Sep 28, 2010 - cal depth and cloud fraction is the 10âmeter wind speed. Constructing the partial correlation between aerosol optical depth and cloud fraction, ...
GEOPHYSICAL RESEARCH LETTERS, VOL. 37, L18814, doi:10.1029/2010GL044361, 2010
Impact of meteorological factors on the correlation between aerosol optical depth and cloud fraction Anders Engström1 and Annica M. L. Ekman1 Received 16 June 2010; revised 10 August 2010; accepted 17 August 2010; published 28 September 2010.
[1] The aerosol optical depth has in several recent studies been found to correlate with cloud fraction. This study examines the global distribution of the total correlation between aerosol optical depth, cloud fraction and meteorological conditions using satellite observations together with atmospheric re‐analysis data from the ECMWF. The results show large regional differences in the correlation between aerosol optical depth and cloud fraction, where a higher correlation is found over remote ocean. The one meteorological variable that correlates significantly with both aerosol optical depth and cloud fraction is the 10‐meter wind speed. Constructing the partial correlation between aerosol optical depth and cloud fraction, with the impact from 10‐meter wind speed removed, yields a significant difference compared to the total correlation. In several regions the remaining partial correlation is reduced from 0.4 to below 0.1. The results highlight the need to investigate all possible correlations between meteorological variables, cloud properties and aerosols. Citation: Engström, A., and A. M. L. Ekman (2010), Impact of meteorological factors on the correlation between aerosol optical depth and cloud fraction, Geophys. Res. Lett., 37, L18814, doi:10.1029/2010GL044361.
1. Introduction [2] Aerosols, both natural and anthropogenic, can impact on cloud properties since they act as cloud condensation nuclei. If the amount of liquid cloud water remains constant, an increase of the number of aerosol particles active as cloud condensation nuclei leads to smaller and more numerous cloud droplets and a higher cloud albedo [Twomey, 1977]. Smaller and more numerous cloud droplets should also, by inhibiting collision and coalescence processes, impact on the ability of clouds to form rain and could consequently increase the cloud fraction [Albrecht, 1989]. However, the effects are highly non‐linear and different for different cloud types, and not all impacts from aerosols on clouds are necessarily microphysical in nature. For instance, direct radiative effects of aerosols may also affect cloud cover and aerosols have also been found to decrease cloud lifetime as a consequence of decreased precipitation [Ackerman et al., 2004]. [3] A number of studies display a correlation between aerosol optical depth and cloud properties [Kaufman et al., 2005; Kaufman and Koren, 2006; Myhre et al., 2007]. The results suggest that variations in cloud fraction can at least to some extent be explained by variations in aerosol 1 Department of Meteorology, Stockholm University, Stockholm, Sweden.
Copyright 2010 by the American Geophysical Union. 0094‐8276/10/2010GL044361
optical depth, which acts as a proxy for the aerosol concentration [Andreae, 2009]. However, the correlation cannot unequivocally be linked to a microphysical impact of aerosols on clouds. It is likely that the observed correlation between aerosols and cloud properties is a combination of many effects, including the microphysical connection. Meteorological conditions, which co‐vary with both aerosols and clouds simultaneously, and measurement biases have been suggested as explanations, or partial explanations, to the observed correlation [Zhang et al., 2005; Charlson et al., 2007; Mauger and Norris, 2007; Stevens and Brenguier, 2009; Twohy et al., 2009; Quaas et al., 2009]. Thus, there exists a need to quantify a correlation between aerosols and clouds that is independent of variations in the meteorological state and measurement biases. The large amount of data available from remote sensing of aerosols and clouds does provide an opportunity to establish a statistical relationship between aerosols and cloud properties. The co‐variation of aerosols and clouds with meteorological variables can be studied using observational data, data from numerical weather prediction models or re‐analysis data. [4] The present study examines the correlation between satellite derived aerosol optical depth and cloud fraction, using re‐analysis data to account for a number of variations in the meteorological state. The aim is to 1) study the correlation on a global scale and 2) to obtain a first‐order estimate of the correlation between aerosol optical depth and cloud fraction that is independent of several influencing meteorological parameters. The analysis is performed globally for all ocean regions between 45S and 45N for a six‐year period between 2003 and 2008. A comparison of the correlation and partial correlation between all variables identifies causal pathways that can be further studied using, e.g., models.
2. Method and Data [5] Aerosol optical depth and cloud fraction observations from the MODerate resolution Imaging Spectroradiometer (MODIS) onboard the Aqua satellite are used [King et al., 2003]. The MODIS L3 products deliver daily information on the average aerosol optical depth and cloud fraction within a 1 × 1 degree grid cell. The MODIS aerosol optical depth retrieval over ocean has been found to meet the expected accuracy, determined by ground based measurements, in 60% of the retrievals [Remer et al., 2008]. The present analysis is focused on warm phase low‐level clouds. Only grid cells where no ice particles are detected and where the mean cloud top pressure higher is than 640 hPa are included. Varying the cloud top pressure limit within 100 hPa does not influence on the presented results. The analysis is further limited to aerosol optical depth values less than 0.5.
L18814
1 of 4
L18814
ENGSTRÖM AND EKMAN: IMPACTS ON AEROSOL‐CLOUD RELATIONSHIPS
Figure 1. Total correlation between aerosol optical depth and cloud fraction. Non‐significant correlations (less than 95% significance by t‐test) and land areas are shown in white. Note that the chosen limit effectively screens out data in the Saharan outflow over the tropical Atlantic, an area heavily influenced by dust aerosols. [6] We have assumed that independent retrieval of aerosol optical depth and cloud fraction is possible, and that they are not correlated due to that clouds are interpreted as aerosols, or vice‐versa. The MODIS cloud screening algorithm, together with the constraint on aerosol optical depth, should ensure that this is a fairly good assumption. Additionally, since collocated observations of aerosols and clouds can not be made using MODIS, we have assumed that the retrieved aerosol optical depth is representative for the entire 1 × 1 degree grid cell. However, if an insufficient amount of cloud free pixels is available in a grid cell, the MODIS aerosol optical depth algorithm does not return a value and the point is excluded from the analysis. [7] The meteorological data are obtained from the ERA‐ Interim re‐analysis at the European Center for Medium Range Weather Forecasts (ECMWF) valid at the time closest to that of the Aqua satellite overpass. We have selected three different meteorological variables that can be assumed to be related to variations in both aerosol optical depth and cloud fraction. These are the 10‐meter horizontal wind speed, the relative humidity and the vertical wind speed. Vertical wind speed and relative humidity are extracted at three different pressure levels: 925, 850 and 700 hPa. The 10‐meter wind speed is chosen on the basis that wind speed can impact emissions of sea spray aerosols [Mulcahy et al., 2008]. Wind and clouds are also simultaneously symptomatic of large‐scale meteorological forcing and changes in wind speed could impact on the boundary layer structure and increase the vertical moisture flux to the cloud layer. One caveat should however be noted for the MODIS retrieval of aerosol optical depth. The retrieval depends on the surface reflective properties which over ocean are determined by the surface wind speed. The surface wind speed is in the retrieval assumed constant at 6 ms−1. Consequently, the aerosol optical depth is underestimated or overestimated if the wind is weaker respectively stronger than 6 ms−1 [Zhang and Reid, 2006] (cf. also discussion in Section 3). Relative humidity is chosen as a variable because in high humidity regions aerosols may take up water vapor which increases the aerosol optical depth. A high relative humidity is also important for cloud formation and thus aerosol optical depth and cloud fraction could be correlated as a result of variations in relative humidity. Lastly we have included vertical wind speed. Vertical wind speed is symptomatic of horizontal divergence/convergence, and has been suggested to affect the presence of both aerosols and clouds [Stevens and Brenguier, 2009]. The analysis data for
L18814
all variables are available every six hours (00Z, 06Z, 12Z and 18Z). [8] We have assumed that errors in the meteorological analysis are not related to variations in aerosol optical depth and cloud fraction. Although aerosols could influence wind patterns locally in the case of thick pollution and dust plumes, we have no reason to believe that this should systematically impact on the presented results. The atmospheric analysis data will however contain errors. In the case of the 10‐meter wind speed these can be assumed relatively small since the surface winds are generally well constrained by scatterometer data. Relative humidity on the other hand does depend largely on boundary layer parameterizations and is generally not well constrained by observations. Additionally, vertical wind speed is a relatively small‐scale variable and it is possible that the 1 × 1 degree average is too coarse to show a relationship between vertical wind, clouds and aerosols. Regardless of this, the atmospheric analysis is our best estimate of the current state of these variables. [9] In addition to examining the total correlation between all variable pairs, we estimate the partial correlation between aerosol optical depth and cloud fraction. The partial correlation is a measure of the linear dependence between two variables where the influence from possible controlling variables is removed. In the case of three variables, a, b and c, the partial correlation is defined as [Hazewinkel, 2002]: rab rac rbc rabc ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; 2 2 1 rac 1 rbc
ð1Þ
where rab, rac and rbc is the total correlation between each variable pair and rab·c is the remaining independent correlation between a and b, assuming no other controlling variables. All correlations and partial correlations in the text are presented together with the lower and upper 95% confidence interval, obtained by two‐tailed t‐test, within parentheses.
3. Results [10] Figure 1 shows the total correlation coefficients between aerosol optical depth and cloud fraction. The global mean correlation coefficient is 0.26 (0.19, 0.33), which suggests, assuming no other controlling variables, that 7% (r2) of the observed variation in cloud fraction can be attributed to variations in aerosol optical depth. The correlation between aerosol optical depth and cloud fraction is found to be lower in near‐coastal regions and higher over remote ocean. The highest correlation is found in the subtropical band between 10–15S and 10–15N, corresponding to areas dominated by shallow convection and marine stratiform clouds. The correlation within the tropical band is mostly non‐significant (less than 95% significance by t‐test). Over the tropical Atlantic this is due to the fact that the aerosol optical depth is often higher than the chosen limit of 0.5. Over the western and eastern Pacific and over Indonesia it is because of the predominance of high clouds which are not included in the analysis. [11] The fact that the correlation between aerosol optical depth and cloud fraction shows such large regional differences is intriguing. The highest correlations are found in regions where the anthropogenic contribution to the variability in aerosol optical depth is relatively low compared to the contribution from natural variability. It is therefore
2 of 4
L18814
ENGSTRÖM AND EKMAN: IMPACTS ON AEROSOL‐CLOUD RELATIONSHIPS
L18814
Figure 2. Total correlation between (a) aerosol optical depth (AOD) and 10‐meter wind speed, (b) cloud fraction (CF) and 10‐meter wind speed, (c) AOD and relative humidity (RH) at 925 hPa, (d) CF and relative humidity at 925 hPa, (e) AOD and relative humidity at 850 hPa, and (f) CF and relative humidity at 850 hPa. Non‐significant correlations (less than 95% significance by t‐test) and land areas are shown in white. worthwhile to examine if variations in the meteorological state covary with both aerosol optical depth and cloud fraction. [12] To examine this we construct similar maps of correlation coefficients between all variable pairs. Figure 2 shows the correlation of aerosol optical depth and cloud fraction with the 10‐meter wind speed and relative humidity at 925 and 850 hPa. The 700 hPa level is not shown but resembles the two lower levels. The results show that the 10‐meter wind speed is correlated with both aerosol optical depth and cloud fraction, where the global mean correlation is 0.41 (0.37, 0.45) and 0.22 (0.17, 0.27), respectively. Relative humidity is generally well correlated with cloud fraction and mainly highlights the presence of clouds at the corresponding pressure level. However, relative humidity is generally uncorrelated with aerosol optical depth. The correlation of both aerosol optical depth and cloud fraction with vertical wind speed is very low or statistically insignificant and is therefore not shown. Since vertical wind speed should to some extent be correlated with cloud fraction this is probably indicative of that the data is too coarse to show any relationship. [13] The correlation between aerosol optical depth and wind speed suggests that aerosols from sea spray emissions is the dominant contributor to the aerosol optical depth in clean remote regions. This corroborates the findings of Mulcahy et al. [2008]. However, it could also be indicative of the wind dependent retrieval of aerosol optical depth from MODIS [Zhang and Reid, 2006]. The fact that both wind and clouds are symptomatic of synoptic forcing, for instance a change in the strength of the sub‐tropical high‐pressure systems, could also explain the correlation between wind speed and cloud fraction. It could also be due to different physical connections, for example a wind‐driven impact on the boundary layer structure or through vertical moist transport to the cloud layer. [14] Two main explanatory pathways can be considered plausible for the correlation between the 10‐meter wind speed, aerosol optical depth and cloud fraction. If aerosols from sea spray emissions are the dominant contributor to the
aerosol optical depth then these aerosols could also increase the cloud fraction. This scenario would not weaken the relationship between aerosol optical depth and cloud fraction, but rather show that one cannot discount natural variability as an important part of the observed correlation since wind acts as an antecedent variable. The second scenario is that aerosol optical depth and cloud fraction are both correlated with wind and not independently with each other. This would for instance be the case if the wind dependent retrieval is dominating the variability in aerosol optical depth, or if aerosol from sea spray simply occur at the same time as a higher cloud fraction due to the correlation with wind. [15] Regardless of which causal pathway is true, wind will in all described scenarios explain a portion of the correlation between aerosol optical depth and cloud fraction. It is therefore of interest to estimate the part of the correlation between aerosol optical depth and cloud fraction which can not be explained by the variance in wind. To this end we construct the partial correlation between aerosol optical depth and cloud fraction. Figure 3a shows the partial correlation and Figure 3b shows the difference between the total and partial correlation. The global mean partial correlation is 0.19 (0.13, 0.25) and the mean difference compared to the total correlation is −27%. The largest difference between the total and partial correlation is found west of South America in the stratiform cloud region, where the partial correlation in some places decreases from 0.4 to lower than 0.1. Large differences are also identified over large parts of the Indian Ocean. [16] Even though causality can not be determined with this method the results show that a large part of the correlation between aerosol optical depth and cloud fraction could be explained by variations in the surface wind speed. It is also interesting to note that the largest difference is found in regions where the total correlation is the highest. It should be pointed out that a correlation with other meteorological variables might persist and consequently that the remaining partial correlation between aerosol optical depth
3 of 4
L18814
ENGSTRÖM AND EKMAN: IMPACTS ON AEROSOL‐CLOUD RELATIONSHIPS
L18814
[18] Acknowledgments. This study was performed within the Bert Bolin Centre for Climate Research. The work of the first author was funded by the Swedish National Space Board.
References
Figure 3. (a) Partial correlation between AOD and CF. Non‐significant correlations (less than 95% significance by t‐test) and land areas are shown in white. (b) Difference between the total correlation (Figure 1) and the partial correlation (Figure 3a). and cloud fraction could be even lower than what is found here.
4. Conclusions [17] The present study examines the correlation between aerosol optical depth and cloud fraction for all ocean regions between 45S and 45N for a six‐year period between 2003 and 2008 using observations from the MODIS instrument onboard the Aqua satellite. We identify large regional differences in the correlation between aerosol optical depth and cloud fraction. By examining the covariation of aerosol optical depth and cloud fraction with different meteorological variables we find that the one variable that correlates significantly with both aerosol optical depth and cloud fraction is the 10‐meter wind speed. While causality can not be determined with the present analysis, we identify and estimate the impact of a number of causal pathways that could explain the correlation between wind, aerosol optical depth and cloud fraction. Controlling for variations in wind reduces the global mean correlation between aerosol optical depth and cloud fraction by 27%. The impact is especially prominent in regions dominated by stratiform clouds in the eastern part of the Pacific, where the correlation between aerosol optical depth and cloud fraction is reduced from 0.4 to less than 0.1. The decrease significantly reduces the validity of a linear independent relationship between the aerosol optical depth and cloud fraction. The results estimate an upper bound for the independent correlation between aerosol optical depth and cloud fraction over ocean and presents a simple technique for studying the impact from meteorological variables. The presented results further highlight the need to examine all possible correlations between meteorological variables, cloud properties and aerosols.
Ackerman, A. S., M. P. Kirkpatrick, D. E. Stevens, and O. B. Toon (2004), The impact of humidity above stratiform clouds on indirect aerosol climate forcing, Nature, 432, 1014–1017, doi:10.1038/nature03174. Albrecht, B. (1989), Aerosols, cloud microphysics, and fractional cloudiness, Science, 245, 1227–1230, doi:10.1126/science.245.4923.1227. Andreae, M. O. (2009), Correlation between cloud condensation nuclei concentration and aerosol optical thickness in remote and polluted regions, Atmos. Chem. Phys., 9, 543–556, doi:10.5194/acp-9-543-2009. Charlson, R. J., A. S. Ackerman, F. A.‐M. Bender, T. L. Anderson, and Z. Liu (2007), On the climate forcing consequences of the albedo continuum between cloudy and clear air, Tellus, Ser. B, 59, 715–727, doi:10.1111/ j.1600-0889.2007.00297.x. Hazewinkel, M. (Ed.) (2002), Encyclopaedia of Mathematics, Springer, New York. Kaufman, Y. J., and I. Koren (2006), Smoke and pollution aerosol effect on cloud cover, Science, 313, 655–658, doi:10.1126/science.1126232. Kaufman, Y. J., I. Koren, L. A. Remer, D. Rosenfeld, and Y. Rudich (2005), The effect of smoke, dust, and pollution aerosol on shallow cloud development over the Atlantic Ocean, Proc. Natl. Acad. Sci. U. S. A., 102, 11,207–11,212, doi:10.1073/pnas.0505191102. King, M. D., W. P. Menzel, Y. J. Kaufman, D. Tanré, B.‐C. Gao, S. Platnick, S. A. Ackerman, L. A. Remer, R. Pincus, and P. A. Hubanks (2003), Cloud and aerosol properties, precipitable water, and profiles of temperature and water vapor from MODIS, IEEE Trans. Geosci. Remote Sens., 41, 442–458, doi:10.1109/TGRS.2002.808226. Mauger, G. S., and J. R. Norris (2007), Meteorological bias in satellite estimates of aerosol‐cloud relationships, Geophys. Res. Lett., 34, L16824, doi:10.1029/2007GL029952. Mulcahy, J. P., C. D. O’Dowd, S. G. Jennings, and D. Ceburnis (2008), Significant enhancement of aerosol optical depth in marine air under high wind conditions, Geophys. Res. Lett., 35, L16810, doi:10.1029/ 2008GL034303. Myhre, G., F. Stordal, M. Johnsrud, Y. J. Kaufman, D. Rosenfeld, T. Storelvmo, J. E. Kristjansson, T. K. Berntsen, A. Myhre, and I. S. A. Isaksen (2007), Aerosol‐cloud interaction inferred from MODIS satellite data and global aerosol models, Atmos. Chem. Phys., 7, 3081– 3101, doi:10.5194/acp-7-3081-2007. Quaas, J., et al. (2009), Aerosol indirect effects—General circulation model intercomparison and evaluation with satellite data, Atmos. Chem. Phys., 9, 8697–8717, doi:10.5194/acp-9-8697-2009. Remer, L. A., et al. (2008), Global aerosol climatology from the MODIS satellite sensors, J. Geophys. Res., 113, D14S07, doi:10.1029/ 2007JD009661. Stevens, B., and J.‐L. Brenguier (2009), Cloud controlling factors: Low clouds, in Clouds in the Perturbed Climate System, edited by J. Heintzenberg and R. J. Charlson, pp. 173–196, MIT Press, Cambridge, Mass. Twohy, C. H., J. A. Coakley Jr., and W. R. Tahnk (2009), Effect of changes in relative humidity on aerosol scattering near clouds, J. Geophys. Res., 114, D05205, doi:10.1029/2008JD010991. Twomey, S. (1977), The influence of pollution on the shortwave albedo of clouds, J. Atmos. Sci., 34, 1149–1152, doi:10.1175/1520-0469(1977) 0342.0.CO;2. Zhang, J., and J. S. Reid (2006), MODIS aerosol product analysis for data assimilation: Assessment of over‐ocean level 2 aerosol optical thickness retrievals, J. Geophys. Res., 111, D22207, doi:10.1029/2005JD006898. Zhang, J., J. S. Reid, and B. N. Holben (2005), An analysis of potential cloud artifacts in MODIS over ocean aerosol optical thickness products, Geophys. Res. Lett., 32, L15803, doi:10.1029/2005GL023254. A. M. L. Ekman and A. Engström, Department of Meteorology, Stockholm University, SE‐106 91 Stockholm, Sweden. (anderse@misu. su.se)
4 of 4
