1 A reevaluation of airborne HO observations from x NASA field campaigns Jennifer R. Olson, James H. Crawford, Gao Chen NASA Langley Research Center, Hampton, VA William H. Brune Penn State University, State College, PA Ian C. Faloona Department of Land, Air and Water Resources, University of California at Davis, Davis, CA David Tan School of Earth and Atmospheric Sciences, Georgia Institute of Technology, Atlanta, GA Hartwig Harder, Monica Martinez Max Planck Institute for Chemistry, Mainz, Germany Submitted to: Journal of Geophysical Research – Atmospheres August 23, 2005 2 ABSTRACT In-situ observations of tropospheric HO (OH and HO ) obtained during four x 2 NASA airborne campaigns (SUCCESS, SONEX, PEM-Tropics B and TRACE-P) are reevaluated using the NASA Langley time-dependent photochemical box model. Special attention is given to previously diagnosed discrepancies between observed and predicted HO which increase with higher NO levels and at high solar zenith angles. 2 x This analysis shows that much of the model discrepancy at high NO during SUCCESS x can be attributed to modeling observations at time-scales too long to capture the nonlinearity of HO chemistry under highly variable conditions for NO . Discrepancies at x x high NO during SONEX can be moderated to a large extent by complete use of all x available precursor observations. Differences in kinetic rate coefficients and photolysis frequencies available for previous studies versus current recommendations also explain some of the disparity. Each of these causes is shown to exert greater influence with increasing NO due to both the chemical nonlinearity between HO and NO and the x x x increased sensitivity of HO to changes in sources at high NO . In contrast, x x discrepancies at high solar zenith angles will persist until an adequate nighttime source of HO can be identified. It is important to note that this analysis falls short of fully x eliminating the issue of discrepancies between observed and predicted HO for high NO x x environments. These discrepancies are not resolved with the above causes in other data sets from ground-based field studies. Nevertheless, these results highlight important considerations in the application of box models to observationally based predictions of HO radicals. x 3 1.0 INTRODUCTION Since Levy [1971] first proposed that the hydroxyl radical (OH) could be photochemically produced within the troposphere, the cycling of odd hydrogen, or HO (OH + HO ), has been a topic of intense study. Oxidation by OH is the major x 2 loss for most trace gases (e.g., carbon monoxide, methane, and non-methane hydrocarbons), and the oxidation of NO by HO is responsible for photochemical 2 production of ozone. The tropospheric distribution of HO is highly variable in time x and space owing to short photochemical lifetimes of a few seconds or less for OH and a few minutes for HO . To enable the study of the natural response of HO to 2 x changes in atmospheric conditions, simultaneous measurements of HO and x species that influence its formation, loss, and cycling are necessary at high temporal and spatial resolution. The first viable techniques for measuring tropospheric HO appeared in x the early 1990’s [Heard and Pilling, 2003 and references therein]. While such measurements are still far from routine, their deployment has become a valuable part of airborne and ground-based studies focused on understanding tropospheric photochemistry. NASA’s airborne field campaigns have deployed HO instruments on several aircraft (i.e., ER-2, DC-8, and P-3B). Here we focus x on observations from NASA’s DC-8 aircraft which has flown the Penn State Airborne Tropospheric Hydrogen Oxide Sensor (ATHOS) in four successive field campaigns (SUCCESS, SONEX, PEM-Tropics B and TRACE-P). These data are highlighted due to the extensive suite of supporting measurements in the 4 DC8 payload and also because of outstanding issues that have resulted from previous analysis. Analysis of HO from these campaigns have highlighted discrepancies x between observed and predicted HO casting doubt on our theoretical 2, understanding of middle and upper tropospheric HO . These include x discrepancies at high NO [Brune et al., 1999; Faloona et al., 2000; Olson et al., x 2004] and at high solar zenith angles (SZA) [Brune et al., 1999; Jaegle et al., 2000]. Evidence for possible uptake of HO on aerosols or liquid cloud water 2 particles has also been discussed [Brune et al., 1999; Jaegle et al., 2000; Olson et al., 2004]. While previous studies from the various investigators have used similar models and approaches, there are differences in model implementation and the use of available observations to constrain model calculations. Additionally, there has been sufficient evolution of recommendations for kinetic rate coefficients and photolysis frequencies to warrant a reexamination of the data. A reevaluation of these data is presented here, with particular emphasis on these discrepancies. 2.0 DESCRIPTION OF DATA AND MODELING APPROACH 2.1 NASA DC-8 Observations. Table 1 lists general information concerning the four NASA DC8 aircraft campaigns addressed here, including locations and deployment dates. For each campaign, HO measurements were obtained from ATHOS which is an in-situ x 5 laser induced fluorescence OH detection system based on fluorescence assay by gas expansion. The absolute accuracy of ATHOS measurements is listed as +/- 40% in earlier studies [Mather et al., 1997], and Faloona et al. [2004] refines this value to +/- 32%. Examples of the lower limits of detection estimated for one- minute integration times and environmental conditions typical of the TRACE-P field campaign (northwest Pacific springtime) are less than approximately 0.003 pptv (OH) and 0.03 pptv (HO ) in the upper troposphere, and 0.01 pptv (OH) and 2 0.1 pptv (HO ) in the lower troposphere [Faloona et al., 2004]. For an in-depth 2 review of characteristics and calibration for ATHOS, see Faloona et al. [2004]. Information on other photochemical measurements for each of the campaigns can be found in Table 1 and in the included references. The typical suite of observations includes meteorological and navigational measurements, radiance and particle measurements, and measurements of basic gas phase species such as O , CO, CH , NO (NO and NO ). Additional measurements may 3 4 x 2 include hydrogen and methylhydrogen peroxides (H O and CH OOH), nitric acid 2 2 3 (HNO ), peroxyacetylnitrate (PAN), a wide array of non-methane hydrocarbons 3 (NMHCs), aldehydes, acetone, and other oxygenated hydrocarbons such as methanol (CH OH). Time-merged data sets are routinely produced for the 3 campaigns whereby the raw data measurements, obtained with a variety of time integrations, are averaged to a common time scale (typically one minute) for purposes of modeling. The base reanalysis presented here uses the publicly available one-minute data merge for each campaign. 6 2.2 Modeling. The modeling approach is based on the assumption of a diurnal steady state, which is typical of most previous analyses of data sets of this type [e.g., Davis et al., 1996; Jacob et al., 1996; Crawford et al., 1997, 1999; Jaegle et al., 1998; 2000; Olson et al., 2001; 2004; Frost et al., 2002]. At a minimum,the model calculation for each individual data point is constrained by coincident observations of O , CO, NO, CH , NMHCs, acetone, temperature, dew point, and 3 4 pressure. With the exception of NO, constraining parameters are held constant throughout the simulated model diurnal cycle. NO varies diurnally; however, total short-lived nitrogen (NO+NO +NO +2N O +HONO+HNO ) is held constant with 2 3 2 5 4 partitioning accomplished by the model. The amount of short-lived nitrogen is determined such that NO matches the observed value at the time of measurement. Other model-calculated species are integrated in time until a reproducible diurnal cycle is achieved, and the predictions at the exact time of day that the observed input data was obtained is used as the instantaneous model predictions for that data point. Several optional constraints may be implemented if desired and if measurements are available, including H O and 2 2 CH OOH, HNO , PAN, methanol, ethanol (C H OH), and acetic and formic acids 3 3 2 5 (HCOOH and CH COOH). Otherwise these species are predicted by the model. 3 7 To maximize the number of points available for modeling, missing data for NMHCs, acetone, and CH OH are filled where possible. During SUCCESS, no 3 measurements of these species were available, so concentrations are assumed to equal linear functions of CO derived from data during PEM-West B [McKeen et al., 1997]. A few missing upper tropospheric acetone and CH OH measurements 3 during TRACE-P are similarly filled using correlations to CO derived from data during that campaign. Otherwise, missing NMHCs, acetone, and methanol during SONEX and TRACE-P are linearly interpolated from adjacent data points within +/- 15 minutes and +/- 500 m altitude. For this analysis, analyzed data are limited to solar zenith angles (SZA) less than 80°. As in previous studies, photolysis frequencies are based on spectroradiometer measurements [Shetter and M(cid:0) ller, 1999]. Diurnal variation of these measured photolysis frequencies are based on clear-sky model calculations using a DISORT four-stream implementation of the NCAR Tropospheric Ultraviolet Visible (TUV) radiative transfer code [Madronich and Flocke, 1998]. The clear sky diurnal variation from TUV is normalized such that it matches the measured photolysis frequency at the time of observation. Unmeasured photolysis frequencies are normalized by the ratio of observed to clear-sky photolysis of NO . Dry and wet deposition for soluble species is 2 implemented as in Logan et al. [1981]. Aerosol and cloud uptake for HO is not 2 directly computed in the model such that potential impacts may be inferred by 8 evaluating correlations between model-to-measurement agreement and aerosol surface area. Current reactions and rates for basic HO -NO -CH -CO chemistry in the x x 4 NASA Langley time-dependent photochemical box model are generally those recommended by Sander et al. [2003]. In addition to these, the model uses the rate for O1D quenching by N given by Ravishankara et al. [2002], temperature 2 dependent quantum yields for acetone photolysis from Blitz et al. [2004], and the parameterization for near-IR photolysis of HNO as described in Roehl et al., 4 [2002]. NMHC chemistry is built from that in Lurmann et al. [1986], with appropriate rate updates from IUPAC recommendations (Atkinson et al, [2003]) when recommendations from Sander et al. [2003] are not available. For sensitivity studies, an earlier version of the model is run to reflect reactions, rates, and photolytic information available circa 1997 [DeMore et al., 1997; Lurmann et al., 1986; Atkinson et al., 1992]. 3.0 Review of previous findings A prominent and often cited uncertainty in HO theory diagnosed from x early analyses of these airborne data sets arises from the tendency for models to underpredict HO under high NO conditions. Figure 1 shows the HO observed- 2 x 2 to-calculated ratio (obs/calc) as a function of NO from SUCCESS and SONEX data originally presented in Faloona et al. [2000], but reproduced here with the Langley model (circa 1997 version). During SONEX, model predictions exceed 9 observations by up to a factor of 5 for NO greater than several hundred pptv x [Brune et al., 1999; Jaegle et al., 1999; Faloona et al., 2000]. A similar but more dramatic correlation is evident from the SUCCESS data, with HO underpredicted 2 by factors of up to several hundred for NO at ppbv levels [Brune et al., 1998; Tan x et al., 1998; Faloona et al., 2000]. Subsequent analyses have revealed similar behavior in other data sets, particularly ground-based studies, including conditions in an urban setting (NY city) [Ren et al., 2003], in the Nashville urban plume [Thornton et al., 2002; Martinez et al., 2003], and in more remote settings [Kanaya et al., 2001; 2002]. Olson et al. [2004] also found during TRACE-P that for the upper 5% of NO observations (i.e., greater than 135 pptv), upper tropospheric data exhibited a tendency for relatively higher obs/calc ratios of HO 2 while lower tropospheric data showed no such tendency. The possibility of instrument artifacts contributing to these discrepancies has been investigated and is described as unlikely [Faloona et al., 2000; Ren et al., 2004]. Potential explanations suggested in the literature include missing or incomplete theory for HO -NO chemistry [Brune et al., 1999; Faloona et al., 2000], and the presence of x x unmeasured HO precursors transported to higher altitudes by convection in x conjunction with high concentrations of NO from convection or lightning [Brune x et al., 1999; Jaegle et al., 2001]. Another discrepancy revealed from earlier analyses of airborne data sets relates to the uptake of HO onto liquid cloud particles or onto ice in cirrus clouds 2 [Faloona et al., 1998; Jaegle et al., 2000]. Jaegle et al. [2000] reported model 10 HO overestimates of factors of 1.5 to 2 times within upper tropospheric cirrus 2 clouds during SONEX, though large overpredictions also extended to areas well outside the cirrus. Laboratory studies measuring the uptake of OH and HO onto 2 ice are sparse, though at least one study reports the measurement of reactive uptake coefficients [Cooper and Abbatt, 1996]. Alternately, Olson et al. [2004] found that evidence of significant in-cloud HO loss during TRACE-P was 2 primarily limited to the lower and middle altitudes. While the median HO 2 overprediction within clouds at middle and lower altitudes (< 6 km) was 31% larger than that for clear air points, the difference was only 9% at altitudes above 6 km. Discrepancies in HO predictions have also been identified at high solar 2 zenith angles, with model underpredictions of up to a factor of 5 during SONEX at both sunrise and sunset [Brune et al., 1999; Jaegle et al., 2000]. Jaegle et al. [2000] determined that while photolysis of HONO produced at night by heterogeneous conversion of NO could account for an additional source of HO 2 2 some time after sunrise, it could not explain the non-zero HO observations prior 2 to, at, and just after sunrise. Other proposed missing sources of HO at high solar 2 zenith angles include photolysis of peroxides and CH O at sunrise, near-IR 2 photolysis of HNO , and uncertainties in HNO formation and loss kinetics [Brune 4 4 et al. 1999; Jaegle et al., 2000]. In this reanalysis, additional explanations for the HO discrepancies are x proposed, such as unsuitable modeling approaches. These include using model