Items

Advanced search

Using neural networks to describe tracer correlations

Neural networks are ideally suited to describe the spatial and temporal dependence of tracer-tracer correlations. The neural network performs well even in regions where the correlations are less compact and normally a family of correlation curves would be required. For example, the CH4-N2O correlation can be well described using a neural network trained with the latitude, pressure, time of year, and CH4 volume mixing ratio (v.m.r.). In this study a neural network using Quickprop learning and one hidden layer with eight nodes was able to reproduce the CH4-N2O correlation with a correlation coefficient between simulated and training values of 0.9995. Such an accurate representation of tracer-tracer correlations allows more use to be made of long-term datasets to constrain chemical models. Such as the dataset from the Halogen Occultation Experiment (HALOE) which has continuously observed CH4  (but not N2O) from 1991 till the present. The neural network Fortran code used is available for download.
Halogen cycling and aerosol pH in the Hawaiian marine boundary layer

Halogen species (HCl* (primarily HCl), Cl* (including Cl2 and HOCl), BrO, total gaseous inorganic Br and size-resolved particulate Cl- and Br -) and related chemical and physical parameters were measured in surface air at Oahu, Hawaii during September 1999. Aerosol pH as a function of particle size was inferred from phase partitioning and thermodynamic properties of HCl. Mixing ratios of halogen compounds and aerosol pHs were simulated with a new version of the photochemical box model MOCCA that considers multiple aerosol size bins. Inferred aerosol pHs ranged from 4.5 to 5.4 (median 5.1, n=22) for super-mm (primarily sea-salt) size fractions and 2.6 to 5.3 (median 4.6) for sub-mm (primarily sulphate) fractions. Inferred daytime pHs tended to be slightly lower than those at night, although daytime median values did not differ statistically from nighttime medians. Simulated pHs for most sea-salt size bins were within the range of inferred values. However, simulated pHs for the largest size fraction in the model were somewhat higher (oscillating around 5.9) due to the rapid turnover rates and relatively larger infusions of sea-salt alkalinity associated with fresh aerosols. Measured mixing ratios of HCl* ranged from <30 to 250 pmol mol-1 and those for Cl* from <6 to 38 pmol mol-1. Simulated HCl and Cl* (Cl+ClO+HOCl+Cl2) mixing ratios ranged between 20 and 70 pmol mol-1 and 0.5 and 6 pmol mol-1, respectively. Afternoon HCl* maxima occurred on some days but consistent diel cycles for HCl* and Cl* were not observed. Simulated HCl did vary diurnally, peaking before dusk and reaching a minimum at dawn. While individual components of Cl* varied diurnally in the simulations, their sum did not, consistent with the lack of a diel cycle in observed Cl*. Mixing ratios of total gaseous inorganic Br varied from <1.5 to 9 pmol mol-1 and particulate Br - deficits varied from 1 to 6 pmol mol-1 with values for both tending to be greater during daytime. Simulated Brt and Br - mixing ratios and enrichment factors (EFBr) were similar to those observed, with early morning maxima and dusk minima. However, the diel cycles differed in detail among the various simulations. In low-salt simulations, halogen cycling was less intense, Br - accumulated and Brt and EFBr increased slowly overnight. In higher-salt simulations with more intense halogen cycling, Br - and EFBr decreased and Brt increased rapidly after dusk. Cloud processing, which is not considered in this version of MOCCA, may also affect these diel cycles (von Glasow et al., 2003). Measured BrO was never above detection limit (~2 pmol mol-1) during the experiment, however relative changes in the BrO signal during the 3-hour period ending at 11:00 local time were mostly negative, averaging -0.3 pmol mol-1. Both of these results are consistent with MOCCA simulations of BrO mixing ratios. Increasing the sea-salt mixing ratio in MOCCA by ~25% (within observed range) led to a decrease in O3 of ~16%. The chemistry leading to this decrease is complex and is tied to NOx removal by heterogeneous reactions of BrNO3 and ClNO3. The sink of O3 due to the catalytic Cl-ClO and Br-BrO cycles was estimated at -1.0 to -1.5 nmol mol-1 day-1, a range similar to that due to O3 photolysis in the MOCCA simulations.
Role of the NO3 radicals in oxidation processes in the eastern Mediterranean troposphere during the MINOS campaign

During the MINOS campaign (28 July-18 August 2001) the nitrate (NO3) radical was measured at Finokalia station, on the north coast of Crete in South-East Europe using a long path (10.4 km) Differential Optical Absorption Spectroscopy instrument (DOAS). Hydroxyl (OH) radical was also measured by a Chemical Ionization Mass-Spectrometer (Berresheim et al., 2003). These datasets represent the first simultaneous measurements of OH and NO3 radicals in the area. NO3 radical concentrations ranged from less than 3x107 up to 9x108 radicals· cm-3 with an average nighttime value of 1.1x108 radicals· cm-3. The observed NO3 mixing ratios are analyzed on the basis of the corresponding meteorological data and the volatile organic compound (VOC) observations which were measured simultaneously at Finokalia station. The importance of the NO3 radical chemistry relatively to that of OH in the dimethylsulfide (DMS) and nitrate cycles is also investigated. The observed NO3 levels regulate the nighttime variation of DMS. The loss of DMS by NO3 during night is about 75% of that by OH radical during day. NO3 and nitrogen pentoxide (N2O5) reactions account for about 21% of the total nitrate (HNO3(g)+NO-3(g)) production.
Effects of the physical state of tropospheric ammonium-sulfate-nitrate particles on global aerosol direct radiative forcing

The effect of aqueous versus crystalline sulfate-nitrate-ammonium tropospheric particles on global aerosol direct radiative forcing is assessed. A global three-dimensional chemical transport model predicts sulfate, nitrate, and ammonium aerosol mass. An aerosol thermodynamics model is called twice, once for the upper side (US) and once for lower side (LS) of the hysteresis loop of particle phase. On the LS, the sulfate mass budget is 40% solid ammonium sulfate, 12% letovicite, 11% ammonium bisulfate, and 37% aqueous. The LS nitrate mass budget is 26% solid ammonium nitrate, 7% aqueous, and 67% gas-phase nitric acid release due to increased volatility upon crystallization. The LS ammonium budget is 45% solid ammonium sulfate, 10% letovicite, 6% ammonium bisulfate, 4% ammonium nitrate, 7% ammonia release due to increased volatility, and 28% aqueous. LS aerosol water mass partitions as 22% effloresced to the gas-phase and 78% remaining as aerosol mass. The predicted US/LS global fields of aerosol mass are employed in a Mie scattering model to generate global US/LS aerosol optical properties, including scattering efficiency, single scattering albedo, and asymmetry parameter. Global annual average LS optical depth and mass scattering efficiency are, respectively, 0.023 and 10.7 m2 (g SO4-2)-1, which compare to US values of 0.030 and 13.9 m2 (g SO4-2)-1. Radiative transport is computed, first for a base case having no aerosol and then for the two global fields corresponding to the US and LS of the hysteresis loop. Regional, global, seasonal, and annual averages of top-of-the-atmosphere aerosol radiative forcing on the LS and US (FL and FU, respectively, in W m-2) are calculated. Including both anthropogenic and natural emissions, we obtain global annual averages of FL=-0.750, FU=-0.930, and DFU,L=24% for full sky calculations without clouds and FL=-0.485, FU=-0.605, and DFU,L=25% when clouds are included. Regionally, DFU,L=48% over the USA, 55% over Europe, and 34% over East Asia. Seasonally, DFU,L varies from 18% in DJF to 75% in SON over the USA. The global annual average contribution from anthropogenic aerosol is FL=-0.314 and FU=-0.404, which yield normalized direct radiative forcings (G) of GL=-205 W (g SO4-2)-1 and GU=-264 W (g SO4-2)-1.
Tethered balloon measurements of biogenic volatile organic compounds at a Boreal forest site

Measurements of biogenic volatile organic compounds (VOCs) were performed at Hyytiälä, a Boreal forest site in Southern Finland as part of the OSOA (origin and formation of secondary organic aerosol) project in August 2001. At this site, frequent formation of new particles has been observed and the role of biogenic VOCs in this process is still unclear. Tethered balloons served as platforms to collect VOC samples within the planetary boundary layer at heights up to 1.2 km above ground during daytime. Mean mixed layer concentrations of total monoterpenes varied between 10 and 170 pptv, with a-pinene, limonene and D3-carene as major compounds, isoprene was detected at levels of 2-35 pptv. A mixed layer gradient technique and a budget approach are applied to derive surface fluxes representative for areas of tens to hundreds of square kilometres. Effects of spatial heterogeneity in surface emissions are examined with a footprint analysis. Depending on the source area considered, mean afternoon emissions of the sum of terpenes range between 180 and 300 mg m-2 h-1 for the period of 2-12 August 2001. Surface fluxes close to Hyytiälä were higher than the regional average, and agree well with mean emissions predicted by a biogenic VOC emission model. Total rates of monoterpene oxidation were calculated with a photochemical model. The rates did not correlate with the occurrence of new particle formation, but the ozone pathway was of more importance on days with particle formation. Condensable vapour production from the oxidation of monoterpenes throughout the mixed layer can only account for a fraction of the increase in aerosol mass observed at the surface.
New insights in the global cycle of acetonitrile: release from theocean and dry deposition in the tropical savanna of Venezuela

Using the proton transfer reaction mass spectrometry (PTR-MS) technique, acetonitrile was measured during the wet season in a Venezuelan woodland savanna. The site was located downwind of the Caribbean Sea and no biomass burning events were observed in the region. High boundary layer concentrations of 211±36pmol/mol (median, ±standard deviation) were observed during daytime in the well mixed boundary layer, which is about 60pmol/mol above background concentrations recently measured over the Mediterranean Sea and the Pacific Ocean. Most likely acetonitrile is released from the warm waters of the Caribbean Sea thereby enhancing mixing ratios over Venezuela. Acetonitrile concentrations will probably still be much higher in biomass burning plumes, however, the general suitability of acetonitrile as a biomass burning marker should be treated with care. During nights, acetonitrile dropped to levels typically around 120pmol/mol, which is consistent with a dry deposition velocity of <IMG WIDTH="15" HEIGHT="15" ALIGN="BOTTOM" BORDER="0" src="acp-4-275-img2.gif" ALT="${sim}$">0.14cm/s when a nocturnal boundary layer height of 100m is assumed.
Retrieval methods of effective cloud cover from the GOME instrument: an intercomparison

The radiative scattering by clouds leads to errors in the retrieval of column densities and concentration profiles of atmospheric trace gas species from satellites. Moreover, the presence of clouds changes the UV actinic flux and the photo-dissociation rates of various species significantly. The Global Ozone Monitoring Experiment (GOME) instrument on the ERS-2 satellite, principally designed to retrieve trace gases in the atmosphere, is also capable of detecting clouds. Four cloud fraction retrieval methods for GOME data that have been developed are discussed in this paper (the Initial Cloud Fitting Algorithm, the PMD Cloud Recognition Algorithm, the Optical Cloud Recognition Algorithm (an in-house version and the official implementation) and the Fast Retrieval Scheme for Clouds from the Oxygen A-band). Their results of cloud fraction retrieval are compared to each-other and also to synoptic surface observations. It is shown that all studied retrieval methods calculate an effective cloud fraction that is related to a cloud with a high optical thickness. Generally, we found ICFA to produce the lowest cloud fractions, followed by our in-house OCRA implementation, FRESCO, PC2K and finally the official OCRA implementation along four processed tracks (+2%, +10%, +15% and +25% compared to ICFA respectively). Synoptical surface observations gave the highest absolute cloud fraction when compared with individual PMD sub-pixels of roughly the same size.
Multi axis differential optical absorption spectroscopy (MAX-DOAS)

Multi Axis Differential Optical Absorption Spectroscopy (MAX-DOAS) in the atmosphere is a novel measurement technique that represents a significant advance on the well-established zenith scattered sunlight DOAS instruments which are mainly sensitive to stratospheric absorbers. MAX-DOAS utilizes scattered sunlight received from multiple viewing directions. The spatial distribution of various trace gases close to the instrument can be derived by combining several viewing directions. Ground based MAX-DOAS is highly sensitive to absorbers in the lowest few kilometres of the atmosphere and vertical profile information can be retrieved by combining the measurements with Radiative Transfer Model (RTM) calculations. The potential of the technique for a wide variety of studies of tropospheric trace species and its (few) limitations are discussed. A Monte Carlo RTM is applied to calculate Airmass Factors (AMF) for the various viewing geometries of MAX-DOAS. Airmass Factors can be used to quantify the light path length within the absorber layers. The airmass factor dependencies on the viewing direction and the influence of several parameters (trace gas profile, ground albedo, aerosol profile and type, solar zenith and azimuth angles) are investigated. In addition we give a brief description of the instrumental MAX-DOAS systems realised and deployed so far. The results of the RTM studies are compared to several examples of recent MAX-DOAS field experiments and an outlook for future possible applications is given.
Particle size distribution and particle mass measurements at urban,near-city and rural level in the Copenhagen area and Southern Sweden

Particle size distribution (size-range 3-900nm) and PM10 was measured simultaneously at an urban background station in Copenhagen, a near-city background and a rural location during a period in September-November 2002. The study investigates the contribution from urban versus regional sources of particle number and mass concentration. The total particle number (ToN) and NOx are well correlated at the urban and near-city level and show a distinct diurnal variation, indicating the common traffic source. The average ToN at the three stations differs by a factor of 3. The observed concentrations are 2500#cm<IMG WIDTH="20" HEIGHT="18" ALIGN="BOTTOM" BORDER="0" src="acp-4-281-img1.gif" ALT="$^{-3}$">, 4500#cm<IMG WIDTH="20" HEIGHT="18" ALIGN="BOTTOM" BORDER="0" src="acp-4-281-img1.gif" ALT="$^{-3}$"> and 7700#cm<IMG WIDTH="20" HEIGHT="18" ALIGN="BOTTOM" BORDER="0" src="acp-4-281-img1.gif" ALT="$^{-3}$"> at rural, near-city and urban level, respectively. PM10 and total particle volume (ToV) are well correlated between the three different stations and show similar concentration levels, in average within 30% relative difference, indicating a common source from long-range transport that dominates the concentrations at all locations. Measures to reduce the local urban emissions of NOx and ToN are likely to affect both the street level and urban background concentrations, while for PM10 and ToV only measurable effects at the street level are probable. Taking into account the supposed stronger health effects of ultrafine particles reduction measures should address particle number emissions. The traffic source contributes strongest in the 10-200nm particle size range. The maximum of the size distribution shifts from about 20-30nm at kerbside to 50-60nm at rural level. Particle formation events were observed in the 3-20nm size range at rural location in the afternoon hours, mainly under conditions with low concentrations of pre-existing aerosol particles. The maximum in the size distribution of the "traffic contribution" seems to be shifted to about 28nm in the urban location compared to 22nm at kerbside. Assuming NOx as an inert tracer on urban scale allows to estimate that ToN at urban level is reduced by 15-30% compared to kerbside. Particle removal processes, e.g. deposition and coagulation, which are most efficient for smallest particle sizes (<IMG WIDTH="15" HEIGHT="29" ALIGN="MIDDLE" BORDER="0" src="acp-4-281-img2.gif" ALT="$<$">20nm) and condensational growth are likely mechanisms for the loss of particle number and the shift in particle size.
Aerosol-cirrus interactions: a number based phenomenon at all?

In situ measurements of the partitioning of aerosol particles within cirrus clouds were used to investigate aerosol-cloud interactions in ice clouds. The number density of interstitial aerosol particles (non-activated particles in between the cirrus crystals) was compared to the number density of cirrus crystal residuals. The data was obtained during the two INCA (Interhemispheric Differences in Cirrus Properties from Anthropogenic Emissions) campaigns, performed in the Southern Hemisphere (SH) and Northern Hemisphere (NH) midlatitudes. Different aerosol-cirrus interactions can be linked to the different stages of the cirrus lifecycle. Cloud formation is linked to positive correlations between the number density of interstitial aerosol (Nint) and crystal residuals (Ncvi), whereas the correlations are smaller or even negative in a dissolving cloud. Unlike warm clouds, where the number density of cloud droplets is positively related to the aerosol number density, we observed a rather complex relationship when expressing Ncvi as a function of Nint for forming clouds. The data sets are similar in that they both show local maxima in the Nint range 100 to 200cm<IMG WIDTH="20" HEIGHT="18" ALIGN="BOTTOM" BORDER="0" src="acp-4-293-img1.gif" ALT="$^{-3}$">, where the SH- maximum is shifted towards the higher value. For lower number densities Nint and Ncvi are positively related. The slopes emerging from the data suggest that a tenfold increase in the aerosol number density corresponds to a 3 to 4 times increase in the crystal number density. As Nint increases beyond the ca. 100 to 200cm<IMG WIDTH="20" HEIGHT="18" ALIGN="BOTTOM" BORDER="0" src="acp-4-293-img1.gif" ALT="$^{-3}$">, the mean crystal number density decreases at about the same rate for both data sets. For much higher aerosol number densities, only present in the NH data set, the mean Ncvi remains low. The situation for dissolving clouds allows us to offer two possible, but at this point only speculative, alternative interactions between aerosols and cirrus: evaporating clouds might be associated with a source of aerosol particles, or air pollution (high aerosol number density) might retard ice particle evaporation rates.