Cloud fraction response to aerosol driven by nighttime processes

1 Supporting Information for2 Cloud fraction response to aerosol driven by nighttime processes3 Geoffrey Pugsley, Edward Gryspeerdt and Vishnu Nair4 Geoffrey Pugsley5 E-mail: [email protected]6 This PDF file includes:7 Supporting text8 Figs. S1 to S79 SI References10 Geoffrey Pugsley, Edward Gryspeerdt and Vishnu Nair 1 of 9 Supporting Information Text11 MODIS Ndcorrelations between subsequent days. The possible combinations of the MODIS Nd observations for two days are 12 shown in S1. Strong correlations are seen between subsequent days, however the relationship is weaker between day 1 and 3 13 as expected. This confirms that if a trajectory has a high Nd at a given time it is likely to be high at other times along the 14 trajectory, and hence using the Ndat a single instance in time along the trajectory is justified.15 10 30 100 300 600 Nd day 1 / cm 3 10 30 100 300 600 Nd day 2 / cm 3 Nd day 1 vs Day 2 (r = 0.63) 10 30 100 300 600 Nd day 1 / cm 3 Nd day 3 / cm 3 Nd day 1 vs Day 3 (r = 0.35) 10 30 100 300 600 Nd day 2 / cm 3 Nd day 3 / cm 3 Nd day 2 vs Day 3 (r = 0.58) 100 101 102 103 log(counts) No data 16 Fig. S1. Ndcorrelation between different MODIS overpass days17 Impact of measurement uncertainties. As the aerosol-cloud system includes many complicated, interacting components, it is 18 essential to determine whether observed relationships could have arisen without an aerosol impact on cloud evolution ( 1 ). A 19 simple model of the system is devised here to assess this possibility.20 In order to study the effect of retrieval errors on the results in the main text, theoretical true values of quantities are calculated, along with the corresponding measured value. The true values are denoted with hats, and the measured values are shown as uppercase assuming that the measurement uncertainty associated with X is distributed ∼ N (0 , σ2 X ). In this context X could take the values: CF, LWP or Ndsuch that X=ˆ X+N(0, σ2 X).[1] A simple model is constructed assuming that a linear relationship between the log of the droplet number concentration ( log ( ˆ Nd )) 21 and the initial cloud fraction (ˆ CFi)exists according to Eq. 2.22 ˆ CFi=βln ˆ Nd ˆ Nd,0+ˆ CF0+N(0, σ2 M1)[2] In equation 2, the constants ˆ CF0 and ˆ Nd,0 are prescribed to be 0.75 and 300 cm−3 and β is 0.2; however the results are not 23 sensitive to these choices. An initial ˆ Nd field is initialised from a log-normal distribution. The corresponding ˆ CFi field is then 24 calculated according to Eq. 2. Additional Gaussian noise ( σM1 =0.2) is added to the ˆ CF field to represent variance in ˆ CFi not 25 explained by ˆ Nd (such as variations driven by other meteorological factors). The initial ˆ CF field is then propagated forwards in 26 time to calculate the CF at the end of the timestep (either dusk or dawn; ˆ CFf ), assuming that the ˆ CF varies sinusoidally in 27 time with a period of 1 day (2) according to Eq. 3.28 ˆ CFf=ˆ CFi+acos(ωt)−bt +c+N(0, σ2 M2) + γˆ Nd[3] In equation 3, (a,b,c) are set to be (0.05,0.003,-0.4) respectively. These values were chosen so that the modelled CF evolution 29 closely resembles observations of the diurnal cycle for marine stratocumulus. The second term ( acos ( ωt )) represents the diurnal 30 variation of the CF due to changes in short wave insolation, the third term ( −bt ) produces a gradual decrease in CF over time, 31 representing the breakup of the stratocumulus cloud deck following advection over warmer waters. Additional meteorological 32 noise ( σM2 =0.05) is added to the CF field once it has been advanced forward in time to represent contributions to the 33 temporal evolution not described by Eq. 3. The choice of the form the temporal development chosen for the CF does not 34 significantly impact the results.35 The final term in Eq. 3represents the aerosol effect on ˆ CF development, referred to as the coupling term. The magnitude of γ36 describes the relative strength of the Nd impact on CF evolution. A positive γ could represent a precipitation suppression 37 effect, with ˆ Nd acting to increase ˆ CF . Setting γ to zero removes the Nd (and hence aerosol) impact on cloud development. For 38 this analysis γ is set to be 0.003 cm3 whilst the coupling is turned on, and zero when coupling is false i.e. no aerosol impact on 39 CF development.40 The model representation is summarised in Fig. S2. In this analysis all changes in CF are calculated for the nighttime rather 41 than the daytime, since it is only the effect of the coupling that is being investigated. The daytime changes would be the same 42 as the night, but with an extra phase of πin Eq. 3.43 2of9Geoffrey Pugsley, Edward Gryspeerdt and Vishnu Nair Fig. S2. The model used to study the effect of retrieval biases. Hatted symbols indicate the true value of the quantities assuming that our model (blue) and some random metrological variance (purple) fully describes these. Blue arrows indicate idealised theoretical relationships between ˆ Nd and ˆ CF ; green arrows indicate where measurement noise has been added and the dashed red arrow indicates the variable effect of Nd on CF development, this connection is not present when γ = 0. Variables in the outer triangle represent the measured satellite product. The shapes in the upper right legend reference the equation from the text that each arrow represents. Model results. ∆CF adj is calculated (Fig. S3), for the cases of γ = 0 and γ = 0.003 cm3 , representing a case with no aerosol 44 impact on CF evolution and one where aerosol acts to increase CF over time respectively. It is shown that ∆CF adj depends on 45 the level of instrument noise ( σCF and σNd ) as well as the strength of the coupling ( γ ). For low levels of CF measurement 46 noise ( . 3%), when γ = 0, a very weak ∆CF adj is observed. Whilst when γ is non-zero significantly stronger gradients in 47 ∆CFadj as Ndis varied are observed.48 Recent work ( 3 ) suggests that σNd∼65 cm−3 using the sampling strategy suggested by ( 4 ); which would correspond most 49 closely to the second row in the figures. The GOES cloud top phase product is supplied at 2 km resolution at nadir, with an 50 accuracy of 80 % for pixels satisfying the highest data quality flag ( 5 ). The CF is calculated at 0.25 ° , therefore it would be 51 expected that approximately 150 GOES measurements are used for each CF calculation. Combining this, the uncertainty on52 each CF measurement would be ∼3%.53 These synthetic data results suggest that for realistic levels of measurement noise in CF and Nd , the observed relationships 54 between Ndand ∆CFadj are driven by an aerosol effect on ∆CF rather than retrieval artefacts.55 Geoffrey Pugsley, Edward Gryspeerdt and Vishnu Nair 3 of 9 Nd / cm 3 0 0.5 1 CF f =0 Nd =0 Nd / cm 3 CF f =0.03 Nd =0 Nd / cm 3 CF f =0.05 Nd =0 Nd / cm 3 0 0.5 1 CF f =0 Nd =50 Nd / cm 3 CF f =0.03 Nd =50 Nd / cm 3 CF f =0.05 Nd =50 10 30 100 300 600 Nd / cm 3 0 0.5 1 CF f =0 Nd =100 10 30 100 300 600 Nd / cm 3 CF f =0.03 Nd =100 10 30 100 300 600 Nd / cm 3 CF f =0.05 Nd =100 0.4 0.2 0.0 0.2 0.4 CF adj CF adj ; = 0 (a) Coupling off Nd / cm 3 0 0.5 1 CF f =0 Nd =0 Nd / cm 3 CF f =0.03 Nd =0 Nd / cm 3 CF f =0.05 Nd =0 Nd / cm 3 0 0.5 1 CF f =0 Nd =50 Nd / cm 3 CF f =0.03 Nd =50 Nd / cm 3 CF f =0.05 Nd =50 10 30 100 300 600 Nd / cm 3 0 0.5 1 CF f =0 Nd =100 10 30 100 300 600 Nd / cm 3 CF f =0.03 Nd =100 10 30 100 300 600 Nd / cm 3 CF f =0.05 Nd =100 0.4 0.2 0.0 0.2 0.4 CF adj CF adj ; = 0.003 (b) Coupling on Fig. S3. ∆CFadj for the nighttime with the coupling turned off (top) and on (bottom) 4of9Geoffrey Pugsley, Edward Gryspeerdt and Vishnu Nair Sensitivity of results to MODIS channel used in reretrieval. The sensitivity of the main results to the choice of MODIS channel 56 used in the re is investigated. The standard MODIS re retrieval uses the 2.6µm channel and is used throughout the main text. 57 It is shown that the results are qualitatively unchanged using the 1.6µm or 3.7µm channels (Figs. S4 and S5 respectively).58 Sensitivity of the results to errors in the advection routine. It is shown that the results in the main text are not sensitive to 59 small errors in the wind field. This is illustrated by S6 where the 850 hPa wind field was used to calculate the trajectories of 60 the air parcels (the main text uses 1000 hPa ) with little difference to the pattern of the results. Throughout this study we only 61 consider changes over a single day or night, therefore the Lagrangian advection is only required to be accurate for up to ∼1262 hours.63 The diurnal cycle of precipitation. There is existing evidence of a diurnal variation in stratocumulus precipitation from the 64 VOCALS campaign ( 6 ), which has some similarities to the region we look at in this work. Here, the diurnal variation of 65 precipitation from the microwave retrievals in Eastman et al. (2019) ( 7 ) collocated with our trajectories is investigated. Fig. 66 S7 provides observational evidence that both the probability of precipitation (Fig. S7 a) and rain rates (Fig. S7 b) are more 67 likely (order a factor of two) during the nighttime. This is in line with previous studies ( 8 – 10 ) and supports the hypothesis 68 that precipitation suppression is more likely to occur during the nighttime.69 References70 1. A Arola, et al., Aerosol effects on clouds are concealed by natural cloud heterogeneity and satellite retrieval errors. Nat. 71 Commun.13, 7357 (2022) Publisher: Nature Publishing Group.72 2. R Eastman, SG Warren, Diurnal Cycles of Cumulus, Cumulonimbus, Stratus, Stratocumulus, and Fog from Surface 73 Observations over Land and Ocean. (2014) Section: Journal of Climate.74 3. E Gryspeerdt, et al., The impact of sampling strategy on the cloud droplet number concentration estimated from satellite 75 data. Atmospheric Meas. Tech.15, 3875–3892 (2022) Publisher: Copernicus GmbH.76 4. DP Grosvenor, et al., Remote Sensing of Droplet Number Concentration in Warm Clouds: A Re77 view of the Current State of Knowledge and Perspectives. Rev. Geophys. 56 , 409–453 (2018) _eprint: 78 https://onlinelibrary.wiley.com/doi/pdf/10.1029/2017RG000593.79 5. NOAA Satellite and Information Service (NESDIS), National Aeronautics and Space Administration (NASA), Goes-r 80 series product definition and users’ guide (2019) Accessed: 2025-04-10.81 6. CD Burleyson, SPd Szoeke, SE Yuter, M Wilbanks, WA Brewer, Ship-Based Observations of the Diurnal Cycle of 82 Southeast Pacific Marine Stratocumulus Clouds and Precipitation. (2013) Section: Journal of the Atmospheric Sciences. 83 7. R Eastman, M Lebsock, R Wood, Warm Rain Rates from AMSR-E 89-GHz Brightness Temperatures Trained Using 84 CloudSat Rain-Rate Observations. (2019) Section: Journal of Atmospheric and Oceanic Technology.85 8. CS Bretherton, ME Peters, LE Back, Relationships between Water Vapor Path and Precipitation over the Tropical Oceans. 86 (2004).87 9. E Serpetzoglou, BA Albrecht, P Kollias, CW Fairall, Boundary Layer, Cloud, and Drizzle Variability in the Southeast 88 Pacific Stratocumulus Regime. (2008).89 10. CD Burleyson, SP de Szoeke, SE Yuter, M Wilbanks, WA Brewer, Ship-Based Observations of the Diurnal Cycle of 90 Southeast Pacific Marine Stratocumulus Clouds and Precipitation. J. Atmos. Sci. 70, 3876–3894 (2013).91 Geoffrey Pugsley, Edward Gryspeerdt and Vishnu Nair 5 of 9 0 0.2 0.4 0.6 0.8 1 CF Morning (a) CF adj Day 0 0.2 0.4 0.6 0.8 1 CF Evening (b) CF adj Night 10 30 100 300 600 MODIS Terra N d / cm 3 CF adj Nd 10 30 100 300 600 MODIS Terra N d / cm 3 CF adj Nd 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0.4 CF adj Nd , r e retrieval taken from 1.6 m channel Fig. S4. As with Fig. 2 in the main text, but using the 1.6µmchannel for the MODIS reretrieval used in the Ndcalculation. 6of9Geoffrey Pugsley, Edward Gryspeerdt and Vishnu Nair 0 0.2 0.4 0.6 0.8 1 CF Morning (a) CF adj Day 0 0.2 0.4 0.6 0.8 1 CF Evening (b) CF adj Night 10 30 100 300 600 MODIS Terra N d / cm 3 CF adj Nd 10 30 100 300 600 MODIS Terra N d / cm 3 CF adj Nd 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0.4 CF adj Nd , r e retrieval taken from 3.7 m channel Fig. S5. As with Fig. 2 in the main text, but using the 3.7µmchannel for the MODIS reretrieval used in the Ndcalculation. Geoffrey Pugsley, Edward Gryspeerdt and Vishnu Nair 7 of 9 0 0.2 0.4 0.6 0.8 1 CF Morning (a) CF adj Day 0 0.2 0.4 0.6 0.8 1 CF Evening (b) CF adj Night 10 30 100 300 600 MODIS Terra N d / cm 3 CF adj 10 30 100 300 600 MODIS Terra N d / cm 3 CF adj 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0.4 CF adj 850hPa winds used for advection Fig. S6. As with Fig. 2 in the main text, but using the 850hPa wind field for the advection of the air parcels. 8of9Geoffrey Pugsley, Edward Gryspeerdt and Vishnu Nair 0 24 h 24 48 h 48 72 h 0.00 0.02 0.04 0.06 0.08 Probability of precipitation (a) Day Night 0 24 h 24 48 h 48 72 h 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 Mean rain rate / mm hr ¹ (b) Time since start of trajectory / hours Fig. S7. AMSR warm rain rates (left) and probability of precipitation (right), co-located with the trajectories and separated by day and nighttime. Geoffrey Pugsley, Edward Gryspeerdt and Vishnu Nair 9 of 9