Comments on ‘‘A CloudSat–CALIPSO View of Cloud and Precipitation Properties across Cold Fronts over the Global Oceans’’

Naud et al. constructed satellite-based composite analyses of clouds and precipitation across cold fronts. However, their approach does not exclude occluded fronts, does not separate anafronts from katafronts, does not separate frontlikephenomena primarily identiﬁedby thermalgradientsfromthose primarily identiﬁedby windchanges,andsmoothsoveralongfrontvariability.Bylumpingthesedisparatefrontalstructurestogether, thefront-centeredcompositecrosssectionsrevealforward-slopingstructuresandweakgradientsacrossthem, raising questions about how to interpret their composite cross sections.

1. Introduction Naud et al. (2015) describe the cloud and precipitation structure of cold fronts over the global oceans from satellites using CloudSat radar data, CALIPSO lidar data, and MERRA reanalyses for temperature and wind.Using a set of automated approaches, more than 30 000 fronts over the global oceans within 308-608N and 308-608S over a 4-yr period were composited to produce cross sections of various properties across the fronts.In this comment, concerns are raised about the approaches used to produce these mean cross sections and how to interpret the resulting cross sections.

Concerns
Four concerns are raised that pertain to the frontal structures going into the composites.1) Naud et al. (2015) state that their study is about cold fronts, but an unspecified number of occluded fronts are included in their composite.''Because we could not differentiate cold and occluded fronts objectively, occluded fronts may be included in our database' ' (p. 6745).No effort is made to quantify to what extent their results are contaminated by occluded fronts.The classic warm-type occluded front is characterized by a forward-sloping frontal structure, cloud pattern, and ascent, which differs from the classic rearward-sloping cold front.Because the majority of occluded fronts are of the warm type (Schultz and Mass 1993;Schultz and Vaughan 2011;Schultz et al. 2014), any occluded fronts in the sample would easily contaminate any composite cold-frontal structure.2) Although commonly depicted as rearward-sloping in textbooks, some cold fronts have the structure of a katafront or split cold front, a forward-sloping cloud structure formed as the dry airstream from behind the cyclone flows over the surface cold front (e.g., Bergeron 1937;Sansom 1951;Browning and Monk 1982;Browning 1986;Young et al. 1987;Mass and Schultz 1993).Katafronts tend to be observed some distance equatorward from the center of the cyclone, which is the region targeted by Naud et al. (2015).Moreover, cold fronts can possess a variety of additional different structures such as the following:  and Browning 1979;Hobbs and Biswas 1979;Hobbs and Persson 1982;Jorgensen et al. 2003).The possibility of these various types of frontal structures in the same composite is also of concern to the quality of the composite.
3) Naud et al. (2015, p. 6745) produce an automated scheme using MERRA reanalyses to identify fronts.This scheme is a result of two different diagnostic approaches to identifying fronts: the thermal front parameter using potential temperature at 1 km AGL (Hewson 1998) and the 6-h change in meridional wind (Simmonds et al. 2012).Fronts identified by either approach are apparently merged into a single dataset.However, previous studies that compared the two approaches showed that they sometimes did not identify the same feature (e.g., Table 4 in Hope et al. 2014;Fig. 1 in Schemm et al. 2015).Specifically, the wind shift and temperature gradient associated with cold fronts are sometimes not coincident, as reviewed by Schultz (2005).Using two different approaches to identify fronts and compositing them risks lowering the quality of the composite.4) Cold fronts can show a substantial alongfront variability (e.g., Jorgensen et al. 2003;Norris et al. 2017), but the approaches employed by Naud et al. (2015) average satellite data along the cold front (pp.6745-6746) and average MERRA output along 1000 km of the front (p.6751).''This compositing technique does not assume a general direction of the cold fronts and averages together information anywhere along and across the cold front.''(p.6746).Thus, any of this alongfront variability is likely eliminated when the composite is constructed.

Results from the compositing procedure
These concerns about how the composite is constructed affect the quality of the composite.Indeed, the composites show a number of unusual properties.
1) The inclusion of occluded fronts in an ostensible composite of cold fronts would result in forward-sloping properties of the composite, as is shown in Naud et al.'s (2015) Figs. 1 and 8c.That 20%-30% of their fronts have a forward-sloping cloud structure is indicative of this potential contamination.2) That the convection lies about 300 km ahead of the cold front suggests that this might be due to some of the fronts having an elevated frontal zone (as might be the case in a warm-type occluded front or katafront), prefrontal trough or wind-shift line.3) Because of the different shapes and structures to the fronts that comprise the composite (i.e., occluded fronts, anafronts, katafronts), the composite will necessarily result from quite a bit of variability of cold-and occluded-frontal structures.By lumping these disparate frontal structures together, the frontcentered composite cross sections smooth out any signal and reveal relatively weak gradients across them, as seen in their Figs. 3,4,5,6,and 7.
For this reason, the approach by Naud et al. (2015) raises questions about how the compositing is done and whether the resulting composite is meaningful.

Conclusions
The remarkable variety of cold-frontal structures observed in nature, plus those from occluded fronts, are lumped together through the Naud et al. (2015) compositing approach.Thus, one cannot generalize about cold fronts in the real world through a compositing process that includes an unspecified number of occluded fronts, katafronts, and anafronts, smoothing out such an observed diversity of frontal structures and alongfront structures.As such, these issues raise questions about the best way to interpret their composite cross sections.