Temperature response measurements from eucalypts give insight into the impact of Australian isoprene emissions on air quality in 2050

Abstract. Predicting future air quality in Australian cities dominated by eucalypt emissions requires an understanding of their emission potentials in a warmer climate. Here we measure the temperature response in isoprene emissions from saplings of four different Eucalyptus species grown under current and future average summertime temperature conditions. The future conditions represent a 2050 climate under Representative Concentration Pathway 8.5, with average daytime temperatures of 294.5 K. Ramping the temperature from 293 to 328 K resulted in these eucalypts emitting isoprene at temperatures 4–9 K higher than the default maximum emission temperature in the Model of Emissions of Gases and Aerosols from Nature (MEGAN). New basal emission rate measurements were obtained at the standard conditions of 303 K leaf temperature and 1000 µmol m−2 s−1 photosynthetically active radiation and converted into landscape emission factors. We applied the eucalypt temperature responses and emission factors to Australian trees within MEGAN and ran the CSIRO Chemical Transport Model for three summertime campaigns in Australia. Compared to the default model, the new temperature responses resulted in less isoprene emission in the morning and more during hot afternoons, improving the statistical fit of modelled to observed ambient isoprene. Compared to current conditions, an additional 2 ppb of isoprene is predicted in 2050, causing hourly increases up to 21 ppb of ozone and 24-hourly increases of 0.4 µg m−3 of aerosol in Sydney. A 550 ppm CO2 atmosphere in 2050 mitigates these peak Sydney ozone mixing ratios by 4 ppb. Nevertheless, these forecasted increases in ozone are up to one-fifth of the hourly Australian air quality limit, suggesting that anthropogenic NOx should be further reduced to maintain healthy air quality in future.



Australian distribution of eucalypt species used in this experiment
From Atlas of Living Australia, accessed 19.3.2020. https://www.ala.org.au/

Figure S1 clockwise from top left. E camaldulensis, E tereticornis, E smithii, E. botryoides.
The occurrences are unfiltered records and include points in arboreta or points over water bodies.

Condition of sapling eucalypt specimens.
The trees are ~1.5 m tall and have produced plenty of leaves prior to the experiment.

Figure S2 photograph of experimental set-up. Left E. camaldulensis, right E tereticornis
There are 5 or 6 replicates in each of the growing treatments. Figure S3 shows the variation in isoprene emission for each species, alongside the default MEGAN response.

ESA to NCAR plant functional type classifications
Constructed for 2010, the ESA uses 36 land cover classes to describe global vegetation. These land cover classes have been processed to the NCAR plant functional types as required by MEGANv2.1 using similar methods as Emmerson et al. (2016). The ESA dataset also describes the percentage tree and shrub cover, so only minor estimation of tree/shrub/grass coverage is required (for example where the percentage of total trees is split between broadleaf evergreen and deciduous species). Any remaining ground is classed as bare soil and contains no vegetation. Table S1 gives the details of how each ESA land cover type has been split into the NCAR plant functional types (PFTs) as required by MEGAN.  Nt Eg = needleleaf evergreen tree, Nt Dc = needleleaf deciduous tree, Bt Eg = broadleaf evergreen tree, Bt Dc = broadleaf deciduous tree, Sb Eg = evergreen shrub, Sb Dc = deciduous shrub, Gs C3 = grass (can be Cd = cold, Cl = cool or W = warm depending on climatic zone).

Climate models used for 2050 temperature projections.
Surface temperature data for eight models considered in the Climate Model Intercomparison Project CMIP5 were downloaded for Australia. https://www.climatechangeinaustralia.gov.au/en/climateprojections/explore-data/map-explorer/ The eight models are: CanESM2, CNRM-CM5, ACCESS1.0, MIROC5, HadGEM2-CC, NorESM1-M, GFDL-ESM2M and CESM1-CAM5. The models were chosen to represent a wide range in climate projections and were judged to have performed well against 20 th century observations.
The period December to February was chosen to correspond with the summer field campaigns used in this work. The change in seasonal temperature is calculated compared to the seasonal average between 1986 and 2005.
A range in warming is predicted by the eight models, with maximums ranging between 1.74°C and 3.83°C. For the purposes of this study, an average change in temperature was calculated from the ensemble. These delta temperatures were re-gridded to suit the four domains used by the C-CTM. Figure S4 shows how often the best performing model in terms of statistical r 2 fit, CC_T+LEF is within =/1 1 standard deviation of the isoprene observations Note the colour of the CC_T+LEF run has changed to red here for visibility.