Review of “Improved method for temporally interpolating radiosonde profiles” by Linus von Klitzing et al.

General Comments: The manuscript describes a new method for interpolating radiosonde observations to provide a high resolution (in time) profile of potential temperature and water vapor mixing ratio for the convective boundary layer. The technique employs a normalization of the height coordinate by the planetary boundary layer height before interpolating in time helping to preserve the inversion-topped structure. Interpolated profiles employing this technique are compared to a case with no height normalization showing marked improvement compared to independent soundings. The products developed using this new technique could have important used for those study boundary layer transport and cloud development at sites where more advanced remote sensing is unavailable. There are several important issues that need to be clarified before the manuscript is ready for publication in AMT. The most significant of these are outlined in my general comments and are related to more clearly defining the targets and the motivations for the products that would be developed using this technique and considering some of the uncertainties related to varying environmental conditions and the instrumentation used for PBLH estimates. With this in mind, I recommend the paper be considerd for publication following major revisions.

Major Scientific/Technical Comments:

• The methodology presented in the manuscript focuses entirely on the convective boundary layer. This should be reflected in the title. I would suggest adding “…in the convective boundary layer” to the end of the current title.
• Lines 21-27: In this introductory paragraph, one of the motivations is the accuracy of the interpolated sounding product as it is used in the ARM MICROBASE product. However, the main use of that thermodynamic profile is in determining the “mixed-phase” layer, i.e. where the temperature is between 0^oC and -16^o The manuscript focuses on an improved methodology in the convective boundary layer as applied to a period from mid-April through mid-July, a time period when I expect the freezing level will generally (if not always) be above the boundary layer and so improvement of the boundary layer thermodynamic profile would have no impact on the MICROBASE product. It might be better to provide a more applicable motivation. Similarly, there is a reference provided to the “MERGED SOUNDING” product (Troyan, 2010) which incorporated model output to improve the interpolation between soundings and might provide improved boundary layer thermodynamics over interpolated sonde. Since MERGED sounding is not available in recent years it may be best to not reference unless it is discussed.
• Line 34-36: Do you consider both well-mixed and decoupled boundary layers in your analysis? In the case of decoupled boundary layers, I believe that even with the normalized height grid there will still be a smoothing of the decoupling layer that may or may not be better captured with the new technique.
• Line 43 – 45: The estimate of the PBLH will have some dependence on the instrument used. Have you considered the impact of these different definitions/measurements of the PBLH? E.g., I am wondering if the increase in standard deviation of the water vapor mixing ratio in figure 3d near the top of the boundary layer is mostly due to varying definitions of the PBL depth? I also wonder, if the technique REQUIRES this separate measurement of PBLH, would there still be value in using the interpolated PBLH depth (as determined from the radiosonde profiles) to normalize the height? I would hypothesize that this would still provide an improvement over the standard grid interpolation.
• Lines 84-91: Thank you for the simple description of the TROPoe data product. It would be helpful to also include the time resolution of the product and a description of how the PBLH is determined, since it is a critical input to the new interpolation method. It would also be helpful to describe why an interpolated sounding product is needed if TROPoe already provides continuous profiles of temperature and water vapor mixing ratio. Also wonder how the TROPoe algorithm is impacted by the presence of boundary layer clouds and how this would impact the interpolation methodology.
• I find Figure 2 very difficult to read and interpret as a scatter plot. Both discerning the different markers used to represent the different interpolation scheme and radiosonde separation, but also the fact that most of the points are clustered in one are with lots of white space among the plot. Have you considered some type of contouring or 2D histogram (or something else) to make this easier to read and interpret? I do find the table communicates the differences much more clearly.
• I think an important point to make, perhaps in the conclusions, is that this height-normalized interpolated sounding has an explicitly different purpose/target compared to the ARM interpolated sounding product. While this new technique focuses on the convective boundary layer structure only, the interpolated sounding product is designed to provide thermodynamic profiles through the depth of the troposphere under all atmospheric conditions (in near real-time). A statement about how these might be combined would be useful.

Specific Comments:

Line 1-2: Suggest adding that this improved technique is for the planetary boundary layer.

Line 2: Not sure there is a need to define or even use acronyms in the abstract, especially if they are not used again within the abstract.

Line 4 (and 17, 56): Minor detail but ARM is a “Facility” rather than a “Program.”

Line 16:  The frequency of launches can vary much more than between 2 and 4 times per day depending on many different parameters. You might say that at many operational sites radiosondes are routinely launched two to four times per day or something similar.

Line 30: Normalizing by the height of the PBL to understand PBL structure has a long history and it should probably be acknowledged here. One of the earliest studies I know of that used this technique was:

Augstein,E., H. Schmidt,and F. Ostapoff, The vertical structure of the atmospheric planetary boundary layer in undisturbed trade winds over the Atlantic Ocean, Boundary Layer Meteorol., 6, 129-150,1974.

Line 32: Replace “level” with “value.” Level is ambiguous because it could refer to the level (i.e. height) in the atmosphere.

Line 41: Suggest adding an “e.g.” ahead of the list of references since there are many, many studies in this subject area.

Line 56: ARM and SGP were already defined.

Line 108: “in-between’ seems too colloquial. Maybe “intermediate” would be a more appropriate word?

Line 109: “an” should be “any”

Line 110: Rather than referring to this as the “standard procedure of the ARM program,” I would suggest using “standard procedure used in the interpolated sonde product.”

Line 110-111: May be better stated as “….and second, using the normalization of the height coordinate using the smooth PBLH estimate.”

Line 117: Can you describe “the characteristics indicative of a convective boundary layer?” Does this include thresholds for the stability? Does it include decoupled cases?

Line 122: I am not sure “normal” or “old” are the appropriate here. “Current” seems like it would better here and elsewhere (but “old is better than “normal”). Also, see comment regarding Line 110.

Line 138: See general comment #3. Are you requiring the CBL be well-mixed throughout its depth?

Line 148-149: Should we expect the TROPoe-derived and the Raman Lidar derived PBLH to agree? Different measurement and retrieval methods, and different resolutions will make the agreement difficult.

Line 195: “constant mixed layer and a sharp gradient at the top” is not clear. I think you are referring to potential temperature or water vapor mixing ratio being constant with height within the mixed layer with a sharp gradient at the top.”

Line 203: I would not refer to these as distributions, rather just scatterplots.

Line 215: What is meant by “introduction of nonphysical atmospheric layers as artifacts of insufficient interpolation?” This is unclear to me.

Line 216-217: Does this mean you do sometimes have decoupled layers that could cause problems for the height normalization method?

Line 257: “level” is ambiguous here, suggest “value instead.” Level could be height in the atmosphere.

Line 280: Can you explain what is meant by “non-physical artifact layers?” Perhaps provide an example?

Line 304: Can you define what is meant by “suitable source?” For example, since you are interpolating thermodynamic variables, should the PBLH be defined from similar thermodynamic variables? i.e. will a lidar-based retrieval that depends on aerosol scattering, or a a retrieval based on turbulence profiles raise some issues with the height normalization?

Line 325: Are the height normalized interpolated soundings available somewhere?

This manuscript presents a new method for interpolating temperature and water vapor mixing ratio (WVMR) profiles between radiosondes. The method uses a height normalized grid interpolation approach to more accurately represent the structure of the boundary layer throughout the day and requires continuous planetary boundary layer height (PBLH) measurements as input to the algorithm.

The normalized height (NH) approach was applied to convective boundary layer cases and compared qualitatively with Raman lidar based measurements of potential temperature and WVMR. The approach shows improvement to the precipitation free boundary layer though with a bias. The mean bias and standard deviation near the surface are very similar for the NH and non-NH methods.

Overall, the paper is well organized, and the writing is clear, though some clarification of the analysis could be improved and are noted below. The NH method demonstrated does provide value in this well constrained example, but the broader applicability in a continuous product is a much farther reach. Despite this I am inclined to accept this manuscript for publication with minor revisions because it is the first demonstrated improvement to interpolated radiosondes that could be valuable for specific applications and case studies. I encourage the authors to pursue further evaluations under more varied conditions, as discussed in the conclusions, so that a more universal application can be realized.

1. Introduction and Motivation

The motivation for this study is to explore ways to improve the interpolation of thermodynamic profiles between radiosonde measurements, which has many positive benefits. This study is focused on thermodynamic structure in the hydrometeor free boundary layer (no fog, precipitation etc.) but much of the discussion in the introduction is around the ARM routine interpolated sonde data and its use in cloud microphysical retrieval products, where the interpolated radiosonde measurements are only used in cloud above the boundary layer. They also demonstrate in their analysis that the NH and non-NH approaches both converge above the boundary layer. This discussion is essentially irrelevant to the paper. I suggest removing this discussion (lines 19-27) or revising it to be more focused on the improvements in the boundary layer and describing applications where an improved product would be beneficial to the community.

I would also note that it is very likely that many more research groups besides ARM use simple interpolations between radiosondes for a variety of applications, and you don’t use the ARM product in the analysis. You may want to make a note of this in the introduction and focus on the different methodologies instead.

1. PBLH estimates

As the authors state, the PBLH is an input to the NH method, and any measurement could be used. But the accuracy of the PBLH measurement is critical to this approach. Deriving a continuous PBLH product over all conditions is not trivial, requires multiple measurement methods, and errors would cause discontinuities in the NH interpolated radiosonde data. The authors should acknowledge this in Sec. 2.2 or the conclusions.

Line 43-45: Can you elaborate on what simple instruments could be used to determine continuous PBLH? The authors use AERI measurements to derive continuous PBLH. The AERI instrument and associated TROPoe algorithm are both complex and expensive.

1. Figure 2

For the discussion regarding Figure 2 you are using cases and profiles interchangeably. I interpret a “case” as being a convective boundary layer event and a profile as representing a 10 min interpolated vertical profile corresponding to the 10 min average Raman lidar profile (based on earlier discussion related to Fig. 1). Please clarify what the 79 profiles represent, how many profiles are compared per event (would it be 2?) and state how many profiles are represented in Fig. 2a and 2b for the HN and non-HN profiles.  

The scatter plot in Figure 2 as presented is hard to interpret, though Table 1 provides quantitative values to support the analysis. A joint PDF might be easier to visualize the differences, though there may not be enough points.  

1. Minor revisions:

Line 26: The Microbase algorithm uses the interpolated radiosonde data and radar reflectivity to determine the cloud phase. The interpolated radiosonde data is not used to determine the radar reflectivity. Please correct this inconsistency in the text if this sentence is not removed (per the discussion above under regarding the introduction).

Lines 122, 313 throughout: Suggest referring to the method without height normalization as non-HN rather than ‘old’ or ‘normal’ as it is more descriptive.

Sec. 4.2.1 You may want to remind readers that the analysis in this section compares the interpolated sondes (both methods) with the radiosonde launches that were not used in the interpolation. This was described in Sec. 3.2 but did not point to which comparison uses the sondes. Anyhow, breadcrumbs are always nice.