mainly sampled Asian outflow in the Alaska phase local forest fire smoke in the Canada phase We expect these phases to represent the two extremes among aerosol environments common over the globe. ARCTAS Instruments relevant to this study are AATS-14 (tracking sunphotometer) for AOD Nephelometers for scattering (dry and humid) PSAP for absorption coefficient COBALT for carbon monoxide Abstract Experiment and Instruments References The properties of tropospheric aerosol and gas vary within a satellite grid cell and between ground-based instruments. This hinders comparison between satellite and suborbital measurements of different spatial scales as well as their applications to climate and air quality studies. This paper quantifies the realistic range of the variability in aerosol optical depth (AOD), its Angstrom exponent, in-situ extinction coefficient and carbon monoxide mixing ratio over horizontal distances of 1-30 km, using measurements from the ARCTAS airborne experiment. The Canada phase in June and July 2008, in which smoke from local forest fires was sampled, likely represents the most heterogeneous of the ambient aerosol environments common over the globe. The relative standard deviation (std rel ) of AOD measured with the 14-channel Ames Airborne Tracking Sunphotometer (AATS-14) has median 19.4% (at 499 nm) among thousands of horizontal 20 km segments. For 6 km segments the analogous median is 9.1%. Another measure of horizontal variability, the autocorrelation (r) of AOD499 across 20 km and 6 km segments is 0.37 and 0.71, respectively. In contrast, the Alaska phase in April 2008, which sampled particles transported from Asia, is presumably among the most homogeneous environments. The median std rel is 3.0% and r is 0.90, both over 30 km, only slightly different from those for 1 km (stdrel=0.4% and r=1.00). in the Canada phase r for in situ extinction coefficient (from a nephelometer and a particle soot absorption photometer) is ~0.2 less than for the AOD. It is ~0.1 less than for the carbon monoxide mixing ratio. The trends of horizontal variability with distance and aerosol environment are different for the wavelength dependence and the humidity response of light scattering. We discuss challenges in estimating aerosol optical properties, particle size and chemical composition from measurements at a distant location. The statistical parameters thus help interpret existing remote-sensing observations and design future ones. Horizontal variability is represented by (relative) standard deviation and autocorrelation calculated for thousands of horizontal flight segments. We describe their changes with distance, environment and property, in an attempt to help interpret and design remote sensing observations. Observed Horizontal Variability and Implications for Remote Sensing Horizontal variability of aerosol optical properties observed during the ARCTAS airborne experiment Y. Shinozuka 1*, J. Redemann 1, P. B. Russell 2, J. M. Livingston 3, A. D. Clarke 4, J. R. Podolske 2 1 Bay Area Environmental Research Institute; 2 NASA Ames Research Center; 3 SRI International; 4 School of Ocean and Earth Science and Technology, University of Hawaii * Anderson, T. L., Charlson, R. J., Winker, D. M., Ogren, J. A., and Holmen, K.: Mesoscale variations of tropospheric aerosols, Journal of the Atmospheric Sciences, 60, , Redemann, J., Zhang, Q., Schmid, B., Russell, P. B., Livingston, J. M., Jonsson, H., and Remer, L. A.: Assessment of MODIS-derived visible and near-IR aerosol optical properties and their spatial variability in the presence of mineral dust, Geophysical Research Letters, 33, L18814, doi: /12006GL026626, Shinozuka, Y., Redemann, J., Livingston, J. M., Russell, P. B., Clarke, A. D., Howell, S. G., Freitag, S., O'Neill, N. T., Reid, E. A., Johnson, R., Ramachandran, S., McNaughton, C. S., Kapustin, V. N., Brekhovskikh, V., Holben, B. N., and McArthur, L. J. B. (2010), Airborne observation of aerosol optical depth during ARCTAS: vertical profiles, inter-comparison, fine-mode fraction and horizontal variability, Atmos. Chem. Phys. Discuss., 10, , doi: /acpd Background and Goal Method - Calculating Standard Deviation and Autocorrelation Summary Forest fires, June 28, 2008 AATS -14 NASA P-3 4 more AERONET fly-over events examined in the paper. Highlights from our previous ARCTAS paper (Shinozuka et al., ACPD, 2010) Å Comparison FMF Comparison Consistency check between the airborne sunphotometer and in situ instruments. Agreement within 3%+0.02 for 2/3 of 55 vertical profiles. Haze, April 6, 2008 Photo by Cameron S. M c Naughton Standard deviation of data, x 1, x 2, x 3, … x N, within a segment can be estimated as. (2) This expression with the sample mean, m, tends to underestimate the standard deviation of population when N is small. Specifically, this estimate is known to be, on average, c times the true value where, (3) for a normal distribution of independent elements and Γ is the gamma function. c is 0.80, 0.89, >0.97 for N = 2, 3, >10, respectively. Accordingly, we divide the estimated standard deviation by this factor:. (4) We divide the standard deviation by the mean to yield relative standard deviation, std rel :. (5) Autocorrelation is the correlation coefficient among all data pairs x j and x j+k that exist at a separation, or lag, of k. That is,, (6) where k indicates the spatial lag (or distance), m +k and std +k denote the mean and standard deviation, respectively, of all data points that are located a distance of +k away from another data point, and m -k and s -k are the corresponding quantities for data points located a distance of -k away from another data point (Redemann et al., 2006; Anderson et al., 2003). In left panel, the black solid curve shows the relative standard deviation of AOD at 499 nm observed during the Canada phase. Cumulative probability for the 20 km (±200 m) segments of horizontal legs below 2 km altitude is plotted on the vertical axis. The fact that this curve goes through a probability of 0.5 (i.e., median) at std rel of 19.4% means that there is a 50% chance that the AOD 499 horizontal variability over a distance of 20 km is ≤19.4% of the average. The numbers in parentheses in the figure legend indicate the number of flight segments that entered the distance category for the (Alaska, Canada) phase. “Cons.” is for consecutive pairs of AATS-14 measurements (4s interval during which the aircraft traveled km). In right panel, the black dots show 499 nm AOD pairs measured 20 km away from each other during horizontal legs below 2 km in the Canada phase. The correlation coefficient, r, of the black dots is This is the autocorrelation coefficient for a lag of 20 km. Relative Standard Deviation Autocorrelation Preliminary Results std rel for extinction coefficients also slowly increases with distance. Its absolute value is inflated by the small average value. The Alaska phase (left panels) sees highly homogeneous horizontal distribution of aerosols after long range transport. Median std rel,AOD499 is as low as 2.1% for 20 km. It slowly increases with distance between 1 km (0.4%) and 30 km (3.0%). Up to 6 km, std med,Angstrom499 is marginally higher for the Alaska phase than for the Canada phase. This may be because the wavelength dependence does not necessarily change with dilution. The horizontal distribution in the Canada phase (right panels) is highly heterogeneous and sensitive to distance. Median std rel,AOD499 is 19.4% for 20 km. It is 10 points lower (9.1%) for 6 km. Little difference between the extinction coefficients due to low ambient RH. Extensive Properties Relative Standard Deviation Autocorrelation Intensive Properties Standard Deviation Autocorrelation All data γ is invariably low in the Canada phase where particles from local fires are hardly responsive to humidity changes. r increases from 0.37 to 0.71 (r 2 from 0.14 to 0.50) as the distance decreases from 20 km to 6 km. To achieve a correlation coefficient of 0.6, for example, a ground measurement site would have to be located within 10 km of the reference point for the retrieval of AOD, and within 3 km for the retrieval of extinction coefficient, in an idealized world represented by our statistical treatment for the Canada phase. r CO is higher than r dryext550 by about 0.1 in the Canada phase. The difference must be explained by phenomena that distinguish aerosols from CO, such as condensation and deposition. r Angstrom499 is not as widely different between the two phases as r AOD499. Knowledge of the horizontal variability of aerosol properties helps Explain part of the difference evident in multi-instrument comparison/validation. (e.g., As one reduces satellite grid size, when does deteriorating signal to noise ratio, surface characterization and cloud screening outweigh the decreased aerosol variability?) Verify whether multiple sensors view similar air masses from an aerosol point of view, as often assumed in inferring aerosol effects. Evaluate whether extrapolation of an observation to larger scales is warranted. (e.g., Can CALIPSO extinction be evaluated adequately with a ground-based lidar 30 km away from its track? What are the uncertainties associated with extrapolating aerosol properties into data sparse regions, e.g. away from the Glory satellite ground track?) Isolate thin clouds from aerosols. These are relevant for both interpretation of existing remote sensing observations and design of future ones. We measured AOD, its Angstrom exponent, in-situ extinction coefficient and carbon monoxide mixing ratio, among other properties, in two contrasting environments. The goal of this study is to quantify the realistic range of the horizontal (over 1-30 km) variability in these properties in aerosol environments common over the globe. Between intensive (the wavelength dependence and the humidity response of light scattering) and extensive (e.g., AOD) properties, the trends of horizontal variability with distance and aerosol environment are distinctly different. This implies unique challenges in estimating particle size and chemical composition from measurements at a distant location. 499 nm AOD likely varies by ~2.1% in an extremely homogeneous environment like the Alaska phase, and ~19.4% in an extremely heterogeneous environment like the Canada phase, both over the course of 20 km. Similar arguments can be made for other properties (Angstrom exponent, extinction, CO) and horizontal scales (1-30 km) based on the data we tabulate. Such arguments are expected to help interpretation and design of remote sensing observations.