The Photolysis Module Jval-14, Compatible with the Messy Standard, and the Jval Preprocessor (jvpp)

We present version 14 of the photolysis module JVAL. Taking atmospheric conditions as input, JVAL calculates photolysis rate coefficients, i.e. the speed of disso-ciation of atmospheric molecules in the sunlight. Computational efficiency is obtained through the use of parameters for polynomial curve fitting and lookup tables. Physical changes compared to the previous version include a parameteriza-tion of the Lyman-alpha absorption, and an update of the UV/VIS cross sections to the most recent recommended values. JVAL also includes the auxiliary program JVPP (JVal PreProcessor) which pre-calculates these parameters based on the absorption cross sections and quantum yields of the atmospheric molecules. It is possible to use JVAL either as a stand-alone program or as a module inside the Modular Earth Submodel System (MESSy). JVAL is a community model published under the GNU General Public License.


Introduction
Many important chemical reactions in the atmosphere are photolysis reactions where molecules dissociate in the sunlight.The rate coefficients of these reactions are called J values.They depend on the actual atmospheric shortwave radiation field and are therefore not constant in time, but have to be reevaluated with changing solar zenith angle, temperature, ozone concentrations, clouds, aerosols, and other atmospheric properties.Landgraf and Crutzen (1998) presented an efficient method for online calculations of J values which has been used in several atmospheric models, e.g.MATCH (von Kuhlmann et al., 2003) and ECHAM5/MESSy (Jöckel et al., 2006(Jöckel et al., , 2010)).For the implementation into the latter, the photolysis code was adapted to the Modular Earth Submodel System (MESSy) interface by Jöckel et al. (2005) and called JVAL.Since the efficiency of JVAL results from a parameterization, it is necessary to recalculate the parameters whenever a new chemical species is added or new spectral data become available.Although straightforward, this has been a tedious procedure, resulting in a time-lag between publication of the latest spectral information and its implementation into JVAL.To automate the generation of these parameters, the JVal PreProcessor (JVPP) has been written.Here we present the new version JVAL-14, which also includes JVPP.

Original code
This section provides a brief summary of the parameterization method used in JVAL.For full details, the reader is referred to the original paper by Landgraf and Crutzen (1998).
The J value for a molecule X (J X ) can be calculated via the integral where λ is the wavelength, σ X the absorption cross section, ϕ X the quantum yield and F (λ) the spectral actinic flux.The photochemically active spectral interval considered here is λ min = 178 nm ≤ λ ≤ 683 nm = λ max .Numerically, this in- tegral can be approximated as the sum where the spectrum is divided into N wavelength bins.Here, σ X (λ i ), ϕ X (λ i ), and F (λ i ) are average values in the bins of size ∆λ i .However, to obtain accurate results, N has to be large, leading to excessive computing times.As an alternative, Landgraf and Crutzen (1998) suggested a method using only the 8 spectral bands shown in Table 1.The basic idea is to calculate the J value not with a fine spectral resolution but with only a few spectral bands in which Eq. ( 1) is approximated by: where J a i,X is the J value for a purely absorbing atmosphere and δ i describes the influence of scattering by air molecules, aerosols, and cloud particles.The J a i,X are precalculated with a fine spectral resolution and are approximated during runtime from lookup tables or polynomial fits.The :::::: lookup ::::::::::: calculations ::: are ::::::::: performed ::: for :: a :::::: purely ::::::::: absorbing :::::::::: atmosphere :::::: using :::: the ::::::::::::: Lambert-Beer ::::: law.:::: For :::: the ::::::: on-line ::::::::::: calculations, :::: the ::::::::::: two-stream ::::::: method :::::: PIFM : (Practical Improved Flux Method, Zdunkowski et al., 1980) : is :::::::::: employed. ::: The : advantage of this procedure is that the fine absorption structures that are present in σ X and ϕ X in integral (1) are considered.Only Rayleigh and cloud scattering, included in F (λ i ), are treated with a coarse spectral resolution.J a i,X is calculated as a function of the slant overhead column of ozone.For the Schumann-Runge and for the Herzberg band (bands 1 and 2 in Table 1), the dependence on the corresponding slant O 2 column is considered as well.Absorption by aerosol and cloud particles and other gases plays only a minor role for most chemical applications in the determination of the actinic fluxes.
-In the original code by Landgraf and Crutzen (1998), the spherical geometry of the atmosphere was taken into account by employing the air mass factor correction of Kasten and Young (1989).In the current code, the correction factors F corr (fj_corr) from Lamago et al. (2003)

Technical changes
The following technical changes have been implemented into the code since the publication of Landgraf and Crutzen (1998): -The syntax was converted from FORTRAN77 to For-tran90.
-In the past, different and incompatible ASCII input files were used for different versions of the code.Now, these molecule-specific parameters are generated with JVPP and included in the Fortran90 code.
-The code was modularized according to the MESSy standard.Briefly, this allows to use exactly the same code for calculating J values in the CAABA/MECCA box model (Sander et al., 2011a), the global ECHAM5/MESSy Atmospheric Chemistry (EMAC) model (Jöckel et al., 2010), the COSMO/MESSy limited area model (Kerkweg and Jöckel, 2012), or any other base model that is MESSy-compatible.For details, see Jöckel et al. (2005).
-A standardized interface via a Fortran90 namelist couples the JVAL module to the base model.This namelist (CPL) comprises switches to select which calculations are to be performed and entries to specify required input data (i.e., the channel objects, see Jöckel et al., 2010) for ozone, the cosine of the solar zenith angle, the distance between Earth and Sun (in AU, astronomical units), and the solar activity.An additional ozone climatology is required, which is used above the model top layer.More details on the namelists are documented in Sect.2.3 of the "JVAL and JVPP User Manual", which is part of the Supplement.

JVPP model description
As described above, JVAL uses parameters in lookup tables and polynomial fits to calculate the values of J a i,X from Eq. (3).To provide these parameters, the JVal PreProcessor (JVPP) has been written.JVPP works in three steps for each molecule X.
First, it reads a data file with absorption cross sections σ X in the UV/visible range at an arbitrary spectral resolution and converts (interpolates) them to a fixed grid of 176 wavelengths.The default method is a linear interpolation between the points of the original spectrum which conserves the integrated value.Alternatively, a variety of spline based interpolation methods can be chosen.For molecules where quantum yields ϕ X and/or the temperature dependence of the cross sections σ X (T ) are available, those are interpolated as well.As an example, the spectrum of OClO is shown in Fig. 1 before and after interpolation.Detailed plots for all molecules are available in the Supplement.
In the second step, a range of typical atmospheric conditions is scanned to obtain effective values for all 8 bands of the spectral range (Table 1).Overhead ::::: Slant :::::::: overhead : ozone columns of up to 7000 Dobson units are used.For bands 1 and 2 (178 to 241 nm), the concentration of oxygen is also varied.For molecules with temperature-dependent cross sections, T is varied in the range from 180 to 320 K.The results are stored in temporary files.
Finally, in the last step, the precalculated values are used to obtain polynomial fits which allow to calculate the J values as a function of temperature, ozone, and oxygen.The fitting parameters are then written into a Fortran90 include file which can be used by JVAL.
Table 2 shows all photolysis reactions that are currently included in JVPP, as well as the references for the UV/VIS spectra used.More information about the JVPP code can be found in the user manual which is in the manual/ directory of the JVAL code in the Supplement.

Model evaluation and applications
Thanks to the MESSy structure, the JVAL photolysis submodel can easily be incorporated into different models.To evaluate the new code, we used a simple column model, an atmospheric chemistry box model, and a global 3dimensional general circulation model.

The simple JVAL column model for testing
The JVAL column model, which is included in the Supplement, can be used for simple tests.As input, the following properties of an atmospheric air column must be defined: temperature, pressure, humidity, ozone concentration, and cloud cover.In addition, a set of solar zenith angles and the surface albedo must be defined.The output contains J values for all specified zenith angles and all altitudes of the column.Detailed information how to run the column model can be found in the user manual which is in the manual/ directory of the JVAL code in the Supplement.

The CAABA box model
The new JVAL submodel has been implemented in the CAABA box model as described by Sander et al. (2011a).Since version 3.3, CAABA uses the updated absorption cross sections as listed in Table 2.

The global EMAC model: J values and atmospheric implications
JVAL has also been implemented in the latest EMAC-2.50model, which is based on version 2.50 of MESSy1 .Global fields of J values from the old and the new code are compared for all photolysis reactions.As an example, the results for ozone are shown in Figs. 2 and 3. Detailed plots for all molecules are available in the Supplement.
To test the atmospheric implications of the updated J values, two model simulations have been performed using the updated and the old J values, respectively.The simulations have been carried out in T42L90MA resolution, i.e., with a triangular truncation at wave number 42 of the spectral ECHAM5 core, which corresponds to an approximate 2.8 • × 2.8 • quadratic Gaussian grid.The vertical model setup comprised 90 layers from the surface up to 0.01 hPa (≈ 80 km) in the middle atmosphere (MA).The model time step was 600 s.A Newtonian relaxation technique ("nudging") of the prognostic variables temperature, vorticity, divergence and the logarithm of the surface pressure towards ERA-Interim reanalysis data from ECMWF (Dee et al., 2011) has been applied, in order to nudge the model dynamics towards the observed meteorology.The nudging has been applied above the boundary layer up to approximately 10 hPa.From the comprehensive MECCA atmospheric chemistry (Sander et al., 2011a), a mechanism for the troposphere, the stratosphere and the lower mesosphere was selected, similar as in Jöckel et al. (2010).It includes ozone-related chemistry of the troposphere and non-methane hydrocarbons (NMHCs) up to isoprene.The chemical mechanism consists of 145 gas phase species and 298 reactions.The aqueous phase chemistry and wet deposition was calculated by the submodel SCAV (Tost et al., 2006), dry deposition by the submodel DDEP (formerly DRYDEP, Kerkweg et al., 2006a) and the prescribed emissions and boundary conditions have been handled by the submodels OFFEMIS (formerly OFFLEM) and TNUDGE (Kerkweg et al., 2006b).Isoprene emissions from plants and NO emissions from soil have been calculated online with the submodel ONEMIS (formerly ONLEM, Kerkweg et al., 2006b), NO from lightning with the submodel LNOX (for the present simulations with the parameterisation of Grewe et al., 2001), and air-sea gas exchange of isoprene, DMS and methanol with the submodel AIRSEA (Pozzer et al., 2006).Boundary conditions for green house gases and anthropogenic emissions have been used according to the definition of the REF-C1SD simulation of the IGAC/SPARC Chemistry Climate Model Initiative (CCMI-1) project2 (Eyring et al., 2013).For the reference simula-tion3 spanning the years 1979 to 2012, the updated J values have been used.
From this reference simulation, a sensitivity simulation with the old J values was branched in January 2000.After 5 years of simulation (August 2005), the differences between the atmospheric compositions (old vs. updated) were already in a dynamical steady state without showing continuously diverging trends.We analyse the differences between the simulations exemplarily for the monthly averages of January and July 2005, respectively.Since the atmosphere is well buffered against small changes in J values, the effect of the new J values on tracer mixing ratios is in general limited.Because of relatively small changes between the old and the new J values, the differences of O 3 and related substances ::: are :: in :::::::: general ::::::: limited :::: and do not show a clear pattern.In Fig. 4, we show absolute and relative differences between the two simulations for O 3 .The relative differences are well below 10 % through the entire atmosphere.For CF 2 ClBr (Halon 1211), CF 3 Br (Halon 1301) and CH 3 Br, the cross section update from DeMore et al. (1997) to Sander et al. (2011b) leads to higher J values of these substances in the troposphere and thus to higher BrO values in this region, as shown in Fig. 5. Detailed comparison plots for a large number of atmospherically relevant species are available in the Supplement.

Summary and outlook
We have presented the current version of the photolysis module JVAL and the JVal PreProcessor (JVPP).Both are community models published under the GNU General Public License4 .The model code and a user manual can be found in the Supplement.Regarding model development, our future plans include to: -Simulate photochemistry under twilight conditions.
An extension of the current JVAL module ::::: based ::: on Williams et al. ( 2006) is under development.Using Eq. ( 2) to infer J values, the actinic fluxes are calculated by a pseudo-spherical radiative transfer solver.Here, the direct solar beam is calculated in spherical geometry but the transport of the diffuse radiation fields is described for a plan parallel model atmosphere.Currently, this JVAL extension is being verified with full-spherical reference simulations.
-Improve the aerosol coupling: It is desirable to implement a generalised coupling (via channel objects, see Jöckel et al., 2010) to the required aerosol properties (extinction coefficient, absorbtion coefficient and asymmetry parameter).This will not only enable the usage of alternative, off-line prescribed aerosol climatologies as replacement for the implemented climatology, but also a direct, consistent coupling to the aerosol properties calculated on-line, if an aerosol submodel is running.
-Compare J values generated by JVAL with those generated by other photolysis models, e.g., Fast-JX5 or TUV6 .
Several additional plots, the model code for JVAL and JVPP, and a user manual are available online as supplementary material at: doi:10.5194/gmd-0-1-2014supplement.(up ::: to :::: 0.01 hPa : ).The first row is plotted on a linear scale, the second on a logarithmic scale.The left column shows the old result, the center column the new result, and the right column the differences.

Figure 1 .
Figure 1.Absorption cross sections of OClO on a linear (left) and logarithmic (right) scale.The red line shows the original data taken from Sander et al. (2011b) and the black line after interpolation by JVPP to a fixed grid of 176 wavelengths.

Figure 2 .
Figure 2. Comparison of old and new J values (in s −1 ) for O3 → O( 1 D), calculated for 2000-01-01, 02:00 UTC at the surface.The first row is plotted on a linear scale, the second on a logarithmic scale.The left column shows the old result, the center column the new result, and the right column the differences.

Figure 3 .
Figure 3.Comparison of the vertical profiles of the old and new J values (in s −1 ) for O3 → O( 1 D), calculated for 2000-01-01, 02:00 UTC at a longitude of 138 • E and 90 model levels ::(up ::: to :::: 0.01 hPa : ).The first row is plotted on a linear scale, the second on a logarithmic scale.The left column shows the old result, the center column the new result, and the right column the differences.

Table 1 .
Subdivision of the spectral range into 8 bands.λini and λfin are the initial and final wavelength.