This manuscript describes interesting measurements of snowpack composition including salinity, bromide, and nitrate. Net deposition trends are observed, as well as profiles of ionic species in the snowpack and small-length scale differences in deposition around the Eureka area. However, the authors continue to try to say their data show that "the release flux of reactive bromine from snow must be a weak process and smaller than the derived bromide deposition flux of ~1×10^7 molecules cm-2 s-1, which flux is smaller than previously estimated flux by a factor of more than an order of magnitude." I believe the arguments they make on this point are flawed, as described below. I agree with the prior reviews that this manuscript is difficult to read and doesn't present a clear story, also making it hard to understand some of their arguments. I would suggest they clarify their description of observations and unless they can make a valid argument that their data actually constrains short-timescale snowpack emissions, remove that point.
Snow sampling and small flux challenges:
Fundamentally, the low concentration of bromide in the surface snow along with variability in ion concentrations related to sampling different surface snowpack (your next day's snow sample will be to the side of the prior day so you don't dig a hole) means that it is difficult to quantify changes in snow composition over time. The authors attempt to make up for this inherent challenge by using a fairly long timeseries. This approach means that the NET long-term deposition trend is quantified. Prior reviews pointed out that analysis errors and variability make it hard to quantify the net deposition trend, and the response was that "...collecting daily snow samples over a relatively long period of 3-4 weeks with an aim of detecting the possible accumulative change." This clearly shows that the study authors understand that they only detect the "accumulative change" = NET deposition.
Net deposition does not constrain short-term snowpack emissions:
Net deposition is measured, but there may be larger fluxes that are bidirectional (snowpack emission and deposition) occurring on shorter timescales. Therefore, the authors cannot say that the long-term trend in NET deposition constrains shorter-term fluxes to be small. If the shorter-term fluxes were unidirectional only (e.g., deposition only), then the net deposition can constrain the process (there is no emission in this presumed unidirectional case), but in the case of bromine, we know that snowpack can both emit reactive halogens and have halogens deposit to it. The short term (days to hours) variability in both BrO and snow Br- also are consistent with significant short term fluxes.
New "Bromine mass balance" approach (section 3.7):
The authors make an equation for the time trend in air column density, equation R7, which says dc_air / dt = P_air - c_air / tau_air. They go on to say: "However, from Figures 5(c) and 6(c), we see a significant decreasing tend of BrO partial column, indicating the input term P_air is much smaller than the loss term, c_air / tau_air." This doesn't make mathematical sense. For equation 7's left side to be negative, P_air should be smaller than c_air / tau_air, but a large value of P_air can be allowed as long as c_air / tau_air is larger.
As an example, let's say that there is no trend in gas-phase BrO (steady state), then the left side of R7 is zero, which means that 0 = P_air - c_air / tau_air, which gives the steady-state result: P_air = c_air / tau_air. Let's take tau_air to be 1 day = 86400s and say BrO is 3e13 molecule cm^-2 (typical value from their plots), and they assume BrO is 0.3 of gas phase Br, so gas-phase c_air = 1e14 molecule cm^-2. Then the production rate is P_air = 1e14 molecule cm^-2 / 86400s = ~1e9 molecule cm^-2 s^-1. This emission flux is within the quoted measurement (in the literature) of "snowpack bromine emission, a direct gradient measurement of Br2 and BrCl above a patch of snowpack was made near Utqiaġvik, Alaska (Custard et al., 2017), who reported emission fluxes of 0.7–12 × 10^8 molecules cm−2 s−1." Instead, the authors decided to choose a lifetime of reactive bromine of 42 days in 2019, and due to this choice, they get a much smaller emission flux. This lifetime of reactive bromine seems unreasonably long given the episodic nature of reactive bromine events, and is discussed below.
If we simply look at the BrO timeseries in Figures 5c and 6c, we can see that BrO varies significantly during individual days. BrO doubles on some days, and there are many instances where there is a factor of two difference in BrO between one day and the next. If we interpret this variability as due to local fluxes, one would clearly accept that BrO lifetime can be on the order of a day, which would allow snowpack emission fluxes comparable to the measured result from Custard et al., 2017.
Complications with their lifetime analysis:
This study was done at a time when the reactive halogen season was declining a bit on a ~month timeframe. This slight decline in net BrO is calculated to be a loss of c_air over time in their equation 7. They then go on to use the same BrO data with an exponential fit to calculate a "loss rate" of reactive bromine. Use of the same data on both sides of equation 7 appears circular. Note that this "loss rate" is a NET loss rate over many weeks, not a loss rate that is specific to shorter term processes. Let's for the sake of argument say that they had stopped their study six days earlier in 2019. It appears that the trend in BrO over time would now be increasing, and if you fitted it to an exponential, you would not have a loss rate, but a growth rate or possibly flat (zero slope, infinite lifetime). They take the exponential loss rate to be be indicative of the lifetime (tau_air) of reactive bromine, but now the loss rate would be very small and the "lifetime" of reactive bromine would be longer than the 42 days they calculate in 2019 -- now reactive bromine might live well into the summer or even over multiple years, which is counter to observations. Is is obvious that you cannot extract the lifetime of reactive bromine in the way they are trying to do it here. Without a constraint on tau_air, they cannot use equation 7 to calculate the magnitudes of the two terms on the right side, and thus they cannot determine the snowpack production flux.
To show that equation 7 cannot be used in this manner, consider this hypothetical situation. Say there was a sealed test tube with some liquid water and vapor in it. We now raise and lower the temperature, which will cause water to evaporate (P_air) and raise c_air or condense and lower c_air. Over a long time (many warming and cooling cycles), if you fit the timeseries of c_air to a slope, dc_air/dt would be close to zero (much like their long-term trend in gaseous bromine is fairly flat). By their method, they would then fit the c_air over time to an exponential, and say that the lifetime of the vapor is long (because the timeseries is on average flat), so tau_air is very long, and the second term (c_air / tau_air) will go to near zero. Equation 7 will then be dc_air/dt = ~0 = P_air - c_air/tau_air (term = ~0), so you find P_air = ~0. Their interpretation is that this system has no evaporation of water (P_air = ~0), while in fact water is evaporating and condensing with potentially large fluxes. The failure is that you cannot use the net flux to constrain faster bi-directional fluxes.
Realistically, over multi day periods, the weather changes, airmass origins change, the sun rises, temperature warms on average. These factors lead to wide variability in BrO as shown on their figures. Yet, they fit a long-term trend through the data and call that the lifetime of reactive bromine as if there were a constant loss rate for the full campaign, nearly a month. It is clear that faster than monthly processes are needed to describe reactive halogen chemistry and that long-term "accumulative changes" do not directly constrain faster underlying bi-directional fluxes.
Summary:
Overall, the central problem with this manuscript is that the authors do not accept the difference between a net deposition rate measured over a month-long period and faster bi-directional fluxes that are occurring (based upon prior literature reports of snowpack emissions). They cannot constrain fast fluxes that happen in bi-directional manners by a long-term deposition flux. The BrO measurements, which vary by factors of two day to day could clearly be consistent with large snowpack emission on one day followed by deposition back to the snow the next day. Alternatively, look at the snowpack Br- timeseries, which shows a lot of variability. It is clear that analysis errors and snow sampling variability can affect variability in measured Br-, but one were to believe that this variability were real, it would indicate large fluxes of reactive bromine out of the snow and re-deposition of Br- back to the snow.
Over the long-term, due to bi-directional fluxes, the net change in snowpack bromine could be small (as is observed), but a lot of chemistry could have happened on shorter timescales than their long-term net trends can capture. Effectively, the approach described in this manuscript is not equipped to put short-term constraints on snowpack emissions fluxes.
If the authors want to report snowpack composition, vertical profiles in pits, and net deposition fluxes over long periods of time, I can see the publication of a manuscript showing those results. However, I see no validity in the attempts they have presented to constrain short-term snowpack emissions fluxes. If the authors seek to maintain that point, I argue for rejection of the manuscript. In this set of comments, numerical examples derived from their figures were shown to be consistent with snowpack emissions fluxes measured by others and reasonable lifetimes for reactive bromine. The lifetimes and fluxes of these faster processes fit with variability observed in both atmospheric reactive bromine and snowpack bromide. |
