Find in Library

<p><span id="page696"/>In situ observations of spectrally resolved aerosol extinction coefficients (300–700 <span class="inline-formula">nm</span> at <span class="inline-formula">∼</span> 0.8 <span class="inline-formula">nm</span> resolution) from the May–June 2016 Korea–United States Ocean Color (KORUS-OC) oceanographic field campaign are reported. Measurements were made with the custom-built Spectral Aerosol Extinction (SpEx) instrument that previously has been characterized only using laboratory-generated aerosols of known size and composition. Here, the performance of SpEx under realistic operating conditions in the field was assessed by comparison to extinction coefficients derived from commercial instruments that measured scattering and filter-based absorption coefficients at three discrete visible wavelengths. Good agreement was found between these two sets of extinction coefficients with slopes near unity for all three wavelengths within the SpEx measurement error (<span class="inline-formula">±</span> 5 <span class="inline-formula">Mm⁻¹</span>). The meteorological conditions encountered during the cruise fostered diverse ambient aerosol populations with varying sizes and composition at concentrations spanning 2 orders of magnitude. The sampling inlet had a 50 % size cut of 1.3 <span class="inline-formula">µm</span> diameter particles such that the in situ aerosol sampling suite deployed aboard ship measured fine-mode aerosols only. The extensive hyperspectral extinction data set acquired revealed that nearly all measured spectra exhibited curvature in logarithmic space, such that Ångström exponent (<span class="inline-formula"><i>α</i></span>) power law fits could lead to large errors compared to measured values. This problem was particularly acute for <span class="inline-formula"><i>α</i></span> values calculated over only visible wavelengths and then extrapolated to the UV, highlighting the need for measurements in this wavelength range. Second-order polynomial fits to the logarithmically transformed data provided a much better fit to the measured spectra than the linear fits of power laws. Building on previous studies that used total column aerosol optical depth observations to examine the information content of spectral curvature, the relationship between <span class="inline-formula"><i>α</i></span> and the second-order polynomial fit coefficients (<span class="inline-formula"><i>a</i>₁</span> and <span class="inline-formula"><i>a</i>₂</span>) was found to depend on the wavelength range of the spectral measurement such that any given <span class="inline-formula"><i>α</i></span> maps into a line in (<span class="inline-formula"><i>a</i>₁</span>, <span class="inline-formula"><i>a</i>₂</span>) coefficient space with a slope of <span class="inline-formula">−2LN(<i>λ</i>_ch)</span>, where <span class="inline-formula"><i>λ</i>_ch</span> is defined as the single wavelength that characterizes the wavelength range of the measured spectrum (i.e., the “characteristic wavelength”). Since the curvature coefficient values depend on <span class="inline-formula"><i>λ</i>_ch</span>, it must be taken into account when comparing values from spectra obtained from measurement techniques with different <span class="inline-formula"><i>λ</i>_ch</span>. Previously published work has shown that different bimodal size distributions of aerosols can exhibit the same <span class="inline-formula"><i>α</i></span> yet have differing spectral curvature with different (<span class="inline-formula"><i>a</i>₁</span>, <span class="inline-formula"><i>a</i>₂</span>). This implies that (<span class="inline-formula"><i>a</i>₁</span>, <span class="inline-formula"><i>a</i>₂</span>) contain more information about size distributions than <span class="inline-formula"><i>α</i></span> alone. Aerosol size distributions were not measured during KORUS-OC, and the data reported here were limited to the fine fraction, but the (<span class="inline-formula"><i>a</i>₁</span>, <span class="inline-formula"><i>a</i>₂</span>) maps obtained from the SpEx data set are consistent with the expectation that (<span class="inline-formula"><i>a</i>₁</span>, <span class="inline-formula"><i>a</i>₂</span>) may contain more information than <span class="inline-formula"><i>α</i></span> – a result that will be explored further with future SpEx and size distribution data sets.</p>