Stratospheric aerosols on Jupiter from Cassini observations

Icarus 226 (2013) 159–171

Contents lists available at SciVerse ScienceDirect

Icarus journal homepage: www.elsevier.com/locate/icarus

Stratospheric aerosols on Jupiter from Cassini observations X. Zhang a,b,⇑, R.A. West c, D. Banfield d, Y.L. Yung a a

Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125, USA Department of Planetary Sciences and Lunar and Planetary Laboratory, University of Arizona, AZ 85721, USA c Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109, USA d Department of Astronomy, Cornell University, Ithaca, NY 14853, USA b

a r t i c l e

i n f o

Article history: Received 27 February 2013 Revised 10 April 2013 Accepted 21 May 2013 Available online 5 June 2013 Keywords: Jupiter, Atmosphere Photometry Radiative transfer Atmospheres, Structure

a b s t r a c t We retrieved global distributions and optical properties of stratospheric aerosols on Jupiter from groundbased NIR spectra and multiple-phase-angle images from Cassini Imaging Science Subsystem (ISS). A high-latitude haze layer is located at 10–20 mbar, higher than in the middle and low latitudes (50 mbar). Compact sub-micron particles are mainly located in the low latitudes between 40°S and 25°N with the particle radius between 0.2 and 0.5 lm. The rest of the stratosphere is covered by the particles known as fractal aggregates. In the nominal case with the imaginary part of the UV refractive index 0.02, the fractal aggregates are composed of about a thousand 10-nm-size monomers. The column density of the aerosols at pressure less than 100 mbar ranges from 107 cm2 at low latitudes to 109 cm2 at high latitudes. The mass loading of aerosols in the stratosphere is 106 g cm2 at low latitudes to 104 g cm2 in the high latitudes. Multiple solutions due to the uncertainty of the imaginary part of the refractive index are discussed. The stratospheric haze optical depths increase from 0.03 at low latitudes to about a few at high latitudes in the UV wavelength (0.26 lm), and from 0.03 at low latitudes to 0.1 at high latitudes in the NIR wavelength (0.9 lm). Ó 2013 Elsevier Inc. All rights reserved.

1. Introduction Aerosols, or hazes in the stratosphere of Jupiter are of particular interest. First, particular absorbers and scatterers affect the radiative heat budget and the solar energy redistribution in the jovian stratosphere (West et al., 1992; Moreno and Sedano, 1997; Zhang et al., 2013). Second, aerosols are involved in the stratospheric chemical cycle. They are one of the end products of the photochemistry or ion-chemistry (Wong et al., 2003). The haze particles shield the UV light and alter the efficiency of photochemistry in deeper layers. Heterogeneous reactions may occur on the particle surfaces. Third, aerosols can serve as ideal tracers for the stratospheric transport (Friedson et al., 1999) and provide valuable information on the stratospheric dynamics. To evaluate the significance of haze, it is important to determine their latitudinal and vertical distribution and optical properties, such as the optical depth, single scattering albedo, and phase function, for the entire wavelength range from ultraviolet (UV) to the near-infrared (NIR) region. Taking advantage of the continuum spectra in NIR wavelengths, two attempts have been made to retrieve the global map of haze and clouds on Jupiter. Banfield et al. (1998) retrieved the latitudinal and vertical distributions of stratospheric and tropospheric ⇑ Corresponding author at: Department of Planetary Sciences and Lunar and Planetary Laboratory, University of Arizona, AZ 85721, USA. E-mail address: [email protected] (X. Zhang). 0019-1035/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.icarus.2013.05.020

hazes covering the entire southern hemisphere and northern equatorial region below 25°N. They discovered that a low-latitude haze layer is located at 50 mbar and its altitude level increases sharply to 20 mbar in the high latitudes (polar hood). The tropospheric haze top is around 0.2 bar and is non-uniform with latitude. Haze density reaches a minimum in the tropopause region, which is unexpected from the previous models (e.g., Kaye and Strobel, 1983). Recently, Kedziora-Chudczer and Bailey (2011) used a line-by-line multiple scattering radiative transfer model to simulate the NIR spectra with a much higher resolution. Their data cover the entire disk of Jupiter. They assume a 1.3 lm particle layer in the troposphere and a 0.3 lm particle layer in the stratosphere. Their results are generally consistent with Banfield et al. (1998), except for an additional distinct haze layer is discovered around 5 mbar in the polar hoods. Many studies focused on the aerosol properties in the UV and visible range, from various data sources such as the intensity measurements from spacecraft (e.g., Pioneer 10 in Tomasko et al., 1978 and Voyager in Hord et al., 1979), space-based telescopes (e.g., Tomasko et al., 1986), and ground-based telescopes (e.g., West, 1979), and polarization measurements (e.g., Smith, 1986). Please see the review in West et al. (2004) for details. Generally speaking, the low-latitude aerosols are composed of small particles with radii between 0.2 and 0.5 lm (Tomasko et al., 1986), while the determination of the high-latitude aerosols is more complicated. Some

160

X. Zhang et al. / Icarus 226 (2013) 159–171

studies (e.g., Moreno, 1996; Barrado-Izagirre et al., 2008) assumed small particles (<0.1 lm) to explain the low phase angle images in the polar region, although small particles in fact are not consistent with high phase angle data (Rages et al., 1999). Alternatively, West and Smith (1991) proposed that the high-latitude particles are fractal aggregates in order to reconcile both the positive polarization (Smith, 1986) and the modest forward scattering (e.g., Tomasko et al., 1978). Details such as the monomer radius, the number of monomers, fractal dimension, and refractive index have not been established. One of the purposes in this study is to test this hypothesis for Jupiter and quantify the aerosol properties. In this study we combine the information from the groundbased NIR spectra and multiple-phase-angle images in UV to NIR wavelengths from the Imaging Science Subsystem (ISS) onboard Cassini during its Jupiter flyby in the late 2000 and early 2001. The ISS acquired 26,000 high-quality time-lapse images of Jupiter during its 6-months-long flyby from 1 October 2000 to 22 March 2001 (Porco et al., 2003). A proper combination of the images from different filters can be used for a specific purpose. For example, the methane channels and corresponding continuum filters (e.g., MT1/ CB1, MT2/CB2, MT3/CB3) provide vertical structure information of the atmospheric aerosols and clouds. The UV1 filter samples the upper troposphere and stratospheric haze layer. Furthermore, Cassini ISS provides images from low to high phase angles. Until now only the low phase angle images have been investigated in the polar region (Barrado-Izagirre et al., 2008). In fact valuable information of the phase functions of the stratospheric particles can be obtained from the high phase angle images, as shown by previous studies, e.g., Tomasko et al. (1978) for the Pioneer data and Rages et al. (1999) for the Galileo data. Therefore, we analyzed the low, middle and high phase angle images together to characterize the size, shape and phase functions of particles on Jupiter. This method also helps distinguish the compact sub-micron (CSM) particles1 and fractal aggregates. This paper is structured as follows. In Section 2 we revisited the data from Banfield et al. (1998) and updated the methane absorption coefficients in the original retrieval model and relaxed the previous assumptions, from which the updated stratospheric aerosol map is obtained. Using a new retrieval model is described in Section 3, the aerosol and cloud properties are retrieved based on the ISS observations, followed by discussions and conclusions in Section 4.

trieval result is called the f value, which is defined as (Banfield et al., 1996):

f ðzÞ ¼

1 pðhÞrxXðzÞ 4mg

ð1Þ

where m is the mean molecular weight of the atmosphere, g is the gravitational acceleration of Jupiter, p(h) is the particle phase function at phase angle h, r is the particle cross section, x is the particle single scattering albedo, and X(z) is the volume mixing ratio of particles at altitude z. Therefore, from the NIR spectra we are only able to retrieve the relative abundance of the stratospheric aerosols instead of the absolute values, and even that assumes that the phase function and single scattering albedo do not change with location. Under those assumptions, the vertical shape of f(z) is similar to that of the aerosol volume mixing ratio profile and therefore it can be incorporated into the ISS data retrieval in Section 3. In the entire spectral region, any pixel with reflectivity (I/F) greater than 0.075 was removed to make sure the single scattering approximation is robust. Note that Banfield et al. (1996) assumed the retrieved f value is constant with wavelength. Banfield et al. (1998) relaxed the assumption by incorporating the spectral shape of the aerosol extinction efficiency, but still assumed a constant particle size of 0.3 lm with latitude and altitude. See Banfield et al. (1996, 1998) for details of the observations, calibrations and the inverse model. In this study, we improved the retrieval technique by (1) updating the CH4 absorption coefficient, and (2) allowing the aerosol size to be varied with latitude. We use the correlated-k method to calculate the atmospheric transmission, which is accurate enough for the NIR low-resolution spectra and broadband filters for Cassini ISS images. We adopt the state-of-art CH4 absorption coefficients from Karkoschka and Tomasko (2010), which are constructed from both visible and NIR CH4 bands, based on the laboratory data and observed spectra from the Huygens probe on Titan and Hubble Space Telescope observations. The correlated-k coefficients are obtained from the calculation by P.G. Irwin (http://www.atm.ox.ac.uk/user/irwin/kdata.html). The upper panel of Fig. 1 shows the total optical depth of CH4 and H2–H2 and H2–He collisional induced absorption (CIA, based on Borysow and Frommhold, 1989; Bory-

10.0

τ

1.0

0.1

2. Retrieval from NIR spectra

1.6

1 In our study we cannot distinguish spherical and compact non-spherical particles in that size regime and so we use the term ‘‘CSM particles’’. The optical properties of the CSM particles small compared to the wavelength can be calculated using Mie theory even if they are not spheres. This is clearly not true for the fractal aggregates and large crystalline ice cloud particles in the troposphere.

1.8

2.0

2.2

2.4

Wavelength (μm)

1.0000

UV1

MT3 CB3

0.1000

τ

The NIR spectra from Banfield et al. (1998) were taken on August 14 1995, from the 200-in. Hale telescope at Palomar Observatory. The spectra were obtained in broadband H (1.45–1.8 lm) and K (1.95–2.5 lm) telluric windows, with the spectral resolution power 100, covering from 25°N to the south pole (80°S) of Jupiter. Since the stratospheric aerosol optical depth is small in the H and K bands, Banfield et al. (1996) developed a direct retrieval technique based on the single-scattering approximation for the NIR spectra, under which the radiative transfer inversion problem is linear. Therefore a simple and effective retrieval technique can be applied to minimize the difference between the simulated spectra and the observations in the least-square sense, with a Tikhonov-type regularization term (a two-point Gaussian correlation matrix) in the cost function to smooth the inverted profiles. The re-

0.0100 0.0010 0.0001 0.2

0.4

0.6

0.8

1.0

Wavelength (μm) Fig. 1. Total gas optical depth including CH4 and H2–H2 and H2–He CIA at 100 mbar. Upper panel shows the difference between the results based on the old correlated-k coefficients (red dashed line) used in Banfield et al. (1998) and the new data (black solid line) from Karkoschka and Tomasko (2010) for the H and K bands in the NIR region. Lower panel shows the comparison between the CH4 optical depth (black) and Rayleigh scattering optical depth (blue) from 0.2 to 1.0 lm. The three vertical dashed lines correspond to the ISS filters used in this study, CB3 (0.938 lm), MT3 (0.889 lm), and UV1 (0.258 lm), respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

161

X. Zhang et al. / Icarus 226 (2013) 159–171

South 70°

0.3μm 0.7μm

0.010

0.25

1

0.2

Pressure (mbar)

sow et al., 1989; Borysow, 1992, 2002) from the top of the atmosphere to 100 mbar. The old CH4 coefficients in Banfield et al. (1998) generally overestimate the CH4 opacity and the band shapes are also slightly different. Banfield et al. (1998) found the spectra are actually not very sensitive to the particle size. This is generally true except for the polar region. Fig. 2 compares the two optimized solutions based on the prescribed 0.3 and 0.7 lm aerosols for 70°S and the equator. The equatorial spectrum is relatively insensitive to the particle size. However the 0.3 lm particle fails to fit the polar region spectrum below 2.1 lm, while the 0.7 lm particle is able to reproduce the observations. Therefore, we improve the retrieval technique by varying the aerosol size with latitude. Through a grid search method, an optimal solution can be obtained for each latitude. The observations require larger particles (>0.6 lm) in the polar stratosphere (poleward of 60°S), but in the other regions the spectra are not able to distinguish the larger and smaller particles. One of the solutions in Rages et al. (1999) from the analysis of the Galileo measurements in the north polar region (60°N) is 1.3 lm particle radius at the 1 mbar level, which seems consistent with our solution for the south polar region, although one can ask whether these large particles are sustainable against sedimentation in the 1 mbar region or higher (Rages et al., 1999). An alternative solution by Kedziora-Chudczer and Bailey (2011) is a two-mode haze model, in which they assume the lower layer (tropospheric haze or cloud) is composed of the 1.3 lm particles and the upper layers (stratospheric haze) are composed of the 0.3 lm particles. Therefore, the 0.7 lm particle size in the polar region may be an average of the mixture of the 1.3 lm tropospheric haze and 0.3 lm strato-

10

0.15

0.1

100

0.05

0 -75

-50

-25

0

25

Latitude (Degree) Fig. 3. Retrieved aerosol map (f value) in the stratosphere and upper troposphere of Jupiter, based on the ground based NIR measurements.

spheric particle size, although our broadband data could not tell the difference. The retrieved aerosol map (f value) is shown in Fig. 3. This map generally agrees with the result from Banfield et al. (1998), except the aerosol layers are shifted slightly downward. This is because the new CH4 absorption coefficients are slightly smaller than the old ones. The pressure level of the haze layer decreases from about 50 mbar at the equator to above 20 mbar at the south pole. The haze layers are found concentrated within one or two scale heights. Note that this map shows a clear region around the tropopause (100 mbar), consistent with Banfield et al. (1998). It is contrary to the hydrazine photochemical model results by Kaye and Strobel (1983). It might be attributed to a deep source in the troposphere without strong upward transport or the fast fallout of heavy particles around the tropopause region, but a satisfactory physical explanation is still lacking (Banfield et al., 1998). The clear region is also found by Kedziora-Chudczer and Bailey (2011). The locations of the haze layers in their selected band and zones are generally consistent with our results. They found a very high stratospheric haze layer above 10 mbar that is beyond our sensitivity region. However, Kedziora-Chudczer and Bailey (2011) does not provide detailed latitudinal and vertical aerosol profiles and therefore we will use the updated aerosol map (Fig. 3) from Banfield’s data to model the haze (Section 3). 3. Retrieval from Cassini images 3.1. ISS data description

0.1000

0.3μm 0.7μm 0.0100

0.0010

0.0001

1.6

Equator 1.8

2.0

2.2

2.4

Wavelength (μm) Fig. 2. Comparison of the best solutions with prescribed 0.3 and 0.7 lm particles, for 70°S (upper panel) and the equatorial region (lower panel), respectively. The observed spectra from Banfield et al. (1998) are shown in black with error bars.

We used three Cassini ISS filters in this study: CB3 (0.938 lm), MT3 (0.889 lm) and UV1 (0.258 lm). The lower panel of Fig. 1 shows the CH4 and Rayleigh scattering optical depth at 100 mbar from UV to NIR region. The Rayleigh scattering optical depth is based on West et al. (2004): s=p ¼ 0:0083ð1 þ 0:014k2 þ0:00027k4 Þk4 , for wavelength k in lm and pressure p in bar. Due to the oblateness of Jupiter, this expression can be corrected by a factor of 24.40/g to any latitude (West et al., 2004). The CB3 filter is the continuum channel sampling the methane window. The MT3 filter is located in a strong methane absorption band. This channel is designed to sample the upper atmosphere. The MT3/CB3 filters are sensitive to the location of the tropospheric haze layer and the aerosol properties at 10 mbar in the high phase angle images. The UV1 filter is not affected by CH4 but by strong Rayleigh scattering. Fig. 1 shows that the Rayleigh scattering optical depth at the UV1 channel is roughly the same as the methane optical depth in MT3 channel. Therefore the UV channel is also more sensitive to the higher hazes.

162

X. Zhang et al. / Icarus 226 (2013) 159–171

We selected 38 Cassini ISS images, covering all the latitudes of Jupiter from 70°S to 70°N, and phase angles from 0.9° to 141°. All images are calibrated according to West et al. (2010). The detailed information of the image index numbers from the Planetary Data System (PDS) and mean phase angles are shown in Table 1. Fig. 4 shows some selected ISS images from the three filters. A significant latitudinal contrast is seen in the MT3 images, revealing brighter bands near the equator and the polar region, and that the northern hemisphere is brighter than the southern hemisphere. The brighter regions imply some combination of higher cloud top and thicker stratospheric haze in the equatorial region and northern high latitudes. The contrast of small-scale features at low and high latitudes in the UV1 images is diminished by Rayleigh scattering above the cloud and haze layers. A darker band in the equatorial region also implies some combination of a higher cloud top and darker particles. The strong evidence for the higher stratospheric haze layer in the polar region comes from the bright polar caps in the MT3 images and corresponding darker polar region in the UV1 images because only the higher stratospheric haze layer can overcome the CH4 absorption and Rayleigh scattering. That the polar haze layers reside in the higher stratosphere is consistent with the results from the NIR retrieval results in Section 2. The spectral information from the three Cassini filters is not enough for retrieving a high-resolution vertical profile of the haze layer. Therefore, we prescribed the shape of the vertical profiles based on the NIR retrieval results in our multiple scattering radiative transfer calculation. For the region northward of 25°N where the NIR retrieval results are not available, we assume that the shape of the vertical aerosol profile is the same as its conjugated latitude in the southern hemisphere. This assumption is justified according to the haze layer locations in zones and bands revealed by Kedziora-Chudczer and Bailey (2011). We divided the whole globe evenly into 29 latitude bins from 70°S to 70°N, with a width of 5° in each bin. For each latitude, we randomly sampled I/F values for 10 pixels spread over longitude to represent the limb darkening profile. Since we focus on the zonally averaged properties of the particles, we removed the significant anomalies such as the Great Red Spot and large ovals. We tested different samples to validate our results.

3.2. Retrieval model description We developed a retrieval model for the ISS data. The model is composed of two parts: the forward module and the optimization module. The forward module consists of a radiative transfer module and an aerosol optical property module. The optimization mod-

ule is based on a nonlinear least square optimization to minimize the difference between model and observations. The details are presented in the following. The radiative transfer module simulates the reflectivity (I/F) for a specific incident angle, viewing angle and phase angle using DISORT (DIScrete Ordinates Radiative Transfer Program for a MultiLayered Plane-Parallel Medium). DISORT (Stamnes et al., 1988) employs the discrete ordinates method and has recently been translated into a C-language version by (Buras et al., 2011). Compared with the original single-precision Fortran code, the new version of the DISORT, called C-DISORT, is using double precision and has removed possible spurious numerical spikes, and its speed is 3–4 times faster than the original version. We use 32 streams to characterize the intensity angular distribution, the results from which display almost no difference from the 64-stream cases. As illustrated by Fig. 5, our forward module includes several atmospheric layers, including the haze layer, the cloud layer and the gas layers, from 1 mbar to the tropospheric cloud top. Typically above the cloud top our model has 12 vertical layers, which are enough to approximate the vertical profile of the stratospheric haze layer from the NIR retrieval. We do not use a CSM particle to approximate the optical properties of the tropospheric cloud layer which is probably consist of large crystalline ice cloud particles such as ammonia ice (West et al., 2004). Instead, since we mainly focus on the stratospheric hazes, for simplicity, we parameterized the CH4 absorption, Rayleigh scattering, aerosols and clouds in the troposphere all together as a semi-infinite ‘‘effective cloud layer’’ (or a bottom scattering layer), which can be characterized by its single scattering albedo (SSA) and a double Henyey–Greenstein (DHG) phase function (Tomasko et al., 1978):

PðhÞ ¼

f1 1  g 21 1  f1 þ 2 4p ð1 þ g  2g cos hÞ3=2 4p 1 1 

1  g 22 ð1 þ

g 22

 2g 2 cos hÞ

ð2Þ

3=2

where P is the phase function and h is the scattering angle. The three parameters in the DHG phase function are: the partition factor f1, the forward asymmetry factor g1, and the backward asymmetry factor g2. We considered two types of clouds: one for the NIR (CB3/ MT3) filters and the other for the UV1 filter. Due to strong tropospheric CH4 absorption in the MT3 filter but little in the CB3 filter, we also retrieved the photon mean free path (PMFP) in the cloud in the MT3 channel, defined as the path length of a photon that travels in the cloud before a scattering event. Besides the optical properties, the effective cloud top is also a free parameter in the model.

Table 1 Selected Cassin ISS images in this work. CB3 filter

MT3 filter

UV1 filter

Image number

Mean phase angle

Image number

Mean phase angle

Image number

Mean phase angle

N1352917174 N1355181340 N1355181726 N1355182081 N1355182442 N1356751773 N1356754443 N1358257928 N1358258182 N1360176531 N1360177530 N1363092160

17.548 3.503 3.458 3.699 3.731 52.915 53.065 119.584 119.580 136.444 136.445 140.988

N1352917145 N1355181377 N1355181763 N1355182101 N1355182462 N1355366470 N1355716697 N1355717439 N1359305173 N1359306172 N1363092096

17.549 3.503 3.457 3.699 3.730 0.936 6.471 6.470 131.917 131.921 140.987

N1352917104 N1355181442 N1355182158 N1355182519 N1355720416 N1355720779 N1355723337 N1357558433 N1358243072 N1358243326 N1358855316 N1358856323 N1358860176 N1358861183 N1363187297

17.548 3.507 3.703 3.731 6.442 6.592 6.568 99.774 119.323 119.318 128.024 128.021 128.066 128.062 141.022

X. Zhang et al. / Icarus 226 (2013) 159–171

163

CB3

MT3

UV1

Fig. 4. Sample images from three Cassini ISS filters. From top to bottom: CB3 (0.938 lm), MT3 (0.889 lm), UV1 (0.258 lm). For each filter, we show a low phase angle (17.5°) image on the left and a high phase angle (141°) image on the right.

For the haze optical property simulator, we choose two models: the ‘‘CSM model’’, in which we assume the aerosols are CSM particles; and the ‘‘AGG model’’, in which we assume the aerosols are fractal particles aggregated from a number of tiny monomers. The phase function and cross sections of CSM particles are calculated on the basis of Mie theory, and that of the fractal aggregates can be accurately calculated from the electromagnetic scattering computation using the multi-sphere methods, for example, by the MSTM code from Mackowski and Mishchenko (2011). However, it is computationally too expensive to meet our retrieval needs even with the help of parallel computing. Instead, we adopt

a useful parameterization method for the aggregates with a fractal dimension of 2 (Tomasko et al., 2008).2 This empirical code has been validated with the rigorous multi-sphere calculations in a large parameter space, spanning the size parameter of the monomer from 104 to 1.5 and number of monomers from 2 to 1024. For the monomer size parameter smaller than 0.5, the model is robust for even larger number of monomers (4028). Based on this empirical code, 2 A typo has been found in their equation (A. 12b). The correct expression should be (personal communication with M. Lemmon): depol_11 = C_p11_ m_3⁄M0⁄taus_outE_p11_t_1.

164

X. Zhang et al. / Icarus 226 (2013) 159–171

Fig. 5. Illustration of the structure of the retrieval model.

Tomasko et al. (2008) were able to fit the descent imager/spectral radiometer (DISR) instrument aboard the Huygens probe for Titan and found that the fractal aggregates in the Titan lower atmosphere are composed of thousands of 0.05-lm radius monomers. Therefore, we adopt the code in Tomasko et al. (2008) as our AGG model. The optimization module is a nonlinear least square optimization package, MPFIT, in Interactive Data Language (IDL) language (Markwardt, 2008). The algorithms were translated from the original Fortran code, MINPACK-1, a minimization routine (Moré, 1978). The algorithm is based on the Levenberg–Marquardt iteration scheme (Levenberg, 1944; Marquardt, 1963), and the errors are calculated from the posterior covariance matrix. This algorithm does not require any a priori knowledge, which is typically suitable for our purpose due to the poor knowledge about jovian aerosols. The IDL package supports the upper and lower constraints for the retrieved parameters that are bounded within their physically allowed region. For instance, the single scattering albedo should be less than or equal to unity. See Moré (1978) and Markwardt (2008) for details of the code and its numerical scheme. The initial guess is sometimes crucial, so we tried different initial guesses and chose the best fitting results. From our synthetic data tests, the minimization package approaches the true values quickly and shows a robust behavior. The detailed computational time depends on the choice of the optical property module, the initial guess, and the number of data points, but a typical retrieval case in this study converges within 20 iterations in several hours, with the help of parallel computing.

haze and cloud, we set up two cases for each retrieved parameter. In one case all the latitudes share the same value of that parameter, and in the other case the parameter is allowed to vary with latitude. If the performances of the two retrieval cases are equally good, we conclude that the retrieved parameter behaves uniformly along the latitudes based on the observations. We organized the retrieved results of our nominal case (the best-fitting case in this study) into two categories: latitudinally invariant and latitudinally varying parameters in Tables 2 and 3, respectively. The latitudinally varying parameters include the cloud top pressure, haze column density above 100 mbar, PMFP in the cloud and single scattering albedos. For the latitudinally invariant parameters, all the latitudes are found to share the same cloud phase functions, and only two types of stratospheric hazes

Table 2 Latitudinally-invariant Parameters in the best-fit model. Mean radius reff (lm)

Low-latitude particle

veff High-latitude particle

Monomer radius (nm) Number of monomers

10 (south) 8 (north) 1000

Imaginary part of the refractive index

NIR (0.9 lm) UV (0.25 lm)

1  103 2  102

Cloud NIR phase function

f1 g1 g2

0.9675 0.6650 0.5954

Cloud UV phase function

f1 g1 g2

0.8303 0.8311 0.3657

3.3. Retrieval results In principle, based on the shape of the vertical profiles from the NIR retrieval results in Section 2, retrieval is performed individually for each latitude bin to obtain the best-fitting parameters from the Cassini data: cloud top pressure, haze column density above 100 mbar, particle shape parameters, particle refractive indices, PMFP in the cloud, cloud phase functions and single scattering albedos. However, multiple solutions exist because there are not sufficient information and constraints to define a unique solution. Furthermore, in order to retrieve the common properties of the

0.3 0.1

Note: There are two types of particles whose properties do not vary with latitude within their own regions: low latitudes (40°S–25°N) and high latitudes (70°S–45°S and 30°N–70°N), respectively. See Section 3.3 for details. We use a two-parameter gamma distribution for the low-latitude particles, characterized by reff and v eff : 1

rv eff NðrÞ ¼

3

  exp  r rv eff eff 1

ðr eff v eff Þv eff

2

;

Cðv1eff  2Þ

where r is the radius and C is the gamma function.

165

X. Zhang et al. / Icarus 226 (2013) 159–171 Table 3 Retrieved latitudinally varying parameters in the nominal case. Latitude

Column density above 100 mbar (106 cm2)

Cloud top (mbar)

CB3 Albedo

MT3 PMFP (mbar)

UV1 Albedo

70 65 60 55 50 45 40 35 30 25 20 15 10 5 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70

9400 4600 2300 1200 790 240 7.9 8.1 7.2 12 19 15 10 16 20 13 13 5.0 16 12 440 720 1300 2200 3200 5400 6000 7500 8000

113 113 239 190 201 194 219 237 213 232 226 235 243 197 144 163 233 189 229 225 131 133 137 175 194 272 291 387 484

0.9990 0.9941 0.9824 0.9810 0.9855 0.9862 0.9848 0.9902 0.9928 0.9937 0.9920 0.9866 0.9849 0.9914 0.9933 0.9897 0.9854 0.9878 0.9933 0.9890 0.9924 0.9889 0.9893 0.9880 0.9848 0.9855 0.9845 0.9841 0.9917

248 141 134 237 230 240 200 153 137 115 109 149 125 112 79 90 121 192 120 143 165 178 210 242 240 240 239 184 140

0.9800 0.9800 0.9800 0.9702 0.9725 0.9370 0.8679 0.8963 0.9168 0.9214 0.9279 0.9396 0.9324 0.9301 0.9124 0.9166 0.9357 0.9470 0.9297 0.9067 0.9472 0.9393 0.9661 0.9800 0.9800 0.9800 0.9800 0.9800 0.9800

Note: The ‘‘albedo’’ in the table refers to the tropospheric cloud single scattering albedo. The uncertainties estimated from the covariance matrix based on Bayesian statistics are on the order of 5–10% of the retrieved values for the column density, cloud top pressure and MT3 PMFP, or 5–10% of ‘‘1  a’’ for the cloud single scattering albedos (a). Note that the uncertainties calculated from the covariance matrix are sensitive to the curvature at a local minimum in the residuals and therefore are very likely to be underestimated because of the existence of multiple solutions.

are needed to explain the ISS data: CSM particles in the low latitudes (40°S–25°N), and fractal aggregated particles in the middle and high latitudes (70°S–45°S and 30°N–70°N). The low latitude boundaries are determined empirically on the basis of best-fit results; a CSM model is appropriate inside but an AGG model is re-

quired outside. In fact the low latitude boundaries can be seen from the MT3 and UV1 images in Fig. 4. Since the ISS data in our study are not very sensitive to the real part of the refractive index, we fix its real part based on Khare et al. (1984). For the imaginary part that determines the haze albedo, multiple solutions corre-

0.7

Reflectivity (I/F)

Reflectivity (I/F)

0.6 0.6

0.4

0.2

0.5 0.4 0.3 0.2 0.1

-1.0

-0.5

0.0

0.5

1.0

0.65

0.70

0.75

sin (Longitude)

0.80

0.85

0.90

0.95

sin (Longitude) 0.7

0.8

Reflectivity (I/F)

Reflectivity (I/F)

0.6 0.6

0.4

0.2

-1.0

0.5 0.4 0.3 0.2 0.1

-0.8

-0.6

-0.4

-0.2

0.0

sin (Longitude)

0.2

0.4

0.75

0.80

0.85

0.90

0.95

sin (Longitude)

Fig. 6. Atmospheric reflectivity (I/F) as a function of longitude in the equatorial region for multiple phase angles indicated in the upper left of each panel. Filled circles are the observations and lines are the model results. Black, orange and blue colors correspond to CB3, MT3, and UV1 filters, respectively. Solid lines: nominal CSM model results (Tables 2 and 3); dashed lines: best-fitting AGG model results; dotted lines: best-fitting CSM model results with the haze optical depth fixed as five times of that in the nominal case. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

166

X. Zhang et al. / Icarus 226 (2013) 159–171

sponding to different choices of the imaginary index are investigated. In our nominal case we chose its values to be 0.001 and 0.02 in NIR and UV filters, respectively.

3.3.1. Low latitudes: CSM model results As in Tomasko et al. (1986), we are not able to determine the stratospheric particle size accurately in the low latitude region because they are optically thin and most of the photons are scattered by clouds in the troposphere. The approximate particle radius range found in this study is between 0.2 and 0.5 lm, consistent with Tomasko et al. (1986). Therefore we fixed the CSM particle properties in the CSM model, in which we use the two-parameter gamma function for the aerosol size distribution. A different type of size distribution does not significantly influence the retrieval results. A typical CSM-model fit with a 0.3-lm radius particle is shown in Fig. 6 (solid lines) for different filters and various low and high phase angles at the equator. For the other intermediate phase angles (not shown in the plot) the fitting is also good. The dashed lines in Fig. 6 shows that the AGG model fails to reproduce the limb-darkening profile. Multiple solutions exist for the stratospheric particle size and refractive index. For example, Table 4 shows two different solutions at the equator. Solution A is our nominal case in Tables 2 and 3, and solution B corresponds to the UV imaginary refractive index 0.18 (Khare et al., 1984) and a particle size of 0.5 lm. Although the aerosol properties are different, both solutions can fit the limb-darkening profiles. But the total aerosol optical depths above 100 mbar are more or less the same, about 0.026 (NIR) and 0.025 (UV) for solution A and 0.034 (NIR) and 0.022 (UV) in

Table 4 Two best-fitted CSM model results for the equator. Solution A (nominal)

Solution B

Mean particle radius, reff (lm) Size distribution parameter, veff NIR imaginary refractive index UV imaginary refractive index Total column above 100 mbar (cm2) Cloud NIR phase function f1 g1 g2 Cloud UV phase function f1 g1 g2 Tropospheric SSA at CB3 PMFP at MT3 (mbar) Tropospheric SSA at UV1

0.3 0.1 1  103 2  102 2.0  107

0.5 0.1 1.6  103 1.86  101 2.4  107

0.9675 0.6650 0.5954

0.9207 0.8507 0.1000

0.8303 0.8311 0.3657 0.9933 79 0.9124

0.7230 0.8574 0.2767 0.9974 49 0.9109

0.030

τ 100 mbar

0.029 0.028 0.027 0.026 0.025 10−4

3.3.2. Middle and high latitudes: AGG model results Outside the low latitude region (40°S–25°N), the CSM model fails (Fig. 8, dashed lines). Although the CSM model with particle radius of 0.1 lm might be able to fit the low phase angle data, as also shown by previous studies (Moreno, 1996; Barrado-Izagirre et al., 2008), it does not fit the high phase angle data simultaneously. On the other hand, fractal aggregates are able to reproduce the limb-darkening profiles in all filters and phase angles (solid lines in Fig. 8). The optical properties of the fractal aggregates depend on the monomer size and number of monomers. Smaller monomer leads to larger UV haze optical depth, and larger number of monomers increases the forward scattering. Provided that the imaginary part of UV refractive index is 0.02, all the observations in the middle and high latitudes can be explained with the same kind of fractal particle that is composed of a thousand 10-nm monomers. The monomer size in the southern hemisphere appears slightly larger than that in the northern hemisphere. Fig. 9 shows that the models with 50-nm monomer (dashed) or 200 monomers with any size (dotted) cannot simulate the mid-latitude observations very well, especially when the phase angles are high. In the nominal case, the total haze optical depths above 100 mbar are about 2–3 in UV and 0.1 in the NIR wavelengths in the high latitudes, with the mass loading on the order of 104 g cm2. Compared with the low latitudes, the high-latitude haze is optically thicker by one to two orders of magnitude. The dotted lines in Fig. 8 shows a sensitivity test case in which we performed the optimization with the optical depth fixed as five times smaller than the best solution, the high phase angle data in the MT3 filter are significantly underestimated. The solutions depend on the choice of imaginary part of the refractive index that affects the haze single scattering albedo. Multiple solutions for 60°S are shown in Fig. 10. When fixing the NIR imaginary index as 0.001, we found decent fits exist if the UV

Mass Loading (10−6 g cm−2)

Retrieved parameters

solution B. The total column density above 100 mbar is 2  107 cm2 for solution A and 2.4  107 cm2 for solution B. Multiple solutions with a CSM particle of 0.3 lm are shown in Fig. 7. We explored the NIR imaginary refractive index from 0.0001 to 0.01 for the NIR filter with the UV imaginary refractive index fixed as 0.02, and the UV imaginary refractive index from 0.002 to 0.01 with the NIR imaginary refractive index fixed as 0.001. Changing the refractive index would only change the single scattering albedo within a factor of 2, so the stratospheric haze optical depth is around 0.02–0.03 and the mass loading is around 2 lg cm2, consistent in all solutions. On the other hand, a CSM model with haze optical depth 0.1 will significantly overestimates the high phase angle reflectivity, as shown by the dotted lines in Fig. 6.

10−3

10−2

Im

10−1

2.5 2.4 2.3 2.2 2.1 2.0 10−4

10−3

10−2

10−1

Im

Fig. 7. Multiple solutions for the equatorial region as functions of the imaginary part of the refractive indices in the NIR (orange) and UV (blue) filters. Each dot represents a solution. Left: total aerosol optical depth at 100 mbar in the NIR (open circle) and UV filters (filled circle); right: total aerosol mass loading (assume the density is 1 g cm3) above 100 mbar. It is important to keep in mind that the UV refractive index was held constant when generating the plotted points for the NIR, and the NIR refractive index was held constant when generating the results for the UV. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

167

0.30

0.5

0.25

0.4

Reflectivity (I/F)

Reflectivity (I/F)

X. Zhang et al. / Icarus 226 (2013) 159–171

0.20 0.15 0.10 0.05 0.00 -1.0

-0.5

0.0

0.5

0.3 0.2 0.1 0.0 0.70

1.0

0.75

0.80

0.85

0.90

0.95

1.00

sin (Longitude)

0.30

0.6

0.25

0.5

Reflectivity (I/F)

Reflectivity (I/F)

sin (Longitude)

0.20 0.15 0.10

0.4 0.3 0.2 0.1

0.05 0.00 -1.0

-0.5

0.0

0.0 0.86 0.88 0.90 0.92 0.94 0.96 0.98 1.00

0.5

sin (Longitude)

sin (Longitude)

Fig. 8. Atmospheric reflectivity (I/F) as a function of longitude at 60°S for multiple phase angles indicated in the upper left of each panel. Filled circles are the observations and lines are the model results. Black, orange and blue colors correspond to CB3, MT3, and UV1 filters, respectively. Solid lines: nominal AGG model results (Tables 2 and 3); dashed lines: best-fitting CSM model results with particle radius fixed at 0.1 lm; dotted lines: best-fitting AGG model results with the haze optical depth fixed at a fifth of that in the nominal case; dash-dotted lines: best-fitting AGG model results with the imaginary part of the refractive index at UV1 filter fixed at 0.2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

0.6

0.4

Reflectivity (I/F)

Reflectivity (I/F)

0.5

0.3 0.2 0.1 0.0 -1.0

-0.5

0.0

0.5

0.4

0.2

0.0 0.6

1.0

0.7

sin (Longitude)

0.4

Reflectivity (I/F)

Reflectivity (I/F)

0.9

1.0

0.6

0.5

0.3 0.2 0.1 0.0 -1.0

0.8

sin (Longitude)

-0.5

0.0

0.5

sin (Longitude)

0.4

0.2

0.0 0.75

0.80

0.85

0.90

0.95

1.00

sin (Longitude)

Fig. 9. Atmospheric reflectivity (I/F) as function of longitude in the middle latitudes (45°N) for multiple phase angles indicated in the upper left of each panel. Filled circles are the observations and lines are the results from the AGG model. Black, orange and blue colors correspond to CB3, MT3, and UV1 filters, respectively. Solid lines: nominal AGG model results (Tables 2 and 3); dashed lines: best-fitting model results with monomer radius fixed at 50 nm; dotted lines: best-fitting model results with monomer number fixed as 200. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

imaginary index is between 0.006 and 0.08 (blue dots in Fig. 10). If we fixed the UV imaginary index as 0.02, our model requires the NIR imaginary index less than 0.005 (orange dots in Fig. 10). In fact both of them can vary and so they do not provide as tight a constraint on the mass loading and microphysical parameters as this figure might imply. Fig. 8 shows that an AGG model with UV imaginary index equal to 0.2 would not be able to explain the limb-darkening profiles. The haze optical depths in UV and NIR wavelengths are roughly consistent in all the solutions. The re-

trieved haze parameters do not vary significantly with the NIR imaginary index because the cloud NIR albedo change is enough to compensate most of the haze absorption change. The monomer radius increases with the UV imaginary index in order to keep the UV to NIR optical depth ratio roughly constant. The number of monomers deceases with the monomer radius increase so that the total particle size and phase function do not change significantly. The total column density and mass loading decease as the UV imaginary index increases due to the scattering efficiency in-

X. Zhang et al. / Icarus 226 (2013) 159–171 40

τ 100 mbar

Monomer Radius (nm)

168

30 20

1.0

10 0.1 0 -4 10

-3

-2

10

10

-1

−4

10

−3

10

10

Im Mass Loading (g cm )

4

10

−2

Number of Monomers

−2

10−1

−2

10

10

Im

3

10

−3

10

−4

10

2

10

-4

10

-3

-2

10

10

-1

−4

10

−3

10

10

Im

10

−1

Im

Fig. 10. Multiple solutions for the south-pole region (60°S) as functions of the imaginary part of the refractive index in the NIR (orange) and UV (blue) filters. Each dot represents a solution. Upper left: monomer radius; upper right: total aerosol optical depth at 100 mbar in the NIR (open circle) and UV filters (filled circle); lower left: number of monomers; lower right: total aerosol mass loading (assume the density is 1 g cm3) above 100 mbar. It is important to keep in mind that the UV refractive index was held constant when generating the plotted points for the NIR, and the NIR refractive index was held constant when generating the results for the UV. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

a

200 300 400 500 −50

Cloud SSA

1.00

0

50 1.00

b

0.99

0.95

0.98

0.90

0.97

τ100 mbar

10.00

0.85 −50

0

50

−50

0

50

-50

0 Latitude (Degree)

50

c

1.00 0.10

Mass Loading (g cm )

0.01

-2

3.3.3. Summary of the ISS retrieval results Fig. 11 shows the latitudinal distributions of the effective cloud top, cloud single scattering albedo, aerosol optical depths, and total mass loading in the nominal case. The effective cloud top is around 200 mbar, roughly consistent with the result from NIR spectrum inversion (Fig. 3). The cloud top does not change dramatically from latitude to latitude, but the equatorial zone and the northern midlatitudes shows a higher effective cloud top. Although the cloud is barely seen in the higher latitudes from the high phase angle images, the low phase angle images seem to imply the northern polar region tends to have lower effective cloud top than the southern polar region, but there is a large uncertainty associated with it. This conclusion is qualitatively consistent with the tropospheric haze tops retrieved from the low phase angle images from other ISS filters and Hubble Telescope observations (Barrado-Izagirre et al., 2008). The cloud tops retrieved by Kedziora-Chudczer and Bailey (2011) also appear lower at high northern latitudes than at high southern latitudes. The belts and zones at low latitudes can be seen from the cloud SSA in all three filters because the clouds contribute the most to the reflectivity. The zones tend to be brighter and the belts tend to be darker in the NIR filter and the opposite behavior exhibits in the UV1 filter. This might be expected because cloud scattering in the UV1 channel is mixed with conservative Rayleigh scattering. The PMFP in the cloud is around 100–200 mbar in the NIR (Table 3). It seems the cloud/tropospheric haze at low latitudes is optically thicker than the high-latitude cloud based on the retrieval results that (1) shorter photon scattering path at low latitudes and (2) higher cloud albedo in the UV wavelengths at high latitudes. This

Cloud Top (mbar)

100

crease with the monomer radius. Therefore, within the uncertainty range of the UV imaginary index, the monomer radius could vary from 5 nm to 40 nm, and accordingly, the number of monomers changes from 10,000 to 100. The total aerosol mass loading in the 60°S ranges from 2  105 g cm2 to 103 g cm2. But the stratospheric haze optical depths only differ less than 30% among all the solutions. The single scattering albedos of the polar haze are 0.8 and 0.8–0.95 in the UV and NIR wavelengths, respectively, consistent in all the solutions.

10

-3

10

-4

d

10-5 10-6

Fig. 11. Summary of important retrieved parameters as function of latitude with a 5° bin width. From top to bottom: (a) effective cloud top in the troposphere. (b) Cloud SSA in the CB3 (black) and UV1 (blue) filters. The left vertical axis corresponds to the CB3 channel and the right axis corresponds to the UV1 channel. The dashed line in (b) indicates the fixed values in the retrieved model because they are less well constrained. (c) Total aerosol optical depth at 100 mbar in the NIR (orange solid) and UV filters (blue solid), with the CH4 optical depth for the MT3 filter (orange dashed) coincidentally overlapping with Rayleigh scattering optical depth at UV1 filter (blue dashed) at 100 mbar. (d) Total aerosol mass loading (assume the density is 1 g cm3) above 100 mbar as function of latitude. Note that there are large uncertainties associated with cloud top pressures in the high latitudes. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

169

X. Zhang et al. / Icarus 226 (2013) 159–171

104

1

Pressure (mbar)

103

Aerosol Phase Function

observations, which were not sampled by Rages et al. (1999). In this study, the aerosol distribution shows a clear region at 100 mbar so the particle number density near the 100 mbar is low (Fig. 3). At 20 mbar, in fact the number density in our equatorial model is also roughly 0.1 cm3. But the peak is actually located around 50 mbar. For the polar region, the required UV optical depth is 10, provided the extinction cross section of the aggregates is of the order of 109 cm2 at the UV wavelength and the total column density is required to be 1010 cm2. Since we have a very concentrated particle haze layer from the NIR retrieval and the layer thickness is usually within one or two atmospheric scale heights (25–50 km), the number density is 103 cm3. The particle density profiles provide useful constraints to chemical and microphysical models. The aerosol and cloud phase functions are plotted in Fig. 13. The aerosol phase functions in low latitudes and high latitudes look similar in the forward peak in both the NIR and UV filters, respectively. The low-latitude haze appears to have stronger back scattering than the high-latitude particle, but note that the phase function of the low-latitude particle depends on the particle size. The tropospheric haze/cloud phase function over the UV–visible wavelengths seems not change too much, as the derived phase functions are roughly consistent with that from Pioneer 10 observations in the RED filter (0.64 lm) at the north component of the

Low-lat IR High-lat IR Low-lat UV High-lat UV

10.0

1.0

0.1 0

50

100

150

Scattering Angle 100.0

NIR UV RED (Pioneer 10)

Cloud Phase Function

is consistent with the NIR retrieval results in Fig. 3. The thickest tropospheric haze (smallest PFMP) is located between 0° and 10°N, consistent with the Galileo observations (West et al., 2004). This high and thick tropospheric haze layer in the equatorial region is likely associated with a strong upwelling from the hydrogen ortho–para fraction data as suggested by Banfield et al. (1998). The aerosol optical depths in the UV and NIR wavelengths at low latitudes in the nominal case are roughly the same. The NIR continuum optical depth increases continuously towards the high latitudes until it is comparable to the CH4 optical depth (0.2) in MT3 wavelengths latitudes ± 70°. The UV optical depth shows discontinuities at about 30°N and near 45°S, where it exceeds the Rayleigh scattering optical depth (0.2 at 100 mbar). The UV/NIR extinction ratio in the middle and high latitudes is roughly constant (35), mainly because they share the same monomer size (10 nm). In our nominal case, the haze column density above 100 mbar is 107 cm2 at low latitudes and ranges from 108 cm2 in the midlatitudes to 109–1010 cm2 in the polar region. Assuming that the mass density is about 1 g cm3, the mass loading of the particles is 106 g cm2 in the low latitudes and 104 g cm2 in the high latitudes. The derived column density and mass loading at mid-latitudes are roughly consistent with Tomasko et al. (1986), who estimated aerosol mass loading on the order of 106 g cm2 in the low latitudes and on the order of 105 g cm2 in the middle latitudes, and the total column density at 45°N is about 5  108 cm2. Our results also generally agree with the global map of aerosol volume density per unit gas abundances in West et al. (1992). But we have shown that the column density and mass loading of the fractal aggregates could vary by one or two order of magnitudes due to the uncertainty of the refractive index in the UV wavelengths, as discussed in previous sections. Fig. 12 shows the aerosol number density map on Jupiter. The maximum aerosol number density also changes dramatically from low latitudes to high latitudes. At equator, the aerosol number density peaks at about 50 mbar, with the value of 10 cm3. In the polar region, for instance, at 65°N, the aerosol number density peaks at about 20 mbar, with the value of 103 cm3. The values are orders of magnitude different from the Galileo high phase angle results (Rages et al., 1999), which show the number density 0.15 cm3 at 100 mbar, and 0.1 cm3 and 0.7 cm3 at 20 mbar for the equatorial region and 60°N, respectively. We attribute the difference to several reasons. First, the grazing limb observations in Rages et al. (1999) can only sample down to optical depths of less than 0.1, which is only the top 1% of the total haze optical depth from Cassini data. Second, we adopted the haze vertical profiles with a peak around 20–50 mbar from the NIR ground-based

10.0

1.0

2

10

10

0.1 101

100

100 -60

-30

0

30

60

Latitude (Degree) Fig. 12. Zonally averaged number density map of stratospheric aerosols (cm3) on Jupiter, as a function of pressure and latitude.

0

50

100

150

Scattering Angle Fig. 13. Retrieved phase functions of the stratospheric aerosol (upper panel) and tropospheric cloud (lower panel) for NIR and UV filters. The dotted vertical lines indicate the scattering angles associated with the ISS images. Blue dashed curve in the lower panel is the cloud phase function derived from Pioneer 10 observations in the RED filter (0.64 lm) at the north component of the South Equatorial Belt of Jupiter (Tomasko et al., 1978). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

170

X. Zhang et al. / Icarus 226 (2013) 159–171

South Equatorial Belt (SEBn) of Jupiter (Tomasko et al., 1978). Detailed analysis of the clouds may require more cloud channel data to separate the tropospheric haze from the bottom layer clouds.

1.75 1.70

n

1.65 1.60 1.55

4. Concluding remarks

0.4

0.6

0.8

1.0

Wavelength (μm) Khare et al. (1984) Khare et al. (1987) This work

0.1000

k

In this study, we analyzed two types of observations to retrieve the jovian aerosols. The spectral shape of the ground-based NIR data in the CH4 bands are used to derive the latitudinal and vertical profiles of the aerosols, from which we can further determine the particle size, shape and optical properties in the optical wavelengths based on the UV and visible-IR limb-darkening profiles at multiple phase angles from Cassini ISS. We obtained an aerosol number density map by combining the two pieces of information. Only one type of tropospheric haze/cloud layer is needed to explain the limb-darkening profiles for all the latitudes. The effective cloud top is located at 200 mbar, consistent in both NIR and Cassini ISS retrievals. The north polar cloud layer appears to be deeper than it is in the south high latitudes. The PMFP in the cloud is around 100–200 mbar in the NIR. It appears that the tropospheric hazes/clouds at low latitudes is optically thicker than the high latitude cloud and the equatorial clouds are the thickest. We distinguished two types of aerosols in the stratosphere of Jupiter. CSM particles are located in the low latitudes between 40°S and 25°N, with a radius between 0.2 and 0.5 lm. The rest of the stratosphere is covered by the fractal aggregated particles composed of a thousand 10-nm monomers, provided that the imaginary part of the UV refractive index is 0.02. The polar haze is one to two orders of magnitude optically thicker than the lower latitude haze. The column density of the aerosols ranges from 107 cm2 at low latitudes to 109 cm2 in the polar region. The mass loading of aerosols in the stratosphere is from 106 g cm2 at the low latitudes to 104 g cm2 in the high latitudes. Multiple solutions exist due to the uncertainty of imaginary part of the refractive index. But it is important to keep the haze optical properties within some range to explain the data. Therefore, the imaginary index requires an upper and a lower limit so that the single scattering albedo is not strongly influenced. For different imaginary indices, the column density of the particles could be varied to keep the haze optical depth at the same level for all the cases, i.e., around 0.02–0.03 at low latitudes and about a few at high latitudes. In order to maintain roughly the same extinction efficiency and phase function, one can adjust the monomer radius and number of monomers of the high-latitude aerosol particles to fit the spectra, and this may lead to multiple solutions. We found that the monomer radius could vary from 5 nm to 40 nm, the number of monomers from 10,000 to 100, and the total aerosol mass loading from 105 g cm2 to 103 g cm2, corresponding to different UV imaginary indices. We also constrained the UV imaginary index within the range of 0.006–0.08 from the high-latitude images. The derived value in the UV wavelengths from Moreno and Sedano (1997) is about 0.02– 0.04 in the high latitudes. Note that their values are based on spherical particles and the low phase angle images only, and we use fractal aggregates and both low and high phase angle images are included. The upper limit of the NIR imaginary index from this study is 0.004. Very few previous laboratory measurements focused on the refractive index for the aerosols in the hydrogen dominant environment. Khare et al. (1987) measured the imaginary part of the refractive index of thin hydrocarbon films produced in the mixture of 7% CH4 and 93% H2 from the charge particle irradiation at 0.13 mbar pressure from 0.4 to 2.5 lm (Fig. 14). Their values in the NIR region are located within the error bar of our retrieval results (Fig. 14). In the UV region, the values derived in this study imply that particles produced in the H2 environment have

1.50 0.2

0.0100 0.0010 0.0001 0.2

0.4

0.6

0.8

1.0

Wavelength (μm) Fig. 14. Real (upper panel) and imaginary (lower panel) part of the refractive index of aerosols from laboratory measurements. Retrieved imaginary refractive index for the ISS NIR and UV filters are plotted for comparison (red). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

weaker absorptivity than that from the N2 environment, as shown by the comparison of our results with the widely used tholin refractive index measured by Khare et al. (1984) in a Titan-like environment. The change of the absorptivity might be related to the unpaired electrons of nitrogen interacting with the delocalized p electrons from aromatics (Imanaka et al., 2004). The similarity between the mid- and high-latitude aerosols suggests that the source of the mid-latitude particles might be in the polar region, possibly due to the complex hydrocarbon synthesis driven by the energetic particle precipitation in the aurora region (e.g., Hord et al., 1979; Pryor and Hord, 1991; Wong et al., 2003). This hypothesis is also consistent with the NIR retrieval results (Fig. 3), which show the polar haze layer is at 10–20 mbar, higher than the middle and low latitudes (50 mbar) in the southern hemisphere, implying an efficient transport from the polar region to the middle latitudes. Another line of evidence is from the boundaries of the CSM particle zone (or the low optical depth zone) at low latitudes, which are not symmetric about the equator. It has been hypothesized to correlate with the hemispheric asymmetry of the auroral precipitation, since the auroral main oval extends to lower latitudes in the northern polar region (West et al., 2004). On the other hand, the difference between the mid-latitude and low-latitude particles, although they reside on the same pressure levels from the NIR observations, suggests that the low-latitude particles might be generated via a different chemical pathway, e.g., by the neutral photochemical processes driven by the UV photons instead of high energetic particles in the high latitudes. The polar particles can be transported by eddy mixing or wind advection. The Stokes sedimentation timescale of the particles at 10 mbar is about an Earth year (Banfield et al., 1998), which is smaller than the horizontal eddy mixing timescales (10–100 years) estimated from the SL-9 debris data (Friedson et al., 1999) and C2H2 and C2H6 distributions from Cassini (Liang et al., 2005). The aerosol heating in the polar region is large, which might induce a circulation from poles to the mid-latitudes. However, the advection by the mean residual circulation from previous studies (timescale 100 years, e.g., West et al., 1992) is not fast enough to transport the polar fractal aggregates to the mid-latitudes. A detail chemical-transport model has yet to be developed to explain the particle transport in the high latitudes. Previous studies on the aerosol solar heating rate are not consistent with each other. Based on the latitudinal distribution of

X. Zhang et al. / Icarus 226 (2013) 159–171

aerosols from the observations by Voyager and International Ultraviolet Explorer (IUE), West et al. (1992) calculated the aerosol heating rate map in the stratosphere of Jupiter and found that the aerosol heating effect is so large, especially at the polar region, that it might drive a circulation from the poles to the mid-latitudes. However, Moreno and Sedano (1997) derived the aerosol properties based on a microphysical model and the Hubble Space Telescope (HST) images. They found that the aerosol heating rate is significantly smaller than that in West et al. (1992), especially in the northern polar region. However, the vertical profile of the jovian stratospheric aerosol was not well determined until the study by Banfield et al. (1998). Now we know the derived aerosol vertical profile from the NIR spectra (Banfield et al., 1998) differs significantly from the microphysical model results in Moreno and Sedano (1997). Since the details of the fractal aggregates were not revealed before, both the sub-micron particles in West et al. (1992) and the tiny particles (<0.1 lm) in Moreno and Sedano (1997) are not consistent with the observations. In light of the aerosol global map and particle properties derived in this study, a renewed effort on aerosol heating in the stratosphere of Jupiter is justified. Acknowledgments We thank M. Lemmon for the parameterization model for the aggregated particles, T. Dowling for the C-DISORT program, P.G. Irwin for the CH4 correlated-k coefficients, M. Line for helpful discussions, and K. Rages and the other reviewer for useful comments. This research was supported by the Outer Planets Research program via NASA Grant JPL 1452240 to the California Institute of Technology. Y.L.Y. was supported in part by NASA NNX09AB72G grant to the California Institute of Technology. X.Z. was supported in part by the Bisgrove Scholar Program in the University of Arizona. References Banfield, D., Gierasch, P.J., Squyres, S.W., Nicholson, P.D., Conrath, B.J., Matthews, K., 1996. 2 lm spectrophotometry of jovian stratospheric aerosols-scattering opacities, vertical distributions, and wind speeds. Icarus 121, 389–410. Banfield, D., Conrath, B.J., Gierasch, Nicholson, P.D., 1998. Near-IR spectrophotometry of jovian aerosols – Meridional and vertical distributions. Icarus 134, 11–23. Barrado-Izagirre, N., Sánchez-Lavega, A., Pérez-Hoyos, S., Hueso, R., 2008. Jupiter’s polar clouds and waves from Cassini and HST images: 1993–2006. Icarus 194, 173–185. Borysow, A., 1992. New model of collision-induced infrared absorption spectra of H2–He pairs in the 2–2.5 lm range at temperatures from 20 to 300 K: An update. Icarus 96, 169–175. Borysow, A., 2002. Collision-induced absorption coefficients of H2 pairs at temperatures from 60 K to 1000 K. Astron. Astrophys. 390, 779–782. Borysow, A., Frommhold, L., 1989. Collision-induced infrared spectra of H2–He pairs at temperatures from 18 to 7000 K. II – Overtone and hot bands. Astrophys. J. 341, 549–555. Borysow, A., Frommhold, L., Moraldi, M., 1989. Collision-induced infrared spectra of H2–He pairs involving 0–1 vibrational transitions and temperatures from 18 to 7000 K. Astrophys. J. 336, 495–503. Buras, R., Dowling, T., Emde, C., 2011. New secondary-scattering correction in disort with increased efficiency for forward scattering. J. Quant Spec. Rad. Trans. 112, 2028–2034. Friedson, A.J., West, R.A., Hronek, A.K., Larsen, N.A., Dalal, N., 1999. Transport and mixing in Jupiter’s stratosphere inferred from Comet SL-9 dust migration. Icarus 138, 141–156. Hord, C.W. et al., 1979. Photometric observations of Jupiter at 2400 angstroms. Science 206, 956–959.

171

Imanaka, H., Khare, B.N., Elsila, J.E., Bakes, E.L.O., McKay, C.P., Cruikshank, D.P., Sugita, S., Matsui, T., Zare, R.N., 2004. Laboratory experiments of Titan tholin formed in cold plasma at various pressures: implications for nitrogencontaining polycyclic aromatic compounds in Titan haze. Icarus 168, 344–366. Karkoschka, E., Tomasko, M.G., 2010. Methane absorption coefficients for the jovian planets from laboratory, Huygens, and HST data. Icarus 205, 674–694. Kaye, J.A., Strobel, D.F., 1983. HCN formation on Jupiter: The coupled photochemistry of ammonia and acetylene. Icarus 54, 417–433. Kedziora-Chudczer, L., Bailey, J., 2011. Modelling the near-IR spectra of Jupiter using line-by-line methods. Mon. Not. R. Astron. Soc. 414, 1483–1492. Khare, B.N., Sagan, C., Arakawa, E., Suits, F., Callcott, T., Williams, M., 1984. Optical constants of organic tholins produced in a simulated titanian atmosphere: From soft X-ray to microwave frequencies. Icarus 60, 127–137. Khare, B., Sagan, C., Thompson, W., Arakawa, E., Votaw, P., 1987. Solid hydrocarbon aerosols produced in simulated Uranian and Neptunian stratospheres. J. Geophys. Res. 92, 15067–15082. Levenberg, K., 1944. A method for the solution of certain problems in least squares. Q. Appl. Math. 2, 164–168. Liang, M.C., Shia, R.L., Lee, A.Y.T., Allen, M., Friedson, A.J., Yung, Y.L., 2005. Meridional transport in the stratosphere of Jupiter. Astrophys. J. 635, L177– L180. Mackowski, D., Mishchenko, M., 2011. A multiple sphere T-matrix Fortran code for use on parallel computer clusters. J. Quant. Spectrosc. Radiat. Trans. 112, 2182– 2192. Markwardt, C.B., 2008. Non-linear least squares fitting in IDL with MPFIT. In: Bohlender, D., Dowler, P., Durand, D. (Eds.), Proc. Astronomical Data Analysis Software and Systems XVIII, Quebec, Canada, ASP Conference Series, 411. Astronomical Society of the Pacific, San Francisco, pp. 251–254. Marquardt, D.W., 1963. An algorithm for least-squares estimation of nonlinear parameters. J. Soc. Ind. Appl. Math. 11, 431–441. Moré, J., 1978. The Levenberg–Marquardt algorithm: implementation and theory. In: Watson, G.A. (Ed.), Numerical Analysis 630. Springer-Verlag, Berlin. Moreno, F., 1996. The structure of the stratospheric aerosol layer in the equatorial and south polar regions of Jupiter. Icarus 124, 632–644. Moreno, F., Sedano, J., 1997. Radiative balance and dynamics in the stratosphere of Jupiter: Results from a latitude-dependent aerosol heating model. Icarus 130, 36–48. Porco, C.C. et al., 2003. Cassini imaging of Jupiter’s atmosphere, satellites, and rings. Science 299, 1541–1547. Pryor, W.R., Hord, C.W., 1991. A study of photopolarimeter system UV absorption data on Jupiter, Saturn, Uranus, and Neptune: Implications for auroral haze formation. Icarus 91, 161–172. Rages, K., Beebe, R., Senske, D., 1999. Jovian stratospheric hazes: The high phase angle view from Galileo. Icarus 139, 211–226. Smith, P.H., 1986. The vertical structure of the jovian atmosphere. Icarus 65, 264– 279. Stamnes, K., Tsay, S.C., Jayaweera, K., Wiscombe, W., 1988. Numerically stable algorithm for discrete-ordinate-method radiative transfer in multiple scattering and emitting layered media. Appl. Opt. 27, 2502–2509. Tomasko, M., West, R., Castillo, N., 1978. Photometry and polarimetry of Jupiter at large phase angles: I. Analysis of imaging data of a prominent belt and a zone from pioneer 10. Icarus 33, 558–592. Tomasko, M., Karkoschka, E., Martinek, S., 1986. Observations of the limb darkening of Jupiter at ultraviolet wavelengths and constraints on the properties and distribution of stratospheric aerosols. Icarus 65, 218–243. Tomasko, M. et al., 2008. A model of Titan’s aerosols based on measurements made inside the atmosphere. Planet. Space Sci. 56, 669–707. West, R.A., 1979. Spatially resolved methane band photometry of Jupiter: II. Analysis of the South Equatorial Belt and South Tropical Zone reflectivity. Icarus 38, 34–53. West, R.A., Smith, P.H., 1991. Evidence for aggregate particles in the atmospheres of Titan and Jupiter. Icarus 90, 330–333. West, R., Friedson, A., Appleby, J., 1992. Jovian large-scale stratospheric circulation. Icarus 100, 245–259. West, R.A., Baines, K.H., Friedson, A.J., Banfield, D., Ragent, B., Taylor, F., 2004. Jovian clouds and haze. In: Bagenal, F., Dowling, T., McKinnon, W. (Eds.), Jupiter – The Planet, Satellites and Magnetosphere. Cambridge University Press. West, R. et al., 2010. In-flight calibration of the Cassini imaging science sub-system cameras. Planet. Space Sci. 58, 1475–1488. Wong, A.S., Yung, Y.L., Friedson, A.J., 2003. Benzene and haze formation in the polar atmosphere of Jupiter. Geophys. Res. Lett. 30, 1447, http://dx.doi.org/10.1029/ 2002GL016661. Zhang, X., Nixon, C.A., Shia, R.L., West, R.A., Irwin, P.G. J., Yelle, R.V., Allen, M.A., Yung Y.L., 2013. Radiative Forcing of the Stratosphere of Jupiter, Part I: Atmospheric Cooling Rates from Voyager to Cassini. Planet. Space Sci. (submitted for publication).