Results from the Intergovernmental Panel on Climatic Change Photochemical Model Intercomparison (PhotoComp)

Jennifer Olson, Michael J. Prather, Terje K. Berntsen, Gregory R. Carmichael, R. B. Chatfield, Peter S. Connell, Richard G. Derwent, Larry W. Horowitz, Shengxin Jin, Maria Kanakidou, P. S. Kasibhatla, V. R. Kotamarthi, Michael Kühn, Kathy S. Law, Joyce E. Penner, Lori M. Perliski, Sanford Sillman, Frøde Stordal, Anne M. Thompson, Oliver WildJournal of Geophysical Research Atmospheres 1997 · DOI

Abstract

The Intergovernmental Panel on Climatic Change (IPCC) conducted the PhotoComp intercomparison to evaluate the consistency of tropospheric photochemical models used to predict ozone ($O_3$) changes. This tightly controlled experiment revealed significant discrepancies in calculated photodissociation rates, with root-mean-square (RMS) errors ranging from ±6–9% for $O_3$ and $NO_2$ to ±15% for $H_2O_2$ and $CH_2O$. A major source of variance was the choice of radiative transfer methods, with multistream models yielding higher rates for $NO_2$ and $CH_2O$. In clean air, $O_3$ tendencies varied by ±10%, primarily due to differences in $HO_2$ chemistry and $H_2O_2/CH_2O$ photolysis. In polluted scenarios, errors rose to ±10–30%, largely attributed to variations in $O_3$ photolysis to $O(^1D)$. The inclusion of non-methane hydrocarbons (NMHC) further doubled the RMS error for $O_3$ concentrations, highlighting the complexity of organic chemistry modeling.

Introduction

Tropospheric ozone is a critical atmospheric component, acting as a primary oxidant and a significant greenhouse gas. Accurate climate predictions depend on the reliability of photochemical models that simulate $O_3$ tendencies. However, these models are developed independently, leading to potential inconsistencies. The PhotoComp experiment was designed to isolate the sources of these differences by providing 21 modeling groups with standardized initial conditions, allowing for a direct comparison of their internal chemical and radiative mechanisms.

Methodology

PhotoComp employed a "blind" intercomparison format. Participants were provided with fixed input parameters—such as temperature, pressure, and initial chemical concentrations—representing six distinct atmospheric environments, from pristine marine air to polluted urban plumes. These cases were derived from synthetic benchmarks based on real-world missions like ABLE 2A and Mauna Loa observations. While inputs were fixed, groups utilized their own chemical mechanisms, numerical solvers, and radiative transfer codes. This setup allowed researchers to correlate specific model architectures, such as "two-stream" versus "multistream" radiation models, with divergent outputs.

Key Findings

The "Light Bias" and Photolysis Rates

A primary driver of model disagreement was the calculation of photodissociation rates. Models employing multistream radiative transfer methods consistently predicted higher rates for $NO_2$ and $CH_2O$ compared to those using simpler two-stream approximations. This systematic difference accounted for approximately one-third of the RMS error for these species.

Figure 1: Calculated tropospheric photodissociation rates for (a) $O_3 \rightarrow O(^1D)$, (b) $H_2O_2 \rightarrow 2OH$, (c) $NO_2 \rightarrow NO + O$, and (d) $CH_2O \rightarrow H + CHO$. The labels '2' and 'M' denote two-stream and multistream radiative transfer methods, respectively.

As shown in Figure 1, the spread in photolysis rates is significant across all altitudes. For $H_2O_2$ and $CH_2O$, the RMS error reached ±15%, which directly impacts the cycling of $HO_x$ radicals.

Ozone Tendencies and $HO_x$ Cycling

In clean, moist surface air, $O_3$ tendencies showed RMS errors of roughly ±10%. These differences were traced back to the destruction of $O_3$ via reaction with $HO_2$. Discrepancies in $HO_2$ concentrations were likely caused by inconsistent reaction rates for the conversion of $HO_2$ to $H_2O_2$ and differences in the calculated photolysis of $H_2O_2$ and $CH_2O$ (Figure 1b, 1d).

In "polluted" middle-tropospheric scenarios ($NO_x >$ few ppb), the RMS error for $O_3$ tendencies increased to ±10–30%. Under these conditions, $O_3$ photolysis to $O(^1D)$ (Figure 1a) provides about 80% of the $HO_x$ source, making it the dominant factor in model variance.

The Impact of Hydrocarbons

The introduction of non-methane hydrocarbons (NMHC) acted as an "error multiplier." NMHCs dramatically increased both the rate of $NO_x$ loss and the associated model-to-model differences. Because there is no standardized method for "lumping" complex organic reactions, the RMS error for $O_3$ concentrations approximately doubled when NMHCs were included in the simulation.

Conclusion

The PhotoComp study demonstrates that while the fundamental physics of atmospheric chemistry are well-understood, significant uncertainties remain in model implementation. The choice of radiative transfer methods and the simplification of hydrocarbon chemistry are major sources of divergence. While most models demonstrated robust numerical conservation of nitrogen, the spread in $O_3$ predictions underscores the need for standardized chemical mechanisms and more sophisticated radiative transfer treatments in future IPCC assessments.

Summary updated

Abstract

Results from the Intergovernmental Panel on Climatic Change (IPCC) tropospheric photochemical model intercomparison (PhotoComp) are presented with a brief discussion of the factors that may contribute to differences in the modeled behaviors of HO x cycling and the accompanying O 3 tendencies. PhotoComp was a tightly controlled model experiment in which the IPCC 1994 assessment sought to determine the consistency among models that are used to predict changes in tropospheric ozone, an important greenhouse gas. Calculated tropospheric photodissociation rates displayed significant differences, with a root‐mean‐square (rms) error of the reported model results ranging from about ±6–9% of the mean (for O 3 and NO 2 ) to up to ±15% (H 2 O 2 and CH 2 O). Models using multistream methods in radiative transfer calculations showed distinctly higher rates for photodissociation of NO 2 and CH 2 O compared to models using two‐stream methods, and this difference accounted for up to one third of the rms error for these two rates. In general, some small but systematic differences between models were noted for the predicted chemical tendencies in cases that did not include reactions of nomnethane hydrocarbons (NMHC). These differences in modeled O 3 tendencies in some cases could be identified, for example, as being due to differences in photodissociation rates, but in others they could not and must be ascribed to unidentified errors. O 3 tendencies showed rms errors of about ±10% in the moist, surface level cases with NO x concentrations equal to a few tens of parts per trillion by volume. Most of these model to model differences can be traced to differences in the destruction of O 3 due to reaction with HO 2 . Differences in HO 2 , in turn, are likely due to (1) inconsistent reaction rates used by the models for the conversion of HO 2 to H 2 O 2 and (2) differences in the model‐calculated photolysis of H 2 O 2 and CH 2 O. In the middle tropospheric “polluted” scenario with NO x concentrations larger than a few parts per billion by volume, O 3 tendencies showed rms errors of ±10–30%. These model to model differences most likely stem from differences in the calculated rates of O 3 photolysis to O( 1 D ), which provides about 80% of the HO x source under these conditions. The introduction of hydrocarbons dramatically increased both the rate of NO x loss and its model to model differences, which, in turn, are reflected in an increased spread of predicted O 3 . Including NMHC in the simulation approximately doubled the rms error for O 3 concentration.
BibTeX
@article{olson1997results,
  author = {Jennifer Olson and Michael J. Prather and Terje K. Berntsen and Gregory R. Carmichael and R. B. Chatfield and Peter S. Connell and Richard G. Derwent and Larry W. Horowitz and Shengxin Jin and Maria Kanakidou and P. S. Kasibhatla and V. R. Kotamarthi and Michael Kühn and Kathy S. Law and Joyce E. Penner and Lori M. Perliski and Sanford Sillman and Frøde Stordal and Anne M. Thompson and Oliver Wild},
  title = {{Results from the Intergovernmental Panel on Climatic Change Photochemical Model Intercomparison (PhotoComp)}},
  journal = {Journal of Geophysical Research Atmospheres},
  year = {1997},
  doi = {10.1029/96jd03380},
}
Figure