Atmospheric Sampling on Ascension Island Using Multirotor UAVs.

As part of an NERC-funded project investigating the southern methane anomaly, a team drawn from the Universities of Bristol, Birmingham and Royal Holloway flew small unmanned multirotors from Ascension Island for the purposes of atmospheric sampling. The objective of these flights was to collect air samples from below, within and above a persistent atmospheric feature, the Trade Wind Inversion, in order to characterise methane concentrations and their isotopic composition. These parameters allow the methane in the different air masses to be tied to different source locations, which can be further analysed using back trajectory atmospheric computer modelling. This paper describes the campaigns as a whole including the design of the bespoke eight rotor aircraft and the operational requirements that were needed in order to collect targeted multiple air samples up to 2.5 km above the ground level in under 20 min of flight time. Key features of the system described include real-time feedback of temperature and humidity, as well as system health data. This enabled detailed targeting of the air sampling design to be realised and planned during the flight mission on the downward leg, a capability that is invaluable in the presence of uncertainty in the pre-flight meteorological data. Environmental considerations are also outlined together with the flight plans that were created in order to rapidly fly vertical transects of the atmosphere whilst encountering changing wind conditions. Two sampling campaigns were carried out in September 2014 and July 2015 with over one hundred high altitude sampling missions. Lessons learned are given throughout, including those associated with operating in the testing environment encountered on Ascension Island.

Model-data synthesis for the next generation of forest free-air CO2 enrichment (FACE) experiments.

The first generation of forest free-air CO2 enrichment (FACE) experiments has successfully provided deeper understanding about how forests respond to an increasing CO2 concentration in the atmosphere. Located in aggrading stands in the temperate zone, they have provided a strong foundation for testing critical assumptions in terrestrial biosphere models that are being used to project future interactions between forest productivity and the atmosphere, despite the limited inference space of these experiments with regards to the range of global ecosystems. Now, a new generation of FACE experiments in mature forests in different biomes and over a wide range of climate space and biodiversity will significantly expand the inference space. These new experiments are: EucFACE in a mature Eucalyptus stand on highly weathered soil in subtropical Australia; AmazonFACE in a highly diverse, primary rainforest in Brazil; BIFoR-FACE in a 150-yr-old deciduous woodland stand in central England; and SwedFACE proposed in a hemiboreal, Pinus sylvestris stand in Sweden. We now have a unique opportunity to initiate a model-data interaction as an integral part of experimental design and to address a set of cross-site science questions on topics including responses of mature forests; interactions with temperature, water stress, and phosphorus limitation; and the influence of biodiversity.

Modelling component evaporation and composition change of traffic-induced ultrafine particles during travel from street canyon to urban background.

We developed a model (CiTTy-Street-UFP) of traffic-related particle behaviour in a street canyon and in the nearby downwind urban background that accounts for aerosol dynamics and the variable vapour pressure of component organics. The model simulates the evolution and fate of traffic generated multicomponent ultrafine particles (UFP) composed of a non-volatile core and 17 Semi-Volatile Organic Compounds (SVOC, modelled as n-alkane proxies). A two-stage modelling approach is adopted: (1) a steady state simulation inside the street canyon is achieved, in which there exists a balance between traffic emissions, condensation/evaporation, deposition, coagulation and exchange with the air above roof-level; and (2) a continuing simulation of the above-roof air parcel advected to the nearby urban park during which evaporation is dominant. We evaluate the component evaporation and associated composition changes of multicomponent organic particles in realistic atmospheric conditions and compare our results with observations from London (UK) in a street canyon and an urban park. With plausible input conditions and parameter settings, the model can reproduce, with reasonable fidelity, size distributions in central London in 2007. The modelled nucleation-mode peak diameter, which is 23 nm in the steady-state street canyon, decreases to 9 nm in a travel time of just 120 s. All modelled SVOC in the sub-10 nm particle size range have evaporated leaving behind only non-volatile material, whereas modelled particle composition in the Aitken mode contains SVOC between C26H54 and C32H66. No data on particle composition are available in the study used for validation, or elsewhere. Measurements addressing in detail the size resolved composition of the traffic emitted UFP in the atmosphere are a high priority for future research. Such data would improve the representation of these particles in dispersion models and provide the data essential for model validation. Enhanced knowledge of the chemical composition of nucleation-mode particles from diesel engine exhaust is needed to predict both their atmospheric behaviour and their implications for human health.

Variation in forest root image annotation by experts, novices, and AI.

BACKGROUND: The manual study of root dynamics using images requires huge investments of time and resources and is prone to previously poorly quantified annotator bias. Artificial intelligence (AI) image-processing tools have been successful in overcoming limitations of manual annotation in homogeneous soils, but their efficiency and accuracy is yet to be widely tested on less homogenous, non-agricultural soil profiles, e.g., that of forests, from which data on root dynamics are key to understanding the carbon cycle. Here, we quantify variance in root length measured by human annotators with varying experience levels. We evaluate the application of a convolutional neural network (CNN) model, trained on a software accessible to researchers without a machine learning background, on a heterogeneous minirhizotron image dataset taken in a multispecies, mature, deciduous temperate forest. RESULTS: Less experienced annotators consistently identified more root length than experienced annotators. Root length annotation also varied between experienced annotators. The CNN root length results were neither precise nor accurate, taking ~ 10% of the time but significantly overestimating root length compared to expert manual annotation (p = 0.01). The CNN net root length change results were closer to manual (p = 0.08) but there remained substantial variation. CONCLUSIONS: Manual root length annotation is contingent on the individual annotator. The only accessible CNN model cannot yet produce root data of sufficient accuracy and precision for ecological applications when applied to a complex, heterogeneous forest image dataset. A continuing evaluation and development of accessible CNNs for natural ecosystems is required.

Machine learning methods for low-cost pollen monitoring - Model optimisation and interpretability.

Pollen is a major issue globally, causing as much as 40 % of the population to suffer from hay fever and other allergic conditions. Current techniques for monitoring pollen are either laborious and slow, or expensive, thus alternative methods are needed to provide timely and more localised information on airborne pollen concentrations. We have demonstrated previously that low-cost Optical Particle Counter (OPC) sensors can be used to estimate pollen concentrations when machine learning methods are used to process the data and learn the relationships between OPC output data and conventionally measured pollen concentrations. This study demonstrates how methodical hyperparameter tuning can be employed to significantly improve model performance. We present the results of a range of models based on tuned hyperparameter configurations trained to predict Poaceae (Barnhart), Quercus (L.), Betula (L.), Pinus (L.) and total pollen concentrations. The results achieved here are a significant improvement on results we previously reported: the average R2 scores for the total pollen models have at least doubled compared to using previous parameter settings. Furthermore, we employ the explainable Artificial Intelligence (XAI) technique, SHAP, to interpret the models and understand how each of the input features (i.e. particle sizes) affect the estimated output concentration for each pollen type. In particular, we found that Quercus pollen has a strong positive correlation with particles of optical diameter 1.7-2.3 µm, which distinguishes it from other pollen types such as Poaceae and may suggest that type-specific subpollen particles are present in this size range. There is much further work to be done, especially in training and testing models on data obtained across different environments to evaluate the extent of generalisability. Nevertheless, this work demonstrates the potential this method can offer for low-cost monitoring of pollen and the valuable insight we can gain from what the model has learned.

Constructing a pollen proxy from low-cost Optical Particle Counter (OPC) data processed with Neural Networks and Random Forests.

Pollen allergies affect a significant proportion of the global population, and this is expected to worsen in years to come. There is demand for the development of automated pollen monitoring systems. Low-cost Optical Particle Counters (OPCs) measure particulate matter and have attractive advantages of real-time high time resolution data and affordable costs. This study asks whether low-cost OPC sensors can be used for meaningful monitoring of airborne pollen. We employ a variety of methods, including supervised machine learning techniques, to construct pollen proxies from hourly-average OPC data and evaluate their performance, holding out 40 % of observations to test the proxies. The most successful methods are supervised machine learning Neural Network (NN) and Random Forest (RF) methods, trained from pollen concentrations collected from a Hirst-type sampler. These perform significantly better than using a simple particle size-filtered proxy or a Positive Matrix Factorisation (PMF) source apportionment pollen proxy. Twelve NN and RF models were developed to construct a pollen proxy, each varying by model type, input features and target variable. The results show that such models can construct useful information on pollen from OPC data. The best metrics achieved (Spearman correlation coefficient = 0.85, coefficient of determination = 0.67) were for the NN model constructing a Poaceae (grass) pollen proxy, based on particle size information, temperature, and relative humidity. Ability to distinguish high pollen events was evaluated using F1 Scores, a score reflecting the fraction of true positives with respect to false positives and false negatives, with promising results (F1 ≤ 0.83). Model-constructed proxies demonstrated the ability to follow monthly and diurnal trends in pollen. We discuss the suitability of OPCs for monitoring pollen and offer advice for future progress. We demonstrate an attractive alternative for automated pollen monitoring that could provide valuable and timely information to the benefit of pollen allergy sufferers.

Quantification and uncertainty of root growth stimulation by elevated CO2 in a mature temperate deciduous forest.

Increasing CO2 levels are a major global challenge, and the potential mitigation of anthropogenic CO2 emissions by natural carbon sinks remains poorly understood. The uptake of elevated CO2 (eCO2) by the terrestrial biosphere, and subsequent sequestration as biomass in ecosystems, remain hard to quantify in natural ecosystems. Here, we combine field observations of fine root stocks and flows, derived from belowground imaging and soil cores, with image analysis, stochastic modelling, and statistical inference, to elucidate belowground root dynamics in a mature temperate deciduous forest under free-air eCO2 to 150 ppm above ambient levels. eCO2 led to relatively faster root production (a peak volume fold change of 4.52 ± 0.44 eCO2 versus 2.58 ± 0.21 control), with increased root elongation relative to decay the likely causal mechanism for this acceleration. Physical analysis of 552 root systems from soil cores support this picture, with lengths and widths of fine roots significantly increasing under eCO2. Estimated fine root contributions to belowground net primary productivity increase under eCO2 (mean annual 204 ± 93 g dw m-2 yr-1 eCO2 versus 140 ± 60 g dw m-2 yr-1 control). This multi-faceted approach thus sheds quantitative light on the challenging characterisation of the eCO2 response of root biomass in mature temperate forests.

Chemistry, transport, emission, and shading effects on NO2 and Ox distributions within urban canyons.

The capacity to predict NO2 and the total oxidant (Ox = NO2 + O3) within street canyons is critical for the assessment of air quality regulations aimed at enhancing human wellbeing in urban hotspots. However, such assessment requires the coupling of numerous processes at the street-scale, such as vehicular emissions and tightly coupled transport and photochemical processes. Photochemistry, in particular, is often ignored, heavily simplified, or parameterized. In this study, MBM-FleX - a process-based street canyon model that allows fast computation of various emission profiles and sun-lit conditions with tightly coupled physical (transport and mixing) and chemical processes and without loss of sufficient spatial resolution - was used to simulate shading effects on reactive species within urban canyons. Driven by pre-generated large-eddy simulation of flow, MBM-FleX results show that shading effects on volatile organic compound (VOC) free-radicals significantly affect the interconversion of odd-oxygen species that cannot be captured by the simple NOx-O3 chemistry, for example, reducing NO2 by limiting the formation of hydroperoxyl radicals. Consistent with previous results in simpler model systems, the inclusion of VOC free-radical chemistry did not appreciably alter the sensitivity of NO2 to shading intensity in regular canyons, but a non-linear relationship between NO2 and shading intensity is found in deep canyons when the air residence time grew. When solar incidence simultaneously passes through multiple vortices in street canyons, VOC chemistry and shade may considerably influence model results, which may therefore affect the development of urban planning strategies and personal exposure evaluation.

Using green infrastructure to improve urban air quality (GI4AQ).

As evidence for the devastating impacts of air pollution on human health continues to increase, improving urban air quality has become one of the most pressing tasks facing policy makers world-wide. Increasingly, and very often on the basis of conflicting and/or weak evidence, the introduction of green infrastructure (GI) is seen as a win-win solution to urban air pollution, reducing ground-level concentrations without imposing restrictions on traffic and other polluting activities. The impact of GI on air quality is highly context dependent, with models suggesting that GI can improve urban air quality in some situations, but be ineffective or even detrimental in others. Here we set out a novel conceptual framework explaining how and where GI can improve air quality, and offer six specific policy interventions, underpinned by research, that will always allow GI to improve air quality. We call GI with unambiguous benefits for air quality GI4AQ. However, GI4AQ will always be a third-order option for mitigating air pollution, after reducing emissions and extending the distance between sources and receptors.

Neighbourhood-scale dispersion of traffic-induced ultrafine particles in central London: WRF large eddy simulations.

Traffic-generated ultrafine particles (UFPs) in the urban atmosphere have a high proportion of their composition comprised of semi-volatile compounds (SVOCs). The evaporation/condensation processes of these SVOCs can alter UFP number size distributions and play an important role in determining the fate of UFPs in urban areas. The neighbourhood-scale dispersion (over distances < 1 km) and evolution of traffic-generated UFPs for a real-world street network in central London was simulated by using the WRF-LES model (the large eddy simulation mode of the Weather Research and Forecasting modelling system) coupled with multicomponent microphysics. The neighbourhood scale dispersion of UFPs was significantly influenced by the spatial pattern of the real-world street emissions. Model output indicated the shrinkage of the peak diameter from the emitted profile to the downwind profile, due to an evaporation process during neighbourhood-scale dispersion. The dilution process and the aerosol microphysics interact with each other during the neighbourhood dispersion of UFPs, yielding model output that compares well with measurements made at a location downwind of an intense roadside source. The model captured the total SVOC concentrations well, with overestimations for gas concentrations and underestimations for particle concentrations, particularly of the lighter SVOCs. The contribution of the intense source, Marylebone Road (MR) in London, to concentrations at the downwind location (as estimated by a model scenario with emissions from MR only) is comparable with that of the rest of the street network (a scenario without emissions from MR), implying that both are important. An appreciable level of non-linearity is demonstrated for nucleation mode UFPs and medium range carbon SVOCs at the downwind receptor site.

Modelling traffic-induced multicomponent ultrafine particles in urban street canyon compartments: Factors that inhibit mixing.

This study implements a two-box model coupled with ultrafine particle (UFP) multicomponent microphysics for a compartmentalised street canyon. Canyon compartmentalisation can be described parsimoniously by three parameters relating to the features of the canyon and the atmospheric state outside the canyon, i.e. the heterogeneity coefficient, the vortex-to-vortex exchange velocity, and the box height ratio. The quasi-steady solutions for the two compartments represent a balance among emissions, microphysical aerosol dynamics (i.e. evaporation/condensation of semi-volatiles, SVOCs), and exchange processes, none of which is negligible. This coupled two-box model can capture significant contrasts in UFP number concentrations and a measure of the volatility of the multi-SVOC-particles in the lower and upper canyon. Modelled ground-level UFP number concentrations vary across nucleation, Aitken, and accumulation particle modes as well-defined monotonic functions of canyon compartmentalisation parameters. Compared with the two-box model, a classic one-box model (without canyon compartmentalisation) leads to underestimation of UFP number concentrations by several tens of percent typically. By quantifying the effects of canyon compartmentalisation, this study provides a framework for understanding how canyon geometry and the presence of street trees, street furniture, and architectural features interact with the large-scale atmospheric flow to determine ground-level pollutant concentrations.