Probing the relationship between formaldehyde column concentrations and soil moisture using mixed models and attribution analysis

In response to changes in environmental factors (e.g., temperature, radiation, soil moisture), plants emit biogenic volatile organic compounds (BVOCs). Once released in the atmosphere, BVOCs influence levels of greenhouse gases and air pollutants (e.g., methane, ozone and aerosols), thus affecting both climate and air quality. In turn, climate change may alter BVOC emissions by modifying the driving environmental conditions and by increasing the occurrence and intensity of severe stresses that alter plant functioning. To understand and better constrain the evolution of BVOC emissions under future climates, it is important to reduce the uncertainties in global and regional estimates of BVOC emissions under present climate. Part of the uncertainty in the estimates of BVOC emissions is related to the impact that water stress might have on BVOC emissions. Field campaign, in-situ and laboratory experiments investigated the effect of different regimes of water stress (short- vs. long-term) on BVOC emissions. However, these studies provide geographically scattered and uneven results. To explore the relationship between BVOC emissions and water stress globally, we use remotely sensed soil moisture and formaldehyde, a proxy of BVOC emissions. As BVOCs include a multitude of gas tracers with lifetime ranging from few hours to days, a fully characterisation of these components is virtually impossible. Nevertheless, in the continental boundary layer, formaldehyde is an intermediate by-product of the oxidation of BVOCs, it thus provides a proxy for probing local BVOC emissions, and in particular isoprene, which accounts for about 50% of the total BVOC emissions.

In the present study, retrievals of formaldehyde from the Ozone Monitoring Instrument (OMI) are combined with observations of soil moisture, biomass, aerosols, evapotranspiration, drought index, temperature and precipitation. Firstly, we look into the linear annual trend of the selected fields. Secondly, assuming formaldehyde as the dependent variable, we apply a linear mixed model analysis that extends the application of a simple linear regression model by accounting for both fixed (i.e., explained by the independent variables) and random (i.e., due to dependence in the data) effects. The analysis of the linear trend of formaldehyde concentrations shows a positive trend over the Amazon and Central Africa and a negative trend over South Africa and Australia. Over the Amazon, formaldehyde is negatively correlated with the Standardised Precipitation-Evapotranspiration Index (SPEI), a drought index that accounts for both changes in temperature and precipitation, with positive and negative values identifying wet and dry events, respectively. The outcomes of this analysis might provide new insights in the relationship between BVOC emissions and water stress and might help in improving parameterizations that link soil moisture to BVOC emissions in numerical models.