Is the biosphere-atmosphere exchange of total reactive nitrogen above forest driven by the same factors as carbon dioxide? An analysis using artificial neural networks
Phase and amplitude of ecosystem-atmosphere fluxes of reactive nitrogen compounds are poorly understood due to a lack of suitable observation methods. Understanding the biophysical controls of the surface nitrogen exchange is essential for the parameterization of process-based and chemical transport models that can be used for the determination of regional or national nitrogen budgets. In this study, we investigated similarities in time series of net total reactive nitrogen (ΣNr) and carbon dioxide (CO2) fluxes above forest with regard to their variability and driving factors. We found corresponding shapes of the mean diurnal summertime patterns of ΣNr and CO2. While ecosystem respiration leads to a net CO2 release at night, ΣNr was on average deposited throughout the entire observation period. Using artificial neural network analysis, global radiation (Rg) was identified to be the main control for both ΣNr and CO2. While the concentration of ΣNr substantially improved the coefficient of determination for ΣNr fluxes when used as a secondary driver, only minor improvements of 2–3% were found for CO2 fluxes when using for example temperature or vapour pressure deficit (VPD) as secondary driver. Considering two dominant drivers, 41 and 66% of the variability in ΣNr and CO2 fluxes, respectively, could be explained. Further data stratification for ΣNr revealed that higher concentrations, higher temperature, and higher VPD as well as dry leaf surfaces tend to favour higher deposition of ΣNr, whereas lower concentrations, lower temperature, and lower VPD as well as wet leaf surfaces mainly correspond to situations when less ΣNr was deposited or even emitted. Our results support the understanding of biosphere-atmosphere interactions, their driving factors, and establish a link between ΣNr and CO2 exchange, which may be beneficial for future developments in state-of-the-art exchange modelling.