The effect of solar zenith angle on MODIS cloud optical and microphysical retrievals within marine liquid water clouds

Abstract

Abstract. In this paper we use a novel observational approach to investigate MODIS satellite retrieval biases of τ and r e (using three different MODIS bands: 1.6, 2.1 and 3.7 µm, denoted as r e1.6 , r e2.1 and r e3.7 , respectively) that occur at high solar zenith angles (θ 0 ) and how they affect retrievals of cloud droplet concentration (N d ). Utilizing the large number of overpasses for polar regions and the diurnal variation of θ 0 we estimate biases in the above quantities for an open ocean region that is dominated by low level stratiform clouds.We find that the mean τ is fairly constant between θ 0 = 50 • and ∼65-70 • , but then increases rapidly with an increase of over 70 % between the lowest and highest θ 0 . The r e2.1 and r e3.7 decrease with θ 0 , with effects also starting at around θ 0 =65-70 • . At low θ 0 , the r e values from the three different MODIS bands agree to within around 0.2 µm, whereas at high θ 0 the spread is closer to 1 µm. The percentage changes of r e with θ 0 are considerably lower than those for τ , being around 5 % and 7 % for r e2.1 and r e3.7 . For r e1.6 there was very little change with θ 0 . Evidence is provided that these changes are unlikely to be due to any physical diurnal cycle.The increase in τ and decrease in r e both contribute to an overall increase in N d of 40-70 % between low and high θ 0 . Whilst the overall r e changes are quite small, they are not insignificant for the calculation of N d ; we find that the contributions to N d biases from the τ and r e biases were roughly comparable for r e3.7 , although for the other r e bands the τ changes were considerably more important. Also, when considering only the clouds with the more heterogeneous tops, the importance of the r e biases was considerably enhanced for both r e2.1 and r e3.7 .When using the variability of 1 km resolution τ data (γ τ ) as a heterogeneity parameter we obtained the expected result of increasing differences in τ between high and low θ 0 as heterogeneity increased, which was not the case when using the variability of 5 km resolution cloud top temperature (σ CTT ), suggesting that γ τ is a better predictor of τ biases at high θ 0 than σ CTT . For a given θ 0 , large decreases in r e were observed as the cloud top heterogeneity changed from low to high values, although it is possible that physical changes to the clouds associated with cloud heterogeneity variation may account for some of this. However, for a given cloud top heterogeneity we find that the value of θ 0 affects the sign and magnitude of the relative differences between r e1.6 , r e2.1 and r e3.7 , which has implications for attempts to retrieve vertical cloud information using the different MODIS bands. The relatively larger decrease in r e3.7 and the lack of change of r e1.6 with both θ 0 and cloud top heterogeneity suggest that r e3.7 is more prone to retrieval biases due to high θ 0 than the other bands. We discuss some possible reasons for this.Our results have important implications for individual MODIS swaths at high θ 0...

Predicting decadal trends in cloud droplet number concentration using reanalysis and satellite data

“…In the present study, level-2 swath data (joint product) from MODIS collection 5.1 (King et al, 2003) is filtered to remove problematic retrievals at a pixel-level following Grosvenor and Wood (2014), including the removal of pixels with a solar zenith angle greater than 65 o . The daily-mean CDNC at 1°x1° resolution is calculated using filtered level 2 swath data and only low (cloud tops below 3.2 km), liquid clouds were used to calculate CDNC.…”

Global and regional estimates of warm cloud droplet number concentration based on 13 years of AQUA-MODIS observations

“…Typically, most of these issues affect the effective radius at 1.6 and 2.1 µm more strongly than the effective radius at 3.7 µm (Z. . However, at high solar zenith angles Grosvenor and Wood (2014) demonstrated that resolved (as opposed to sub-pixel) threedimensional radiative effects are likely to cause the effective radius to be biased high, with larger biases expected for the 3.7 µm retrieval compared to the 1.6 µm one. Retrieved R e,2.1…”

“…A variety of observational case studies and process studies were also published using similar approaches for deriving CDNC (Boers et al, 2006;George and Wood, 2010;Painemal and Zuidema, 2010;Rausch et al, 2010;. Various authors have addressed shortcomings and issues related to CDNC climatologies (Merk et al, 2016;Grosvenor and Wood, 2014) as well as issues related to the cloud retrievals underlying the CDNC climatologies (Zhang and Platnick, 2011;Nakajima et al, 2010;Maddux et al, 2010;Horvath et al, 2014). Painemal and Zuidema (2011) validate MODIS-derived CDNC against in situ observations taken in the South Pacific during VOCALS-Rex (Wood et al, 2011).…”

The aerosol-cyclone indirect effect in observations and high-resolution simulations

“…One approach would be to use the observed cloud droplet number concentration (CDNC). This provides a good proxy for CCN (Wood, 2012), but it is potentially problematic because retrieval errors relating to overlying ice cloud (Sourdeval et al, 2016), cloud heterogeneity (Grosvenor and Wood, 2014;Sourdeval et al, 2016), and low sun-angle (Grosvenor and Wood, 2014) may spuriously bias the measurements making 30 it difficult to interpret any observed covariation between cyclone properties and CDNC. To avoid these ambiguities we take a similar approach to previous studies and use sulfate mass concentration at the surface from reanalysis as a proxy for CCN (McCoy et al, 2017;.…”

“…Cyclone compositing was performed following the algorithm in Field and Wood (2007). Supporting observations of 20 CDNC from Moderate Resolution Imaging Spectroradiometer (MODIS) (King et al, 2003) were also composited on MERRA2 SLP and were filtered as dictated by previous studies to reduce biases associated with inhomogeneous clouds and low sun angle and cloud tops were restricted to be below 3.2km (Grosvenor and Wood, 2014). CDNC from MERRA2 was calculated following the regression model relating log sulfate mass concentration to log CDNC presented in previous studies (McCoy et al, 2017).…”