Heterogeneity and chemical reactivity of the remote troposphere defined by aircraft measurements

Abstract. The NASA Atmospheric Tomography (ATom) mission built a
photochemical climatology of air parcels based on in situ measurements with
the NASA DC-8 aircraft along objectively planned profiling transects through
the middle of the Pacific and Atlantic oceans. In this paper we present and
analyze a data set of 10 s (2 km) merged and gap-filled observations of the
key reactive species driving the chemical budgets of O3 and CH4
(O3, CH4, CO, H2O, HCHO, H2O2, CH3OOH,
C2H6, higher alkanes, alkenes, aromatics, NOx, HNO3,
HNO4, peroxyacetyl nitrate, other organic nitrates), consisting of
146 494 distinct air parcels from ATom deployments 1 through 4. Six models
calculated the O3 and CH4 photochemical tendencies from this
modeling data stream for ATom 1. We find that 80 %–90 % of the total
reactivity lies in the top 50 % of the parcels and 25 %–35 % in the top
10 %, supporting previous model-only studies that tropospheric chemistry
is driven by a fraction of all the air. In other words, accurate simulation
of the least reactive 50 % of the troposphere is unimportant for global
budgets. Surprisingly, the probability densities of species and reactivities
averaged on a model scale (100 km) differ only slightly from the 2 km ATom
data, indicating that much of the heterogeneity in tropospheric chemistry
can be captured with current global chemistry models. Comparing the ATom
reactivities over the tropical oceans with climatological statistics from
six global chemistry models, we find excellent agreement with the loss of
O3 and CH4 but sharp disagreement with production of O3. The
models sharply underestimate O3 production below 4 km in both Pacific
and Atlantic basins, and this can be traced to lower NOx levels than
observed. Attaching photochemical reactivities to measurements of chemical
species allows for a richer, yet more constrained-to-what-matters, set of
metrics for model evaluation.


(30-90 sec), but these can be mapped onto the 10 s parcels with loss of the higher frequency 62 variability found in the 10 s measurements. The frequent profiling of the DC-8 gives us both 63 vertical and horizontal scales: the vertical extent of a 10 s parcel is 50-110 m (55 %-95 % of all 64 parcels, with < 50% having near level flight) and the horizontal extent is typically 1.4-2.5 km (10 65 %-90 % of all parcels). A few key species have 1 Hz measurements, and, as a case study, we 66 examine the time series of O3 and H2O measured during one of the profiles of ATom-1 RF 3 in  Table   80 S2. Every species in each air parcel is now flagged so that the instrument is clearly identified (in 81 the case that two instruments measure the same species) and the type of the gap filling (dependent 82 on the length of the gap) is denoted so that the users can develop their own criteria for including,  NaN values with flags = 1, 2 or 3, is shown (the remaining % has flags = 4, 5 or 6). These data 142 correspond to a primary or secondary direct measurement (1 or 2) or else short-gap interpolation 143 (3, see text below). Missing data for an entire flight (0%) has shaded cells.

145
Mor.all combined species and fixes. The

185
Pressure and temperature. P and T have 5 very small gaps of length ~6 (# of 10s parcels 186 missing) plus a longer gap of length 28. All gaps occurred during smooth descent or ascent and 187 so were filled using linear interpolation. These are denoted by flag_M(:,10) = flag_M(:,11) = 3.

188
In this document we are careful to give measured species a suffix that denotes their provenance, 189 and thus the MDS variables denoting the combined, continuous data are labeled P_M and T_M.

198
CO. In our first attempts to produce a gap-filled record for chemical modeling, we sought a 199 species with continuous measurement that could be used as a proxy for unusual or polluted air 200 during the gaps in other species. CO was the obvious species because it is indicative of biomass 201 burning or industrial pollution, and ATom has two well calibrated, nearly continuous 202 measurements: CO_NOAA and CO_QCLS. The primary CO data are from QCLS because it has 203 higher precision and the secondary are from NOAA which has fewer gaps. Unfortunately, after 204 creating this gap-filled CO data and applying it as a proxy for MDS versions 0 and 1, we found 205 that CO had little skill in filling the gaps in other species. We use this method to generate our 206 CO_M record for the MDS, but do not use it for other species. This processing of the CO data 207 was done with the full 149,133-parcel dataset, and not the airport-collapsed data set.

232
Short-gap simple interpolation for remaining species. It was decided that the least intrusive 233 method for filling short data gaps was to simply interpolate using only the instrument data. In 234 MDS versions 0 and 1, CO was used as a proxy to fill these gaps, but later analysis showed little 235 correlation with absolute CO or even the short-term variability in CO. We examined the typical

255
Long-gap interpolation -Stratosphere. We find the most robust definition of stratospheric-like 256 air to be based primarily on H2O rather than O3, because O3 abundances > 200 ppb are often seen 257 in large, clearly tropospheric air masses with H2O > 50 ppm. Based on percentiles of O3 at 258 different values of H2O (see Fig. S4a) we pick <30 ppm as the criteria for being stratospheric, 259 with the secondary requirements that O3 > 80 ppb and CO < 120 ppb (see Fig. S4b). For the 260 stratospheric air we create mean 'profiles' in terms of six O3 bins (< 200, 200-300, 300-400, 400-261 500, 500-700, > 700 ppb) use this as a lookup table for gap filling. There are many fewer 262 stratospheric parcels, and the stratosphere tends to be similar across latitudes, and so we create a 263 single lookup tables from all research flights at all latitudes. In general, these near tropopause air 264 parcels are cold and dry and not highly reactive; however when partitioning the chemistry model 7 calculated reactivities between stratosphere and troposphere, these criteria may need to be re-266 investigated.

268
As a measure of the error in this long-gap interpolation, we randomly select 10% of the air 269 parcels from data stream before calculating the long-gap interpolation, interpolate those 10% 270 points, and calculate the mean bias and root-mean-square error (rmse). This is repeated 10 times 271 and we show the average results in Table S5 below. We find these results acceptable, and better 272 than the multiple linear regressions we tried. There may be a better way to do this in future 273 versions MDS-2, perhaps with a machine-learning approach. Gaps interpolated in this way are 274 given flag = 4 (troposphere tables) and flag = 6 (stratosphere tables).

276
Missing data for an entire flight.  Table S6.

298
From the reactivity results for ATom-1 shown in this paper, the lack of ATom-3 NOx 299 observations in the Eastern Pacific (RF 1) mean that the P-O3 statistics there (not calculated in 300 this paper) will not be useful.  anomalies. There is no evident bias, but some scatter, and so the NaNs in the primary record 311 (which first has had short-gap interpolation as noted above) are simply filled with the secondary 312 record (also with short-gap interpolation).

314
CH2O: (1) CH2O_ISAF, (2) CH2O_TOGA. Formaldehyde is a key reactive species and TOGA 315 provides a secondary record for the 2 nd half of ATom-4 when ISAF failed. The overlapping data 316 with both instruments is plotted in below (Fig. S5). The mean difference in overlapping 317 observations is very small (-1 out of a mean of 134 ppt), but the rms is larger (75 ppt). ISAF has 318 a number of values > 1000 ppt, while TOGA has none. A linear fit gives a slope of 0.8 with R 2 = 319 0.59, but a 1:1 slope has only slightly smaller R 2 = 0.55. We do not attempt to rescale the TOGA 320 data in this case and just replace any NaNs remaining in the short-gap-interpolated ISAF record 321 (particularly flights 42:48) with TOGA data (also short-gap interpolated).  Table S4) and so we did not scale WAS.    In this paper, we use 6 global atmospheric chemistry models for their August chemical statistics.