. In this paper, we report further observations on the second release and improved data reduction since the original COHRS release. We discuss the characteristics of the COHRS data and present integrated-emission images and a position–velocity (PV) map of the region covered. The PV map shows a good match with spiral-arm traces from existing CO and H i surveys. We also obtain and compare integrated one-dimensional distributions of 12CO (1–0) and (3–2) and those of star-forming populations.


INTRODUCTION
Carbon monoxide (CO) is essential for tracing the molecular interstellar medium (ISM) and understanding its physical properties.Molecular hydrogen (H 2 ) is the most abundant molecule in the interstellar medium, but due to its symmetry and low molecular weight, there is a lack of transitions with energies that are excited under standard molecular ISM conditions with T ∼ 10-100 K.As the second most abundant molecule, CO is the best tracer of the bulk of the molecular ISM.Its ro-tational (J) transitions trace different conditions in the ISM.Cold gas preferentially emits mostly in the J = (1-0) line, and warmer, denser gas radiates more in the higher-J lines.Different CO isotopologues reflect opacity effects. 12CO is the most abundant CO isotopologue, which produces strong emission lines from most of the molecular gas in the ISM. 13 CO and C 18 O are less abundant and therefore fainter than 12 CO, but trace optically thinner emission in most cases of low excitation temperature.So 12 CO traces where the molecular gas is, while 13 CO and C 18 O provide a view of the optically thickest gas.Thus, observations of different CO isotoplogues and transitions are complementary.
While continuum tracers blend emission along the line of sight, the high spectral resolution and Doppler effect of CO line emission allow us to identify discrete features in velocity that map to distance, although there is kinematic ambiguity within the Solar circle in positionposition-velocity space.Also, CO is often used as a mass tracer, even though the emissivity per unit mass (that is, X-factor) is variable (Bolatto et al. 2013).Therefore, CO is useful for investigating detailed information about the morphological, physical, and kinematic properties of molecular clouds.
Molecular gas is predominantly distributed in the galactic disks of late-type galaxies, including the Milky Way.It traces large-scale galactic structures such as spiral arms, giant molecular clouds, and star-forming regions.Significant efforts have long been made to provide targeted surveys of the Galactic plane in several CO isotopologues and transitions by using a variety of facilities.Since the molecular gas in the Milky Way is centrally concentrated, CO surveys in the Milky Way usually focus on the first and fourth Galactic quadrants.Table 1 presents examples of Galactic CO surveys in which the first quadrant is within the scope of the survey.The early observations from the University of Massachusetts-Stony Brook (UMSB) survey (Sanders et al. 1986), the CfA survey (Dame et al. 2001), and the Boston University-Five College Radio Astronomy Observatory Galactic Ring Survey (Jackson et al. 2006) provide 12 CO (1-0) or 13 CO (1-0) emission data.In particular, since the CfA survey covers the entire Galactic plane, it is still important for understanding the largescale distribution of molecular gas, although the angular resolution of the survey data is low by modern standards.
More-recent surveys simultaneously observe multiple lines of CO isotopologues at moderate or much-higher angular resolutions.For example, there are the FOR-EST Unbiased Galactic plane Imaging survey (FUGIN; Umemoto et al. 2017); the Milky Way Imaging Scroll Painting (MWISP) I and II projects (e.g., see Su et al. 2019;Yuan et al. 2022); and the structure, excitation, and dynamics of the inner Galactic1 interstellar medium (SEDIGISM) survey (Schuller et al. 2017).In addition, two surveys that used the JCMT measure the (3-2) transition lines of three CO isotopo-logues ( 12 CO/ 13 CO/C 18 O).For 12 CO, the CO high-Resolution Survey (COHRS), of which the first release (R1) was presented by Dempsey et al. (2013, hereafter Paper I) and the second release (R2) is described in this paper, has been carried out.For 13 CO and C 18 O, the CO Heterodyne Inner Milky Way Plane Survey (CHIMPS; Rigby et al. 2016) mapped part of regions covered by the COHRS.The extension of this survey, CHIMPS2, is now proceeding to extend into lower galactic longitudes, and to cover the Galactic center and an outer region (Eden et al. 2020).These recent surveys provide significantly improved CO data in angular resolution compared with the CfA survey.Specifically, about 1000 times the area of the COHRS beam is equal to one CfA beam area.
COHRS observes 12 CO (3-2), which is more optically thin compared with the lower transition lines of the same isotopologue and is seen at a higher frequency, resulting in higher-resolution data given the same telescope diameter.Compared with 12 CO (1-0), 12 CO (3-2) is excited in a warmer (energy above ground (E u /k B ): 5.5 K for J = 1, 33 K for J = 3) and denser (critical density: ∼ 2 × 10 3 cm −3 for (1-0), ∼ 5 × 10 4 cm −3 for (3-2) in the optically thin regime) environment.This transition traces molecular clouds, particularly gas that is likely to be more strongly associated with star-forming regions.It is also an excellent tracer of outflow activity, which generally indicates the very early stages of star formation (Banerjee & Pudritz 2006).Using the COHRS R1 data, Li et al. (2018) established a catalog of high-mass outflows associated with ATLASGAL clumps, resulting in the detection rate of 22%.Colombo et al. (2019) analysed integrated properties of molecular clouds by applying the Spectral Clustering for Interstellar Molecular Emission Segmentation algorithm (Colombo et al. 2015).They identified 85,020 clouds and found that 35% of the classified clouds are located within spiral arms, assuming arm widths of 600 pc (Vallée 2017).They derived mass and size distributions showing the power-law relationship with spectral indices of −1.75 and −2.80, respectively, and the distributions are truncated at ∼ 3 × 10 6 M and ∼ 70 pc, respectively.
COHRS can be compared with other Galactic largescale surveys at submillimeter and infrared wavelengths to study detailed structures of individual star-forming regions and large-scale structures in the Galactic plane.The existing continuum surveys covering the first Galactic quadrant are, for example, the APEX Telescope Large Area Survey of the Galaxy at 870 µm (ATLAS-GAL; Schuller et al. 2009), JCMT Galactic Plane Survey at 850 µm (JPS; Moore et al. 2015;Eden et al. 2017), the Bolocam Galactic Plane Survey at 1.1 mm (BGPS;    Benjamin et al. 2003;Churchwell et al. 2009), and MIPSGAL at 24 and 70 µm (Carey et al. 2009).This paper presents the full COHRS data covering 52.8 square degrees, which is almost twice the first release, by completing the overall planned latitudes in a more extensive longitude range than the previous one.The R2 data are provided by mitigating off-position contamination mentioned in Paper I and improved data reduction.In Section 2, we explain the COHRS observations and the general data-reduction procedure.Section 3 provides the information of the second release of the full COHRS data and how to access them online.Section 4 describes the noise characteristics.We present integrated position-position or position-velocity maps and descriptions of one-dimensional distributions in Section 5. Section 6 shows examples of COHRS data for three star-forming regions.In Section 7, we analyze the one-dimensional distribution of COHRS data and compare it with those of other data, such as the lowest 12 CO transition and star-forming population.We summarize main results in Section 8.
2. OBSERVATIONS AND DATA REDUCTION

Observations
COHRS is a spectral line survey mapping a strip of the Galactic plane in the first quadrant in 12 CO (3-2) line (345.786GHz).The survey covers a total of approximately 52.8 square degrees in the region between 9. • 5 ≤ l ≤ 62. • 3 and |b| ≤ 0. • 5.The observations were carried out with the Heterodyne Array Receiver Program (HARP; Buckle et al. 2009) on the 15 m James Clerk Maxwell Telescope (JCMT) in Hawaii.
The target longitude range for the original survey was 10 • ≤ l ≤ 65 • .This was ultimately extended down to l = 9. • 5 to follow the interesting structure seen in R1 around l ∼ 10 • .The upper end of the range was truncated to l ∼ 62 • ; this decision was made in order to concentrate our remaining time maximizing the overlap between COHRS and the JPS, as well as re-observing the noisiest tiles from R1.We chose not to re-observe those tiles affected by contamination in the off-position.This decision was driven by time constraints and the desire to complete the survey area.We were confident that the effect of the contamination could be mitigated in post-processing.We ultimately approached the issue of off-position signals by determining modal spectra and removing them during data reduction (see Section 2.2.1 for details).
The observations were taken over four semesters at JCMT during 2013-2014 and 2017-2018.As in previous observations, the observing time had been allocated as a mixture of PI time which includes some of CHIMPS2 time (proposal ID: M17BL004), Directors Discretionary Time, empty-queue or poor-weather backup time (Panel for the Allocation of Telescope Time (PATT) numbers: M13AU41, M13BN02, and M14AU09).The data were collected in opacities ranging from τ 225 ∼ 0.06 to τ 225 ∼ 0.31.The observational strategies follow those described in Paper I.
HARP is a 4×4 array receiver with 16 superconductorinsulator-superconductor (SIS) heterodyne detectors2 arranged at intervals of 30 .At the observing frequency, HARP has an angular resolution of 14 and a main-beam efficiency of η mb = 0.61.The Auto-Correlation Spectral Imaging System (ACSIS; Buckle et al. 2009) was used as a setting for an 1 GHz bandwidth and 2048 frequency channels, providing a frequency resolution of 0.488 MHz or a velocity resolution of 0.42 km s −1 .The bandwidth was set to a velocity coverage of −400 km s −1 < v LSR < +400 km s −1 .The observations were taken in a position-switching raster (on-the-fly) mode with 1/2 array spacing.The bulk of the off-positions were measured above the Galactic plane with a latitude offset of +2.• 5.

Data Reduction
As with the original COHRS release, the observational data were reduced with the ORAC-DR pipeline software (Jenness & Economou 2015), specifically the REDUCE SCIENCE NARROWLINE recipe.This recipe invoked applications from the KAPPA (Currie & Berry 2013), SMURF (Chapin et al. 2013), and CU-PID (Berry et al. 2007) packages, all from the Starlink collection (Currie et al. 2014).While the applications called by the recipe were from the Starlink 2018A release3 , the ORAC-DR code included improvements made since that 2018A release4 to address specific survey requirements.Jenness et al. (2015) provided a detailed description of the workings of the heterodyne recipes in ORAC-DR at the time of its writing.
Since the R1 data processing was finished, the reduction recipe has undergone many improvements, yielding better-quality products.The highlights include flatfielding; masking of time intervals of spectra in Receptor5 H07 affected by correlated noise called ringing (Jenness et al. 2015); and automated removal of emission from the reference (also called off-position) spectrum that appear as absorption lines in the reduced spectra, which can bias baseline subtraction and affect flux measurement of the emission.Since the removal of offposition signal was not written before Jenness et al. (2015), we describe the procedures that we adopted in Section 2.2.1.
Reductions were performed at least twice, the first automated attempt permitted visual inspection to assign recipe parameters, the most important being the baseline and flat-field regions, whether or not there is off-position signal or ringing spectra to remove.The recipe parameters used to control RE-DUCE SCIENCE NARROWLINE are available at the COHRS repository. 6An annotated example parameter file is given in Appendix B.
In its quality-assurance (QA) phase, the recipe used statistics to reject outlier spectra arising from nonastronomical sources, be it alternating bright and dark spectral channels or deviant baselines, both transient and persistent.The former typically rejected about 0.2 per cent of COHRS spectra.Due to ringing in Receptor H07 in about 5 per cent of observations, 5-10 percent of the spectra therein were rejected.In about 65 per cent of cases one to three entire receptors were expunged because of highly non-linear baselines or overwhelming external noise sources.A small number of observations were further excluded because they failed to meet quality-assurance thresholds, such as a maximum permitted T sys .The QA parameters for COHRS are listed in Appendix A.
The second phase of the recipe converted the timeseries spectra into position-position-velocity (PPV) spectral cubes, grouping both components of the weave pattern (Buckle et al. 2009) and for some regions it incorporated data from multiple nights in order to yield a tolerable signal-to-noise ratio.PPV cubes were regridded to 6-arcsec spatial pixels, convolved with a 9arcsec Gaussian beam, resulting in 16.6-arcsec resolution.The released data were also re-gridded along the velocity axis such that three raw channels became two channels, resulting in ∆V = 0.635 km s −1 .As can be seen in Table 1, this higher resolution is more comparable to other surveys than the ∆V = 1.0 km s −1 of R1.This improvement will make structural analyses more effective.For example, the cloud catalog generated by Colombo et al. (2019) using the SCIMES catalog was limited in its ability to recover low-mass molecular clouds by the coarse velocity resolution.It also generates less channel-to-channel covariance than for a velocity width of an arbitrary round number.
In the second reduction run, only one iteration to refine baseline regions was necessary, aided by the chosen recipe parameters.R2 retained linear baseline fitting to avoid the creation of a small artificial dip across the emission regions excluded from the fitting process.Note that this contrasts with the fourth-order polynomial adopted by CHIMPS (Rigby et al. 2016).A firstorder polynomial proved sufficient for the vast majority of COHRS observations.
The receptor-to-receptor responses were removed using a variant of the Curtis et al. (2010) summation method with user-defined velocity limits in which the median intensity approximately exceeded 0.2 K.In under ten per cent of observations, mostly for l > 50 • , there was insufficient signal to determine a reliable flatfield; the errors in the response ratios being comparable or larger than typical ratios themselves.Normalisation was with respect to Receptor H05, except in the 14 cases where this receptor had failed QA, whereupon the reserve H10 became the reference receptor.On average H10 was 1.0% more sensitive than H05 for the COHRS data.
Mosaics of width 5 • were formed from groups of PPV cubes with the PICARD (Gibb et al. 2013) recipe MO-SAIC JCMT IMAGES in ORAC-DR-combined with task WCSMOSAIC from the KAPPA package.Spectral alignment was in the kinematic local standard of rest (LSRK).7It allocated pixel contributions using a Gaussian distribution with a full-width at half maximum of one pixel.The most-central cube in each mosaic was assigned to be the reference for the world coordinate system, so as to minimise the distortions mapping from the galactic coordinates to a rectilinear pixel array.Then the tiles that form R2, which abut their neighbors, were extracted from the mosaics.

Correction of off-position signal
A significant deficiency with Release 1 was the presence of absorption features in 72 per cent of observations due to the existence of emission in the off-position (also known as the reference) spectrum, itself still at relatively low galactic latitudes.For Release 2 we have attempted to remove all the detectable off-position features.
Hitherto, when circumstances make retrospective direct observations of the off position impossible, a com-mon approach to dealing with a source in the reference position is to interpolate across the locations of such absorption.However, should these interpolated locations overlap a narrow emission line, the emission line could be erroneously weakened, or even eradicated.Likewise if the emission is varying rapidly downward, mere interpolation may over-compensate for the reference signal.Since the off-position features appear in every spectrum of an observation, we concluded that a better approach would be to determine the modal spectrum over regions devoid of emission or have minimal flux, then interpolate.Further, as the receptors look at slightly different spatial locations, a modal spectrum should be derived for each receptor.The difference between each interpolated modal spectrum and its original modal spectrum derives an estimate of the reference signal for the corresponding receptor.
In outline, the off-position contamination mitigation operated as follows.For each receptor, the algorithm formed a pair of approximate reference spectra.The first of the pair was derived from the time-series cube where detected astronomical emission had been masked, whereas the other originated from the raw time series.
At first glance, using the emission-masked data ought to be sufficient, as it provides better discrimination between genuine dips arising from multiple-source emission at different velocities and ones due to off-position absorption lines.In practice, however, incomplete emission-line detection, due to noise dominating in the line wings, often left the off-position lines in steep-sided valleys, and as a result were underestimated -typically by 0.1-0.2K -the depth of these lines, thus left residual absorption lines.In contrast the unmasked spectrum offered better estimates of the depth of reference spectral lines.However, the unmasked modal spectrum sometimes had difficulties discriminating between weaker source emission and a reference line with spectrally extended emission.The aim of using both modal spectra was to combine their assets: locate the lines with the masked version, and determine the line strengths from the unmasked version.Then the algorithm refined each of the approximate reference spectra to exclude source emission and background, to form an estimated reference spectrum.
Since the estimated reference spectrum should have a flat baseline at zero, accurate removal of the baseline is desirable.In practice this proved difficult in the presence of source emission, which might be extended and weak over a wide velocity range.Our algorithm took an iterative approach of twice measuring and masking lines from the off-position then from sources.The line masking yielded better estimates of the background signal, which after subtraction led to better estimates of the line strengths, and an improved baseline.The derived reference spectrum is subtracted from every spectrum in the time-series cube.Details of the automated algorithm are provided in Appendix C.
The automated algorithm left no perceptible absorption feature for about a third of the lines, but over half of the lines were only partially corrected, mostly caused by the noise raising the subtracted base level, thus not quite removing all of the reference line.A typical residual was 0.04 K.The method was also less reliable for reference lines located where there was varying and much broader source emission.In about a tenth of cases the reference line was not removed at all or left a prominent line, albeit much weaker, but still could be several tenths of a kelvin deep in the most extreme cases.
For the cases where reference lines were still present after the automated filtering, an additional processing step was performed.This required the velocity limits of the residual reference lines to be supplied via a recipe parameter.First, these line regions were masked in the modal spectra for each receptor.Then a smooth function based on iterative approximate solutions to Laplace's Equation filled the gaps.The difference between the interpolated and original spectra then yielded estimates of the residual reference line.
Despite this further attempt, off-position lines stubbornly remained in seven observed sections of the survey.The cause was the presence of a sheet of emission at the velocity of each absorption line, where the sheet spanned the bulk of the spatial pixels.As a consequence, the median or modal spectrum was representative of the emission, rather than near or at the baseline.
To circumvent this obstacle, for each survey section we manually extracted a polygonal spatial area devoid of emission at the line velocity.For each area we computed the median spectrum.This was smoothed with a Gaussian kernel of 25 channels full width at half maximum to determine the residual baseline.The smoothing was barely affected by the off-position lines that only spanned a few channels.Subtraction of the smoothed spectrum from the original median spectrum resulted in a flat baseline at zero.Although in four cases, where the off-position line was located in strong emission, a small offset (ranging from 0.003 to 0.03 K) correction was applied to bring the neighboring baseline to zero.Spectral channels beyond the off-position absorption line were set to zero.The name of this estimated reference spectrum was passed to REDUCE SCIENCE NARROWLINE through a recipe parameter, so that it could be added to the reference spectrum formed by the previous method when the observations were reduced again.
We performed sanity checks of our off-position corrections by comparing the median spectra of the same overlapping regions of adjacent observations.In order to compare like with like, undefined spectra in either observation were masked in both regions, and the spectra were aligned along the spectral axis.Each overlap region typically contained 20,000 spectra.Besides giving confidence in the existing corrections, these revealed many weak off-position signatures, as evidenced by similar dips in one observation compared with all of its neighbors.For some overlaps, both neighbors exhibited a co-located residual absorption feature, thus the procedure required a few iterations.This was particularly evident for l = 22-33 • where the Aquila Rift gave rise to a common 8 km s −1 off-position line spanning a sequence of regions.These were corrected in 72 cases by using the second semi-automated method described earlier, usually with adjustments to the previously estimated line bounds.If that failed wholly or partially, the manual approach was adopted in 59 cases.For this final resort variance-weighted average displacements from the neighboring median spectra within the velocity limits of the off-position signal were used to form a spectrum to be subtracted.

DATA RELEASE 2
We provide 106 tiles in FITS format with a 0. • 5longitude width and a full latitude range (0. • 5 × 1. • 0 per one tile).All of them are perfectly contiguous.Each tile is named in the form of central Galactic coordinate values and suffixes indicating the nature of the file, e.g., COHRS 09p50 0p00.Exceptionally, the first tile with the phrase '09p50' in its name is in the range of l = 9. • 5 to 9. • 75 with a longitudinal width of 0. • 25.All intensities are in units of T * A .The corrected antenna temperatures T * A can be converted to main beam brightness temperatures T mb by dividing η mb = 0.61.Along the velocity axis, the data cubes in the second release have been cropped to a range of −200 km s −1 < v LSR < +300 km s −1 , which is extended compared with that of the first release (−30 km s −1 < v LSR < +155 km s −1 ).The R2 data cubes can be obtained online from the CANFAR data archive ( doi:10.11750/22.078).We note that the comparison plots for R1 and R2 are also stored in the same archive.
Figure 1 shows a histogram of the corrected antenna temperature (T * A ) for all voxels.The values can be modelled as a normal distribution with a mean value of 0.036 K and a standard deviation of 0.49 K.The distribution shows a strong positive tail and a relatively weak negative tail.While the former is primarily due to voxels containing 12 CO (3-2) emission, the latter is sig- Histogram of the corrected antenna temperature for all voxels in COHRS, displayed on a logarithmic scale.The bin width is 0.1 K.The red line shows a Gaussian fit to the distribution described by 1.14 × 10 9 exp(− 1 2 (T * A − 0.032) 2 /0.49 2 ).The inset show the same distribution on a linear scale.
nificantly affected by random noise fluctuations in voxels with much higher-than-average noise levels (e.g., many of them at |b| 0. • 3).The tile name and the mean rootmean-square (RMS) noise for each tile are tabulated in Table D1 of Appendix D.

NOISE CHARACTERIZATION
A histogram and two-dimensional map of RMS noise levels of the spectra across the COHRS areas are shown in Figures 2 and 3, respectively.The noise levels were measured by taking the standard deviation of a baseline over a velocity range, v LSR < −100 km s −1 or v LSR > +200 km s −1 , in which no astronomical signal is visible.The histogram peaks at about 0.3 K, close to the standard deviation of the normal distribution of all voxels shown in Figure 1, and contains a fat tail, mainly contributed from noisiest areas at relatively high latitudes (see Figure 3).The variation across the noise map results from the combined effect of weather conditions, observing elevations, and the number of HARP receptors included in the reduction.We found that 27% of the pixels have RMS noise levels < 0.4 K, 60% < 0.6 K, and 80% < 0.8 K.The mean and median values of noise levels are 0.6 K and 0.5 K, respectively.Figure 4 shows a histogram of the signal-to-noise ratio for all voxels indicating the noise is Gaussian (normal distribution).The positive tails in the log plot are due to real astronomical signals.
The average of the RMS noise values of all pixels for each tile is given in Table D1 of Appendix D. For tiles with the full latitude range of R1 shown in Table 3 of Paper I, the RMS noise level of those tiles in R2 is reduced to 13-55% (36% on average).As mentioned in Paper I, some of the final tiles contain maps observed on different nights and under different conditions, resulting in some significant changes in RMS noise levels in one tile.Therefore, in such a case, it is not appropriate to say that the average RMS is a representative value for an individual map.Instead, Figure 3 is helpful for visualizing the spatial noise distribution throughout the survey and within each tile.

Integrated position-position maps
Figure 5 shows channel-maps of 12 CO (3-2) emission with the velocity interval of 20 km s −1 from v LSR = −50 to 150 km s −1 but also an integrated map over the whole velocity range.
Most of 12 CO (3-2) emission in the ranges of l 10 • to 22 • and 38 • to 45 • appears in negative latitudes.Such a tendency between l = 12 • and 22 • was reported by Umemoto et al. (2017) using FUGIN 12 CO/ 13 CO/C 18 O (1-0) maps.They explained that this is a distance effect that occurs because the Sun is not located at the true Galactic midplane.The Sun is located slightly above the true Galactic midplane, and an estimated offset is ∼ 10-30 pc (e.g., Anderson et al. 2019;Karim, & Mamajek 2017, and references therein).Many of the previous studies to estimate the Sun's height used the star-counts methods.Recent studies by Su et al. (2016Su et al. ( , 2019) used a large-scale CO gas distribution to derive the position of the Sun, yielding ∼ 17 pc, which is similar to the median of the published estimates.For objects located  Histogram of the signal-to-noise ratio for all voxels in COHRS, displayed on a logarithmic scale.The bin width is 0.1 K.The blue line is a Gaussian with a width (standard deviation) of 1 centered at 0. The inset show the same distribution on a linear scale.
at the true Galactic midplane, those closer to the Sun will appear at higher negative latitudes while those at a large distance will converge to b = 0 • .On the other hand, the remaining 12 CO (3-2) emission regions are roughly distributed around b = 0 • .But we note that positive velocity channels contain 12 CO (3-2) components at two distances along line-of-sight, the near and far.Therefore, 12 CO (3-2) emissions originating from two very difference distance ranges are stacked up, and the distance effect of this discrepancy for each line-ofsight should be carefully considered.
Many prominent bright regions appear across the COHRS area while significant faint extended emission is detected.Some of the bright regions include well known star-forming regions, such as W31 (l = 10.• 3), W33 (12.• 8), W42 (25.• 3), W43 (30.• 8), W47 (37.• 6), W49A (43.• 2), and W51 (49.• 4).Among them, three massive and luminous star-forming regions, W43, W49A, and W51, are presented in Section 6.  and the 12 CO (3-2) distribution related to the H ii regions (see Figure 7).We extract the H ii region-related 12 CO (3-2) from the (d hii , d hii , 15) pixel bin at each H ii region location.d hii is the diameter of the H ii region given in the WISE catalog, and 15 pixels at velocity are arbitrarily chosen to have about 10 km s −1 .The PDF of H ii region-related 12 CO (3-2) has a lower peak, which is mainly contributed by noise, and a fatter positive tail, which indicates stronger CO signals, than the PDF of all 12 CO (3-2).8 show a double peak feature due to self-absorption at v LSR ∼ 7 km s −1 , which is stronger in the (1-0) line than in the (3-2).On the other hand, the two transitions of 13 CO have different velocity profiles, indicating that the two transitions trace different internal conditions.

Integrated position-velocity map
W51 is also a particularly prominent massive and luminous Galactic star-forming complex (M gas ∼ 1.2 × 10 6 M , L bol ∼ 4.68 × 10 6 L , Carpenter & Sanders 1998;Urquhart et al. 2014b).It is estimated to be at a distance of 5.4 kpc from the Sun (Sato et al. 2010) and located near the tangent point of the Sagittarius 12 CO profiles show a clear double peak due to self-absorption around v LSR = 65 km s −1 .The self-absorption feature is stronger in the (3-2) compared with the (1-0), indicating foreground gas is likely colder than the gas associated with the W51 complex.However, the self-absorption situation can occur in subthermal excitation, where the density is smaller than the effective critical density, even if the gas is not colder.Therefore, additional analysis is needed to understand the actual situation.The FUGIN survey mapped part of the first quadrant of the Galactic plane at the lowest rotational transition (1-0) of three CO isotopologues ( 12 CO/ 13 CO/C 18 O) with an angular resolution comparable to that of the COHRS survey (see also Table 1 for detailed survey information). 12CO (1-0) is a fundamental rotation transition expected to be excited even in the coldest and most diffuse molecular ISM.Its critical density at 10 K is ∼ 10 3 cm −3 while that of 12 CO (3-2) is ∼ 10 4 cm −3 .However, for 12 CO, radiative trapping causes the molecule to emit at a density an order of magnitude lower than its critical density.In any case, the (3-2) emission line is emitted in relatively warmer and denser ISM conditions than (1-0).
Figure 11 presents longitudinal (l), latitudinal (b), and velocity (v LSR ) distributions of normalized integrated intensity for COHRS 12 CO (3-2) (blue profiles) and FUGIN 12 CO (1-0) (black profiles).Each profile was obtained by integrating over the two orthogonal axes, and then the intensity was normalized to the peak in- Both l-profiles integrated over latitude and velocity for the two transitions tend to decrease in general above l = 30 • .It is because the distribution of molecular ISM is not uniform in the Galactic disk and the Sun is far from the Galactic center.First, most of the molecular ISM is concentrated in the inner Galactic disk, and the distance (∆d los ) from the Sun and the far side of the inner disk steeply decreases with increasing longitude (∆d los ∼ 6 kpc for two lines of sight, l = 60 • and 30 • , while ∆d los ∼ 2 kpc for two lines of sight, l = 30 • and 10 • ).In other words, more CO emission lines usually accumulate along the line of sight towards lower Galactic longitudes.Second, the molecular ISM is strongly associated with Galactic spiral arms and the arrangement of the spiral arms along with a line of sight affects the shape of the l-profiles (see Figure 6 for an example).Above l = 30 • , there are fewer spiral arms lying close to the Sun.On the other hand, l−profiles with l < 30 • show large fluctuations rather than a smooth change in profile strength.This is mainly caused by how many luminous velocity components (or GMCs) overlap in a line of sight.In comparison between the two transition profiles, the peak locations are generally equal to each other.The strengths of the two profiles are also similar in some longitudes (for example, at l ∼ 30.• 5 and 35 • ), but their distinction is clearly seen in many other longitudes.While (3-2) emission becomes more strongly peaked than (1-0) emission, for example, at l ∼ 13 • , 23. • 5, 24.• 5, 43 • , and 49.• 5, while the predominance of (1-0) emission appears mainly at weak peaks or between peaks.The three prominent star-forming regions men-  tioned in Section 6 are also well located at CO peaks.The comparison with such star-forming population will be discussed in the next section.
The b-profiles integrated across longitude and velocity for the two CO transitions have a shape close to a normal distribution since lots of CO-emission components are integrated over a wide (l, v LSR ) range.Least-square Gaussian fitting gives best fit functions of 0.940 exp(− 1 2 (b − 0.115) 2 /0.428 2 ) and 0.948 exp(− 1 2 (b − 0.105) 2 /0.365 2 ) for (1-0) and (3-2), respectively.They have a nearly equal peak position and a slightly-stronger negative wing than a positive one, while the normalized intensity of the (3-2) profile decreases more rapidly than that of the (1-0) profile.
As shown in the l-and b-profiles, the v LSR -profiles integrated over longitude and latitude for the two transitions look similar each other.In other words, the locations of peaks normalized by maximum intensity of each profile are almost the same.However, except for the two peaks at v LSR ∼ 50-60 km s −1 , (1-0) emission is always stronger than (3-2) emission.The difference is relatively more pronounced at v LSR ∼ 5-15 km s −1 , indicating the presence of more diffuse local emission in the (1-0).

12 CO Emission versus Star-forming Population
The detection of H ii regions is the clearest evidence for ongoing massive star formation.The WISE catalog provides the entire sky Galactic H ii regions identified using mid-infrared data.In the area where COHRS and FUGIN overlap (l = 10 • -50 • and b ≤ 0. • 5), we found 2179 WISE H ii regions were found, except for one source without radio data.These are all WISE objects in the area except for one source without radio data.On the other hand, as the densest parts within GMCs are where star formation can take place, dense molecular clumps can be at various early evolutionary stages of star formation, from starless to early embedded stages.The ATLASGAL compact-source catalog provides about 10,000 dense clumps in the range of |l| < 60 • and |b| < 1. • 5 (Contreras et al. 2013;Urquhart et al. 2014a) found by using submillimeter survey data.For about 8000 dense clumps of them in 5 • < |l| < 60 • , Urquhart et al. (2018) investigated their detailed properties including velocities and distances, luminosities, and masses and inferred evolutionary stages using mid-and far-infrared survey data.Their classification scheme divides clumps into four groups: massive star-forming (MSF) clumps, YSO-forming clumps, protostellar clumps, and starless or pre-stellar clumps.In the area where COHRS and FUGIN overlap, 2178 AT-LASGAL clumps with signs of star-formation (i.e., except for those classified as a quiescent phase or unclassified) were identified: including 455 MSF clumps, 1222 YSO-forming clumps, and 501 protostellar clumps.
Figure 12 shows almost the same normalized l-profiles as shown in Figure 11 for the star-forming populations of WISE H ii regions and ATLASGAL clumps together with FUGIN 12 CO (1-0) and COHRS 12 CO (3-2).The histograms of the star-forming populations are obtained by counting the number of each catalog source in the moving bin with a bin size of 1 • and a step size of 12 .These moving-bin histograms avoid bias due to a specific bin size.For statistical comparative analysis, the two CO-emission profiles are interpolated are using the IDL INTERPOL function to have a bin size (12 ) equal to that of the star-forming population profiles.In the given longitude range, all profiles reach their largest peak at the longitudes in the W43 direction (30.• 8).At the longitudes in the W51 direction (49.• 4), all of them also show a distinct peak.However, near the W49A (43.• 2) direction, the peak height decreases significantly with smoothing.At l = 49.• 5, 12 CO (3-2), WISE H ii regions, and ATLASGAL clumps show a huge excess compared to 12 CO (1-0), which can indicate a very high temperature of CO due to active star-forming activities, while at l = 30 • , all four distributions exhibit similar intensities.
In addition, a broad hump centered around l ∼ 24 • stands out.The distribution of 12 CO (3-2) shows three thin and sharp peaks, one of which coincides with the peak of the ATLASGAL clumps.However, the distributions of 12 CO (1-0) and WISE H ii regions are relatively smooth.The G24 • -hump is seen as a combination of WISE H ii regions/GMCs close to each other in the longitude direction as well as those in the same line of sight: for example, GMCs with massive star-forming activities such as G23.01−0.41 at ∼ 77 km s −1 , G23.44−0.18 at Figure 8. W43 star-forming region within the COHRS survey area.From the top left to the bottom right, each panel displays 12 CO (1-0) FUGIN data, 12 CO (3-2) COHRS data, 13 CO (1-0) FUGIN data, 13 CO (3-2) CHIMPS data, 8 µm GLIMPSE data, 850 µm JPS data, and examples of CO spectra, respectively.The CO maps are velocity-integrated over the vLSR range of (80, 110) km s −1 (Nguyen Luong et al. 2011), and the units on the intensity scale of the integrated main beam temperature are K km s −1 .The units on the intensity scale of the GLIMPSE and JPS data are MJy/sr and Jy/beam, respectively.The CO spectra are obtained at the position closest to the velocity-integrated 12 CO (3-2) emission peak.The offsets of 10K, 20K, and 30K to the spectra have been added for better visualization.   .Each profile is obtained by integrating across the two orthogonal axes.Note that b-and v-profiles are used data within 10 • ≤ l ≤ 50 • since available FUGIN data are limited to the Galactic longitudinal range.CO intensity is normalized to the peak value of each profile.The two CO l-profiles are smoothed to have a bin size of ∼ 60 while their b-and v-profiles have a bin size corresponding to a latitudinal pixel size and a velocity channel width of each CO data respectively.On the top panel, three vertical dotted lines are drawn to help locate three star-forming regions: W43 (l = 30.• 8), W49A (43.• 2), and W51 (49.• 4).∼ 100 km s −1 , and G25.38−0.18(W42) at ∼ 65 km s −1 (e.g., Brunthaler et al. 2009;Ohishi et al. 2012;Su et al. 2015;Dewangan et al. 2015).This incredibly rich line-of-sight is being targeted for the GASTON Galactic plane survey (Rigby et al. 2021).Interestingly, the distribution of WISE H ii regions shows a strong peak at about 12. • 5, but not the rest.The peak does not appear when counting only objects, which have a single measured velocity, marked in Figure 6.About half of the WISE H ii regions contributing to the peak do not have measured velocity information and most of them are classified as radio-quiet sources.This area might contain many old WISE H ii regions that have dispersed the molecular clouds they were formed in.
We estimated the line ratio of the two CO transitions, i.e., R 31 ≡ 12 CO (3-2)/ 12 CO (1-0).R 31 l-distribution with two different bin sizes is displayed in the bottom panel of Figure 12 and compared to the WISE H ii regions or ATLASGAL clumps.For R 31 , a gray profile (R 31,1 ) has the same 60 bin size as the CO l-profile in Figure 11, and a green profile (R 31,2 ) has the same 12 bin size as the other profiles in Figure 12.We find a median R 31 = 0.27, which is similar to the mean value of 0.31 for nearby galaxies (Leroy et al. 2022).Compared to the CO profiles normalized by maximum shown in the upper panels, the R 31 profile shows less dramatic variation except near the W51 direction (See Figure 13 also).At l = 48.• 2, a deep valley is visible, with the higher-longitude side increasing steeper than the lower one.At l = 49.• 4, W51's line of sight, the R 31 profile is peaked (R 31,2 = 0.50).
A scatterplot was drawn for each pair, as shown in Figure 13, and the Spearman correlation test was applied to evaluate the relationship of results.We used pymccorrelation8 of Privon et al. (2020), which is a Python implementation and expansion of a Monte Carlo error-analysis procedure described by Curran (2014).We computed the Spearman correlation coefficient (ρ) using 1000 bootstrapping iterations, and the median and 16/84 percentile ranges are listed in Table 2. ρ = +1 or −1 is a perfect positive or negative correlation while ρ = 0 is no correlation between the data.For the two CO transitions, ρ is 0.94.It is not surprising that they show a very strong positive correlation.There is also a positive correlation between CO and star-forming population.As the ATLASGAL clumps used in this paper were selected only for those that form stars, it is natural that there is a strong positive correlation between WISE and ATLASGAL (WISE-ATLASGAL).On the other hand, ATLASGAL clumps shows a stronger correlation with CO or R 31,2 than WISE H ii regions.That is because that the ATLASGAL catalog contains sources in earlier evolution stages than the WISE catalog, which are still deeply embedded in their natal molecular cloud.Comparing the relationship between the two transitions and ATLASGAL clumps, the COHRS-ATLASGAL correlation coefficient is greater than the FUGIN-ATLASGAL correlation coefficient.Also, the difference in correlation coefficient between COHRS-WISE and COHRS-ATLASGAL is larger than that between FUGIN-WISE and FUGIN-ATLASGAL.These are explained by 12 CO (3-2) being more sensitive to dense gas than 12 CO (1-0).In addition, the COHRS-ATLASGAL correlation is slightly stronger than ATLASGAL-WISE.Thus, the J = (3-2) transition is a better tracer of star-forming gas.

SUMMARY
We present the full data of COHRS, which is a survey mapping a region of the Galactic plane, covering 9. • 5 ≤ l ≤ 62. • 3 and |b| ≤ 0. • 5, in 12 CO (3-2) using HARP on the JCMT.Since the initial public release of Paper I, further observations have been made to reach the full scope of the survey, and improved data-reduction processes have been applied, including steps to mitigate off-position contamination effects.The COHRS data are publicly accessible at doi:10.11570/22.0078.The released data have an angular resolution of 16. 6 and a velocity resolution of 0.635 km s −1 with a velocity coverage of −200 km s −1 < v LSR < +300 km s −1 .The data are sampled on 6 pixels and achieve a mean RMS of 0.6 K.
We investigate integrated one-dimensional distribution of COHRS 12 CO (3-2) and compare with those of FUGIN 12 CO (1-0) or star-forming population (WISE H ii regions and ATLASGAL star-forming clumps).When comparing them in the integrated longitudinal space, the peak locations are generally similar to each other, but differences in peak intensity can be seen in Figure 12.Integrated l-distribution of 12 CO (1-0) (FUGIN; black), 12 CO (3-2) (COHRS; blue), Galactic H ii regions (WISE; red in the top and bottom panel), star-forming clumps (ATLASGAL; orange in the middle and bottom panel), and the line ratio R31 (gray and green in the bottom panel).Each profile is integrated over b ≤ 0. • 5 and −60 km s −1 < vLSR < +170 km s −1 .The star-forming clumps contain all except starless clumps among the ATLASGAL compact-source catalog, that is, massive star-forming clumps, young stellar object-forming clumps, and protostellar clumps.CO histograms are drawn in the same way as in Figure 11, but interpolated to have the same bin size (12 ) and abscissa values as those of H ii regions or star-forming clumps.For H ii regions and star-forming clumps, each histogram is obtained by counting the number of sources in a moving bin with a population-counting bin width of 1 • and a bin-moving interval of 12 .The green profile (R31,2) is the ratio of the two CO-transition profiles in the upper panels, but on an absolute scale that is not normalized by maximum.Likewise, the gray profile (R31,1) is the ratio on the absolute scale of the two CO profiles shown in Figure 11.Three vertical dotted lines are drawn to help locate three star-forming regions: W43 (l = 30.• 8), W49A (43.• 2), and W51 (49.• 4).
Figure 13.Scatter plots between two targets using the normalized histogram values shown in Figure 12: clockwise from top left, 12 CO (1-0) vs. 12 CO (3-2), WISE H ii regions (red) or ATLASGAL star-forming clumps (orange) vs. R31, WISE H ii regions vs. CO emission, and star-forming clumps vs. CO emission.In the two bottom panels, 12 CO (1-0) and 12 CO (3-2) are displayed with black and blue circles, respectively.A solid line is the best result of least-squares fit in a linear model using the IDL LINFIT procedure.many longitudes.For example, the distinct peak of l = 12.• 5, visible only in WISE H ii regions, suggests that old star-formation regions are distributed in the line of sight and the surrounding molecular gas has already blown away.All available pairs ( 12 CO (1-0) vs. 12 CO (3-2) and WISE H ii regions or ATLASGAL starforming clumps vs. R 31 or CO emission) represent a positive correlation.The relationship between 12 CO (3-2) and ATLASGAL clumps is slightly stronger than that between 12 CO (3-2) and WISE H ii regions, while the relationship between 12 CO (1-0) and the two starformation tracers is relatively similar.This can happen because the higher CO transition traces denser areas within molecular clouds and are more closely related to early star-formation stages.
The COHRS data will be complemented with existing and upcoming CO and continuum surveys to study statistical properties of molecular gases along the Galactic plane as well as detailed structures and properties of individual objects.These high-resolution data of molecular gas will also help to investigate outflow activities in star-forming regions.Methanol masers are an unambiguous indicator of massive star formation, and Green et al. (2010) and Breen et al. (2015) provide unbiased 6 GHz class II methanol maser surveys in the COHRS area.In future work, we will investigate outflow features toward massive star-formation regions where methanol masers are detected.
HIGHFREQ INTERFERENCE THRESH CLIP set the number of standard deviations at which to threshold the noise profile of raw spectra above its median level, in order to decide whether to reject spectra with high-frequency noise.
# # Flatfield receptors # FLATFIELD = 1 FLAT_METHOD = sum FLAT_REGIONS = 12.0: 25.2,27.2:29.1,35.7:41.5These parameters defined whether or not to flatfield (FLATFIELD); always with the summation method FLAT METHOD, which proved to be the most stable; and the list of velocity ranges over which to integrate the fluxes for each receptor.If FLATFIELD = 0, the subsequent flat-field parameters were ignored.
-Manual location # SUBTRACT_REF_SPECTRUM = 1 REF_SPECTRUM_COMBINE_REFPOS = 1 REF_SPECTRUM_REGIONS =-1.5:0.1,2.5:4.0,7.0:11.1 The reference (off) position for the majority of the observed regions contained emission that appears as absorption features in all spectra.When detected after inspection of the first-pass reductions PPV cubes, the removal techniques were enabled by switching on SUBTRACT REF SPECTRUM.We did not want any unnecessary modification of spectra where no evident offposition was visible.
An outline of the methods used can be found in Section 2.2.1.The first stanza defined parameters for the automated method.REF EMISSION MASK SOURCE used not only used the source-masked spectrum to locate the lines, but also the unmasked modal spectrum to determine the line strengths.The emission was located with the ClumpFind algorithm (Williams et al. 2011) applied in one dimension by FINDCLUMPS from the CU-PID (Berry et al. 2007)  In the seven cases where even the manual guidance did not remove all the absorption lines, the name of a manually determined off-position residual spectrum was supplied through REF SPECTRUM FILE (not shown above).The velocity limits of the PPV cubes were set by the first two recipe parameters.These limits were further trimmed during mosaic formation in order to prevent exceeding the maximum number of array elements.REBIN assigned velocity resolutions for re-gridded PPV cubes, generating one at the R2 width of 0.635 km s −1 , and the other at 1.0 km s −1 width for comparison with R1.
For completeness, the final set of parameters asked for the creation of a longitude-velocity (LV) map, summing over galactic latitude.These LV maps were for a quick inspections of the reductions, and do not form part of the release.The released LV maps were derived from the mosaics.The first two parameters restrict the velocity range when computing the moments map, and were used for efficiency.

C. AUTOMATED ALGORITHM FOR REMOVAL OF OFF-POSITION SIGNAL
This appendix expands on the outline, presented in Section 2.2.1, of the automated algorithm to remove offposition signals.
1. Data observed at different reference positions are processed separately.
2. The initial step is to collapse the time axis by forming the mode at each spectral channel.Those modal spectra are mildly smoothed with a 1.5channel full-width half-maximum Gaussian pointspread function in order to define the extents of the absorption lines better.The mode at each spectral channel was determined by an iterative maximum-likelihood function, for which the data were inversely weighted by their deviations from the current mean.At each iteration outliers at 3.0 standard deviations from the current mean were clipped.Iterations proceeded until convergence to a stationary point.
3. Refinement of the modal spectra occurs for each receptor as follows.
(a) The lines under analysis are always in emission, as required by the clump-finding software.
(b) Before the locations of reference-spectrum emission lines are determined within the masked-source modal spectrum, an attempt to remove residual source emission is made.Its steps are: subtract a 75-pixel mediansmoothed version, then mask channels that fall below a −3 * rms threshold, then repeat the first step but having the kernel reduced to 41 pixels.
(c) In the search for off-position lines the background is not initially subtracted.While this choice may lose weaker reference emission embedded in extended source signal, it compensates by not regarding dips in the source signal as reference emission.
(d) Line properties come from CUPID's FIND-CLUMPS with a tuned ClumpFind method, with 2 * rms minimum detection level.Consequently, to allow for the wings near the baseline, an additional three pixels either side of the line masked.A fixed 19-channel smoothing kernel is used to determine the background for the line finding (but there is an option to measure the widest line iteratively in order to set the smoothing kernel).
(e) The masked channels for the reference and the source are applied to each unmasked modal spectrum, which is analysed in the same fashion as for the masked modal spectrum.
(f) Any residual background from spectrally broad source emission is removed with FINDBACK from the CUPID package once the masked channels are filled using an iterated solution to Laplace's Equation.The revised background is more accurate as the bulk of the emission and off-position lines have been excised.
(g) Measure the properties of reference lines once again, now improved by the more-accurate background.
(h) Remove any varying residual background to cater for spectrally extended source emission.Use a narrow (9-channel) kernel to track the background more precisely.
(i) A bias remains in the background subtraction and a 1.5 * rms empirical correction is added.
(j) Masked values beyond the spectral lines in the estimated reference spectrum are set to zero.
4. For data taken at different epochs, the mapping from pixel to velocity is likely to be different, so they are aligned to the first epoch.This permits pixel-by-pixel subtraction.
5. The estimated reference spectrum is expanded to the bounds of the raw time series, from which the spectrum is subtracted.b The mean T * A RMS noise in the tiles rebinned to 0.635 km s −1 channel width.
Figure 1.Histogram of the corrected antenna temperature for all voxels in COHRS, displayed on a logarithmic scale.The bin width is 0.1 K.The red line shows a Gaussian fit to the distribution described by 1.14 × 10 9 exp(− 1 2 (T * A − 0.032) 2 /0.49 2 ).The inset show the same distribution on a linear scale.

Figure 2 .
Figure 2. Histogram of the noise for all pixels in COHRS, displayed on a logarithmic scale.The bin width is 0.02 K.A dotted vertical line indicates the mean value of noise levels across the survey.The inset shows the same data on a linear scale.

Figure 3 .
Figure 3. Noise maps for the COHRS data.The noise level is obtained by the conventional way that calculates RMS values over velocity ranges where is no astronomical signal (see text for details).The intensity scale is in T * A (K).
Figure 4.Histogram of the signal-to-noise ratio for all voxels in COHRS, displayed on a logarithmic scale.The bin width is 0.1 K.The blue line is a Gaussian with a width (standard deviation) of 1 centered at 0. The inset show the same distribution on a linear scale.
A position-velocity (l-v LSR ) map of 12 CO (3-2) emission integrated over the whole latitude range is shown in the top panel of Figure 6.The bottom panel presents the same 12 CO (3-2) l-v LSR map, but known H ii regions and spiral arm loci overlaid.H ii regions are obtained from the catalog (V2.2; doi:10.26131/IRSA146) of the all-sky Wide-Field Infrared Survey Explorer (WISE; Anderson et al. 2014).Note that H ii regions shown here are those given a single measured velocity, but the original WISE catalog lists numerous H ii regions with no available velocity measurement or many H ii regions with multiple velocities measured.The traces of spiral arms are derived from Reid et al. (2016) and updated in Reid et al. (2019).In the figure, main spiral arms (Scutum, Sagittarius, Perseus, and Norma-Outer arm) and interarm features (Local Spur, Aquila Spur, Aquila Rift, and 3 kpc arm) are drawn. 12CO (3-2) emission, in general, agrees well with the spiral-arm traces although those associated with the Outer arm are extremely faint and sparsely distributed.The Aquila Rift ranges from about 17 • -43 • , from which bright parts near l = 18 • -22 • and 32 • -36 • are clearly seen.Also, in the remaining sections, weak 12 CO (3-2) emission appears along where other CO isotopologue emissions are detected (cf. Figure 5 of Dame & Thaddeus 1985 and Figure 3 of Jackson et al. 2006).Well-studied star-forming regions such as W43 (l, v LSR = 30.• 9, 95 km s −1 ) and W51 (49.• 4, 60 km s −1 ) show highly peaked emission with complex velocity structures.The locations of H ii regions generally coincide with CO-bright regions.This characteristic is illustrated by comparing the probability distribution functions (PDFs) for the entire 12 CO (3-2) distribution

Figure 5 .
Figure 5.The maps of velocity-integrated emission (T * A ) in COHRS.This map is obtained by integrating over the velocity range written at the top of each panel.The units on the intensity scale are K km s −1 .

6.
EXAMPLE COHRS DATAExamples of COHRS data for active star-forming regions, such as W43, W49A, and W51, are displayed in Figures8-10.These figures present integratedintensity maps from two rotational transitions of two CO isotopologues ( 12 CO/ 13 CO (1-0) and (3-2)) from FU-GIN, COHRS, and CHIMPS data together with 8 µm (GLIMPSE; doi:10.26131/IRSA210)and 850 µm (JPS orEden et al. (2018)) continuum maps, if available.Also, examples of available CO spectral lines obtained from the nearest position to the velocity-integrated 12 CO (3-2) emission peak are shown.That is, spectra of different CO isotopologues and transitions are extracted from the matching positions within the onepixel size of each survey.COHRS emission maps show well clumpy or filamentary structures and in addition extended diffuse CO gas compared with other CO maps.Also, the bright clumpy or filamentary features appear to be closely associated with star-forming regions seen in the continuum maps.W43 is one of the most massive molecular complexes (total gas mass (M gas ) of ∼ 7.1 × 10 7 M , bolometric luminosity (L bol ) of ∼ 8.5×10 5 L , Nguyen Luong et al. 2011; Urquhart et al. 2014b) in the Galaxy.The distance of W43 is estimated to be 5.5 kpc from the Sun (Zhang et al. 2014).The region is located near the tangential point of the Scutum arm, where the spiral arm meets the Galactic bar (Nguyen Luong et al. 2011).It appears between l = 29 • and 32 • and b = −1 • and +1 • in the velocity range v LSR = 80-110 km s −1 (e.g., Nguyen Luong et al. 2011; Kohno et al. 2020).W43 extends beyond the

Figure 5 .
Figure 5. Continued latitude coverage of the COHRS survey, but the brightest parts such as W43-main (l ∼ 30.• 8) and W43-south (l ∼ 29.• 9), which are very active star-forming regions, are covered, as seen in Figure 8. W43-main is considered as a mini-starburst region, which contains fifty-one protocluster candidates (Motte et al. 2003).Kohno et al. (2020) suggests that a supersonic cloud-cloud collision causes the local mini-starbursts in W43.W49A is another one of the most well-known massive and luminous Galactic star-forming complexes (M gas ∼ 1.1 × 10 6 M , L bol ∼ 3.63 × 10 6 L , Galván-Madrid et al. 2013; Urquhart et al. 2014b), despite being located at a large distance of 11.1 kpc from the Sun (Zhang et al. 2013).It is centered at (l, b) = (43.• 15, −0.• 01) and appears in the LSR velocity range from −20 to +30 km s −1 .The region is lying on the Perseus arm in the inner Galaxy.The giant molecular cloud (GMC)of W49A extends more than 100 pc (∼ 30 ) in longitudes(Simon et al. 2001), while active star formation occurs mainly in the central area within ∼ 20 pc (∼ 6 )(Welch et al. 1987;Alves & Homeier 2003).As shown

Figure 5 .
Figure 5. Continued arm.The region is centered on (l, b) ≈ (49.• 4, −0.• 3), and appears as a long filamentary stream with a length of ∼ 100 pc, which is mostly covered by COHRS data (see Figure 10).W51 GMC is distributed in a broad velocity range of v LSR = 30-85 km s −1 (Kang et al. 2010), and embeds two star-forming regions, W51A and W51B, and host a supernova remnant W51C.The brightest COemission region near (l, b) ≈ (49.• 5, −0.• 4) shown in Figure 10 is associated with W51A.12 CO profiles show a clear double peak due to self-absorption around v LSR = 65 km s −1 .The self-absorption feature is stronger in the (3-2) compared with the (1-0), indicating foreground gas is likely colder than the gas associated with the W51 complex.However, the self-absorption situation can occur in subthermal excitation, where the density is smaller than the effective critical density, even if the gas is not colder.Therefore, additional analysis is needed to understand the actual situation.

Figure 5 .
Figure 5. Continued tensity in the profile.Note that FUGIN was mapped over a more-limited longitudinal range (l ≤ 50 • ) than COHRS.Since intensity variations along longitudes are much larger than the pixel size of the survey data (6 for COHRS and 8. 5 for FUGIN), the original l-profiles are smoothed to have a bin size of 60 using the IDL INTERPOL function to provide a better visual exhibition.On the other hand, the b-and v LSR -profiles are displayed without smoothing.Periodic oscillations in the FUGIN b-profile exhibit a horizontal stripe pattern caused by instrumental artifacts.Both l-profiles integrated over latitude and velocity for the two transitions tend to decrease in general above l = 30 • .It is because the distribution of molecular ISM is not uniform in the Galactic disk and the Sun is far from the Galactic center.First, most of the molecular ISM is concentrated in the inner Galactic disk, and the distance (∆d los ) from the Sun and the far side of the inner disk steeply decreases with increasing longitude (∆d los ∼ 6 kpc for two lines of sight, l = 60 • and 30 • , while ∆d los ∼ 2 kpc for two lines of sight, l = 30 • and

Figure 6 .
Figure 6.Position-velocity (l − vLSR) map for the 12 CO (3-2) emission (T mb ) in COHRS.This map is obtained by integrating over the latitude axis.The map is drawn on a square-root scale.The units on the intensity scale are K degrees.The bottom image is the same as the top, but is shown for a narrower velocity range overlaid with known H ii regions and spiral-arm loci.Cross symbols indicate WISE H ii regions.The traces of main spiral arms (Scutum, Sagittarius, Perseus, and Norma-Outer arms) and interarm features (Local Spur, Aquila Spur, Aquila Rift, and 3 kpc arm) from Reid et al. (2016, 2019) are overlaid using black curves.

Figure 10 .
Figure10.Same as Figure8, but for W51 star-forming region.The CO maps are velocity-integrated over the vLSR range of (30, 85) km s −1(Kang et al. 2010).There are no available CHIMPS data for this region.

Figure 11 .
Figure11.Integrated (one-dimensional) l-, b-, and vLSR-distributions of 12 CO (1-0) (FUGIN; black) and 12 CO (3-2) (COHRS; blue).Each profile is obtained by integrating across the two orthogonal axes.Note that b-and v-profiles are used data within 10 • ≤ l ≤ 50 • since available FUGIN data are limited to the Galactic longitudinal range.CO intensity is normalized to the peak value of each profile.The two CO l-profiles are smoothed to have a bin size of ∼ 60 while their b-and v-profiles have a bin size corresponding to a latitudinal pixel size and a velocity channel width of each CO data respectively.On the top panel, three vertical dotted lines are drawn to help locate three star-forming regions: W43 (l = 30.• 8), W49A (43.• 2), and W51 (49.• 4).

Table 1 .
Examples of CO Surveys that Includes the First Galactic Quadrant Resolution * Sensitivity # Ref. †

Table 2 .
Spearman Correlation Coefficients package.In rare circumstances where repeat observations had switched reference positions, each reference position was analysed separately.(REFEMISSION COMBINE REFPOS).The second stanza was to deal with residual offposition signal, that the automated method left, being either untouched lines or, most commonly, lines reduced in depth but not eliminated.Application of this algorithm was enabled by SUBTRACT REF SPECTRUM.REF SPECTRUM COMBINE REFPOS performed the equivalent action as REF EMISSION COMBINE REFPOS.The list of the line extents were supplied through REF SPECTRUM REGIONS.

Table D1 continued
Table D1 (continued) The numbers in the tile name give the central longitude and latitude of the tile.