GENERAL COMMENT:

This study applies a regional vegetation-climate-chemistry model to investigate the influence of water stress on isoprene emissions and surface ozone over Europe in 1992-2016. This is done by coupling the land module and biogenic emission module to derive the soil water stress function, which is then used to determine soil moisture activity factor in the parameterization of isoprene emission. Simulation results show that water stress reduces summertime isoprene emissions on average by nearly 6%, and by -20 to -60% in extreme dry summers, but influence on ozone is relatively small. This study is well-designed, easy to follow, and the results are useful for the community. It has room to be improved by addressing the following comments.

Authors’ response: We thank Reviewer #3 for this positive evaluation of the study and for the insightful comments. Below, we separately reply to each comment and, when necessary, we precise where and how the manuscript has been modified. Supplementary figures produced to answer to reviewers’ comments have been gathered in a separated document uploaded as a supplement (“Responses_to_Reviewers_fig.pdf”). To avoid confusion with figures in the manuscript and in the Supplementary Material, we customized the counter as Figure R.*. While reviewing the manuscript, we found an error in the conversion of ozone and formaldehyde mixing ratios that has been corrected in the revised manuscript.

SPECIFIC COMMENTS

1. My major concern is that this study lacks direct evaluation with observed isoprene emissions. Does the inclusion of water stress effect improve the simulation of isoprene emissions? Does the new scheme outperform the old scheme?

Authors’ response: We agree with the Reviewer that direct evaluation with observed isoprene emissions would be the best choice to assess the model performance and compare the old and new schemes that link the effect of soil moisture on isoprene emissions. However, since there is no network over Europe routinely measuring isoprene emissions, or isoprene concentrations, in vegetated areas, we focused the model evaluation on a proxy of isoprene emissions such as formaldehyde (HCHO). Nevertheless, we have now also a limited evaluation of isoprene concentrations as simulated by the RegCM4chem-CLM4.5-MEGAN2.1 model in the GAMMA-SMoff simulation. The model output have been compared against observations collected during two field campaigns. Figure R.1 shows the comparison against isoprene concentrations measured in south-eastern France (site: La Verdière; Latitude: 43.63° N, Longitude: 5.93° E) during the summer 2000 (from June 21 to July 6) in the framework of the ESCOMPTE field campaign (Cros et al., 2004) when isoprene concentrations had been measured every 30 minutes using a Fast Isoprene Sensor. In Figure R.2, model output have been compared against data collected in Cyprus (site: Ineia; Latitude: 34.96° N, Longitude: 32.39° E) during the summer 2014 (from July 7 to August 3; data collected nearly every 30 minutes) using techniques of gas chromatography - mass spectrometry.

At both sites, RegCM underestimates isoprene concentrations, which is consistent with RegCM underestimating concentrations of a proxy of isoprene such as formaldehyde, as shown in Section 3.1.3 in the manuscript. Sometimes, the model reproduces a delayed peak in isoprene concentrations compared to observations. Differences between observations and model output could result from multiple factors, for example:

1. The cold and wet model bias (see Fig. 1 in the manuscript) that limits isoprene emissions;

2. Differences between the dominant vegetation types on the field and in the model grid-cell. For example, in La Verdière, vegetation is mainly characterized by Mediterranean oak forest (more than 80% of Quercus Pubescens, which is a deciduous tree), while the RegCM grid-cell is mainly covered with needle-leaf evergreen temperate trees (36%), and C3 grass (37%), which are both low isoprene emitters, and has only a small amount of broadleaf deciduous trees (6%).

3. Different scales: a model grid-cell spans over a surface of around 25x25 km², while station measurements have a footprint of a few hundreds of meters, depending on the terrain where the observations have been collected. For example, Ineia is located close to the seacoast leading to use a model grid-cell that is located over the sea.

Based on these results, we have added the comparison between model output and in-situ measurements of isoprene concentrations in Section 3.1.3 in the revised manuscript.

Cros, B., et al.: The ESCOMPTE Program: an overview. Atmospheric Research, 2004, 69 (3-4), pp.241-279, DOI: 10.1016/j.atmosres.2003.05.001_xffff_.

2. I also wonder how the water stress effect changes the relationship between isoprene emissions and temperature. Previous studies have revealed the decrease in isoprene emission in extreme high temperature, but would the water stress effect further change the turning point of the T-emission curve? Some discussions would be useful.

Authors’ response: This is an interesting question.

The version of the MEGAN model (2.1) implemented in the RegCM4.7 model accounts for the effect of past temperatures on isoprene emissions (over the past ten days): the warmer the temperatures, the higher the emissions (see Fig. 4 in Guenther et al., 2006). However, this version does not account for the decrease in isoprene emissions due to extreme high temperatures.

Extreme high temperatures often co-occur with droughts. It is, then, not trivial to separate the effect of these climate extremes on isoprene emissions. In our simulations, we observed the largest decreases in isoprene emissions in the summers 2003 and 2012 (see Table 4) when the observation-based summer temperatures are nearly 4–5 standard deviations above (warmer than) the 1970–1990 climatology, as shown by Figure S.6 in the Supplementary Material. These results and those mentioned by the Reviewer, which reveal the decrease in isoprene emissions under extreme high temperatures, suggest that isoprene emissions would be strongly reduced when heat wave and drought co-occur. In the revised manuscript, we added the following comment in Sect. 3.2:

“The largest decreases in isoprene emissions occur in the summers 2003 and 2012 (Table 4), when the observation-based summer temperatures are nearly 4–5 standard deviations above (warmer than) the 1970–1990 climatology (Fig. S.6). These results suggest that isoprene emissions would be strongly reduced when heat wave and drought co-occur.”

3. Related to point 2, I think some discussions on the extreme ozone episodes would be helpful to evaluate the effect of water stress on ozone concentration.

Authors’ response: Extreme high temperatures often exacerbate ozone pollution and lead to extreme ozone episodes. In a second paper (in preparation), we investigated the ozone climate penalty by performing simulations under both present-day (1990–2004) and future climates (2035–2049). To assess the impact on ozone concentration of both the direct effect of high temperatures and the indirect effect of water stress on isoprene emissions, we designed two sensitivity simulations over the summer 2003 with and without the soil moisture activity factor activated. In these simulations, we artificially increased air temperature across the whole atmospheric column by a fixed amount varying with atmospheric levels. Results show that an average increase of 2.5°C in air temperature leads to an increase in surface-ozone level smaller than 2 ppbv (1%), regardless the soil moisture activity factor is activated or not. Although the paper will be submitted soon, we added the following comment in Sect. 4:

“In a future study, we aim to explore the ozone climate penalty over the Euro-Mediterranean region under both present-day and future climates and to assess the impact on ozone concentration of both the direct effect of high temperatures and the indirect effect of water stress on isoprene emissions.”

4. I feel that the model evaluation of chemical fields is rather insufficient. It only shows the mean magnitude of observed and simulated HCHO column and ozone concentrations. How well does the model capture the spatial and temporal pattern of HCHO and ozone? It might be also important to evaluate the ozone chemical regime (NOx-limited or VOCs-limited) somewhere.

Authors’ response: We agree with the Reviewer that the model evaluation for chemical species could be extended. In our study, we focused on the analysis of differences between sensitivity simulations, therefore possible systematic biases in the model would cancel out, thus giving robust results. However, we performed some additional evaluation for both ozone and formaldehyde by comparing model output against re-analyses from the Copernicus Atmosphere Monitoring Service (CAMS) for the period 2003–2007. In addition, surface ozone concentrations have been evaluated against observations collected in La Verdière (France) in the summer 2000 (from June 21 to July 6) using an conventional UV absorption ozone analyzer (Environement S.A., Poissy, France, model O3 41M). Ozone concentrations have been measured during the ESCOMPTE field campaign, together with isoprene measurements shown in Figure R.1.

For near surface ozone, model output are lower than CAMS re-analyses with differences between 10 and 20 ppbv over the Mediterranean Basin, with some summers and few grid-cells showing differences between 20 and 30 ppbv (Fig. R.3). The model underestimates near-surface ozone as well when compared to in-situ measurements (Fig. R.4); in particular, model output shows a smaller variability than in-situ measurements. For formaldehyde, model outputs near the surface are also lower than CAMS reanalyses with differences between -1 and -4 ppbv (Fig. R.5).

We also assessed the ozone chemical regime using the ratio between formaldehyde (HCHO) and nitrogen dioxide (NO₂) as presented in Duncan et al. (2010). Results show that the model reproduces a VOC-limited regime over the whole domain, with a HCHO/NO₂ ratio lower than 1 (Fig. R.6). Based on this analysis, in the revised manuscript we updated all discussions about the ozone regime over the model domain.

Duncan, B. N. et al.: Application of OMI observations to a space-based indicator of NOx and VOC controls on surface ozone formation, Atmospheric Environment, Volume 44, Issue 18, 2010, Pages 2213-2223, DOI: https://doi.org/10.1016/j.atmosenv.2010.03.010.

5. The figure quality can be improved. For example, the size of the figure is often too small compared to the colorbar (Figs 11 and 12), and in some cases the label is missing (Fig.5).

Authors’ response: We improved all figures. We added axis labels to Figure 5 (now Fig. 6), we increased the size of figures compared to the colorbars, and we implemented colorblind-friendly colormaps in Figures 6 (now Fig. 7) and 11 (now Fig. 12), as requested by the EGU journals.