230 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

Comparing -Track Droplet Sizes Inferred from Terra and Aqua MODIS Data

BURCU KABATAS Eurasian Institute of Earth Sciences, Istanbul Technical University, Istanbul, Turkey

W. PAUL MENZEL Space Science and Engineering Center, University of Wisconsin–Madison, Madison, Wisconsin

ATA BILGILI Maritime Faculty, Istanbul Technical University, Istanbul, Turkey

LIAM E. GUMLEY Space Science and Engineering Center, University of Wisconsin–Madison, Madison, Wisconsin

(Manuscript received 7 November 2011, in final form 3 July 2012)

ABSTRACT

In this study of ship tracks, Moderate Resolution Imaging Spectroradiometer (MODIS) measurements from late-morning (Terra) and early-afternoon (Aqua) Earth Observing System platforms are analyzed in five separate geographically distributed cases to compare estimates of the sizes (and their changes in time) of droplets associated with ship exhaust. Ship tracks are readily detected in near-infrared imagery as bright features, especially in 2.13-mm observations. The Terra ‘‘MOD06’’ and Aqua ‘‘MYD06’’ products are used to determine the effective radius of the ship-track droplets; droplet age (time in the atmosphere) is estimated as a function of the distance from the ship. Terra and Aqua MODIS estimates of droplet sizes in ship-track plumes are found to be in agreement, with a correlation greater than 0.90; for the cases studied, droplet sizes in the ship plumes are between 6 and 18 mm. Moreover, the droplets’ size growth rates inferred 2 from the length of the ship track were found to average between 0.5 and 1.0 mmh 1.

1. Introduction balance (Durkee et al. 2000a). The exhaust released by has been found to increase the number of cloud The effect of anthropogenic aerosols on the earth’s droplets, to reduce the droplet size (Coakley et al. 1987), energy budget can either be direct through scattering and thus to contribute to a cooling effect on the earth’s and absorbing the incoming solar and infrared radiation surface (Twomey 1974; Albrecht 1989). Platnick and or indirect by acting as cloud condensation nuclei (CCN). Twomey (1994) used ship tracks to define cloud suscep- Ship tracks, which were first observed by the Television tibility and explained how the increase in cloud-droplet and Infrared Observation Satellite VII (TIROS VII)in number also increases the cloud reflectivity (). 1965 (Conover 1966), are linear seen in near- In this paper, Moderate Resolution Imaging Spec- infrared images of marine stratocumulus that are caused troradiometer (MODIS) cloud products (King et al. by emissions from ships (Segrin et al. 2007). The Monterey 1997) derived from near-infrared (NIR) MODIS spectral- Area Ship Track experiment studied ship tracks to un- band measurements are used to study ship-track droplet derstand the role of anthropogenic aerosols in modifying sizes and their changes over time in five case studies. the cloud reflectivity and hence the earth’s radiation MODIS cloud-particle-size estimates from the Terra and Aqua platforms of the National Aeronautics and Space Administration (NASA) Earth Observing System Corresponding author address: W. Paul Menzel, Space Science and Engineering Center, University of Wisconsin—Madison, 1225 (EOS) are compared. The following sections present a West Dayton St., Madison, WI 53706. summary of detection of ship tracks in MODIS imagery E-mail: [email protected] (section 2), a description of the data and algorithms

DOI: 10.1175/JAMC-D-11-0232.1

Ó 2013 American Meteorological Society Unauthenticated | Downloaded 10/02/21 05:56 PM UTC JANUARY 2013 K A B A T A S E T A L . 231

(section 3), results from the five case studies (section 4), between reflection from a ship track’s smaller droplets and some conclusions (section 5). and nearby cloudy larger droplets. How much greater depends on droplet size and optical depth, which are inferred in the MOD06 algorithm (King et al. 1997; 2. Ship-track formation and detection Platnick et al. 2000). Ship tracks form in the low stratus and stratocumulus clouds off the western coasts of large continents. The dominant areas for stratiform-cloud formation are the 3. Data and algorithms eastern basins between 208 and 508 latitude and at a. MODIS data overview high latitudes above about 608 (Durkee et al. 2000a). Schreier et al. (2007) reported on ship tracks in the sub- MODIS is a scanning radiometer on the EOS Terra tropical latitudes along the west coasts of southern platform (equator crossing at ;1030 local standard time) Africa, South America, and North America. In the and EOS Aqua platform (equator crossing at ;1330 local Northern Hemisphere, cold upwelling ocean currents standard time) that has a long-term science mission to cause a stable atmospheric layer that is due to Ekman study global changes in land, ocean, and atmosphere pumping by northerly winds along the coast; in the (King et al. 1992). Each MODIS provides worldwide Southern Hemisphere, the necessary winds are south- datasets in 36 spectral bands every 2 days from a polar- erly winds. This causes the marine boundary layer to orbiting, sun-synchronous, platform at an altitude of become saturated and thus provides an ideal environ- 705 km with a 2330-km swath width. ment for the formation of stratiform clouds (Evans 1992; The MODIS cloud products (denoted by ‘‘MOD06’’ Klein and Hartmann 1993). When particles from ship for Terra and ‘‘MYD06’’ for Aqua) have been described engine exhaust enter a stratiform cloud layer in the by King et al. (1997); these include the effective particle boundary layer, they can act as CCN and form cloud radius, which is used to examine the droplet size along droplets (Hobbs et al. 2000). the lengths of ship tracks. Not every ship produces a ship track. Ambient con- b. Estimating droplet size ditions of a well-mixed boundary layer, low numbers of CCN, and near-constant surface temperature and rela- The dependence of cloud reflectance on droplet size is tive humidity in the marine atmosphere enhance the related to the ratio of total volume of the drops to their probability of ship-track formation (Conover 1966). The total surface area (Rosenfeld and Woodley 2001). The high static stability associated with these atmospheric incoming radiation is scattered from the surface of conditions confines the clouds to the boundary layer, the droplets, and therefore the scattered radiation is enabling the ship-emitted aerosols to reach the cloud proportional to the total surface area of the droplets base. The boundary layer depth must be low; ship tracks (proportional to the droplet radius squared: ;r2). The rarely occur in low-level clouds having altitudes of absorbed radiation is proportional to the total volume greater than 1 km. A small rise in low-level cloud alti- of the droplets (proportional to the droplet radius cu- tude can cause the disappearance of ship tracks from one bed: ;r3) since it occurs inside the droplet. Hence, in day to the next (Coakley et al. 2000). Fuel type also plays approximately constant amounts of liquid water, the a significant role in whether a ship track is produced. It scattering-to-absorption ratio increases in ship-track has been reported (Hobbs et al. 2000; Noone et al. 2000) clouds with smaller and more numerous droplets; the that diesel ships burning low-grade marine fuel oil emit increased scattering is evident in the higher reflection larger particles than do ships burning navy distillate fuel, from the ship-track plumes than from the surrounding and these particles serve as CCN at lower supersatura- clouds with larger and fewer droplets. tions (and will therefore be more likely to produce ship Reflection in the NIR wavelengths is sensitive to tracks). both cloud optical thickness and cloud-particle size In ship tracks, the smaller and more numerous droplet (Nakajima and King 1990; Platnick et al. 2000). MOD06 sizes make the plume brighter and more reflective to and MYD06 use the bands at 0.86 and 2.13 mm. For incoming sunlight, especially in the NIR part of the a given optical depth, the 2.13-mm reflected radiance is spectrum at 1.64 and 2.13 mm (King et al. 1992). These more sensitive to the cloud-particle size; conversely, NIR spectral bands exhibit little water vapor absorption for a given particle size, the reflected radiance of the and reflect more from smaller droplets. Moreover, the 0.86-mm band is more sensitive to the cloud optical decrease in reflectivity with increased droplet radius depth. (ranging from 5 to 20 mm) is more pronounced at longer In this work, the droplet effective radius at a given wavelengths; hence, 2.13 mm shows greater contrast location within a ship-track plume is estimated from

Unauthenticated | Downloaded 10/02/21 05:56 PM UTC 232 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

FIG.1.Terra MODIS 2.13-mm reflectance for a scene exhibiting ship tracks off the coast of California at 1920 UTC 30 Sep 2005. Five plumes are identified with numbers. Two transects are indicated across exhaust plume 1, one nearer the ship (in red) and another farther away (in green).

MODIS Terra MOD06 and Aqua MYD06 determi- Figure 1 shows the reflectance values for a scene nations along a transect perpendicular to the ship track, with several ship tracks off the coast of California on away from ship-track intersections. Working from the 30 September 2005 in Terra MODIS band 7 (2.13 mm). 2.13-mm radiance data, one selects a given part of the The reflectance values within a plume are larger (;0.20– ship track, and a pixel near the center of the plume, 0.25) than the values outside the plumes (;0.10–0.15). where a maximum reflectance value is found, is trans- As an example of droplet-size differences along the plume, lated to the corresponding MOD06 (or MYD06) cloud two transects across exhaust plume 1, one near the ship (in product. Single pixel values might seem to be too sub- red in Fig. 1) and another farther away (in green in Fig. 1), jective. A more objective way of calculating the droplet are presented. The reflectances and the associated MOD06 size is to use an average value from a number of pixels droplet sizes along each transect are shown in Fig. 2. Near around the centerline (at least five but increasing with the ship a droplet size of 9.0 mm is associated with the re- position along the ship track as the plume gets wider flectance maximum of 0.27 occurring at 35819936.060N, with dispersion). Comparison of the two approaches 12484294.530W. Farther down the plume at 3384197.730N, reveals that the single pixel value is representative; the 127819938.340W, the droplet size inferred from the transect maximum difference in droplet size between the single reflectance maximum of 0.23 has increased to 13.2 mm. The pixel value and the average transect value is found to sharpness of the reflectance maximum has been dispersed be 5%. noticeably farther down the plume.

Unauthenticated | Downloaded 10/02/21 05:56 PM UTC JANUARY 2013 K A B A T A S E T A L . 233

FIG. 2. (top) Transects of MOD06 particle sizes and reflectances near the ship where a droplet size of 9.0 mmis associated with the reflectance maximum of 0.27 occurring at 33819936.060N, 12484294.530W. (bottom) Transects of MOD06 particle sizes and reflectances farther down the plume at 3384197.730N, 127819938.340W, where the droplet size inferred from the reflectance maximum of 0.23 has increased to 13.2 mm.

Regular spacing of transects along the ship track is Note that the pixels near but not in the ship-track desirable so that the droplet changes as a function of plume may be in broken clouds; retrieved droplet radii time can be studied. The crossings of two separate ship are less reliable as the cloud-cover fractions decrease tracks must be avoided, however, since the droplets (Zhang and Platnick 2011). As a consequence, the nearby from two different ships have been mixed. The selection cloud-droplet radii show relatively large variability in of the exact location for a given transect introduces comparison with the droplet radii in the overcast ship variability in the droplet-size estimate; droplet sizes are track plume. It can reliably be said that the ship-track- found to change by 60.5 mm when the associated tran- plume droplets are consistently smaller than the nearby sect is moved one or two pixels nearer or farther along cloud droplets, however. the plume. c. Estimating changes over time The variability of droplet sizes along the ship-track plume and in the nearby clouds is demonstrated in Fig. 3. In straight-line plumes, the length of time that an ex- The top panel in Fig. 3 shows the droplet size as a func- haust particle and the associated droplet have been in tion of distance within plume 2 off the coast of California the atmosphere can be estimated using the droplet dis- (in Fig. 1); droplet sizes fluctuate between 11 and 15 mm tance from the ship divided by the speed of the combined for the first 150 pixels (km) as earlier ship plumes are ship and wind velocity vectors (Durkee et al. 2000b). encountered and then settle to a gradual increase from Given the respective latitude–longitude locations of the about 12 to 18 mm for pixels 150–350. Growth of the ship and the droplet, the great-circle distance between droplet with distance from the ship is evident. The bot- the two is calculated with a Haversine formula. The ship tom panel in Fig. 3 shows the associated cloud-droplet velocity is calculated from the great-circle distance be- sizes found alongside plume 2 but in the nearby clouds tween the plume origins in the Terra and Aqua images between plumes 2 and 3 (see Fig. 1); droplet sizes start at divided by the Terra and Aqua overpass time difference. 28 mm, fluctuate between 12 and 28 mm in the first 200 The wind velocity is inferred from Quick Scatterometer pixels where clouds and plumes are mixing, and then (QuikSCAT) data. Ship and wind velocities are assumed stay near 25 mm except when crossing one more ship to stay constant during ship-track-plume formation. In track (ship-track plume crossings are evident near pixels Fig. 1, the time difference for the droplets near to the 60, 135, 160, and 290). The nearby cloud droplets remain ship and far from the ship is estimated to be 5.2 h. larger than the ship-track-plume droplets for the length The origin of the plume is somewhat displaced from of the plume. the exact ship location. The higher reflectance in the ship

Unauthenticated | Downloaded 10/02/21 05:56 PM UTC 234 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

FIG. 3. Off the coast of California on 30 Sep 2005, (top) the MOD06 droplet radius along the centerline of ship-track plume 2 (see Fig. 1) for 350 pixels (km) going southwest away from the ship starting near 358N, 1258W and ending near 328N, 1288W, and (bottom) the MOD06 ef- fective droplet radius associated with nearby clouds between ship-track plumes 2 and 3 going southwest parallel to plume 2. Location denotes the pixel number (and distance in kilometers from the ship) along the path.

2 exhaust occurs some minutes after (Durkee Near-constant winds of 10–15 kt (1 kt ’ 0.5 m s 1) et al. 2000b), introducing a small offset in our time es- were noted in our case studies. timates. Given the uncertainty in the distance between two pixel locations (roughly 2 km) divided by the un- certainty in the ship and wind velocity estimates (roughly 4. Case studies 21 3ms ), we believe our time estimates have a relative Droplet sizes and their growth with time are estimated accuracy of about 10 min. for ship tracks identified in 2.13-mm imagery in both Terra MODIS and Aqua MODIS data for five different d. Role of regional winds parts of the world: (a) California (30 September 2005), QuikSCAT data are used for estimation of near-surface (b) North Pacific Ocean (10 February 2003), (c) Alaska wind velocity. QuikSCAT obtains wind speed and direc- (4 March 2009), (d) Kuril Islands (2 July 2003), and (e) tion by combining measurements of radar backscatter Europe (12 February 2005). The droplet sizes are deter- from a given location on the sea surface at multiple mined from MOD06 and MYD06 collection 5 (except for antenna look angles. The accuracy of the wind re- Alaska, for which collection-5.1 data were used). Only trievals is best characterized in terms of vector com- the California case study is described in detail; results ponent errors (Freilich and Dunbar 1999); the QuikSCAT from all five cases are summarized in Table 1. 2 speed accuracy is between 0.75 and 1.5 m s 1 for along- a. California coast case study wind and crosswind components. Direction accuracy at 2 wind speeds of higher than about 6 m s 1 is about 148. Ship tracks were seen in satellite images off the coast The wind map is acquired from SeaWinds, which is at- of California on 30 September 2005 (see Fig. 1). These tached to QuickSCAT (Remote Sensing Systems 2010). stringlike clouds form when water molecules condense

Unauthenticated | Downloaded 10/02/21 05:56 PM UTC J ANUARY 03KABATASETAL. L A T E S A T A B A K 2013

TABLE 1. Summary of all cases. Note that the correlation between Terra and Aqua particle-size estimates for each common plume is calculated by comparing particle sizes as a function of time in the atmosphere for each sensor. Aqua estimates are linearly interpolated in time to match Terra times.

Case name California North Pacific Alaska Kuril Islands Europe Date 30 Sep 2005 10 Feb 2003 4 Mar 2009 2 Jul 2003 12 Feb 2005 Data Terra MODIS; Terra MODIS; Terra MODIS; Terra MODIS; Terra MODIS; Aqua MODIS Aqua MODIS Aqua MODIS Aqua MODIS Aqua MODIS Time between two 95 105 110 100 100 sensors (min) Resolution (km) 1 1 1 1 1 Wind speed (kt) 15 10 15 15 15 Vessel speed (kt) ;21 6 4 ;31 6 3 ;20 6 4 ;22 6 4 ;23 6 4 Duration of plume (h) 5–9 (according 7–19 (according 8–10 (according 13–16 (according 5–16 (according to Aqua) to Aqua) to Aqua) to Aqua) to Aqua) Droplet size (mm) at beginning of plume and end of plume, and plume length (in pixels) Terra plume 1 10.2, 13.0, 428 6.9, 8.0, 874 8.3, 14.0, 382 6.6, 11.5, 716 10.3, 15.5, 385 Terra plume 2 10.8, 16.7, 488 5.9, 9.1, 570 9.1, 16.4, 340 8.1, 12.2, 548 7.1, 11.6, 367 Terra plume 3 12.4, 14.4, 396 5.9, 9.0, 396 9.7, 14.0, 536 8.6, 12.7, 750 10.6, 14.8, 488 Terra plume 4 6.8, 10.2, 477 7.0, 10.3, 432 10.6, 15.8, 511 9.1, 17.0, 820 7.2, 14.7, 575 Terra plume 5 11.1, 13.3, 367 8.2, 12.7, 826 12.8, 18.9, 454 — — Aqua plume 1 10.8, 13.3, 500 7.0, 8.0, 936 8.3, 14.8, 441 6.3, 10.4, 773 8.5, 13.5, 350 Aqua plume 2 11.4, 15.4, 538 6.6, 9.1, 619 8.2, 16.4, 384 7.1, 11.5, 685 8.1, 11.4, 377 Aqua plume 3 11.1, 14.6, 430 5.9, 9.3, 420 10.1, 18.7, 573 9.5, 14.7, 806 7.3, 14.6, 405

Unauthenticated |Downloaded 10/02/21 05:56 PMUTC Aqua plume 4 8.8, 12.1, 527 6.9, 10.8, 490 10.6, 19.1, 601 10.0, 15.1, 940 7.2, 13.9, 660 Aqua plume 5 9.4, 12.7, 486 8.2, 13.5, 940 12.4, 19.8, 500 — — Correlation of droplet-size ;0.91 6 0.07 for ;0.93 6 0.02 for ;0.91 6 0.07 ;0.91 6 0.07 ;0.91 6 0.06 for determinations between sensors 47 samples 66 samples for 53 samples for 52 samples 45 samples Avg droplet-size growth with 0.3–0.8 for Terra and 0.5–1.0 0.1–0.5 for Terra 0.6–1.2 for Terra and 0.8–1.2 0.4–0.8 for Terra 0.8–1.0 for Terra 2 time (mmh 1 ) for Aqua and 0.1–0.6 for Aqua for Aqua and 0.3–0.5 for Aqua and 0.6–1.2 for Aqua 235 236 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

FIG. 4. Image of effective cloud-particle-size determinations from (a) Terra MOD06 and (b) Aqua MYD06 off the coast of California on 30 Sep 2005. Black indicates clear sky.

2 on the particles released from ship exhaust and are ex- and 1.0 mmh 1 for Aqua. Figure 6 shows the Terra and tended by the wind over a long, narrow path. As noted Aqua effective radius values for droplets of the same age earlier, since the particles emitted by ship are smaller within each plume. They are found to be in agreement, than those in the nearby atmosphere, their albedo is with a correlation of 0.91 for the 24 Terra and 23 Aqua higher than the surrounding clouds. values (where the Aqua values have been linearly inter- The ambient conditions of a stable atmospheric layer polated to the Terra droplet ages) for the five plumes. The and a northerly wind are present. According to the wind lengths of the plumes range from roughly 360 to 540 map (not shown), the wind is blowing from the north pixels (where 1 pixel at nadir is 1 km). with a speed of ;15 kt. b. Other case studies Terra and Aqua measurements, covering the same region with a time difference of 95 min, offer an oppor- The same analysis described for the case study of the tunity to compare droplet estimates from the two in- California coast is applied for the remaining case studies struments and to estimate droplet growth as a function of in the North Pacific on 10 February 2003, off the coast time. Figure 4 shows images of the effective particle sizes of Alaska on 4 March 2009, near the Kuril Islands on for the ship-track plumes in the Terra (MOD06) and 2 July 2003, and off the coasts of France and Spain on Aqua (MYD06) data separately; the droplet sizes within 25 February 2005. Figure 7 shows the MODIS detection the ship track range from 6 to 18 mm. Outside the ship of the ship tracks in 2.13-mm images for the additional tracks the droplet sizes in the nearby clouds are estimated four cases. Table 1 shows a summary of the findings for all to be larger, between 20 and 25 mm (not evident in Fig. 4 five cases; the time, location, ambient conditions, plume as the enhancement has been tuned for the droplets in duration, droplet size at beginning of plume, droplet size the plumes). Figure 5 shows plots of the droplet size as at end of plume, plume length, Terra and Aqua correla- a function of droplet age for each plume individually for tion for droplet size as a function of age, and average Terra and Aqua. The growth of the droplet size (rate of droplet-size growth with time are listed. change) in the ship-contaminated clouds is estimated to Ambient conditions were noted to have steady wind 2 be between 0.4 and 0.8 mmh 1 for Terra and between 0.5 speeds of 10–15 kt from the north. Droplet sizes are as

Unauthenticated | Downloaded 10/02/21 05:56 PM UTC JANUARY 2013 K A B A T A S E T A L . 237

FIG. 5. Image of effective cloud-particle-size determinations off the coast of California on 30 Sep 2005, showing estimated ship-track droplet size (radius) as a function of droplet age for five plumes in the (a) Terra MOD06 and (b) Aqua MYD06 data.

small as 6 mm near the start of the plumes and increase the end of the track and thus are optically thin; the to as much as 20 mm far from the plume. Schreier et al. droplet radii are overestimated because they are in (2006) studied the ship track plumes in the North Pacific pixels that are only partly cloudy (Coakley et al. 2005; with Terra MODIS and found droplet sizes of about Zhang and Platnick 2011). This explanation is less likely 10 mm, which falls within our range of 6–13 mm from since most of our observations of droplet-size increase plume beginning to end. Plume lengths are between 340 were made in the nondecaying portions of the ship-track and 940 pixels (km). Correlations between Terra and plumes; inspection of visible imagery (not shown) re- Aqua droplet-size estimates for individual plumes are all veals that the plume transects, in which the droplet radii greater than 0.85; overall, for all of the plumes the cor- were estimated, were chosen in optically thick portions relation is 0.92. of the plume. A third possible explanation is offered by Terra in a descending (morning) orbit and Aqua in an noting that all of the ship tracks are associated with ships ascending (afternoon) orbit have reversed viewing ge- traveling to higher latitudes (northward). As the tails of ometry for each case study, and therefore Terra MODIS the plumes in all of the tracks are at lower latitudes, they could be in the forward-scattered view-zenith-angle di- are probably under the influence of somewhat warmer rection while Aqua MODIS could be in a backscattered sea surface temperatures, which are often associated direction. We find that the two sensors’ estimates of with deeper boundary layers. It is possible that the droplet effective radius are highly correlated and have clouds are thicker at these latitudes and the cloud the same increasing trend regardless of view-angle dif- droplets in the ship tracks are presumably growing be- ferences. Christensen et al. (2009) noted that sun–target– cause the cloud layers are deepening in response to the satellite geometries have a negligible impact on the aerosol plumes (Christensen and Stephens 2011). retrieval of cloud-droplet sizes. Christensen et al. (2009) found ship tracks for which 2 Droplet-size growth between 0.3 and 1.0 mmh 1 was the droplet effective radii decreased significantly in the found in all plumes. One possible explanation for the same clouds tracked from morning (Terra) to afternoon droplet growth is continued moisture accretion to the (Aqua). This was attributed to the sun ‘‘burning off’’ the ship-exhaust CCN as they are dispersed. The ship tracks marine layer during the day. This diurnal cycle of droplet are expanding with distance from the ship, suggesting effective radius was not observed in the ship tracks pre- that dispersion is reducing the concentration of particles sented here—perhaps because most of the droplet sizes in the ship exhaust, some of which served as CCN. Con- were investigated in their first 6 h. sequently, the droplets grow because there are fewer Effective droplet size at the beginning of each ship- particles that are being activated, thereby leaving fewer track plume is found to range from 6 to 12 mm. Figure 8 droplets competing for the available moisture. Another shows a histogram of initial ship-exhaust particle sizes possible explanation is that the clouds are decaying near (droplet effective-radius values at the beginning of the

Unauthenticated | Downloaded 10/02/21 05:56 PM UTC 238 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

FIG. 6. Droplet size as a function of droplet age from Terra MOD06 and Aqua MYD06 overlaid for each of the individual plumes.

Unauthenticated | Downloaded 10/02/21 05:56 PM UTC JANUARY 2013 K A B A T A S E T A L . 239

FIG. 8. Histogram of initial ship-exhaust particle sizes [droplet effective-radius values (mm) at the beginning of the plume] for the 23 plumes studied in the Terra and Aqua data (all of the plumes in the five case studies). Here, 6 indicates 5.51–6.5 mm, 7 indicates 6.51–7.5 mm, etc.

plume) for all of the plumes studied. Droplet sizes tend to cluster into two groups: less than and greater than 9 mm; the latter are possibly connected with ships burning low-grade marine fuel oil. Plumes with initial droplet sizes of less than 9 mm are more likely (40% vs 13%) to extend beyond 600 km in this study. Figure 9 shows the comparisons of droplet-size change with time between Terra and Aqua MODIS for each case. Plumes with larger initial droplet sizes show smaller 2 values of droplet-size growth (0.3–0.5 vs 0.6–1.0 mmh 1).

5. Conclusions MODIS observations of ship tracks were analyzed for five geographically distributed cases using the MODIS cloud-product data. The ship tracks were selected be- cause they were evident in the 2.13-mm imagery of both Terra and Aqua MODIS. This enables comparison of the same droplet sizes from the Terra MOD06 and Aqua MYD06 estimates (based on calculations of ship- exhaust CCN hours in the atmosphere); they were found to agree, with a correlation greater than 0.9. Droplet sizes clustered into two groups: those larger than and those smaller than 9 mm. Ship tracks with smaller initial droplet sizes were usually found to persist for a longer time (as indicated by the longer plume lengths). Cloud-droplet growth with time was tracked in all of the ship-track plumes out to more than 6 h. A possible explanation for the droplet growth along the plume is that the ship tracks are dispersing with distance from FIG. 7. Ship tracks seen in the 2.13-mm images for the remaining four case studies (a) in the North Pacific Ocean on 10 Feb 2003, (b) the ship, the concentration of particles in the ship ex- off the coast of Alaska on 4 Mar 2009, (c) near the Kuril Islands on haust is decreasing, and the droplets grow because there 2 Jul 2003, and (d) off the coast of Spain and France on 12 Feb 2005. are fewer particles competing for the available mois- Plume numbers are indicated in coordination with Table 1. ture. Plumes with larger initial droplet size exhibited

Unauthenticated | Downloaded 10/02/21 05:56 PM UTC 240 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

FIG. 9. Droplet-size growth with time for Terra (black bars) and Aqua (gray bars) MODIS for each case. distinctively smaller values of droplet-size growth (;0.3– Evans, M. E., 1992: Analysis of ship tracks in cloudiness transition 2 0.5 vs ;0.6–1.0 mmh 1). regions. M.S. thesis, Dept. of Meteorology and Physical Oceanography, Naval Postgraduate School, 102 pp. Freilich, M. H., and R. S. Dunbar, 1999: The accuracy of the NSCAT 1 vector winds: Comparisons with National Data Acknowledgments. This study was supported by Buoy Center buoys. J. Geophys. Res., 104, 11 231–11 246. NASA Contract NNX09AO74G, and the work was ac- Hobbs, P. V., and Coauthors, 2000: Emissions from ships with re- complished at the Space Science and Engineering Cen- spect to their effects on clouds. J. Atmos. Sci., 57, 2570–2590. ter of the University of Wisconsin—Madison. King, M. D., Y. J. Kaufman, W. P. Menzel, and D. Tanre´, 1992: Remote sensing of cloud, aerosol, and water vapor properties from the Moderate Resolution Imaging Spectrometer (MODIS). IEEE Trans. Geosci. Remote Sens., 30, 2–27. REFERENCES ——, S. Tsay, S. E. Platnick, M. Wang, and K. Liou, 1997: Cloud retrieval algorithms for MODIS: Optical thickness, effective Albrecht, B. A., 1989: Aerosols, cloud microphysics and fractional particle radius, and thermodynamic phase, MOD06—Cloud cloudiness. Science, 245, 1227–1230. product. NASA Goddard Space Flight Center MODIS Al- Christensen, M., and G. Stephens, 2011: Microphysical and mac- gorithm Theoretical Basis Doc. ATBD-MOD-05, version 5, rophysical responses of marine stratocumulus polluted by un- 79 pp. derlying ships: Evidence of cloud deepening. J. Geophys. Res., Klein, S. A., and D. L. Hartmann, 1993: The seasonal cycle of low 116, D03201, doi:10.1029/2010JD014638. stratiform clouds. J. Climate, 6, 1587–1606. ——, J. A. Coakley Jr., and W. R. Tahnk, 2009: Morning to af- Nakajima, T., and M. D. King, 1990: Determination of the optical ternoon evolution of marine stratus polluted by underlying thickness and effective particle radius of clouds from reflected ships: Implications for the relative lifetimes of polluted and solar radiation measurements. Part I: Theory. J. Atmos. Sci., unpolluted clouds. J. Atmos. Sci., 66, 2097–2106. 47, 1878–1893. Coakley, J. A., Jr., R. L. Bernstein, and P. A. Durkee, 1987: Effect Noone, K. J., and Coauthors, 2000: A case study of ships forming of ship-stack effluents on cloud reflectivity. Science, 237, 1020– andnotformingtracksinmoderatelypollutedclouds. 1022. J. Atmos. Sci., 57, 2729–2747. ——, and Coauthors, 2000: The appearance and disappearance of Platnick, S., and S. Twomey, 1994: Determining the susceptibility ship tracks on large spatial scales. J. Atmos. Sci., 57, 2765– of cloud albedo to changes in droplet concentration with the 2778. Advanced Very High Resolution Radiometer. J. Appl. Me- ——, M. A. Friedman, and W. R. Tahnk, 2005: Retrievals of cloud teor., 33, 334–347. properties for partly cloudy imager pixels. J. Atmos. Oceanic ——, and Coauthors, 2000: The role of background cloud micro- Technol., 22, 3–17. physics in the radiative formation of ship tracks. J. Atmos. Sci., Conover, J. H., 1966: Anomalous cloud lines. J. Atmos. Sci., 23, 57, 2607–2624. 778–785. Remote Sensing Systems, cited 2010: Description of scatterometer Durkee, P. A., K. J. Noone, and R. T. Bluth, 2000a: The Monterey data products. [Available online at http://www.ssmi.com/ Area Ship Track Experiment. J. Atmos. Sci., 57, 2523–2541. qscat/qscat_description.html.] ——, and Coauthors, 2000b: Composite ship track characteristics. Rosenfeld D., and W. Woodley, 2001: Pollution and clouds. Phys. J. Atmos. Sci., 57, 2542–2553. World, 14, 33–37.

Unauthenticated | Downloaded 10/02/21 05:56 PM UTC JANUARY 2013 K A B A T A S E T A L . 241

Schreier, M., A. A. Kokhanovsky, V. Eyring, L. Bugliaro, Segrin, M. S., J. A. Coakley Jr., and W. R. Tahnk, 2007: MODIS H. Mannstein, B. Mayer, H. Bovensmann, and J. P. Burrows, observations of ship tracks in summertime stratus off the west 2006: Impact of ship emissions on the microphysical, opti- coast of the United States. J. Atmos. Sci., 64, 4330–4345. cal and radiative properties of marine stratus: A case study. Twomey, S., 1974: Pollution and the planetary albedo. Atmos. Atmos. Chem. Phys., 6, 4925–4942. Environ., 8, 1251–1256. ——, H. Mannstein, V. Eyring, and H. Bovensmann, 2007: Global Zhang, Z., and S. Platnick, 2011: An assessment of differences ship track distribution and radiative forcing from 1 year of between cloud effective particle radius retrievals for marine AATSR data. Geophys. Res. Lett., 34, L17814, doi:10.1029/ water clouds from three MODIS spectral bands. J. Geophys. 2007GL030664. Res., 116, D20215, doi:10.1029/2011JD016216.

Unauthenticated | Downloaded 10/02/21 05:56 PM UTC