An advanced method of contributing emissions to short-lived chemical species (OH and HO2) The TAGGING 1.1 submodel based on the Modular Earth Submodel System

. To mitigate the human impact on climate change, it is essential to determine the contribution of emissions to the concentration of trace gases. In particular, the source attribution of short-lived species such as OH and HO 2 is important as they play a crucial role for atmospheric chemistry. This study presents an advanced version of a tagging method for OH and HO 2 (HO x ) which attributes HO x concentrations to emissions. While the former version (V1.0) only considered 12 reactions in the troposphere, the new version (V1.1), presented here, takes 19 reactions in the troposphere into account. For the ﬁrst time, the main chemical reactions for the HO x chemistry in the stratosphere are also regarded (in total 27 reactions). To fully take into account the main HO 2 source by the reaction of H and O 2 , the tagging of the H radical is introduced. In order to ensure the steady-state assumption, we introduce rest terms which balance the deviation of HO x production and loss. This closes the budget between

Abstract. To mitigate the human impact on climate change, it is essential to determine the contribution of emissions to the concentration of trace gases. In particular, the source attribution of short-lived species such as OH and HO 2 is important as they play a crucial role for atmospheric chemistry. This study presents an advanced version of a tagging method for OH and HO 2 (HO x ) which attributes HO x concentrations to emissions. While the former version (V1.0) only considered 12 reactions in the troposphere, the new version (V1.1), presented here, takes 19 reactions in the troposphere into account. For the first time, the main chemical reactions for the HO x chemistry in the stratosphere are also regarded (in total 27 reactions). To fully take into account the main HO 2 source by the reaction of H and O 2 , the tagging of the H radical is introduced. In order to ensure the steady-state assumption, we introduce rest terms which balance the deviation of HO x production and loss. This closes the budget between the sum of all contributions and the total concentration. The contributions to OH and HO 2 obtained by the advanced tagging method V1.1 deviate from V1.0 in certain source categories. For OH, major changes are found in the categories biomass burning, biogenic emissions and methane decomposition. For HO 2 , the contributions differ strongly in the categories biogenic emissions and methane decomposition. As HO x reacts with ozone (O 3 ), carbon monoxide (CO), reactive nitrogen compounds (NO y ), non-methane hydrocarbons (NMHCs) and peroxyacyl nitrates (PAN), the contributions to these species are also modified by the advanced HO x tagging method V1.1. The contributions to NO y , NMHC and PAN show only little change, whereas O 3 from biogenic emissions and methane decomposition increases in the tropical troposphere. Variations for CO from biogenic emissions and biomass burning are only found in the Southern Hemisphere.

Introduction
The radicals hydroxyl (OH) and hydroperoxyl (HO 2 ) are crucial for atmospheric chemistry. Both radicals are very reactive and have a lifetime of only a few seconds. OH is frequently converted to HO 2 and vice versa. Thus, OH and HO 2 radicals are closely linked and often referenced together as the chemical family HO x . The ratio of OH to HO 2 in an air parcel strongly depends on the chemical background, in particular on the composition of nitrogen oxides NO x (= NO + NO 2 ) and non-methane hydrocarbons (NMHC) (Heard and Pilling, 2003). HO x impacts global warming and local air quality in various ways: by reacting with greenhouse gases such as methane (CH 4 ) and ozone (O 3 ), OH reduces their atmospheric residence time (e.g. Stevenson et al., 2006;Voulgarakis et al., 2013;Righi et al., 2015). Hence, HO x controls the impact of CH 4 and O 3 on global warming. Moreover, being the main oxidizer in the troposphere, OH is involved in the decomposition of pollutants and in the production of ground-level ozone, photochemical smog and secondary organic aerosols (e.g. Lawrence et al., 2001;Heard and Pilling, 2003). Consequently, to quantify the human impact on cli-Published by Copernicus Publications on behalf of the European Geosciences Union. 2050 V. S. Rieger et al.: The TAGGING 1.1 submodel mate and air quality, it is essential to understand the distribution and variability of OH and HO 2 in the atmosphere.
However, the determination of OH and HO 2 concentrations in the atmosphere is still challenging due to their short lifetimes. In field campaigns HO x concentrations are measured on a local scale, which is generally difficult to compare with global models (e.g. Ren et al., 2003;Olson et al., 2006). For certain environments, such as the marine boundary layer, model studies compare well with measurements. Other regions, such as unpolluted forest areas, show large discrepancies (Heard and Pilling, 2003;Stone et al., 2012). On regional and global scales, no direct HO x measurements are available. So far, OH concentration and its inter-annual variability can only be estimated indirectly by measurements and emission rates of methyl chloroform (CH 3 CCl 3 ) (Prinn et al., 2005;Montzka et al., 2011). As emissions of CH 3 CCl 3 steadily decline, Liang et al. (2017) suggest an alternative method: they combine several trace gases such as CH 2 F 2 , CH 2 FCF 3 , CH 3 CHF 2 and CHClF 2 in a gradient-trend-based two-box model approach to derive a global OH concentration of 11.2 × 10 5 molec cm −3 . Overall, global chemistry climate models estimate a tropospheric OH concentration of around 11 × 10 5 molec cm −3 , which compares well with the observation-based results from Prinn et al. (2005) and Liang et al. (2017).
To mitigate the human impact on climate change or pollution in general, it is crucial to determine the contribution of an emission sector to the concentration of certain chemical species (Grewe et al., 2012;Clappier et al., 2017). To do so, we use a "tagging" method: the theoretical framework of this tagging method is given in Grewe et al. (2010) and Grewe (2013), and the implementation is described in Grewe et al. (2017). This method splits up all chemical species which are important for O 3 production and destruction into 10 source categories: emissions from anthropogenic non-traffic (e.g. industry and households), road traffic, shipping, aviation, biogenic sources, biomass burning, lightning, methane (CH 4 ) and nitrous oxide (N 2 O) decompositions and stratospheric ozone production. Subsequently, the contributions of these sources to the concentrations of O 3 , CO, OH, HO 2 , peroxyacyl nitrates (PANs), reactive nitrogen compounds (NO y , e.g. NO, NO 2 , HNO 4 ) and non-methane hydrocarbons (NMHC) are diagnosed. The contribution calculations are based on chemical reaction rates, online emissions (e.g. lightning), offline emissions (e.g. road traffic) and deposition rates. Emissions of NO and NO 2 contribute to the NO y concentration, while emissions of e.g. C 2 H 4 , C 3 H 6 and HCHO contribute to the NMHC concentration. This tagging method considers the competition of NO y , CO and NMHC in producing and destroying O 3 .
The tagging method of the long-lived species O 3 , CO, PAN, NO y and NMHC and of the short-lived species OH and HO 2 is based on the same principles of apportioning the contributions. (In this study, O 3 , CO, PAN, NO y and NMHC are denoted as long-lived species because their atmospheric life-time is significantly longer then the lifetime of OH and HO 2 .) However, the implementation for long-lived and short-lived species differs. For the long-lived species, each source tracer is transported, receives the corresponding online or offline emissions, is deposited and reacts with other species. Based on these processes, the tagging method determines the concentration of the source tracers. A detailed description of the implementation of the tagging method for long-lived species is given in Grewe et al. (2017).
However, the short-lived species HO x are not transported and experience neither emission nor deposition. Thus, the same implementation of the tagging method as for long-lived species is not possible. Tsati (2014) and Grewe et al. (2017) introduced a modified approach for tagging HO x : since the lifetime of OH and HO 2 is very short, a steady state between the production and destruction of OH and HO 2 is assumed. Using the main chemical reactions of HO x chemistry, the contributions of each source category to OH and HO 2 are determined.
The contributions to long-lived and short-lived species are closely linked (see Fig. 1). For example, the reaction involves the long-lived species O 3 and the short-lived species OH and HO 2 . Hence, this reaction is considered in the implementation of the tagging method for long-lived and shortlived species. The contribution of, for example, shipping emissions to O 3 influences the contribution of shipping emissions to HO 2 : the higher the contribution to O 3 , the more HO 2 is attributed to shipping emissions. Furthermore, OH from shipping emissions destroys O 3 and thus reduces the contribution of shipping emissions to O 3 .
The implementation of the tagging method for the shortlived species HO x , presented by Grewe et al. (2017), is referred to as the HO x tagging method V1.0. It did not consider all relevant reactions for the production and loss of HO x . In particular, the reactions which are important in the stratosphere were not taken into account. Moreover, the steadystate assumption between HO x production and loss was not fulfilled. In this study, we present a revised version V1.1 of the HO x tagging method, largely improving these shortcomings. It includes the main chemical reactions of HO x chemistry in the troposphere and stratosphere. This is enabled by introducing the tagging of the hydrogen radical (H). Special care is taken for the steady-state assumption.
The paper is structured as follows: after introducing the model set-up in Sect. 2, we present the advanced HO x tagging method V1.1 in Sect. 3. In Sect. 4, the results are compared with the tagging method V1.0 by Grewe et al. (2017). Finally, Sect. 5 concludes the methods and the results of this study.

Model description of EMAC and MECO(n)
To evaluate the further developed HO x tagging method we use the same model set-up as Grewe et al. (2017). A global climate simulation is performed with the ECHAM/MESSy Atmospheric Chemistry (EMAC) chemistry climate model. EMAC is a numerical chemistry and climate simulation system that includes submodels describing tropospheric and middle atmosphere processes and their interaction with oceans, land and human influences . It uses the second version of the Modular Earth Submodel System (MESSy2.53) to link multi-institutional computer codes. The core atmospheric model is the 5th generation European Centre Hamburg general circulation model (ECHAM5; Roeckner et al., 2006). For the present study we apply EMAC in the T42L90MA resolution, i.e. with a spherical truncation of T42 (corresponding to a quadratic Gaussian grid of approx. 2.8 by 2.8 • in latitude and longitude) with 90 vertical hybrid pressure levels up to 0.01 hPa. For the simulation presented in this study, the time span of July 2007 to December 2008 is considered: half a year as a spin-up and 1 year for the analysis. For the chemical scheme, we use the submodel MECCA (Module Efficiently Calculating the Chemistry of the Atmosphere), which is based on Sander et al. (2011) andJöckel et al. (2010). The chemical mechanism includes 218 gasphase, 12 heterogeneous and 68 photolysis reactions. In total 188 species are considered. It regards the basic chemistry of OH, HO 2 , O 3 , CH 4 , nitrogen oxides, alkanes, alkenes, chlo-rine and bromine. Alkynes, aromatics and mercury are not considered.
Total global emissions of lightning NO x are scaled to approximately 4 Tg(N) a −1 (parameterized according to Grewe et al., 2001). The submodel ONEMIS (Kerkweg et al., 2006) calculates NO x emissions from soil (parameterized according to Yienger and Levy, 1995) and biogenic C 5 H 8 emissions (parameterized according to Guenther et al., 1995). Direct CH 4 emissions are not considered, and instead pseudoemissions are calculated using the submodel TNUDGE (Kerkweg et al., 2006). This submodel relaxes the mixing ratios in the lowest model layer towards observations by Newtonian relaxation (more details are given by . To show the effect of the HO x tagging method on a regional scale, a further simulation with the coupled model system MESSyfied ECHAM and COSMO models nested n times (MECO(n)) is performed. The nested system couples the global chemistry climate model EMAC online with the regional chemistry climate model COSMO/MESSy (Kerkweg and Jöckel, 2012a, b). To test the HO x tagging in MECO(n), we conduct a simulation using one COSMO/MESSy nest over Europe with a resolution of 0.44 • . EMAC is applied in a horizontal resolution of T42 with 31 vertical levels. The period from July 2007 to December 2008 is simulated. The set-up of the simulation is identical to the one described in Grewe et al. (2017). A detailed chemical evaluation of the set-up is given in Mertens et al. (2016).
Both model simulations are based on the quasi chemistrytransport model (QCTM) mode in which the chemistry is decoupled from the dynamics (Deckert et al., 2011). The anthropogenic emissions are taken from the MACCity emission inventory (Granier et al., 2011). The TAGGING submodel (as described by Grewe et al., 2017) is coupled to the detailed chemical solver MECCA from which it obtains information about tracer concentrations and reaction rates. Based on this information, it calculates the contributions of source categories to O 3 , CO, NO y , PAN and NMHC concentrations. The contributions of OH and HO 2 are calculated with the advanced method V1.1 presented in the next section. The implementation is based on MESSy2.53 and will be available in MESSy2.54.
3 Tagging method of short-lived species 3.1 Tagging method V1.0 The tagging method V1.0 described by Grewe et al. (2017) determines the contribution of source categories to O 3 , NO y , CO, NMHC, PAN, OH and HO 2 concentrations. A total of 10 source categories are considered, and every species included in the tagging method is decomposed into these categories: for example, the concentration of O 3 is split up into O 3 produced by anthropogenic non-traffic (e.g. industry) emissions ( (O CH 4 3 ), nitrous oxide decomposition (O N 2 O 3 ) and stratospheric ozone production (O str 3 ). These tagged species go through the same chemical reactions and the same deposition loss processes as O 3 . The tagging method uses a combinatoric approach to determine the contributions: it redistributes the production and loss rates of each species to the 10 source categories according to the concentrations of the tagged species. Details on the tagging theory and implementation in EMAC and MECO(n) are found in Grewe (2013) and Grewe et al. (2017), respectively.
For the first time, V1.0 determined the contribution of source categories to OH and HO 2 concentrations. The tagging method V1.0 was based on 12 reactions for the HO x chemistry (reactions marked with "o" in last column of Table 1). It included the main production and loss reactions of HO x with O 3 , NO y , NMHC, CO and CH 4 . V1.0 only regarded reactions which are important in the troposphere. Reactions which mainly occur in the stratosphere were not taken into account. However, the main HO 2 production by the Reaction (1) H + O 2 −→ HO 2 (see Table 1) was not regarded. It was combined with Reaction (11), CO + OH −→ H + CO 2 (see Table 1), to But not all H radicals in the troposphere are produced by the reaction of CO + OH. Reactions (7) OH + O( 3 P), (10) H 2 + OH and (28) HCHO + hv also produce H (Table 2). These reactions were neglected in V1.0. Thus, only 80 % of the H production and therefore only 80 % of the HO 2 production by Reaction (1) was considered in the troposphere. In the stratosphere, the reaction of CO + OH becomes less important and most H is produced by Reactions (7) and (28). Consequently, only 6 % of the H and thus of the HO 2 production by Reaction (1) was regarded in this approach. (Numbers are derived from an EMAC simulation as described in Sect. 2.) In the troposphere, the most important reactions not covered in V1.0 are Reaction (1) H + O 2 , Reaction (15) NO 2 + HO 2 and Reaction (18) for the decomposition of HNO 4 . In the stratosphere, Reactions (1) H + O 2 , (5) HO 2 + O( 3 P) and (7) OH + O( 3 P) play a leading role and were not included in V1.0.
Most reaction rates used in the tagging method correspond to the production and loss rates directly provided by the chemical scheme MECCA of EMAC. However, for reactions with NMHC, the reaction rates were obtained indirectly. The reaction rate of OH with NMHC (Reaction 21, Table 1) was determined via the production rates of CO by assuming that each reaction of OH with NMHC produces one CO molecule. This method neglects all intermediate oxidation reactions of NMHC and considers only these reactions when NMHC is finally oxidized to CO. For the reaction rates of NO y and HO 2 with NMHC (Reactions 22 and 23), only the reaction of HO 2 with the methylperoxy radical (CH 3 O 2 ) was considered.
To derive the contributions to OH and HO 2 , a steady state between HO x production and loss was assumed. However, the steady-state assumption was not completely fulfilled for V1.0 (see Sect. 3.4). Moreover, the sum of the contributions of the 10 source categories to the OH and HO 2 concentrations did not equal the total OH and HO 2 concentrations. It deviated by about 70 %.
3.2 Reduced HO x reaction system V1.1 OH and HO 2 react with many chemical species. To reduce the calculation time of a simulation, we reduce the HO x chemistry used in the chemical scheme MECCA to the most important reactions which occur in the troposphere and stratosphere. We consider only reactions with a tropospheric or stratospheric annual mean reaction rate larger than 10 −15 mol mol −1 s −1 (see Table 1). Hence, we increase the number of reactions from 12 (V1.0) to 27 (V1.1), which still constitutes a reduced set of reactions compared to the full chemical scheme MECCA used in EMAC. In the following, we call this set reduced H O x reaction system V1.1.
The reactions which are important in the troposphere are indicated in Table 1. As stated above, Reaction (1) of H and Table 1. The reduced HO x reaction system V1.1 describes the main reactions of HO x chemistry in the troposphere and stratosphere. These 27 reactions are used for the tagging method V1.1. In the column "tropos." ("stratos."), reactions which are important in the troposphere (stratosphere) are marked. In the column "V1.1", reactions marked with "o" were already included in V1.0. Reactions marked with "x" are added in V1.1. Reactions marked with "(x)" were only partly taken into account in V1.0. The numbers of reactions are referenced in the text.

Reaction
Rates Tropos. Stratos. V1.1  (14), which produces 21 % of OH and destroys 24 % of HO 2 . Since large quantities of O 3 are found in the stratosphere, O 3 or the excited oxygen radical (O( 3 P)) destroys about 62 % of HO 2 . Reactions with NMHC, CO and CH 4 play only a minor role in the stratosphere.
Reactions of OH and HO 2 with chlorine and bromide were not considered in V1.0. We add these reactions, which take place only in the stratosphere, to the tagging method V1.1. Reactions (21) to (25) involve the chemical family NMHC, which contains several species such as formaldehyde (HCHO), ethylene (C 2 H 4 ) and propane (C 3 H 8 ). The rate for Reaction (21) is determined by adding up the rates of all reactions of OH with each single species of the family NMHC. The reaction rate (23) contains all rates of the reactions between the species of the chemical families NO y and NMHC. All reaction rates are directly derived by the MECCA mechanism of EMAC. Table 1 does not consider all reactions with annual reaction rates larger than 10 −15 mol mol −1 s −1 . The photolysis of hydrogen peroxide (H 2 O 2 ), hypochlorous acid (HOCl) and hypobromous acid (HOBr) is excluded from the reduced HO x reaction system V1.1 as the tagging method cannot be applied. The specific reasons are explained in Appendix A.

Deductions of tagged species
To derive how much OH and HO 2 is produced and destroyed by a source category i, the tagging approach described in Grewe et al. (2010Grewe et al. ( , 2017 is used. In general, bimolecular reactions with two chemical species A + B −→ C are tagged as follows: each tagged species is split up into its contribution from n source categories A = n i=1 A i , B = n i=1 B i and C = n i=1 C i . These contributions (A i , B i , C i ) go through the same reactions as their main species (A, B, C). If A from category i reacts with B from category j , then the resulting species C belongs half to the category i and half to the category j : Consequently, the production P and loss L of a species from the category i (here LossA i , LossB i and ProdC i ) are determined by regarding all possible combinations of the reaction between A i and B j : with k being the reaction rate coefficient and R = k A B being the respective reaction rate. For unimolecular reactions A −→ B + C, the distribution of categories from the educts is completely passed to the products: with the reaction rate R = kA.
As described above, the long-lived species O 3 , CO, NO y and NMHC are tagged according to the tagging method described in Grewe et al. (2017). To limit memory demand, other species such as H 2 , H 2 O 2 , CH 4 , ClO and BrO are not tagged (as in V1.0). Here, different approaches are derived to retain the ratio of the contribution to total concentration A i A .
1. If a tagged species reacts with a non-tagged species, the non-tagged species does not contribute and the tagging method for a unimolecular reaction is applied (see Eq. 3). Examples are Reactions (9), (10) and (13).
2. Using the family concept as described in Grewe et al. (2017) allows for the assumption that all tags are distributed equally among the species within the same chemical family.
As mentioned in Grewe et al. (2017), all species which are frequently converted back and forth to ozone are considered as an "ozone storage" (Crutzen and Schmailzl, 1983). These species together with O 3 are lumped into one chemical family: ozone. Both O( 1 D) and O( 3 P) belong to this chemical family. Hence, as in Grewe et al. (2017), we apply the family concept and set 3. In Reaction (1), neither H nor O 2 is tagged. To obtain the ratio HO 2 i HO 2 , we set up an extra tagging of H itself. As the H radical is very reactive, we assume that H production balances H loss (see Sect. 3.4). Table 2 presents the main reactions for H, which still constitute a subset of full H chemistry implemented in MECCA. Based on Table 2, we set up the H production ProdH i and H loss LossH i for the contribution of a source category i.

Steady-state assumption
The steady-state assumption of the HO x chemistry is the basic principle of the tagging method for short-lived species (Tsati, 2014;Grewe et al., 2017). In steady state, the production and loss of OH and HO 2 balance each other. Table 3 shows the annual means of the HO x and H production and loss rates of the reduced reaction system for the tagging methods V1.0 and V1.1 as well as the total production and loss rates derived from the complete chemical scheme MECCA in EMAC. The production and loss rates are obtained from an EMAC simulation following the set-up described in Sect. 2. Note that for V1.0 no values for the H production and loss are available since the tagging of H was not considered in V1.0.
In general, total OH production (derived by MECCA) equals total OH loss in the troposphere and stratosphere. The same holds for HO 2 and H. In the troposphere, the OH loss of V1.1 and V1.0 represents the total OH loss in the troposphere well. However, the OH production for V1.1 and V1.0 differs by 12 % from the total OH production. Considering HO 2 in the troposphere, the total production and loss rates are well reflected by V1.1. In contrast, the HO 2 production and loss of V1.0 differs by 14 and 41 % from the total rates.
In the stratosphere, V1.1 represents the total rates very well. However, the OH production of V1.1 misses 10 % of the total OH production. Since V1.0 was only developed for the troposphere, not all reactions which are important in the stratosphere were considered. Thus, the OH and HO 2 production and loss rates of V1.0 considerably underestimated the total production and loss rates.
The reduced H reaction system in V1.1 (Table 2) represents the total H production and loss in the troposphere very well. However, in the stratosphere H loss in V1.1 deviates by 17 % from the total H loss.
Summing up, the reduced HO x reaction system V1.1 represents the total HO x production and loss in the troposphere and stratosphere well. V1.1 reproduces the HO x chemistry better than V1.0. However, OH production in the troposphere and stratosphere as well as H loss in the stratosphere of V1.1 deviate from the total rates derived by MECCA. Thus, the steady state for the reduced HO x and H reaction system (Tables 1 and 2) is not completely fulfilled.
But steady state between production and loss is crucial for the tagging method for short-lived species. To re-establish steady state, it would be necessary to include the complete HO x and H chemistry in the tagging method. However, this is not possible as the tagging method of short-lived species does not apply to all reactions of the HO x and H chemistry (for examples see Appendix A). Moreover, tagging all chemical species of the HO x and H chemistry with the implementation of long-lived species would significantly increase the memory demand of a climate simulation (for a detailed discussion see Sect. 6 in Grewe et al., 2017). Consequently, we introduce the rest terms resOH, resHO 2 and resH for OH, HO 2 and H to compensate for the deviations from steady state. Each rest term is calculated by subtracting the production rate of the reduced reaction system from the loss rate (Tables 1 and 2). The resulting rest terms are shown in the Supplement (Fig. S1).
Considering the rest terms resOH, resHO 2 and resH leads to the closure of the budget. In V1.0, the sum of the contributions from all source categories did not balance the total concentration. The averaged deviations for OH and HO 2 in the troposphere were about 70 % of the total concentrations. Since the stratosphere was not considered in V1.0, the deviations were even larger (104 % for OH and 89 % for HO 2 ). In V1.1, the sum of OH and HO 2 now balances the total OH and HO 2 concentrations. The deviations are negligible (below 10 −3 %). Consequently, including the rest terms in the tagging method is mandatory for the steady-state assumption and also closes the budget.

Determination of HO x contributions
Taking the above considerations into account, we finally derive the OH and HO 2 production and loss terms per source category i. In the reduced HO x reaction system V1.1 (Table 1), OH is produced by the Reactions (2) (1) is not tagged. To be able to determine the HO 2 production by Reaction (1) The HO 2 loss is determined by Reactions (3) HO 2 + HO 2 , (4) HO 2 + O 3 , (5) HO 2 + O( 3 P), (8) HO 2 + OH, (14) NO + HO 2 , (15) NO 2 + HO 2 , (22) NMHC + HO 2 , (26) ClO + HO 2 and (27) Section 3.4 shows that the steady-state assumption for OH and HO 2 is justified when the rest terms resOH, resHO 2 and resH are regarded. Therefore, the rest terms are divided by the number of source categories n to add them to the contributions of a category i. In steady state, production of OH i and HO 2 i equals the loss.
Equations (13) and (14) are rewritten as follows: with the variables P OH , L OH , P HO 2 , L HO 2 , A i and B i as follows (compare to Grewe et al., 2017 Eqs. 25 to 28).
By solving Eqs. (15) and (16), we finally obtain the contributions of a source category i to the OH and HO 2 concentration (same equations as Eqs. 29 and 30 in Grewe et al., 2017, but with differently defined coefficients).
These equations are implemented in the TAGGING submodel, and EMAC and MECO(n) simulations according to Sect. 2 are performed. The results for the OH and HO 2 contributions are analysed and compared with V1.0 in the following section.  (Figs. B1, B2). First, the OH and HO 2 contributions of V1.1 are described in the following. For the categories which are determined by anthropogenic emissions, such as shipping, road traffic and anthropogenic non-traffic, the maximum values of OH and HO 2 contributions occur in the lower troposphere in the Northern Hemisphere. This clearly shows that for anthropogenicdominated categories the OH and HO 2 contributions are caused by anthropogenic emissions. The contributions vary among these categories of surface emissions as not only the amount but also the composition of the emissions differs. For the category aviation, maximum OH contributions are found in the Northern Hemisphere between 200 and 250 hPa. However, the HO 2 contribution has a minimum in this region and a maximum in the lower troposphere. The OH values for the categories CH 4 decomposition, N 2 O decomposition, lightning and biogenic emissions are largest in the upper troposphere. Most OH contributions of biomass burning are found in the lower tropical troposphere. In contrast, negative values occur in the upper tropical troposphere. Concerning the HO 2 contribution, the residual categories show a maximum in the tropical lower troposphere. In addition, the category lightning shows a strong HO 2 loss in the upper tropical troposphere, which is caused by Reaction (14).
The results obtained by V1.1 are compared to the OH and HO 2 zonal profiles of V1.0 only in the troposphere (Figs. 2  and 3). The HO x tagging method V1.0 was only developed for the troposphere. Hence, a comparison in the stratosphere is not reasonable. In general, contributions to OH and HO 2 concentrations of V1.1 are larger in the troposphere compared to V1.0. This overall shift towards larger values is explained by the re-establishment of the steady state and thus the closure of the budget in V1.1. In V1.0 the budget was not closed and thus the contributions were underestimated.
For OH, the categories lightning and aviation show no large changes in the general pattern of the zonal means between V1.0 and V1.1. Considering the HO 2 contributions, no large changes are found for the categories biomass burning, anthropogenic non-traffic, road traffic and shipping.
The contribution of the category aviation to HO 2 in V1.1 shows roughly the same pattern compared to V1.0. However, the HO 2 destruction along the flight path is no longer as pronounced, which is caused by the inclusion of Reactions (15) and (18) Figure 5. Annual mean contributions of 10 source categories to O 3 concentration in percent. tions 7, 10, 11 and 28) leading to a larger HO 2 production. Also the addition of Reactions (15) and (18) (for an explanation see above) as well as the addition of Reaction (23), which considers more reactions than in V1.0, increases the HO 2 contribution of the categories N 2 O decomposition and lightning.
Large changes in pattern are observed for the contributions of biogenic emissions and CH 4 decomposition to OH and HO 2 as well as for the contributions of biomass burning and anthropogenic non-traffic to OH. In V1.1, these categories mainly constitute a source of OH and HO 2 in the troposphere. The addition of Reactions (24) and (25) to the reduced HO x reaction system V1.1 presents an HO 2 source increasing OH and HO 2 contributions. Furthermore, reactions of NMHC with OH, HO 2 and NO y (Reactions 21, 22 and 23) are important throughout the whole troposphere. In contrast to V1.0, V1.1 considers all reactions of NMHC with OH, HO 2 and NO y (see Sect. 3.2), significantly changing the pattern of biogenic emissions, CH 4 decomposition, biomass burning and anthropogenic non-traffic.
To demonstrate the impact of the advanced HO x tagging method on a regional scale, Fig. 4 shows the contributions of ship emissions to OH and HO 2 in the boundary layer simulated with the high-resolution model MECO(n) (see Sect. 2). The ship paths in the Atlantic, Mediterranean and Red Sea are clearly visible and lead to OH and HO 2 production along these paths. In the polluted area at the coast of Marseille the OH and HO 2 contributions are reduced. In this region NO y from shipping emissions is larger than in the Mediterranean Sea, causing a reduction of OH and HO 2 by Reactions (14) to (17).
The tagging method V1.0 (Grewe et al., 2017, their Fig. 6) showed negative HO 2 shipping contributions along the ship paths. This was explained by Reaction (14): NO destroys HO 2 and leads to negative contributions. However, in V1.1 HO 2 shipping contributions are positive. The change in sign is caused by the addition of Reactions (15) and (18) to the reduced HO x reaction system V1.1, which constitutes a net HO 2 production, leading to positive HO 2 contributions (for an explanation see above). The comparison shows that HO 2 contributions in V1.0 were systematically and erroneously underestimated.
To summarize, the contributions to OH and HO 2 concentrations show larger values in V1.1 compared to V1.0. This is explained by the re-establishment of the steady state. For OH, no large changes are found in the categories lightning and aviation. However, large changes are found for biomass burning, CH 4 decomposition and biogenic emissions. For HO 2 , no large differences occur in the categories biomass burning, anthropogenic non-traffic, road traffic and shipping. In comparison, the categories biogenic emissions and CH 4 decomposition differ strongly. The differences between the contributions of V1.1 and V1.0 are traced back to the addition of certain reactions to the reduced reaction system considered in the HO x tagging method.

Effects on long-lived species
The tagging of short-lived and long-lived species closely intertwines (see Fig. 1). Changes in the contributions to OH and HO 2 influence the contributions to the long-lived tracers O 3 , NO y , CO, NMHC and PAN. For example, Fig. 5 shows the zonal mean of the contributions of the 10 source categories to O 3 . Grewe et al. (2017) present the same figure for the HO x tagging method V1.0 (their Fig. 4). For consistency, we compare our results with the results of Grewe et al. (2017) only for the year 2008. In general, no large differences between V1.1 and V1.0 for long-lived species are found. The categories biogenic emissions and CH 4 decomposition show an O 3 increase in the tropical troposphere. Stratospheric O 3 production slightly increases in the Southern Hemisphere. Small O 3 changes are found for the categories lightning and N 2 O decomposition. Regarding the remaining long-lived species (see Figs. S3-S6), the contribution of biomass burning to CO decreases, while the contributions of biogenic emissions to CO increase in the Southern Hemisphere. The remaining sectors stay rather unchanged. NO y , NMHC and PAN show only minor changes. Even though major differences in OH and HO 2 occur between V1.0 and V1.1, these do not have a large effect on the long-lived species.

Discussion and conclusion
We present an extension of the HO x tagging method described by Grewe et al. (2017). A total of 15 new reactions producing and destroying HO x are added to the tagging mechanism. In Grewe et al. (2017), the HO x tagging method V1.0 was restricted to the troposphere only. We further include the reactions which are essential for HO x production and loss in the stratosphere. Moreover, we introduce an equivalent tagging method to obtain the contributions to the H radical. This step is mandatory to fully account for the main HO 2 source: the reaction of H with O 2 .
In V1.0, the steady-state assumption was not completely fulfilled, resulting in an unclosed budget: the sum of the HO x contributions and the total HO x concentration deviated by about 70 %. To re-establish steady state, we add more reactions to the reduced HO x reaction system and introduce rest terms to balance the deviation of HO x production and loss. This leads to the closure of the budget. Thus, the tagging mechanism introduced by Grewe et al. (2010) operates not only for long-lived but also for short-lived species.
The advanced HO x tagging method V1.1 was implemented in the global chemistry climate model EMAC and in the regional model MECO(n). A 1-year simulation was performed in both model systems and compared to V1.0. For most categories, the general zonal pattern of the contributions to OH and HO 2 show minor differences. In contrast, large changes are observed in the category biogenic emis-sions and CH 4 decomposition, which are traced back to the addition of certain reactions to V1.1. Although the contributions of long-lived and short-lived species influence each other, no large changes are found for long-lived species.
The mechanism presented in this study (and introduced by Tsati, 2014, andGrewe et al., 2017) is the first method for tagging short-lived species. Other studies quantify the source attributions of chemical species with a significantly longer lifetime. The idea of source attribution is applied to attribute CO to different emission types and regions (e.g. Granier et al., 1999;Pfister et al., 2004Pfister et al., , 2011, to attribute NO x concentrations to emission sources (Horowitz and Jacob, 1999) or to trace stable isotopic compositions . Also for the source attribution of tropospheric O 3 , there are several tagging approaches attributing tropospheric O 3 only to NO x sources (Lelieveld and Dentener, 2000;Grewe, 2004;Grewe et al., 2012;Emmons et al., 2012), only to NMHC sources (Butler et al., 2011;Coates and Butler, 2015) or to NO y , CO and NMHC emissions simultaneously (Grewe et al., 2017).
A common technique to quantify the impact of emissions on OH is the so-called perturbation method, which compares two simulations: one simulation with all emissions and one simulation with reduced emissions (e.g. Niemeier et al., 2006;Hoor et al., 2009). However, if the underlying chemical processes are non-linear (as is the case for OH), the perturbation method largely underestimates the contribution (Grewe et al., 2012;Emmons et al., 2012;Mertens et al., 2018). Consequently, the tagging approach presented in this study delivers the actual contribution of the emission source, while the perturbation method displays the impact of the emission reduction.
To conclude, the further developed HO x tagging method can be used to identify the contribution of anthropogenic emissions to the atmospheric composition. In particular, the contribution of emission sectors to the concentrations of OH and HO 2 in the troposphere and stratosphere can be measured. This method will be applied for re-evaluating the impact of the traffic sector on climate.
Code availability. The Modular Earth Submodel System (MESSy) is continuously further developed and applied by a consortium of institutions. The usage of MESSy and access to the source code is licensed to all affiliates of institutions which are members of the MESSy Consortium. Institutions can become a member of the MESSy Consortium by signing the MESSy Memorandum of Understanding. More information can be found on the MESSy Consortium website (http://www.messy-interface.org, last access: 22 May 2018). The submodel TAGGING 1.1 will be included in MESSy version 2.54. The code being used to obtain the presented results is available upon personal request.