Chasing down OH with the help speech recognition techniques

Detergent for air

OH radicals are very reactive molecules and important oxidants in the atmosphere. We sometimes describe them as a detergent for air, since OH reacts with many trace gases, thus oxidizing and removing them from the atmosphere. Among those trace gases are emissions from human activities and natural sources, including greenhouse gases and substances that are potentially harmful to human health.

However, direct measurements of OH are rare and difficult to achieve. However, knowing the OH concentration and its variability throughout the day is key to understanding the air chemistry taking place in the interface region of the rainforest and the atmosphere.

OH radicals are produced when solar radiation breaks down ozone molecules to produce atomic oxygen. This then further reacts with water molecules to form OH. In the tropics the solar radiation is high and it is very humid. This means the requirements for the formation of OH are well met. Thus, many OH radicals are produced above the Amazon rainforest.

Turning a problem into a solution

At the same time, this region is a huge source of biogenic volatile organic compounds (BVOC). They react predominantly with OH radicals, removing both OH and BVOCs from the atmosphere through the formation of other compounds. This process creates problems for global atmospheric models in this region as they tend to overestimate the BVOC mixing ratios.

To address this, Akima Ringsdorf and Achim Edtbauer from the working group of Jonathan Williams at MPI-C took advantage of the reaction of OH with BVOCs. Measurements of BVOCs are easier to achieve than those of OH, and commonly collected at ATTO. They measured BVOC concentrations at three heights (80, 150 and 325 m) along the ATTO tower with a Proton Transfer-Mass Spectrometer. This instrument is capable of detecting many BVOC continuously in real time. This includes isoprene, the most abundant BVOC emitted from forest environments.

They observed that isoprene concentrations decrease with increasing height. This makes sense since further up means further away from the source, mainly the forest canopy. In addition, as air-carrying BVOCs move upwards from the canopy, isoprene is also removed by the reaction with OH radicals. Therefore, you can calculate the amount of OH necessary to obtain the observed decrease of isoprene. But you also need to know the time that isoprene is exposed to OH during the transport between 80 and 325 m, i.e. the time it took for the air to turbulently travel to the top of the tower.

Akima Ringsdorf at the top of the ATTO tall tower.
© Carolina Monteiro

Getting help from speech recognition

Akima Ringsdorf and her colleagues need to create a new method to estimate this reaction. It takes into account the warm-up and cold downward motions that result from the heating of the low atmosphere during daytime. For this new method, they looked to the field of speech recognition. They adapted a technique called ‘Dynamical Time Warping’ for the use on atmospheric parameters. Dynamical Time Warping allowed them to look at the time shift between the diurnal evolution of the potential temperature at 80 and 325 m and quantify the time that it takes to propagate the heat upwards.

The calculation of the OH concentration from the fate of isoprene within the derived mixing time then holds another challenge. The decrease of isoprene with height results not only from the removal by OH. The strength of the emission of isoprene at the canopy level also influences this, as well as entrainment from isoprene-free air from layers above and turbulent mixing of those air masses. To determine the influence of those dynamic processes on the observed decrease of isoprene they used a turbulence-resolving model.

The resulting diurnal time series of OH shows the highest concentrations around 14:00 local time with concentrations up to 2.2 million molecules per cubic centimeter. This is a good agreement with measurements of OH in similar environments. Thus, for tower-based measurement sites, the developed method presents a good substitution if direct OH measurements are not available.

So finally, Akima and her colleagues compared their OH values calculated from isoprene measurements to values predicted by atmospheric models. They find huge differences between the model results in the first 380 m above the canopy, and a better agreement at higher levels. They therefore conclude that the region characterized by a tall canopy is still difficult to model accurately due to complex atmospheric chemistry and dynamics.

Akima Ringsdorf et al. published the study “Inferring the diurnal variability of OH radical concentrations over the Amazon from BVOC measurements” Open Access in Nature Scientific Reports.

Similar articles