Island cloud trails represent only a small fraction of ocean boundary layer clouds. However, they can be an important key to understanding how marine boundary layer clouds respond to perturbation. During the La Niña period of 1999–2001 the island cloud of Nauru demonstrated daytime persistence over 100 kilometers. Our results show that boundary layer clouds over the eastern margin of the warm pool are affected by regional subsidence and lack of CCN (Cloud Condensation Nuclei) during this active warm pool convection phase of the ENSO (El Nino Southern Oscillation). The increased persistence of boundary layer clouds during the day, as reflected by the island cloud trails, combined with the dissipation of boundary layer clouds at night contribute to cooling the ocean surface. During La Nina, boundary layer cloudiness around Nauru decreased by a factor of 1.9 and SSTs decreased by about 3°C.
 In this paper we discuss the relationship between three physical phenomena observed over the Tropical Western Pacific ocean near the island of Nauru during the 1999–2000 period: 1) boundary layer clouds, 2) island cloud trails and 3) SSTs (Sea Surface Temperatures). The study of the relationship of cloud radiation forcing and SSTs in the Tropics has a long history and has been thoroughly described [e.g., Ramanathan et al., 1989]. These studies show that an increase in cloudiness in the Tropics is associated with an increase in SSTs due to the strong water vapor downward infrared emission warming effect dominating the solar reflection cooling effect. Most of these studies have focused on cirrus clouds rather than boundary layer clouds. However, based on satellite derived cloud liquid water (CLW) and SSTs, Wong et al.  found that SSTs decreased with increased cloudiness during the SST warming associated with the 1998 El Nino from 30N to 30S. Satellite CLW is a measure of both boundary layer and mid-level liquid water clouds. The focus in this paper is a region near the boundary of the Tropical Warm Pool and a La Nina period between 1999 and 2000. During this period and in this region there were relatively few mid-level or precipitating clouds. The second phenomenon (equatorial island cloud trails) has only been studied recently after their discovery by Nordeen et al. [2001a, 2001b]. McFarlane et al.  have described the effect of these cloud trails on surface measurements made at the island of Nauru as part of the ARM (Atmospheric Radiation Measurement) program. The purpose of this paper is to show how the boundary layer cloud persistence demonstrated by island is both related to large-scale subsidence during the strong La Nina period of 1999–2000. These results suggest that Tropical boundary layer clouds may be more important than previously thought in cloud/SST feedbacks during the La Nina phase of the ENSO.
 The Republic of Nauru is a small island (∼21 km2) located at the eastern edge of the TWP (Figure 1). Figure 1 shows the difference in SSTs between 2000 and 2002. Island cloud trails were observed in satellite data in 2001. Island cloud trails are similar to ship trails [Porch et al., 1999] in form and have been observed over other islands as well, for example, Guadalupe Island in the Pacific Ocean about 300 km west of Baja California [Dorman, 1994]. Though the precise mechanisms of cloud trail formation are not well understood and likely depend on local circumstances, two general conditions are necessary for island cloud trails to form and persist over long distances. First, lifting of air must occur through heating of the air as it passes over the island surface. Second, weak regional subsidence (a few cm/s) is necessary to keep the clouds from dispersing in the vertical. Yamada and Kao  showed that weak negative buoyancy (subsidence) near the top of the mixed layer was necessary to maintain tropical boundary layer clouds using a numerical tropical marine boundary layer model. It should be noted that these small vertical velocities are extremely difficult to measure in clouds as they are on the order of the settling velocity of cloud droplets.
 The focus of this paper is to show that the boundary layer cloud properties are strikingly different during the La Nina phase of the Southern Oscillation from observations taken during the El Nino phase and are directly related to regional SSTs.
2. Observational Data Sets
 The ARM site on the island of Nauru has operated since 1998. Cloud coverage and frequency were inferred from the CLW content as measured by the microwave radiometer (MWR). Here we define a cloud as having liquid water path greater than 0.006 cm as measured by the MWR. The MWR receives microwave radiation from the sky at 23.8 GHz and 31.4 GHz. These frequencies allow simultaneous determination of water vapor and liquid water burdens along a selected path. The MWR does not respond to cloud ice. Rawinsondes are launched at least twice daily at 0 and 12 GMT. Aerosol optical thickness at different visible wavelengths can be obtained from the sun-photometer.
 We examine the cloud trails using data from two satellites. The first satellite is the DOE high spatial resolution Multi-Spectral Thermal Imager (MTI). MTI began its mission in March 2000 and provides data in 15 spectral bands ranging from infrared to visible wavelengths (11 μm to 0.4 μm) [Henderson and Chylek, 2005]. The MTI spatial resolution at nadir is 5 m in the visible and 20 m in the infrared. We also use data from the Japan Meteorological Agency Geostationary Meteorological Satellite (GMS-5) that has a field of view capable of imaging entire cloud trails. The GMS-5 records data in one visible (0.7 μm) and three infrared channels (6.8 μm to 12 μm). Its ground resolution is approximately 1.25 km (visible) and 5 km (infrared) [Nordeen et al., 2001a]. The MTI satellite allowed us to view the island and the surrounding clouds both during the day and at night.
 We examine SSTs around Nauru from 1990 to 2004 using data from the National Oceanic and Atmospheric Administration (NOAA) optimal interpolator for SST, a 1° latitude by 1° longitude monthly mean SST data set derived from satellite and in situ observations [Reynolds et al., 2002]. We also use SSTs inferred from MTI satellite images.
3. ENSO and Cloudiness Over Nauru
 Nauru is more cloudy during the El Nino (2002) than La Nina (2000) (Figure 2). Clouds are observed over Nauru roughly 38% of the time during the El Nino ENSO phase but only 25% of the time during the La Nina. The cloud frequency during both ENSO phases peaks shortly after noon. The El Nino to La Nina cloud frequency ratio is about 1.9 consistently with no diurnal characteristics of note (Figure 2).
McFarlane et al.  show that wind directions during the La Nina phase (1999–2000) were almost exclusively from the East while they varied more uniformly from West and East during the El Nino phase (2002). Observations of mineral dust at Nauru were made by Prospero in the early 1990's during periods when the winds were blowing into the island from the East. These observations [Zender et al., 2003] show that Nauru has very little mineral aerosol when the winds blow from the East. Our analysis of the sun-photometer data showed that the Angstrom exponent (alpha) of aerosol light scattering from the sun-photometer at the Nauru ARM Site is averaged near zero between.440 and.675 micrometers (0.311 in 1999 and 0.132 in 2000), but much higher in 2002 (1.21) when the wind direction was not persistently from the East. The lower value of alpha implies clean background aerosol [Porch et al., 1973]. There is a relative lack of CCN when the wind blows continually from the East compared to conditions where mineral aerosol may be transported from Australia and Asia and Organic aerosol from phytoplankton rich regions around Indonesia. This lack of CCN affects the boundary layer cloud properties during the La Nina phase by decreasing overall cloudiness and increasing the persistence of clouds once formed.
4. Cloud Trails and the Southern Oscillation
 In Figure 3 we compare satellite images of Nauru and its surroundings on two days, one without a cloud trail (12 December 2000) and one with (13 December 2000). On 12 December (the day with no cloud trails) the area around Nauru is generally cloud free. There are only a few small clouds visible in the high-resolution MTI image (Figure 3). Subsidence on this day was strong enough to suppress nearly all convection in the region. On 13 December a cloud trail was present extending southwest from the island for approximately 100 km (Figure 3). Even though there is more large-scale cloudiness around Nauru on this day the cloud trail is well defined in both the larger GMS image as well as the high resolution MTI image. On this day the subsidence was just strong enough to allow a cloud trail to form but not so weak as to allow strong convection to dissipate the trail. An interesting feature visible in the MTI image and to some extent in the GMS image is the cloud free region along the side boundaries of the cloud trail.
 The potential temperature profiles from the daytime rawinsondes that were launched within an hour of when these images were taken are shown in Figure 4. The gap above 500 m is due to a relatively warm dry layer on 12 December (strong subsidence) that was replaced on 13 December with a cooler moist region more typical of the regional profiles. The warm dry layer on 12 December is likely a case where the subsidence was high enough to inhibit cloud formation. NCEP (National Center for Environmental Prediction) analysis show that Omega (vertical velocity positive downward) at 850 mb over a 1 degree grid over Nauru was positive 0.25 to 0.50 Pa/s during 2000 and −0.05 to −0.15 during 2002 implying subsidence during the La Nina period over Nauru. This is supported by a positive pressure difference between Nauru and the island of Manus 2200 km West of Nauru during 2000 (0.21 hPa) and a negative difference in 2002 (−0.51 hPa). Analysis of 1944 soundings from Manus and Nauru in 2000 and 2002 showed that the altitude of the minimum difference between the air temperature and the dew point was over 500 m lower in 2000 at Nauru than in 2002 which also support the influence of large scale subsidence.
 Satellite observations at night show that the island clouds detach from the island in the evening and dissipate. We could find no evidence of an island cloud trail persisting at night. The dissipation of island cloud trails at night is also consistent with an overall cooling effect of boundary layer clouds during the La Nina period.
 We examined approximately 1200 GMS satellite images from the La Nina year (2000), and the El Nino year (2002). During the El Nino phase of ENSO there were far fewer cloud trails observed (Figure 5). During the La Nina cloud trails were observed about 66 percent of the time (18 percent of the time the trails were less than 50 km in length, 48 percent greater) while during the El Nino they were observed only about 21 percent of the time (10 percent less than 50 km, 11 percent greater than 50 km). There does not appear to be any seasonal character to the cloud trail frequency (Figure 5).
 This inverse relationship between cloud frequency and cloud trail frequency during the El Nino – La Nina cycle suggests that subsidence at the margins of the TWP is affected by the ENSO cycle; that is, during La Nina overall subsidence is stronger creating better conditions for cloud trails but this subsidence suppresses convection sufficiently to decrease overall cloudiness.
5. SST and Cloudiness
 The SST near Nauru decreased slightly from 1990 to 1998 but generally stays within ∼1°C of about 29.5°C. There is then a sharp decline in SST leading into the 2000 La Nina followed by a SST increase to the El Nino of 2002 when temperatures are slightly greater than during the 1990 to 1998 period (Figure 6). The minimum temperature occurs in 1999 leading to an overall difference between the La Nina and El Nino period of nearly 3°C (Figures 1 and 6). Note that the data in Figure 6 are monthly average temperature that are smoothed with a three-month running mean filter while in Figure 1 the data are annual averages of the years 2000 (which does not actually include the minimum temperature in 1999) and 2002. The SST increase is also evident in the MTI inferred temperatures (Figure 6). It is worth noting that the La Nina – El Nino temperature difference would be greater if Nauru were located further to the East (Figure 1). Over the period of the SST increase there is also a notable increase in boundary layer cloudiness inferred from the MWR. The percent cloudiness increases from ∼15% during La Nina to ∼45% during El Nino (Figure 6). The correlation coefficient between the time-coincident cloudiness and SST data is 0.71, suggesting they are interlinked. There are likely feedbacks involving an increase in SST that leads to higher water vapor concentrations and increased convection and cloudiness. The increased cloudiness then helps to warm the sea surface.
6. Summary and Conclusions
 Large-scale subsidence at the margins of the convective center of the TWP is the major driver in maintaining island cloud trails for distances of hundreds of kilometers. Observations of island cloud formation frequency allow characterization of the subsidence in this region. Also, because of the high frequency of cloud trail occurrence during La Nina periods, the island of Nauru is an ideal location for testing the influence of aerosols (natural and seeded) on tropical cloud radiative properties and island cloud trail formation.
 The observed correlation between boundary layer cloudiness and SST is intriguing particularly given the feedbacks in the cloudiness/SST system. It is possible that increased convective activity in the Tropical Warm Pool region leads to increased subsidence outside of the Tropical Warm Pool. A connection between convection in the warm pool and subsidence outside the warm pool would be analogous to the Iris Hypothesis [Lin et al., 2002] where heating in one region leads to cooling over a larger region (only in this case, the controlling mechanism is boundary layer cloud properties rather than upper level clouds and water vapor). Determining the relative importance of these mechanisms and attributing cause and effect is not possible with these data sets but would require a combination of observational data and detailed model calculations. Future studies to address these issues are urgently needed.
 This work was funded by the Los Alamos National Laboratory's Directed Research and Development Project entitled “Resolving the Aerosol-Climate-Water Puzzle (20050014DR).” The authors thank S. Winiecki for her preliminary analysis, P. Minnis and L. Nguyen for their satellite analysis, and L. Jones and the ARM scientific community for support and helpful discussions.