Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 Over water, infrared radiometers on satellites measure radiation leaving from the surface skin layer and therefore the retrieved temperature is representative of the skin layer. This is slightly different from the bulk layer deeper in the water where various floating thermometers take temperature measurements to validate satellite measurements. The difference between the bulk and skin temperature (skin effect) must be understood to properly validate schemes that use surface skin temperature to infer bulk temperatures. Further skin temperatures retrieved over inland waters may show different patterns to those retrieved over oceans due to differences in conditions such as wind speed, aerosols, and elevation. We have analyzed the differences between the skin and bulk temperatures at four permanent monitoring stations (buoys) located on Lake Tahoe since 1999 and compared the results with similar studies over the ocean typically obtained from boat cruises. Skin effect distributions were found to be consistent across the buoys; however, the diurnal behavior of the skin effect was slightly different and shown to be related to wind speed measured at an individual buoy. When wind speed was less than 2 m s−1, the skin temperature osclillated and greatly increased the uncertainty in the skin effect reported over Lake Tahoe. When downwelling sky radiation was increased from clouds or high humidity, this led to nighttime skin temperatures that were warmer than bulk temperatures by as much as 0.5 K. The size of the warm skin effect is larger than other ocean studies that observed warm nighttime skin values around 0.1 K. The nighttime skin effect was seen to be more consistent with a smaller standard deviation compared to the daytime skin effect. The nighttime skin behavior had a mean and standard deviation that ranged between 0.3 and 0.5 K and between 0.3 and 0.4 K, respectively. In contrast, daytime skin effect was strongly influenced by direct solar illumination and typically had a mean of 0.5 K in the morning that decreased to 0.1 K by midday. The standard deviation of the daytime skin effect ranged from 0.3 in the morning to 0.8 by midday. As the solar heating reduces later in the day the skin effect increases to a 0.3 K mean with a standard deviation of 0.4 K. The results for Lake Tahoe clearly demonstrate that validating satellite-derived skin measurements or merging multiple satellites data sets together would be most successful when using nighttime data at wind speeds greater than 2 m s−1 with greater uncertainties expected when using daytime measurements. Further, the assumptions used for the skin effect behavior over oceans may not be appropriate over lakes because of the greater range of environmental conditions that affect lakes.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Thermal infrared (TIR) satellite data are invaluable for studying climate change by routinely measuring sea surface temperatures (SST) and inland water surface temperatures (IWST) [Donlon et al., 2002; Hook et al., 2007; Schneider et al., 2009]. They provide high spatial resolution, routine observations, with an accuracy on the order of 0.1–0.5 K, allowing them to be used in climate change studies of oceans and lakes on the global scale [Schneider and Hook, 2010]. Surface temperature is a critical parameter for both oceans and lakes in understanding the exchange of heat, momentum, and trace gases with the atmosphere. Surface temperatures are commonly classified as skin or bulk temperatures depending on their effective measurement depth [Jessup and Branch, 2008; Minnett et al., 2011]. Satellite infrared sensors and radiometers aboard ships or buoys measure the radiation from the surface skin layer of water [Donlon et al., 2002; Schluessel et al., 1990]. The thickness of the skin layer is less than a millimeter and is dependent upon the local energy flux through the water surface [Schluessel et al., 1990]. At just a few centimeters of depth, the bulk temperature is consistently warmer than the skin by a few tenths of a Kelvin [Saunders, 1967]. At depths of a meter or more a bulk temperature measurement can differ from the skin by a few degrees depending on the diurnal thermocline [Minnett, 2003]. This difference between the skin and bulk temperature is commonly known as the skin effect, cool skin effect, or cool skin [Fairall et al., 1996; Minnett et al., 2011]. We point out that some studies classify surface temperatures as skin, subskin, and depth measurements. Here the subskin is a laminar sublayer below the skin, and the depth measurement is similar to the bulk measurement with its depth specified [Donlon et al., 2007]. Knowing the depth of the bulk measurement is beneficial when accounting for the uneven vertical heating due to sunlight. However, in this study we follow [Minnett et al., 2011] and define the skin effect as the bulk minus skin temperature because our bulk measurements vary between 2 and 5 cm below the surface.
 Understanding the physical processes that influence the skin effect over inland waters and being able to confidently parameterize its magnitude is critical for several reasons: to properly validate satellite skin temperatures with in situ bulk temperatures, to determine intersensor biases for satellite data used in merged data sets, for converting skin temperatures to bulk temperatures to study physical lake processes, and parameterizing surface exchange processes through converting bulk temperatures to skin temperatures. For example, in a recent study skin temperatures from Along-Track Scanning Radiometer (ATSR) were merged with bulk temperatures from the advanced very high resolution radiometer series of instruments flown on NOAA satellites to look at nighttime global lake temperature trends from 1985 to 2009 [Schneider and Hook, 2010]. A skin effect value of 0.1 K was used to convert the ATSR skin temperatures to bulk temperatures to create a merged global inland water data set. Comparisons between in situ buoy bulk temperature measurements and the satellite-derived bulk temperatures in this study showed good agreement and were similar to observations over the oceans that found a nighttime effect of 0.17 K [Donlon et al., 2002]. However, the size, depth, and environmental conditions surrounding inland waters are very heterogeneous, and skin effect values will vary with different conditions. For example, a recent study over a volcanic lake found a skin effect around 1.5 K [Oppenheimer, 1997]. The study also demonstrated that not accounting for the skin effect can lead to approximately a 10% overestimation of evaporative and sensible heat fluxes.
 Accurately estimating climate change from SST measurements requires long and temporally complete time series of in situ data for use in validating the satellite SSTs and the merging of complimentary satellite data sets [Donlon et al., 2002]. Similarly, estimating IWST trends requires spatially and temporally robust geographic sampling of in situ sites that represent the range of conditions observed. The ideal in situ measurement to validate a satellite SST or IWST is a TIR radiometer capable of making highly accurate measurements of the skin temperature [Minnett et al., 2011]. Over the oceans in situ TIR measurements of skin temperature are typically made from ships during extended campaigns [Donlon and Nightingale, 2000; Horrocks et al., 2003; Minnett et al., 2011]. Currently, because radiometric in situ measurements are limited both geographically and temporally, they are unable to provide a global validation data set [Donlon et al., 2002; Kilpatrick et al., 2001]. As a result satellite-derived SST measurements are validated against bulk temperatures made from buoys that are far more ubiquitous [Donlon et al., 2002; Kilpatrick et al., 2001]. Other than at Lake Tahoe and more recently the Salton Sea (from 2008), long regular time series of in situ IWSTs exist only as bulk temperatures and mostly reside in North America and Europe [Schneider and Hook, 2010].
 While numerous studies have analyzed the skin effect over the oceans [Minnett, 2003; Minnett et al., 2011; Paulson and Simpson, 1981; Saunders, 1967; Schluessel et al., 1990], far fewer studies have investigated the skin effect over inland waters. The characteristics of inland waters are more heterogeneous and dynamic than ocean waters since they exist in a wider range of environmental conditions. For example, inland water have a wide range of salinity (and other dissolved substances) and typically do not have the swell and large waves found in the ocean [Kurzeja et al., 2005] and can have a wide annual range in surface temperature and atmospheric conditions compared with oceans. Moreover, the atmospheric boundary layer above inland waters is more variable than over the oceans, due to the proximity of boundaries, leading to higher uncertainties in the heat and momentum fluxes [Mahrt et al., 1998]. Clarifying similarities and differences between the skin effect of the oceans and inland waters should improve future inland water validation campaigns that utilize in situ bulk temperatures due to their greater availability.
 This study analyzes in situ observations of radiometric skin temperature and bulk temperature at Lake Tahoe, California/Nevada, to determine the skin effect over a large, high-altitude freshwater lake. The in situ measurements at Lake Tahoe provide the most comprehensive set of measurements for any inland body of water [Hulley et al., 2011]. Typical skin effect ocean studies last for a few weeks onboard cruise ships where corrections for ship effects to wind measurements must be applied. A recent ocean study used six ocean cruises to examine approximately 2900 skin and bulk measurements [Donlon et al., 2002]. In contrast this study examines data from four buoys measuring skin and bulk temperatures at 5 min intervals continuously since 1999, leading to approximately 300,000 data points per buoy. Colocated meteorological data are also available from each buoy starting in 2001. We present the annual and diurnal distributions for the Lake Tahoe skin effect. Histograms and wind speed distributions for the skin effect are compared to previous ocean studies in order to examine the effect of low speeds on the skin effect and to compare the nighttime negative skin effect between Lake Tahoe and the ocean. Relationships are developed linking nighttime skin effect to longwave radiation. In the context of these results recommendations for validation campaigns and methods to construct merged satellite data sets for global trend analysis are discussed.
2 Background on Skin Effect
 The skin layer of the water surface, roughly 10–500 µm in thickness [Donlon et al., 2007; Donlon et al., 2002; Fairall et al., 1996], is regulated by conductive and diffusive heat processes [Saunders, 1967]. As illustrated in Figure 1a, a temperature gradient is formed across the skin layer, the magnitude of which is determined by the heat flux at the water-atmosphere interface [Donlon et al., 2002]. Skin temperature (Ts) measurements are made by infrared radiometers onboard ships, floating on buoys, and flown on satellites [Donlon et al., 2002; Hook et al., 2003]. Measurements of water temperature below the skin are made at a few centimeters by floating thermistors or at a few meters by ship intakes and floating buoys [Fairall et al., 1996]. In this layer the bulk temperature (Tb) is determined by turbulent heat transfer processes and solar heating [Donlon et al., 2002]. As illustrated in Figure 1b, a warm layer is created during the daytime as solar heating warms the upper 2–20 m of the water and suppresses shear-induced mixing [Fairall et al., 1996].
 In the absence of solar radiation (i.e., at night), the magnitude of the skin effect is determined by the sum of sensible and latent heat combined with the net longwave radiation flux between the water surface and atmosphere. Typically, the heat flux is from the water to the air and causes a temperature depression in the skin in the range of 0.1–0.5 K [Fairall et al., 1996]. The skin effect is defined as the difference between the bulk and skin temperature:
 Numerous studies have attempted to model the skin effect over the oceans [Fairall et al., 1996; Saunders, 1967; Zeng et al., 1999]. A simple physical model quantifying the bulk skin temperature difference ΔT was first proposed by Saunders 
where k is the thermal conductivity of water, Q is the net heat flux out of the water, and δ is the thickness of the molecular sublayer that defines the thickness of the cool skin gradient. The thickness of the molecular sublayer was given in Saunders  as
where λ is an empirical coefficient known as the Saunders constant [Saunders, 1967], ν is the kinematic viscosity of water, and u* is the friction velocity at the water surface. Combining equations ((2)) and ((3)) yields the following relation for the skin effect
Saunders  originally estimated λ to be between 5 and 10. However, the data set used was limited to a limited skin effect that ranged between 0.2 and 0.3 K. Then Grassl  took observations from a tropical Atlantic region and suggested that λ is a function of wind speed. This study found that λ varied between 2.2 and 5.5 for wind speeds ranging between 1 and 10 m s−1. Aboard a German cruise in the northeast Atlantic between October and November, Schluessel et al.  found that λ ranged between 1.1 and 8.4 for wind speeds between 1 and 11 m s−1. The colder and drier atmosphere from Schluessel et al.  resulted in larger skin effect values compared with those of Grassl . A more recent study by Fairall et al.  estimated that λ should range from 4 to 8 and is consistent with the typical nighttime cool skin values of 0.3 K. It has been shown that constant or linearly regressed values for λ cannot adequately model the skin effect [Coppin et al., 1991; Schluessel et al., 1990], but parameterizations for the skin effect based solely on wind speed do exist [Donlon et al., 2002; Minnett et al., 2011].
3.1 Lake Tahoe Study Site
 The following description of the Lake Tahoe site is summarized from Hook et al. [2003, 2007]. Lake Tahoe is a large freshwater lake located in the eastern Sierra Nevada along the California and Nevada border. Due to Lake Tahoe's extreme depth (maximum depth is 501 m) and associated high thermal capacity, it never freezes in winter. The lake level is located at 1898 m above sea level and has a surface area of 496 km2. Figure 2 shows the location of Lake Tahoe and its bathymetry.
3.2 Buoy Data
 There are four permanently moored buoys on Lake Tahoe operated by the Jet Propulsion Laboratory in conjunction with University of California (UC) Davis that have provided simultaneous measurements of skin and bulk temperatures since 1999. Beginning in 2001, equipment was installed to measure standard meteorological variables. The buoys define a square centered on the lake with each buoy being approximately 5 km from its nearest neighbors and 3–5 km from the shore. The buoys are all moored over deep water (> 450 m) as shown in Figure 2.
 Each buoy is equipped with a custom-built, self-calibrating radiometer that measures skin temperature with an uncertainty less than ±0.1 K [Hook et al., 2007]. Each radiometric measurement is converted to a skin temperature by subtracting the reflected sky radiance and correcting for emissivity effects [Hook et al., 2007]. Reflected downwelling sky radiance is calculated by a radiative transfer model driven by National Centers for Environmental Prediction reanalysis profiles [Kalnay et al., 1996]. Uncertainty from this method of atmospheric correction was found to be less than 0.01 K [Hook et al., 2007].
 Attached to each buoy is a ring-float which measures the bulk temperature at approximately 2–5 cm below the surface [Hook et al., 2003]. There are also thermistor chains that measured the bulk temperature at multiple depths beneath each buoy. The bulk temperature measurements are calibrated against a National Institute of Standards and Technology-certified water bath and typically have errors less than ±0.1 K [Hook et al., 2003]. Combining the accuracy of the bulk and skin measurements suggests that the best estimate of ΔT is less than ±0.2 K. Figure 3 displays the floating buoy along with all the equipment attached to it.
 Meteorological data including wind speed and direction, air temperature, relative humidity, and atmospheric pressure are recorded at 5 min intervals. Wind speed is measured at heights of 3.6 m (TB1), 3.75 m (TB2), and 3.17 m (TB3 and TB4). Sunrise and sunset times are calculated using equations from Doggett et al. . Every observation is classified as night, daybreak, day, and twilight. Table 1 describes how these time periods are defined.
Table 1. Standard Definitions of Time Periods Used in This Study
Sunset + 2 h to sunrise − 1 h
Sunrise − 1 h to sunrise + 2 h
Sunrise + 2 h to sunset − 1 h
Sunset − 1 h to sunset + 2 h
3.3 Radiation Station
 UC Davis maintains a meteorological station at the United States Coast Guard base near Tahoe City on the northwest side of Lake Tahoe (UCMS in Figure 2) approximately 8 km to the northwest of TB4. This study uses radiation data (longwave and shortwave radiation up and down) from 2011 to correlate with the skin effect. The full radiation station includes the CNR1 net radiometer package, the CM21 Kipp and Zonen (KZ) pyranometer measuring downwelling solar radiation, and the Eppley pyrgeometer measuring downwelling terrestrial radiation. The Eppley pyrgeometer and the CM21 KZ pyranometer are used in this study because shadowing from a large radio antenna affects the CNR1 net radiometer package. During daytime measurements longwave measurements are correlated to the downwelling shortwave radiation. Following [Michel et al., 2008], 1.5% of the shortwave radiation is subtracted from the longwave measurement to decorrelate the longwave measurements from the shortwave. Radiation measurements that are taken when the temperature is within 90% of the dewpoint temperature are removed from the data set. This assures dew buildup has not interfered with the longwave measurement.
4 Observations and Results: Skin Effect at Lake Tahoe
 The skin effect, calculated using equation ((1)), is similar at all the buoys and has a unimodal distribution that is primarily contained between 0 and 1 K (Figure 4). Daytime distributions have a larger scatter than nighttime distribution due to solar heating. The median nighttime distribution has a larger skin effect than during daytime because of daytime solar heating which reduces the skin effect, especially on calm days when the skin warms more rapidly than the bulk temperature [Hook et al., 2003]. All the distributions show significant populations of negative skin effects. This is very different from the ocean where negative skin effects are primarily seen during the day [Donlon et al., 2002; Minnett et al., 2011]. Daybreak skin effect values are slightly higher than the nighttime values, while the twilight values are slightly lower. Daytime skin effect distributions have tails that are largely negative and extend well beyond −2 K. This is more in agreement with the results from Donlon et al.  rather than Minnett et al. . Neither daybreak nor twilight distributions show the long negative tail of the daytime distribution. As shown in the red curve of Figure 1, daytime solar heating stratifies the temperature profile of the surface. With this in mind the negative skin effect results from two separate processes: (1) intense daytime solar heating overcomes the net upward longwave energy flux and warms the skin, or (2) the right combination of low wind and solar heating creates a warm layer of water above the floating thermistor.
 The results of Figure 4 are summarized in Table 2, which displays the statistics for all the buoys at Lake Tahoe. The results are divided into four categories: night, daybreak, day, and twilight, as described in Table 1. Median nighttime and daytime skin effect values range from 0.34 to 0.46 K and 0.24 to 0.34 K, respectively. Nighttime median skin effect values in this study are consistent with the 0.3 K value reported in Fairall et al. . Table 2 indicates that the skin effect standard deviations for nighttime and daytime measurements range from 0.32 to 0.36 K and 0.41 to 0.64 K, respectively. Nighttime distributions for the skin effect are positive within one standard deviation of the mean for three of the four buoys (TB1, TB3, and Tb4); Tb2 is slightly negative one standard deviation away from the mean. Of note is that TB2 and TB4 have lower nighttime skin effect values compared with TB1. Mean and median skin effect values are higher for daybreak compared with their associated nighttime values. This indicates that the skin effect has its maximum right after sunrise. Table 2 also indicates that daytime skin effect values are several hundred millikelvin negative one standard deviation away from the mean. This behavior persists into twilight, but the negative skin effect is not nearly as pronounced.
Table 2. The Summary Stats for the Skin Effect Calculated With Equation ((1)) for Lake Tahoea
aDay is defined to be 2 h after sunrise till 1 h before sunrise. Night is defined to be 2 h after sunset till 1 h before sunrise. Twilight is defined to be 1 h before sunset till 2 h after sunset. Daybreak is defined to be 1 h before sunrise till 2 h after sunrise. All is defined as every observation for the buoy where a skin and bulk measurement took place.
Tahoe Buoy # 1
Tahoe Buoy # 2
Tahoe Buoy # 3
Tahoe Buoy # 4
 Diurnal variation of the skin effect is presented in Figures 5 and 6. Figure 5 displays the skin effect as a function of time after sunset that continues to sunrise; points later than 13 h only occur during the winter. The solid trace in each figure is the mean skin effect in bins of 10 min, while the standard deviation of the skin effect is presented as a dashed line. Buoys 1, 2, and 4 show average skin effect drops a few millikelvin immediately after sunset. This drop in skin effect only lasts for about 30 min. Buoy 3 does not experience the same drop in temperature immediately after sunrise; rather it holds a steady value of around 0.28 K. After 1 h the skin temperature drops faster than the bulk temperature, and the skin effect increases over time for all buoys. Toward the end of the night the average skin effect decreases between 0.10 and 0.50 K across all four buoys. The drop in average skin effect begins around 9 h after sunset for TB2 and TB3, while the drop occurs much later for TB1 and TB4, starting around 11–12 h past sunset.
 The standard deviation (SD) of the skin effect (dashed) for each buoy is presented in each subfigure of Figure 5 to describe the variance across the average skin effect. The skin effect SD drops 0.05 to 0.10 K after sunset over approximately 1 h. The size of the drop over this time is nearly the same magnitude as the drop in average skin effect for the same time. SD remains relatively constant for all buoys over the next 2–5 h and then increases as the night goes on.
 Figure 6 shows the daytime diurnal variations of skin effect as a function of time since sunrise. Immediately after sunrise the skin effect is at its maximum for the daytime, decreases for an hour, and then increases approximately to its original value. The average skin effect remains constant for 2.5 to 3 h until the sun gets high enough in the sky to heat the skin. At this point the skin warms up faster than the bulk, and the skin effect drops many tenths of a degree. For each buoy, the skin effect reaches its minimum at midday and ranges between 0.05 and 0.23 K. TB1 experiences the largest drop of 0.40 K, and TB2 decreases the least at 0.18 K.
 The dashed line of each subfigure in Figure 6 shows the daytime SD of the skin effect. The SD of the skin effect for all buoys steadily rises after sunrise until it reaches its maximum around 7 h past sunrise. The maximum SD ranges between 0.8 (TB1) and 0.5 K (TB2) and occurs midday. The standard deviations are much larger than the average at midday and results in many instances where the skin is warmer than the bulk temperature. As the sun begins to set later in the day all buoys show an increasing average skin effect indicating the skin is cooling faster than the bulk. Standard deviation in all buoys lowers as the day moves on. Compared with nighttime skin effect SD, the daytime skin effect is much larger.
 Equation ((4)) demonstrates that the skin effect is inversely related to wind speed. Therefore, higher wind speeds correlate to lower skin effects. To understand the diurnal skin effect shown in Figures 5 and 6, we plot each buoy's average and SD wind speed in Figures 7 and 8. All the skin and bulk measurements are matched up to wind speed data within 5 min. The wind speed is linearly interpolated in time to match the time of the skin and bulk measurements. Then the skin effect is calculated using equation ((1)) and placed into 10 min bins for average and SD calculations.
 Figure 7 shows that immediately after sunset the wind speed decreases between 1 and 1.5 m s−1 depending on buoy location. The decrease in average speed occurs over a course of a few hours, and then a period of relatively constant wind speed persists until around 10 h past sunset. The period of constant low wind speed allows for little mixing and helps explain the average skin effect increases shown in Figure 5. (We define low wind speed as less than 3 m s−1 because a similar study suggests this is approximately the speed for ubiquitous wave formation [Donlon et al., 2002].) Also, the drop in average skin effect shown in Figure 5 begins around the 10 h mark and coincides with the increase in wind speed shown in Figure 7. All four buoys show an increase in wind speed between 2.5 and 4 m s−1 after the 10 h mark. TB4 shows the latest rise in wind speed, occurring around 12 h into the night and rapidly rises past 3 m s−1. Wind speed begins increasing earliest at TB2, where the wind speed increases beginning around the 8 h mark. One possible explanation for the late night wind speed increase is a terrain-driven valley flow that has strengthened enough to penetrate the nighttime temperature inversion down to the lake surface. The late night wind speed increase is likely due to the sun beginning to heat the nearby land around sunrise, creating a thermal gradient between the lake and the land, and setting up a local lake breeze. The easternmost buoys (TB1 and TB2) experience the late night increase in wind speed around 3 h earlier than the western buoys. It is important to point out that these average points occur across the whole year, where the length of the night varies. This means that points later than 12 h past sunset have fewer points and are primarily wintertime measurements.
 Figure 8 shows the same measurements from Figure 7, but daytime values are used instead of the nighttime values. All four buoys' wind speeds decrease slightly for the first 2.5 h after sunrise. This decrease is on the order of 0.5–1.0 m s−1. Around 5 h after sunrise the wind speed begins to increase and rises to 2–3 m s−1 over the next 6 h. Depending on the time of year, local noon is around 5–6 h after sunrise. It is around this time of day that the lake breeze begins and combines with the synoptic flow that has mixed to the surface once the nighttime temperature inversion has mixed out. Unlike the nighttime measurements, the decrease in daytime skin effect occurs before the increase in wind speed. This comes from the sun heating the skin faster than the bulk and reducing the skin effect. Around noon, the wind picks up and mixes the skin and bulk layers, keeping the skin effect between 0.05 and 0.25 K until solar intensity decreases later in the day. The SD of the wind speed stays relatively constant from 5 to 10 h after sunrise. Like the magnitude in the SD of the skin effect, the wind speed SD is about the same magnitude as the average wind speed. Combining the SD of the wind speed, typical wind speed values during the daytime range from 0 to 8 m s−1.
 To better understand the relationship between skin effect and wind speed the average skin effect for all buoys versus wind speed is plotted in the top of Figures 9 (night) and 10 (daytime), respectively. In these plots skin and bulk measurements were matched with a wind measurement when they were within 5 min of each other. Data from all buoys were put together and binned in 0.5 m s−1 increments. The top of Figures 9 and 10 are the average skin effect for night and day, respectively, while the bottom of Figures 9 and 10 are SD. Nighttime average skin effect is at its maximum at 0 m s−1. Average skin effect decreases until approximately 4 m s−1, where it reaches a constant value of approximately 0.2 K. Both the critical value of 4 m s−1 and the threshold skin effect value of 0.2 K are similar to values reported in [Donlon et al., 2002]. As stated in [Donlon et al., 2002], this is the speed where wave breaking begins to mix skin and bulk layers together and reduces skin effect. At 0–2 m s−1, the SD is at its maximum. At greater than 4 m s−1, the skin effect SD is approximately 0.25 K. Therefore, at night many values of the skin effect are negative in this range. Figure 11 shows a scatter plot that compares the skin effect to measured downwelling longwave radiation. This relationship shows that higher downwelling longwave radiation corresponds to lower and in some cases negative skin effect values. In order for equation ((2)) to be negative, then Q (sum of latent, longwave, and sensible heating) must be negative. Therefore, the longwave radiation is enough to overcome the latent and sensible heating fluxes to the atmosphere.
 An analysis of time series of skin effect has shown that the SD is a result of two different processes. In one case, the environmental factors that control the energy balance of the atmosphere lake exchange control much of the variation in skin effect greater at wind speeds greater than 3 m s−1. However, when sufficiently long periods with wind speeds less than 3 m s−1 exist, the skin temperature can change as much as a few degrees in a period of 10–20 min. This is shown in Figure 12, which is a time series of wind speed (top) and skin temperature and buoy temperature (bottom). The figure shows that on 4 January the wind speed was less 3 m s−1 for some time, and the result was to have a rapidly changing skin temperature. The same effect on the bulk temperature is usually not seen. Instrumental error was ruled out by comparing buoys during times of low wind. To better understand how this process relates to wind speed, a rolling 20 min mean and SD are computed for every skin and bulk temperature for all buoys. As was done in Figures 9 and 12, the SD values are matched to wind speed measurements and put into wind speed bins of 0.5 m s−1. Figure 13 shows the average 20 min skin and bulk SD measurements for nighttime. Between 0 and 2 m s−1, the average 20 min skin SD is between 0.10 and 0.15 K. In this same wind speed regime the average 20 min bulk SD is an order of magnitude smaller and negligible. At low wind speeds the rapidly changing skin temperature is of similar size to changes due to other environmental conditions that change on time scales of hours to days. As wind speed increases past 4 m s−1 the average 20 min skin SD decreases down to approximately 0.06 K, which is much smaller than the total SD of 0.2 K. In this regime mixing leads to a well-established skin effect and is determined almost entirely by environmental conditions.
 Daytime measurements comparing the skin effect to wind speed constructed the same as nighttime measurements are shown in Figures 10 and 14. The top of Figure 10 shows the daytime average skin effect versus wind speed, where daytime is defined in Table 1. In contrast to the nighttime skin effect which is at its maximum when there is no wind, the daytime is approximately 0 K at 0 m s−1. However, the SD shown in the bottom of Figure 10 is 1.0 K at zero wind speed. Based on these data, prediction of the daytime skin effect using wind speed alone is not possible, as the range of values could lie between −1 and 1 K for low speeds. The range decreases as wind speed increases but is still −0.15 to 0.45 K for wind speeds around 10 m s−1.
 Average skin effect rises to a local maximum of 0.25 K at 2 m s−1 and then decreases to approximately 0.15 K when wind speeds greater than 12 m s−1 occur. SD of the average skin effect decreases rapidly from 0 to 2 m s−1 and then decreases more slowly afterward. This divide probably marks a region where waves can substantially mix the skin and bulk layers. Figure 14 shows that the average 20 min skin and bulk both are substantial compared to the overall SD at zero to low wind speeds. When combined together they explain almost half of the total SD shown in Figure 13. On average the skin effect during the daytime is smaller than nighttime; however, the variance is much larger. This is because of the skin effects' strong dependence on solar heating, which varies based on time of year, time of day, and cloud cover.
 The average bulk temperature minus air temperature is plotted as a function of time in Figure 15 to understand the sensible heat flux's contribution to the skin effect. At the start of the nighttime the air is warmer than the bulk but cools more rapidly. Just past sunset the nighttime sensible heat flux is into the lake but switches sign around 7–8 h after sunset with the behavior consistent among all buoys. The magnitude of the temperature difference between the nighttime air and bulk temperatures is a magnitude larger than the nighttime skin effect and follows the same slope until 10 h past sunset at which time the wind speed increases. Therefore, nighttime skin effect is largely determined by the sensible heat flux until late nighttime strong winds mix it out near sunrise.
 Daytime average bulk minus air temperature difference starts near zeros and slopes negative throughout the day indicating that the air is warming faster than the bulk temperature. The daytime skin effect does not exhibit the similar behavior because the shortwave radiation flux is much stronger.
 Figure 16 shows the average seasonal variability of the skin effect for individual buoys and the overall average. The top of Figure 16 displays the daytime seasonal cycle, and the bottom of Figure 16 shows the nighttime seasonal cycle. Daytime skin effect has a local maximum in the early spring which decreases to its minimum in June. The daytime skin effect rises after June to its maximum in September and slightly decreases to December. Nighttime skin effect is at its minimum from March to May and then rises to its maximum in September. Interbuoy seasonal variation of the skin effect is much smaller for nighttime compared to daytime.
 The presence of large negative daytime skin effect values are more consistent with Donlon et al.  and less consistent when compared to Minnett et al. . At first glance this may be surprising given that the depth of the bulk measurement in this study is much closer to that of Minnett et al. ; however, a close examination of Figure 6 from Minnett et al.  shows only a couple of points with a solar insolation value greater than 700 Wm−2. Because of the high altitude of Lake Tahoe, many solar noon insolation values are 1000 Wm−2 or greater. This illustrates a clear different skin behavior for this high-altitude lake versus typical ocean skin behavior.
 Figure 6 showed that daytime skin effect decreases approximately for 1 h after sunrise, and then rises for 2–3 h before falling off to the midday minimum. The likely mechanism is that the large solar zenith angle of early morning solar radiation is absorbed more in the skin compared with the rest of the water column for the first hour. As the sun continues to rise in the sky, solar heating can penetrate deeper and increases the temperature of the bulk water column. The decreasing skin effect at 4 h after sunrise occurs because increased winds due to daytime heating cause more mixing between the surface and the water below.
 The night skin cools faster than the bulk and explains why the average skin effect grows over time. The winds are high enough to overcome small-scale convection but not so high as to be dominated by turbulent mixing. This results in the low standard deviation for the first few hours in the night. As the night progresses the increase in wind speed provides more mixing, and subsequently the skin effect decreases. This turbulent mixing likely causes the increase in the standard deviation later in the nighttime. The low wind speed at night results from temperature inversions that are common year round and is a key difference in meteorology for Tahoe when compared to oceans.
 The late night warming was different across the buoys. It was shown that the eastern buoys' skin effect increased at earlier times compared with that of the western buoys. This is due to spatial and diurnal variations resulting from local thermal and terrain-driven wind systems on top of the large-scale synoptic flows from passing weather systems.
 Temperature measurements from satellites are normally combined with other satellites when temperature trends are analyzed. This is done because long time series are needed to look at climate signals. In order to assure no biases are being introduced when satellite measurements are combined into one time series, one satellite's measurements are usually homogenized to match the other satellites' measurements on average. One concern of homogenizing satellite instruments is the changing structure of the skin effect behavior with time.
 This study suggests that homogenizing satellite measurements with nighttime measurements will be much more successful than with daytime measurements. Standard deviation in the skin effect during the daytime is consistently higher, and parameterizing the effect of sunlight is difficult [Wick et al., 2005]. Figure 5 showed that Buoys 2 and 3 (Tb2 and Tb3) have very consistent average skin effect with time, changing less than 0.1 K; buoys 1 and 4 (Tb1 and Tb4) were not as consistent over the night; however, their largest changes occurred after 10 h past sunset. This length of time past sunset exists only in wintertime. Assuming that most lakes have similar nighttime summertime skin effect behavior, then this study supports using multisatellite data sets that have different overpass times for building long-term temperature trends.
 This study recommends the construction of a global network of inland water validation sites with enough sampling variability to represent all lakes measured by current satellites. Such a network may be possible if combined with other activities underway to develop global inland water buoy networks such as Global Lake Ecological Observatory Network [Esa, 2012]. Like the oceans, this will likely be in the absence of highly accurate skin measurements, and bulk temperatures will have to be used. A validation strategy, similar to the oceanic strategy from Donlon et al. , should include an indirect validation strategy of using bulk temperature measurement combined with a wind-based skin effect correction to maximize the usefulness of the available data. At Lake Tahoe wind speeds greater than 4 m s−1 can be used if radiometers are unavailable; this is lower than the oceanic value of 6 m s−1 found in Donlon et al. . With topography, elevation, depth, and atmospheric composition varying widely between lakes, it is not appropriate to assume that the skin effect at Lake Tahoe will be applicable for all other inland waters. Workshops such as the Global Lake Temperature Collaboration are creating a global database of all in situ lake measurements, with nearly all in situ data sets existing as bulk measurements [Lenters et al., 2012]. Determining the skin effect at different lake locations would allow for a more global in situ data set capable of validating satellite IWSTs.
 Lake Tahoe has lower wind speeds, lower wave fetch, lower cloud cover, and higher water clarity and is higher in altitude than the ocean, resulting in a more negative skin effect being observed in this data set compared to that of Donlon et al.  or Minnett et al. . The general low humidity and low winds also lead to large positive skin effects over Lake Tahoe; however, high humidity can produce enough longwave radiation at night to heat the skin more than the bulk. The reason this is seen over Lake Tahoe and not the oceans is the consistent lower wind speeds (and associated wave fields) at Lake Tahoe compared to those at oceans. It should be noted that Lake Tahoe's high water clarity likely results in less warming immediately beneath the skin compared to most lakes. A study of skin effect distributions for a lower clarity lake could provide results more general to average lakes around the globe. One possible site would be to use data taken at the Salton Sea that contains similar equipment used in this study.
 Differing wind speeds drive diurnal variation across the buoys at Lake Tahoe. The different offshore wind speeds result from local thermal and terrain-driven features as well as the shape of the lake. In general, lakes have much more heterogeneous wind fields compared to oceans due to the complex topography and land use features surrounding lakes. Therefore, when determining the skin temperature from bulk temperature data at other lakes, this study suggests that concurrent wind data are needed to accurately determine the size of the skin effect. It is not likely that reanalysis wind data will have the spatial resolution needed to accurately measure the skin effect for an individual measurement site; however, a high-resolution numerical model could be used to resolve the wind fields in a manner to accurately model the skin effect. The spatial resolution of the model likely would change on a lake-by-lake basis depending on the complexity of the offshore topography. In some cases the findings from this study could be applicable to coastal waters where the atmospheric boundary layer and wind speed variations could be considered a hybrid of lake and ocean cases.
 This study recommends using nighttime only measurements to construct climate data records because while nighttime measurements, on average, have a higher skin effect the standard deviation of those measurements is far less than during the daytime. A recent study of global lake temperature trends used merged satellite data sets but did not include global validation and used a single skin effect correction [Schneider and Hook, 2010]. With recent studies suggesting that lake temperatures may be rising faster than ambient air temperatures, more accurate lake temperature trends are required [Austin and Colman, 2007]. Utilizing in situ data sets from groups such as the Global Lake Temperature Collaboration and doing a validation campaign for satellite IWST retrievals that includes an accurate parameterization of the skin effect will yield a more confident trend analysis for global lake temperatures.
 We thank Brant Allen of the UC Davis Tahoe Environmental Research Center for his efforts in physically maintaining the Lake Tahoe buoys. We also thank Evan Fishbein for his editing assistance. The research described in this paper was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. The work was funded by a NASA award to Simon Hook as part of the NASA EOS program.