Model evaluation

3.3.1 Stratosphere-troposphere exchange

Initially, a 38-level version of the model was developed. This version is still em-ployed at the UK Met Office, and is described briefly in O’Connor et al. [2009]. I found ozone concentrations that were too high in remote marine areas when using this version of the model, especially in the northern hemisphere during March-April and the southern hemisphere during September-October. This seasonal cycle is suggestive of too strong stratosphere-troposphere exchange (STE), which is a result of the Brewer-Dobson circulation [Dobson et al., 1946; Brewer, 1949]. The seasonality of the circulation leads to a strengthening of the downward motion in the winter hemisphere [Holton et al., 1995]. The Brewer-Dobson circulation is stronger in the northern hemisphere, due to to increased wave activity in the region, and this may contribute to observed maxima in spring concentrations of ozone in the region [e.g. Monks, 2000].

The introduction of an on-line STE diagnostic¹reinforced our hypothesis. Sea-sonally varying annual STE [Tg O3 yr⁻¹] is shown in Figure 3.3 for two model

1The diagnostic was implemented by Dr. N. Luke Abraham.

STE(Tg/year)

D J F M A M J J A S O N

1600

1400

1200

1000

800

600

400

Figure 3.3. Stratosphere-troposphere exchange [Tg O3 yr⁻¹] for two model runs.

The 38 level version is shown in pink, and the 60 level version is shown in green (the

‘BASE’ run).

runs. The 38-level version of the model simulates the atmosphere up to 56 km, while the 60-level version captures the whole stratosphere and simulates up to 84 km. This extension into the mesosphere meant that the Brewer-Dobson overturn-ing could be fully and more accurately modelled. Indeed, the 60-level version has an average annual STE is 424 Tg yr⁻¹, which lies within the first standard devi-ation of the model average from the recent comprehensive model assessment by Stevenson et al.[2006] of 556 ± 154 Tg yr⁻¹. This was a marked improvement over the 38 level version, which had an average annual STE of 1324 Tg yr⁻¹. The estimate from the 60-level version of the model is similar to that reported in Mur-phy and Fahey[1994], who used observed ratios to estimate the flux of ozone from the stratosphere to be 450 Tg yr⁻¹. Gettelman et al. [1997] estimated STE to be 450-590 Tg yr⁻¹from satellite data.

J F M A M J J A S O N D

Figure 3.4. Comparison of climatological ozone (black diamonds) as described in [Stevenson et al., 2006] with the BASE model run (green line). Lines show the stan-dard deviation of the measurements and the lighter green shows the stanstan-dard devia-tions for the model. Both model and measurements are sampled for a series of latitude bands and three pressure levels identically. Data are from Logan [1999] and Thomp-son et al.[2003].

3.3.2 Comparison to measurements

Ozone is of particular chemical importance for tropospheric chemistry (see section 2.1.1), and also has an impact on the radiative balance of the atmosphere, human health, and crop health. Throughout this thesis, I use ozone as a sort of exemplar trace constituent in the troposphere, and examine the sensitivity of ozone to vari-ous perturbations. Though a more exhaustive analysis of the model’s performance can be found in a forthcoming paper by Fiona O’Connor (to be submitted as Part II of Morgenstern et al. [2009]), here I briefly describe the model’s ability to sim-ulate ozone. Chapter 4 elaborates on the model chemical mechanism’s ability to reproduce measurements.

Figure 3.4 shows the performance of UKCA against the climatological ozone fields described in Logan [1999] and Thompson et al. [2003] as shown in Steven-son et al. [2006]. Data are broken down into four latitude bands—the southern high latitudes, the southern tropics, the northern tropics, and the northern high latitudes—and three pressure levels—750 hPa, 500 hPa, and 250 hPa. Near the surface, UKCA compares well with measurements, capturing both the seasonal variation and the magnitude of the climatologies. One key weakness is apparent:

the northern tropics and northern high latitudes are slightly overestimated in mag-nitude (by approximately 8 and 6 ppbv, respectively). The overprediction is also evident in the multi-model assessment of Stevenson et al. [2006], which uses the same emissions, and could be an artefact thereof.

In the middle of the troposphere at 500 hPa, the seasonal variation in the clima-tology continues to be reproduced by the model, though the tropical measurements are slightly underestimated in magnitude. The opposite occurs in the northern high latitudes, where the model slightly overpredicts ozone concentrations; this is visi-ble in the multimodel comparison, and again could be an artefact of the emissions [Stevenson et al., 2006]. The comparison is poorest in the the upper tropical tropo-sphere, where the underprediction is exacerbated and reaches up to 15 ppbv. This is not the case for the multi-model average of Stevenson et al. [2006], where upper troposphere ozone is overpredicted by approximately 7 ppbv in the northern hemi-sphere during winter. The negative bias is only partially (up to -2 ppbv) explained by the positive bias in the model humidity in the same region [O’Connor et al., 2009]. It could also be linked to lightning NOx, for which the uncertainty has re-cently been estimated at 60% [Schumann and Huntrieser, 2007]; perturbations to lightning NOxled to up to 10% changes in the ozone burden in the work of Wild [2007]. Another contributing factor could be the underestimation of HOxfrom the photolysis of formaldehyde, for which the absorption cross section has recently been found to be greater than previously thought [Carbajo et al., 2008].