Abstract South-east Australia, characterized by arid and semi-arid climate, has experienced large-scale rainfall reductions in recent decades. Larger temporal and spatial drought conditions are predicted in future. The temperate south east coastal zone is characterized by dense forests of Eucalyptus. Drought conditions have implications for the functioning of these indigenous ecosystems, and for emissions of reactive gases that are upwind of major metropolitan regions along the eastern coast. Here, we focus on the impact of drought on the emission of isoprene, a volatile organic compound emitted by a range of trees and shrubs. Previous model calculations grossly overestimate observations of the isoprene mixing ratio, potentially due to overestimated emission factors for native vegetation, but could also be due to drought-induced isoprene emission reductions. We develop the implementation of the Model of Emissions of Gases and Aerosols from Nature (MEGAN) within the CSIRO Chemical Transport Model to include a parameterization of drought using soil moisture. We test this parameterization using two approaches. First, we drive MEGAN using soil moisture fields from two Australian land surface models achieving reductions in isoprene emissions of 24–52% during summer. Second, we use a simple statistical approach to nudge model soil moisture towards satellite observations from the Soil Moisture and Ocean Salinity (SMOS) instrument. This work is the first application of SMOS towards isoprene emission modelling, providing a constraint on surface soil moisture which extends to global scales. Applying these soil moisture approaches, domain average isoprene emissions reduce by 38–58% in summer. Using these results, errors in basal emissions are likely in the region of 40%. Comparison of modelled soil moisture to observations using root density weighted averages across depths of 4 m showed minor differences of up to 0.04 m³ m⁻³. We find that the choice of land surface model used in the SMOS assimilations has a greater impact on isoprene emissions than adjusting either the nudging strength or the number of soil levels these satellite data may influence. However there are only small differences of 3% when using hourly or 24-hourly soil moisture input data to drive the emission calculations, suggesting that the lower temporal availability of satellite data does not reduce model quality. As the spatial resolution of satellite observations of atmospheric composition and land surface properties begin to approach the resolution of regulatory air quality models, we anticipate that these data will improve model predictive skills further.

Graphical abstract Display Omitted

Highlights • Decreased isoprene emissions in arid Australia using soil moisture function. • Developed technique to nudge model soil moisture towards satellite SMOS data. • Isoprene emissions decrease by 38–58% and reduce modelled bias. • 3% difference incurred when using daily (rather than hourly) soil moisture input. • Remaining error in basal isoprene emission factors approximately 40%.