We describe atmospheric temperature inversions and height-resolved water vapor fields over the wintertime subtropical northeastern Pacific Ocean in observations by the satellite-borne Atmospheric Infrared Sounder (AIRS) experiment. A comparison with model analyses shows good agreement in temperature. Water vapor comparisons with operational radiosondes at four sites in California and Hawaii during December 2002–January 2003 have low biases in the 1000–700 and 700–500 hPa layers. Maps of inversion frequency, and, water vapor at 1000–700 and 700–500 hPa over the subtropical northeast Pacific during 1–16 January 2003–when high pressure and clear conditions prevail–show inversions occurring at a local minimum in water vapor at 1000–700 hPa. Water vapor at 700–500 hPa has a broad minimum extending from Baja California to Hawaii, with inversions found on its eastern half. These observations illustrate the potential of the AIRS data for describing a climatology of temperature and water vapor in subtropical oceanic regions.
 Low level atmospheric temperature inversions prevail in subtropical ocean regions where cool eastern boundary currents are overlain by warm air associated with high pressure systems. A cool, humid layer typically one kilometer or less in thickness can cover significant areas of subtropical eastern oceans. Stratocumulus clouds commonly occur within the inversion layer, with a nighttime cloudiness increase due to radiative cooling to space through the very dry intervening atmosphere. Several interesting dynamical features, including bores, Kelvin waves, internal gravity waves, and interactions between the inversion layers and synoptic features have been observed in these shallow, dense layers [Ralph et al., 2000; Mass and Steenburgh, 2000]. The high albedo of stratocumulus clouds makes them an important component of the planetary radiative balance [Klein and Hartmann, 1993]. However, details of the convective and radiative processes leading to stratocumulus formation are poorly represented in numerical models. Consequently, a realistic representation of stratocumulus boundary layers in global models has been a major challenge in climate and weather prediction [Duynkerke and Teixeira, 2001].
 We present here the thermal and humidity structure of the atmosphere where temperature inversions are detected under predominantly clear conditions. This study complements many earlier studies of shallow convection describing its most obvious manifestation: clouds. Because stratocumulus clouds are readily observed from satellite in the visible and infrared, detailed climatologies are available [e.g., Rossow and Schiffer, 1991]. Also, several field campaigns have observed the heavy cloud cover associated with temperature inversions [see Stevens et al., 2003 and references therein].
2. Instruments, Retrieval System and Analyses
 The Atmospheric Infrared Sounder (AIRS) experiment on the EOS-Aqua platform, launched in May 2002, includes the AIRS infrared spectrometer and companion Advanced Microwave Sounding Unit and Humidity Sounder for Brazil (see Aumann et al.  and associated articles). AIRS shares its input optics with a four-channel Visible/Near Infrared (Vis/NIR) imager. The combination of infrared and microwave radiances allows retrieval of high resolution temperature and humidity profiles for infrared cloud fraction (the product of emissivity and coverage) up to about 70% [Susskind et al., 2003]. Information content is limited for higher cloud fraction. The AIRS retrievals have nominal 45 km horizontal spacing, and an orbit track thirty retrievals across. Standard retrieved products include surface temperature, infrared cloud fraction, cloud top temperatures and pressures, and profiles of temperature and water vapor. The vertical resolution of the AIRS system is specified as 1 km for temperature and 2 km for water vapor in the troposphere. The actual resolution may better than this, and results shown here suggest sensitivity to sub-kilometer temperature structure under some conditions. Retrievals are being validated against a variety of in situ and model-generated data [Fetzer et al., 2003]; results therein are broadly consistent with ours. In this study we use a criterion of agreement with the Real Time Global sea surface temperature of ±3 K described by Fetzer et al. . Retrievals version is 3.1.9, similar to publicly available v3.0.
 Resolving near-surface structure is possible with AIRS because its retrieval algorithm deconvolves spectrally adjacent lines whose weighting function peaks are spaced closely in the vertical, although individual weighting function scales are larger than the inversion depth [Susskind et al., 2003]. Roughly twenty channels peaking near the surface [Fishbein et al., 2003b] provide about three pieces of information each for temperature and water vapor below 700 hPa. In this study we consider AIRS data reported at 1000, 925, 850, 700, 600 and 500 hPa; higher vertical resolution profiles contain little additional information. An examination of AIRS retrievals for January 2003 reveals regions of near-surface temperature inversions, predominantly on the eastern edges of ocean basins: off California and South America in the Pacific, off North and South Africa in the Atlantic, and off Western and Southern Australia in the Indian Ocean. This is consistent with climatologies of marine temperature inversions and associated stratus clouds [Klein and Hartmann, 1993; Norris, 1998].
 Despite the plausibility of this distribution, we consider here a limited area off southern and Baja California in January 2003. We examined a variety of observations to establish the validity of these results. Maps of AIRS retrieval type and retrieved quantities suggest prevailing high pressure, and attendant clearer skies, conducive to both inversion formation and high AIRS information content. These conditions were corroborated by an examination of geosynchronous imagery, that also shows a series of synoptic storms affecting the Pacific Northwest. We also examined European Center for Medium-range Weather Forecasts (ECMWF) model analyses for January 3 and 4, and their differences with AIRS. Finally, we compared AIRS retrievals with operational rawinsondes from Quillayute WA, Oakland CA, Vandenberg CA, San Diego CA, Hilo, HI and Lihue, HI.
3. Case Study
Figure 1a shows part of an observing swath from the AIRS experiment off Southern and Baja California on 3 January 2003. Locations with retrieved temperature inversions are indicated as inversion strength in false color. Figure 1b shows the simultaneous Vis/NIR image. Retrieved inversions are seen predominantly in the cloud-free regions in the Vis/NIR image. Nearly cloud-free conditions in the southern half of Figure 1 are confirmed by agreement between retrieved infrared cloud fraction and separately inferred Vis/NIR cloud fraction, to within the estimated uncertainties of both quantities. All inversion tops in Figure 1 are at or below 850 hPa, with a rapid return to more nearly adiabatic profiles above. The streak of cloud to the north of Point Conception has top heights of 5–10 km, and is associated with a weak frontal system. Locations without black crosses in Figure 1a indicate fully converged retrievals, implying infrared cloud fractions less than about 70%. Note that AIRS detects some inversions off central California through this high cloud layer.
Figure 2a shows the retrieved temperature profile averaged over all points in Figure 1a, and the strongest and most isothermal inversions circled in Figure 1a. Figure 2b shows the profiles of differences between those in Figure 2a and the analogous profiles from the ECMWF model [Fishbein et al., 2003a]. As shown by Teixeira , temperatures in the dry boundary layer are well represented in the ECMWF model, and the mean difference profile in Figure 2b is less than 0.5 K; the extrema show local variations between AIRS and model. The rawinsonde launched from San Diego on 4 January at 0Z did not show an inversion but instead a nearly adiabatic profile, weak low-level easterly winds, and low relative humidity. These factors, and regionally high temperatures at 700 hPa, indicate mild Santa Ana conditions with prevailing high pressure. Figure 2c shows average differences for water vapor. Note the very good agreement in the 1000–700 hPa layer (−1.4% bias), but differences of 25–30% at higher resolution.
4. Humidity Comparisons With Rawinsondes
 AIRS retrieved water vapor profiles are compared against operational rawinsondes at San Diego, CA, Vandenberg, CA, Hilo, HI and Lihue, HI during December 2002 and January 2003. Table 1 shows the AIRS mean water vapor in the 1000–700 hPa layer, and differences with operational rawinsondes. Note that biases in Table 1 are a few percent or less. We use 2 km layer averages in this comparison for several reasons: this is the system specification, operational sondes are launched from coastal and island sites where the boundary layer may be different from purely oceanic conditions, and, the differences for higher resolution in Figure 2c could be due to either AIRS or the model. Retrievals in this comparison are confined to ocean-only within 200 km of the sonde sites; results are similar using a 100 km limit. The sun-synchronous Aqua orbit overpasses lead sonde launches by 2 hours in San Diego, while near-synchrony holds in Hawaii.
Table 1. Summary of AIRS-radiosonde comparisons at four locations in California and Hawaii for the period December 2002 and January 2003 in the 1000–00 hPa layer
AIRS Precipitable Water Vapor
Bias and Standard Error
−4.9 ± 24%
3.1 ± 19%
−0.3 ± 14%
2.7 ± 14%
 Correlative information about the 700–500 hPa layer over the northeast Pacific is limited to Vaisala sondes launched at Lihue. (The 700–500 hPa bias corresponding to Figure 2c is −8.4%.) All other sites in Table 1 launch Sippican sondes, with wet biases under cold or dry conditions [da Silviera et al., 2003]. The AIRS mean 700–500 hPa precipitable water at Lihue is 1.95 mm, and the difference is −4.6% ± 44%. Though this uncertainty is higher than the globally average values of 0.0 ± 26% against Vaisala sondes reported by Fetzer et al. , it is not surprising. The high dynamic range at Hawaii at 700–500 hPa (from 0.3 to about 10 mm in our study, due primarily to large changes in the depth of the moist boundary layer) can lead to higher standard errors without affecting the bias. Dedicated sondes with high vertical resolution launched in synchrony with AIRS overpasses at other locations show much better agreement. These considerations suggest that the ±44% random error in water vapor at 700–500 hPa is an upper bound to the true uncertainty. The low biases between AIRS and Vaisala sondes shows that AIRS may be useful as a global standard to reconcile observations from dissimilar sonde humidity sensors.
5. Inversion Frequency and Associated Water Vapor
Figure 3a gives the average inversion occurrence frequency over ocean at 10–35N and 105–160W during 1–16 January 2003. Figure 1 epitomizes the synoptic conditions during this period: high pressure and clear conditions to the south of about 30 N with a series of storms to the north. Inversions in Figure 3a occur as frequently as half of the time adjacent to the coast, with a local maximum off Baja California. Figures 3b and 3c show the associated height-resolved precipitable water vapor fields. The broad maximum in water vapor at 700–500 hPa extending from north of Hawaii to northern California in Figure 3c is from zonal transport by midlatitude storms manifested as sporadic, highly localized water vapor structures. The maximum to the southeast in Figure 3c is decoupled from the secondary maximum to the west of central California. Inversions are more frequent in Figure 3a south of the storm track, and are associated with low water vapor amounts of about 10 mm at 1000–700 hPa (Figure 3b), and 1–2 mm at 700–500 hPa (Figure 3c). No inversions are detected to the west of about 140 W, shown dashed in Figure 3a. Figures 3a and 3b are consistent with a well-accepted view of the subtropics. Off California the boundary layer is well defined, shallow, and often has a temperature inversion. This layer transitions to a weaker inversion capping a deeper moist layer in the vicinity of Hawaii. A realistic doubling of the boundary layer height from California to Hawaii can explain the distribution of the values shown in Figure 3b.
 The collocation of inversion frequency and near surface water vapor minimum is apparent in Figures 3a and 3b. The relationship between inversions and 700–500 hPa water vapor is not as obvious. A broad minimum in Figure 3c extends from Hawaii to Baja California. An examination of individual AIRS overpasses shows this to be a persistent, largely stationary feature of the region during the first half of January 2003. Inversions in Figure 3a are most frequent at the eastern boundary of this minimum. Note, however, that inversions are very infrequent near the minima in Figure 3c of less than 1 mm. The significant increase in humidity at the southern edge of Figure 3c may be associated with outflow from the Intertropical Convergence Zone, and can be seen to move day-to-day in the AIRS observations.
6. Summary and Conclusions
 We present a case study of remotely sounded atmospheric boundary layer temperature inversions off Baja California in January 2003 under prevailing local high pressure and clear skies. A comparison with ECMWF shows agreement in near-surface temperature at scales less than about 1 km, and in water vapor averaged in 2 km layers. A comparison of AIRS retrievals and rawinsondes launched from four sites in southern California and Hawaii show small biases in the 1000–700 hPa layer at all sites. Biases are also small at in the 700–500 hPa layer over Lihue, HI, which launched sondes with better sensitivity at lower relative humidities. Maps of inversion occurrence frequency show inversions confined to the eastern half of the area examined, associated with a local minimum in moisture at 1000–700 hPa. The relationship between water vapor at 700–500 hPa and inversion frequency is less definite, with humidity in this layer showing a broad minimum between Hawaii and Baja California.
 These results show the potential of AIRS observations for resolving near-surface temperatures and associated humidity. Two issues complicate the creation of an inversion climatology with AIRS. First is the sensitivity of combined infrared and nadir sounding as the infrared cloud fraction threshold of 70% is exceeded. Second is the resolution of fine-scale vertical structure. Note that both these issues are fundamental to the nadir sounding methodology employed by AIRS, regardless of the phenomenon of interest. These issues are being addressed through retrieval algorithm improvements, including more realistic retrieved error estimation as a function of increasing infrared cloud fraction up to and past 70%. Also, comparisons with in situ observations–such as from field campaigns in oceanic inversion regions, in the case of this study– are needed to validate these algorithm improvements. In addition, the information content of AIRS retrievals can be better understood through numerical simulations. For example, simulations will help resolve the discrepancy between AIRS and ECMWF water vapor shown in Figure 2c. Since AIRS takes more than 300,000 profiles daily the potential benefits of these effort are significant. As we address issues of resolution and sensitivity described above, we can also examine in detail the wide variety of phenomena seen in the AIRS observations.
 George Aumann, Sung-Yung Lee, Moustafa Chahine, and Joel Susskind all provided valuable guidance in this study. Mike Gunson, Brian Kahn, and especially Barney Farmer, made many helpful comments about the manuscript. The comments of an anonymous reviewer were very useful. We utilized radiosondes available online from the University of Wyoming. This is work was performed at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA.