Improving estimates of the atmospheric oxidative capacity and Amazon fire emissions

Stijn Naus

Research output: Thesisinternal PhD, WU


The composition of the air outside, the atmosphere, is the end-result of complex chemistry, dynamical transport and emissions of gases. Anthropogenic activity contributes to a rapidly changing composition of the atmosphere, often with adverse effects: air quality issues, global warming and depletion of stratospheric ozone are some of the most complex and impactful problems that society must face. Many gases are removed from the atmosphere through the process of oxidation and the atmosphere’s primary oxidant is the hydroxyl radical (OH). Gases removed by OH include the potent greenhouse gas methane (CH4) and urban pollutants such as CO and NOX. Therefore, the oxidation capacity of the atmosphere, as determined by the "cleansing-agent” OH, is an important quantity that is central to this thesis.

Since OH is very reactive, it has an atmospheric lifetime of seconds and a low atmospheric abundance. This makes it difficult to measure OH concentrations directly and impossible to extrapolate those measurements to larger scales. Consequently, despite the importance and omnipresence of OH, there are still open questions concerning its distribution around the globe and how this distribution has varied over past decades. For example, we know that OH concentrations are highest in the tropics, but the degree of hemispheric symmetry is uncertain. Such uncertainties limit our ability to interpret atmospheric budgets of a variety of pollutants, such as that of CH4.

In Chapters 2 and 3, we investigate indirect observational constraints on the atmospheric oxidative capacity using the trace gas methyl chloroform (MCF). MCF is an anthropogenically produced gas, that was used mainly as a solvent in paints and degreasers. The production of MCF was phased out in the Montreal protocol and subsequent amendments (1987-1999), because its emissions harm the stratospheric ozone layer. Fortuitously, MCF is removed from the atmosphere mainly through oxidation by OH, resulting in an atmospheric lifetime of 5 to 6 years. Therefore, the production phase-out resulted in a rapid atmospheric decline in MCF abundance, and variations in this rate of decline provide a proxy for large-scale OH variations.

The atmospheric decline of MCF has been monitored mainly from remote surface sites that represent the background atmosphere. Comprehensive interpretation of such observations requires the use of atmospheric models. Forward models simulate emissions of gases into the atmosphere, and their subsequent transport and chemistry. Such models can vary in complexity from a global one-box model to a complex 3D transport mode. In addition to forward models, we develop and use inverse models, which use observations of a gas such as MCF to estimate, for example, those MCF emissions and OH variations.

In Chapter 2, we use a two-box model inversion of the troposphere to explore the constraints that MCF surface observations place on global-scale OH variations over the 1994-2014 period. The two-box model set-up incorporates important aspects of the real atmosphere, because the atmosphere is mixed significantly faster within hemispheres than between them and because most anthropogenic emissions enter the atmosphere in the Northern Hemisphere. Therefore, a similar set-up was applied to the MCF-OH problem in two previous studies. However, we find that the two-box model approach is susceptible to biases that we quantify with a 3D transport model.

Firstly, we find, that the interhemispheric exchange rate that is used in two-box models is not only reflective of physical transport changes, but is also highly dependent on a tracer's distribution within hemispheres. The distribution of MCF has changed strongly over the 1994-2014 period that we investigate, due to the rapid drop in its emissions. We find that this change resulted in large, multi-annual variations in the interhemispheric exchange of MCF, that are very different from the variations derived for CH4. The same redistribution of MCF drove a rapid decrease in loss of tropospheric MCF to the stratosphere by up to 70%, as well as a shift in the effective interhemispheric OH ratio that MCF is exposed to.

To test the impact of the identified biases on the OH variations and CH4 emissions derived in a tropospheric two-box model inversion, we perform two sets of inversions: one in which we correct for the biases, and one in which we do not. Notably, we find that the inversion that includes bias corrections produces OH variations that show a positive trend, while the standard inversion does not. However, we also find that the uncertainties driven by other two-box parameters that cannot be constrained from a 3D model simulation, for example related to MCF emissions, remain large. The implication for the CH4 budget is that it remains difficult to attribute variations in the atmospheric CH4 abundance to either emission or OH changes.

In Chapter 3, we investigate how MCF constraints on OH change if we move the inversion to the 3D transport model TM5. In a TM5-4DVAR inversion, we can include observed MCF gradients within hemispheres and the tropical maximum in OH: examples of advantages over the two-box model approach. We co-optimize MCF emissions and the latitudinal distribution of OH over the 1998--2018 period, and we find that small interannual variations in OH (<3%) without a longterm trend already result in a good match with NOAA surface observations at most sites. The timing and sign of interannual OH variations are found to be robust with respect to the choice of prior OH and MCF emission distribution, while the amplitude of the variations depends on the degree of convergence and, relatedly, on assumed uncertainties in observations.

However, we also find that, to reproduce observed intrahemispheric gradients of MCF, large adjustments in the latitudinal OH distribution are required, the amplitude of which we consider to be physically unrealistic. From a series of sensitivity tests we conclude that the most likely explanation for these MCF gradients includes a changed ocean flux. Specifically, while the ocean is principally a sink of MCF, earlier work has hypothesized that, in response to the emission drop, oceans at high latitudes have become a source of MCF, which fits the sign and approximate magnitude of the intrahemispheric bias we find for MCF. Positively, our study presents the first evidence from MCF surface observations for an ocean source. Negatively, the existence of an uncertain ocean source further complicates the derivation of OH variations from MCF observations. However, we consider that the most likely driver of interannual variations in MCF is still its OH oxidation sink. Therefore, we conclude that the reference twenty-year timeseries of OH that we have derived is worth including in future work, for example in global CH4 inversions.

In Chapter 4, we move away from a global perspective to zoom in on one of Earth's most precious ecosystems: the Amazon basin. The Amazon is home to the world's largest rainforests and to a rich biodiversity, but large-scale deforestation and agricultural expansion threaten the ability of this ecosystem to survive in a rapidly changing climate. Every year, during the local dry season, fires burn through the Amazon forests and savanna. It is crucial to understand and monitor these fires, because they are driven by direct, local anthropogenic activity, but their extent is also sensitive to drought intensity and frequency that might increase due to climate change. Fires emit vast amounts of pollutants into the atmosphere, and we use satellite observations of one such trace gas, CO, to constrain fire emissions over South America. Specifically, we use the TM5-4DVAR inverse system, together with satellite-observed CO columns of the Measurements of Pollution in the Troposphere (MOPITT) instrument, to optimize reported fire emissions of CO (from the Global Fire Assimilation System; GFAS) over the 2003--2018 period.

MOPITT CO columns over South America display strong seasonality and interannual variations that are matched in the optimized CO emissions. Additionally, a simulation with optimized fire emissions better reproduces independent aircraft profiles, which were sampled over five sites in the Brazilian Amazon, compared to a simulation with GFAS emissions. This confirms the skill of the inverse system to estimate emissions at sub-Amazon scales. Similarly, we find that we can firmly constrain emissions also at the level of individual Brazilian states, and interannual variations in emissions at state-level correlate well with local soil moisture anomalies. Superimposed on the interannual variations, we find that emissions have decreased between 2003 and 2012, and stabilized afterward. The decrease is especially strong over forest-covered areas (55%), and the timing and magnitude of this decrease is confirmed in deforestation rates reported by INPE. Optimized emissions are additionally found to be robust with respect to the input fields: for example, inversions based on a climatological fire prior retrieve largely the same interannual variations, even at state-level, as the standard inversions.

These results demonstrate that our inverse system provides strong constraints on Amazon fire emissions that are a product of local anthropogenic activity and natural variability. In principle, our approach can straightforwardly be adapted into an operational system that would be a valuable addition to the existing palette of fire monitoring systems. Due to the integrated signal that CO observations provide, such a system will be less hampered by cloud occurrence or missed understory fires than existing systems that are based on land remote sensing. Moreover, constraints from CO can help better quantify the carbon release during fires.

Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Wageningen University
  • Krol, Maarten, Promotor
  • Peters, Wouter, Promotor
  • Montzka, S.A., Co-promotor, External person
Award date3 Feb 2021
Place of PublicationWageningen
Print ISBNs9789463956420
Publication statusPublished - 2021


Dive into the research topics of 'Improving estimates of the atmospheric oxidative capacity and Amazon fire emissions'. Together they form a unique fingerprint.

Cite this