Field Measured Spectral Albedo–Four Years of Data from the Western U.S. Prairie


  • Joseph J. Michalsky,

    Corresponding author
    1. Earth System Research Laboratory, National Oceanic and Atmospheric Administration, Boulder, Colorado, USA
    • Corresponding author: J. J. Michalsky, National Oceanic and Atmospheric Administration, 327 Broadway GMD Boulder, CO 80305, USA. (

    Search for more papers by this author
  • Gary B. Hodges

    1. Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado, USA
    Search for more papers by this author


[1] This paper presents an initial look at four years of spectral measurements used to calculate albedo for the Colorado prairie just east of the Rocky Mountain range foothills. Some issues associated with calculating broadband albedo from thermopile sensors are discussed demonstrating that uncorrected instrument issues have led to incorrect conclusions. Normalized Difference Vegetative Index (NDVI) is defined for the spectral instruments in this study and used to demonstrate the dramatic changes that can be monitored with this very sensitive product. Examples of albedo wavelength and solar-zenith angle dependence for different stages of vegetative growth and senescence are presented. The spectral albedo of fresh snow and its spectral and solar-zenith angle dependence are discussed and contrasted with other studies of these dependencies. We conclude that fresh snow is consistent with a Lambertian reflector over the solar incidence angles measured; this is contrary to most snow albedo results. Even a slope of a degree or two in the viewed surface can explain the asymmetry in the morning and afternoon albedos for snow and vegetation. Plans for extending these spectral measurements for albedo to longer wavelengths and to additional sites are described.

1 Basic Considerations

[2] It appears to the casual observer that albedo is a straightforward calculation (‘albedo measurement’ will be used for shorthand). It is the ratio of the upwelling irradiance from a surface to the downwelling irradiance incident on that surface measured simultaneously, i.e.,

display math(1)

[3] Albedo can be calculated over a narrow wavelength interval, or integrated over an action spectrum such as photosynthetically active radiation (PAR), or over the complete solar spectrum; broadband solar is the most common albedo calculation.

[4] There are a number of reasons to measure spectral albedo. In a broadband shortwave clear-sky model-measurement closure study Michalsky et al. [2006] found that using measured spectral albedos rather than a simple broadband albedo that was assumed constant for all wavelengths provided lower calculated diffuse irradiance that improved closure with six different models. Albedo is also critical as an input for retrieving cloud optical depth from the surface (or from space). The radiation that reaches the surface has traversed the entire cloud, top to bottom. However, to calculate the optical depth correctly, it is necessary to take into account the reflections between the surface and cloud in radiative transfer calculations. Accurate albedo measurements, therefore, are essential for this retrieval. In cloud-free skies Michalsky and Kiedron [2008] have shown that in the ultraviolet (UV) it is possible to estimate the single scattering albedo of aerosols accurately even in low aerosol optical depth conditions. This retrieval, which uses the diffuse irradiance transmission, depends on correctly assessing the effect of the surface reflection contribution to the diffuse irradiance. In the UV, vegetation albedos are low, but in the visible spectrum vegetation can have high albedos, and it will be necessary to measure spectral albedo accurately to successfully retrieve single scattering albedo at these longer wavelengths. In section 3 the normalized difference vegetative index (NDVI) using multi-filter rotating shadowband radiometer (MFRSR) wavelengths is defined and its sensitivity to the condition of the Colorado piedmont vegetation is demonstrated to be useful in assessing stress on the prairie vegetation.

[5] Some recurring problems with albedo measurements are sensors that are not level, offset corrections not made, and the angular response of the sensor not considered. Dirmhirn and Eaton [1975] noted this latter issue in an early albedo paper, and it has been pointed out by other authors [Wiscombe and Warren, 1980; Feister and Grewe, 1995], but continues to be an issue as is demonstrated in this paper. Sensor leveling is obviously important and is of greatest concern for the measurement of downwelling irradiances in clear conditions when direct solar irradiance is the significant component of the downwelling irradiance.

[6] Offsets are a common problem in pyranometers that use a thermopile sensor [Gulbrandsen, 1978; Dutton et al., 2001] and need to be understood and corrected. Thermal offsets arise when the viewed target is at a significantly different temperature than the radiometer. For example, if an upward pointing pyranometer views the clear sky, the effective temperature of the sky is typically much colder than the pyranometer. This results in a thermal offset that superimposes a negative signal on the larger, positive signal that is generated as the sensor heats by absorbing direct beam and diffuse sky radiation. The dome of the pyranometer cools by radiating to the cooler sky, which in turn cools the detector surface relative to the pyranometer's ambient temperature. On the other hand, a low overcast day produces a condition where the effective temperature of the sky is closer to the pyranometer temperature resulting in a smaller offset. The lower signal in the denominator of equation (1) artificially increases the albedo. If the thermopile sensor views the surface from a 10-m tower for an upwelling measurement, the temperature of the target and pyranometer do not typically differ much with the result that there is little or no offset. Estimating and correcting offsets is discussed in Dutton et al. [2001], for example.

[7] All irradiance sensors have less than perfect angular responses. A perfect angular response is one that decreases exactly as the cosine of the angle of incidence. It is all but impossible to achieve this perfection, but some pyranometers perform better than others. The greatest divergence from true cosine response is typically at the largest angles of incidence, generally beyond 60º, and often, but not necessarily, this response is progressively lower than the true cosine response as the incident angle increases. If a sensor with this typical cosine response is calibrated at 45º incidence angle and used when the direct solar beam is incident at an angle of 75º, the irradiance measurement of the downwelling radiation will be too low resulting in an albedo that is too high (see equation (1)).

[8] To illustrate the offset and angular response issues, Figure 1 contains global horizontal solar irradiance (GHI) data from the SURFace RADiation (SURFRAD) site near Sioux Falls, South Dakota, USA [Augustine et al., 2005]. Either value of GHI could be used for the denominator in equation (1). The black dotted plot is based on a thermopile pyranometer that has a negative offset and a cosine response that is lower than the cosine response at 45° for all solar-zenith angles (same as incidence angle) in this plot. The green dotted plot is based on a sum of the direct component that falls on the horizontal (measured direct normal irradiance multiplied by the cosine of the calculated solar-zenith angle) plus the diffuse irradiance measured by a pyranometer under a tracking ball that blocks the direct solar; this pyranometer has almost no offset for any conditions and is calibrated at 45°. This component summed GHI is the recommended Baseline Surface Radiation Network (BSRN) [Ohmura et al., 1998] measurement for the determination of GHI. The black points are lower than the green points because the measured GHI pyranometer has a negative offset and a cosine response that is too low. Pyranometers have a less than perfect cosine response. In Chapter 6 of Vignola et al. [2012] an argument is made that the optimum angle for calibrating pyranometers that are used to measure diffuse sources, such as the sky and the surface, is 45°. SURFRAD pyranometers are all calibrated at 45°. Based on the above arguments, the best broadband solar albedo measurement should ratio the measured upwelling irradiance by the component summed GHI. Figure 2 illustrates the differences in calculated solar albedo for a snow-covered surface on a clear day at Sioux Falls. The albedo that uses the GHI pyranometer data is high compared to the GHI from the component sum. Further, the sense of the solar-zenith angle dependence changes from an upturn to an almost negligible downturn with increasing solar-zenith angle. The solar-zenith angle near solar noon is 57.4°, and at the morning and afternoon end points in this plot it is 80°. The solar-zenith angle dependence is only shown to 80° in this and the plots that follow because most pyranometers have very poor responses at larger angles of incidence. Solar-zenith angle dependence and the tilt of the plot will be discussed in a later section.

Figure 1.

Global horizontal irradiance (GHI) for three consecutive clear days measured using a single pyranometer lacking offset and cosine corrections (black) versus GHI calculated by adding measured direct beam irradiance multiplied by the cosine of the solar-zenith angle and the diffuse horizontal irradiance measured with a zero-offset pyranometer under a tracking disk (green). Points are three-minute averages of one-second samples.

Figure 2.

Differences in albedo for the first day in Figure 1 using two different measures of downwelling irradiance in the denominator of equation (1). Using summation the albedo is lower and consistent with Lambertian reflection. Tilt is discussed in the final section of the paper. Points are three-minute averages of one-second samples.

[9] Figure 3 is the albedo for the snow-covered surface at the Sioux Falls site three days earlier when the skies were overcast throughout the day. As there was no direct sun during the day, the skylight distribution of incident irradiance remained unchanged leading to a nearly constant value for the albedo that uses the component sum for the downwelling GHI (green points). In this case the GHI is solely the diffuse measurement with the zero-offset pyranometer as the direct beam is zero. The measurement of GHI using the single pyranometer with an assumed constant offset led to a higher albedo at solar noon with even higher values (black points) at the end points (80º solar-zenith angle). Assuming a similar offset throughout the day, higher irradiances near solar noon and lower irradiances near the end points explains the bowl-shape of the albedo plot. Note the albedo is higher in Figure 3 than in Figure 2 possibly caused by the metamorphosis of the snow grains from small to larger sizes that will be discussed in section 4. As a side note Wang and Zeng [2008] used these SURFRAD data in their analysis. It should be clear from the shape of the plot in their Figure 4 that they used the measured downwelling GHI rather than the component sum for the albedo calculation yielding results that need to be reconsidered. Schaaf et al. [2011] had earlier pointed out that the Wang and Zeng [2008] paper had also used MODIS data incorrectly.

Figure 3.

Albedo on a cloudy day should be constant (green curve) because the downwelling effective angle of incidence does not change appreciably over the day. A constant offset in a thermopile pyranometer with cloudy downwelling horizontal irradiance that peaks near solar noon, which it did on this day, would produce the upswing in albedo (black) in the morning and afternoon as well as slightly higher albedo all day compared to component sum (green curve). Points are three-minute averages of one-second samples.

Figure 4.

PAR, UVB, and MFRSR radiometers horizontally mounted looking at the surface from the 9-m level of the 10-m tower.

[10] Albedo measurements that are described in this paper are not separated into white-sky and black-sky albedos [Schaepman-Strub et al., 2006]. These could be approximated using a fairly straightforward and reasonably robust procedure. It requires that we obtain albedo measurements on a cloudy day that is near in time to a day with clear skies and a high direct solar irradiance. It could even be the same day if part of the day is overcast. The key assumption is that there is not an appreciable difference in the albedo caused by the difference in the spatial distribution of skylight on a clear and on a cloudy day. An overcast sky produces a white sky albedo Awhite

display math(2)

[11] The black sky albedo Ablack could then be calculated using

display math(3)

[12] In words, the fraction of the upwelling that is due to the downwelling diffuse reflecting upwards is removed from the upwelling signal and divided by the direct normal component on the horizontal surface, where sza is the solar-zenith angle.

[13] In the next section the venue for four years of albedo measurements is discussed along with the spectral and broadband measurements made there. This is followed in section 3 by a discussion of the vegetative spectral albedos and their daily and annual variability; the latter is demonstrated using the NDVI as defined for the instruments used for the spectral albedo measurements. A section specifically on snow albedos concludes the measurements with a final section that discusses these results and our plans for future measurements.

2 The Table Mountain SURFRAD Site Near Boulder, Colorado USA

[14] The Table Mountain SURFRAD site is located at 40.125° N, 105.237° W at an elevation of 1689 meters. It is on a mesa not far from the foothills of the Rocky Mountains. Native grasses, generally less than a meter in height, cover the mesa. The site beneath the 10-m tower is undisturbed year round. The gentle slope of the area surrounding the albedo tower is between 1 and 2° rising from east to west. The tower's instruments are mounted near the 9-meter level on horizontal booms and down facing as seen in Figure 4. The instruments on the left measure PAR (smaller instrument) and UVB. The multi-filter radiometer (MFR) [Hodges and Michalsky, 2011] sensor on the right measures spectral radiation near 415, 500, 615, 673, 870, and 940 nm with 10-nm band passes every 20 seconds. The open silicon channel in the MFR is not used for albedo measurements. These upwelling filter measurements are cosine corrected. Another 10-m tall tower 70 m to the northwest measures upwelling broadband solar and broadband infrared irradiances. Comparable instrumentation to that on the two towers is located on a deck that is about 40 m south of the spectral albedo tower. Broadband solar downwelling is measured with an un-shaded pyranometer as well as by component summation of measured direct and diffuse. The un-shaded pyranometer is part of the SURFRAD instrument suite, but it is generally only used as a quality assurance check on the component sum. PAR and UVB sensors on the deck are un-shaded radiometers with no provision for component summation for these wavelength bands. The multi-filter rotating shadow-band radiometer (MFRSR) [Harrison et al., 1994], which contains the 10-nm filters specified above, samples every 20 seconds and effectively corrects for the cosine response of the direct normal and diffuse horizontal spectral irradiance used to form the component summed spectral GHI. Further details on the downwelling measurements made on the deck are available at

[15] Measurements to be discussed in this paper cover the dates from March 2008 until April 2012 with two multi-month gaps caused by instrument issues. Quality control measures included regular calibrations, offset corrections as a function of temperature, sensor-leveling checks, and cosine response corrections for the downwelling direct and diffuse and for the upwelling diffuse. All irradiance sensors are calibrated at an incident angle of 45° except for the PAR sensors that were calibrated at normal incidence. PAR measurements are not discussed in this paper. Offsets as a function of sensor temperature are illustrated in Figure 5 for the 415-nm MFR, This channel has the largest offsets and a significant temperature dependency. For example, June dark counts averaged -23 while a signal at noon on a clear day may be 27 counts, i.e., nearly half the signal is dark. As another example, physically impossible snow albedos exceeding 100% were calculated when offset corrections were not applied. Note that the offsets and temperature dependence of the other channels are much smaller; nonetheless, these correction procedures were applied to every measurement.

Figure 5.

415-nm (dark) offset signal for the down-looking multi-filter radiometer as a function of head temperature determined over the annual temperature cycle using nighttime data with sun well below the horizon. For example, June darks average -23 counts while a signal at noon on a clear day may be 27 counts, i.e., nearly half the signal is dark. Other channels have smaller, but still significant issues, and this correction is imperative.

3 Vegetation Albedos

[16] Short grasses mainly cover the albedo site. Figure 6 contains a spectral albedo for green grass from Bowker et al. [1985]. All green vegetation is similar in wavelength dependence when it is near its maximum greenness. Carotenoids and chlorophyll absorb strongly in the visible below the 550-nm peak and chlorophyll absorbs strongly above the 550-nm peak before the red edge that begins to rise at 700 nm. The 550-nm peak results from a minimum in chlorophyll absorption between the strong absorption features in the blue and red regions of the visible spectrum. Cell wall reflectance of plants is high in the near infrared out to about 1300 nm. Rouse et al. [1974] developed the normalized difference vegetative index (NDVI), although they called it the vegetative index (VI), based on the visible and near infrared reflectance contrast in satellite measurements made with a broad visible and a broad near-infrared channel. For this paper we define NDVImfrsr as

display math(4)
Figure 6.

Green grass (green) and dry grass (brown) albedos from Bowker et al. [1985]. MFRSR channels are located at vertical line positions. The thicker lines represent the two filters used for the NDVI calculations in the next figure.

[17] From Figure 6, it is obvious the six MFRSR filters, shown as vertical lines, provide a clear contrast in the visible and near infrared albedos. As vegetation senesces, the differences in the visible and near infrared albedos decrease (brown trace in Figure 6). The 673-nm albedo shows the largest fractional increase of the visible channels, and the 870-nm decreases leading to a smaller NDVImfrsr (shortened hereafter to NDVI). Figure 7 is a plot of all available NDVI calculations in the four-year record. The data gaps are apparent, however, the annual behavior is clear. The lowest values of NDVI occur in the last two weeks of February when the grasses are not only dead, but also flattened by snowfall during the winter. The peak NDVI each year occurs during the last two weeks of May and the increase of NDVI to this peak is very steep. The senescence phase proceeds differently each year. In 2008 the decrease in NDVI was much steeper after the May peak than in the succeeding years due to a rainfall deficit, but the NDVI recovered in late summer because of higher than normal August rains that raised NDVI values to those typical of other years. In 2009 there was above normal rainfall the entire summer and fall seasons. In 2010 rain almost stopped falling after mid August for the rest of the year, and in 2011 the rains were again above normal throughout the summer and fall. Note that during the winter season there are drops in the NDVI to values near zero; these are caused by snow cover, which we shall see in the next section results in albedos at 673 and 870 nm that are similar thus yielding an NDVI close to zero. A final comment on Figure 7 concerns the black versus the red points on the plot. In order to calculate an index every day of the year only data with solar-zenith angles between 64° and 75° are used; these are the black points. The red points are those that fall only between 64° and 65°. This demonstrates that NDVI works well as a normalized index that indicates vegetative condition with little dependence on solar-zenith angle. The measured albedo as a function of solar-zenith angle can vary by a factor of two as will be discussed next.

Figure 7.

NDVI calculated using equation (4). Green vegetation has the largest NDVI. Dry and flattened vegetation has the lowest winter values of NDVI except for snow-covered surfaces when the NDVI is near zero. Note summer and fall NDVI behavior after peak May values each year show variability associated with rainfall patterns; for example, the summer of 2008 was dry until abundant mid-August rainfall. The black points include all measurements between 64 and 75º solar-zenith angle and the red points are only between 64 and 65º solar-zenith angle demonstrating that NDVI is nearly independent of solar-zenith angle.

[18] Figure 8 is a plot of albedo in late spring about one month after the peak value of NDVI in Figure 7. At solar noon on this day the sun is about 17° from the zenith; the end points in the morning and afternoon are at 80° solar-zenith angle. The skies were practically clear all day with only a few passing thin clouds early in the morning and after solar noon. We see that higher solar-zenith angles produce larger albedos that depend on the wavelength. The wavelength dependence at any point in time is in general agreement with the generic green grass plot of Figure 6 with the exception that the 615 and 673-nm albedos are about the same value rather than 615 being higher than 673 as in Figure 6. As the data were taken about one month after the highest NDVI value, this higher value for 673 could be a signature of the early onset of senescence. Also note the asymmetry in the albedo with higher morning than afternoon values at 80°. In Figure 9 we plot the albedos normalized to the solar noon values to emphasize the solar zenith angle dependency differences among the wavelengths. The 415-nm normalized albedo is not shown because of the data's low signal-to-noise. A factor of two enhancement at 80° relative to solar noon at 17° in the near-infrared channels is obvious when plotted this way. The enhancement in the visible channels is about one-half that of the two near-infrared channels.

Figure 8.

Albedo at six wavelengths one month after peak NDVI. The wavelength pattern is similar to the green grass pattern in Figure 6 except 615 and 673 albedos are similar rather than having 615 higher as in Figure 6. This may indicate early browning associated with beginning senescence. Notice the morning-afternoon asymmetry that is explained in the paper.

Figure 9.

Normalized (using low solar noon) albedos of Figure 8 indicate a doubling of albedo between solar zenith angles of 17° and 80°; the effect is larger in near-infrared than in visible.

[19] Figure 10 is a plot for the data one year later that has nearly the same albedos with small shifts in the wavelength dependence compared to the previous year that show more separation in wavelengths, perhaps indicating senescence is further along. Figure 11 is the same as Figure 9 except one year later. There is a clear difference in this plot in that the shorter wavelengths are in better agreement with the near infrared ones in solar-zenith angle dependence than they were one year earlier. This behavior is discussed in the last section of the paper. Figure 12 is for October where the grasses are very dry. The wavelength dependence qualitatively resembles the dry grass plot wavelength dependence in Figure 6 with its monotonic increase with wavelength. In this plot the solar-zenith angle varies between 51° at solar noon and 80° so the change with solar-zenith angle is much smaller than in the June plots. This is confirmed in Figure 13 where the enhancement is about 60% rather than over 100% as in Figures 9 and 11. Note that in Figure 13 the normalized albedos are in better agreement across all wavelengths than in the previous normalized plots (Figures 9 and 11).

Figure 10.

Similar to Figure 8 except measurements are one year later. Albedos at 615 and 673 have greater separation indicating senescence may be further along.

Figure 11.

Normalized (using low solar noon) albedos of Figure 10 are similar to Figure 9 except the visible wavelength behavior is closer to that of the near infrared than it was in Figure 9.

Figure 12.

Albedos for drying fall vegetation monotonically increase with wavelength.

Figure 13.

Normalized albedos of Figure 12 are now nearly the same at every wavelength.

[20] Although asymmetry exists in Figures 8 through 13, the normalized plots better demonstrate this asymmetry with higher morning values for the same solar-zenith angles. The simplest explanation could be misaligned instruments, but all instruments are routinely aligned; since this asymmetry is always present on clear days, sensor misalignment is not likely the cause. Minnis et al. [1997] suggested several possible explanations for albedo asymmetry. In their case the albedos were consistently higher in the morning than in the afternoon at the same solar-zenith angles. They suggested that dew on the leaves of the vegetation in the morning yielded higher albedos than in the afternoon. The other possibility that Minnis et al. [1997] proffered was a difference in morning and afternoon aerosol loads. Differences in aerosols would change the partitioning of the direct and diffuse irradiance and, therefore, the balance between black and white-sky albedos (discussed in section 1) before and after solar noon. Examination of the broadband diffuse irradiance on the days included in this paper, which is very sensitive to the aerosol loading, indicated either no asymmetry between morning and afternoon, or some days with higher mornings and some with higher afternoons, but with all days showing similar asymmetric albedos. While these may explain albedo asymmetries at some locations, the Table Mountain asymmetric albedos are likely caused by the slope of the site with the surface slightly higher in the west than in the east. The slope is estimated to be only between 1° and 2°. The rising sun better illuminates the surface than the setting sun leading to an asymmetry in the amount of reflected light. This surface slope effect is discussed further in the final section.

[21] Under overcast conditions the radiance distribution is more uniform than when the solar disk can be seen. The average direction of the incident radiation on the surface does not differ appreciably over the course of the day resulting in an albedo that does not noticeably change. Figure 14 is a plot of albedo just after the clear day in Figure 8. Figure 14 is noisier because the cloud cover reduces the signal to noise. As expected, the albedos appear constant throughout the day. Comparing this to Figure 8, note that the albedos are consistent with those measured on the clear day near 170.36 and 170.64; At these times the solar zenith angle was about 45°. This is consistent with the arguments in Chapter 6 of Vignola et al. [2012] that the effective angle of incidence for a diffuse source such as the cloudy sky is never far from 45°.

Figure 14.

Cloudy day albedos are constant over the day since the effective angle of incidence does not change appreciably. Albedo values are consistent with clear-day values in Figure 8 (one day earlier) when the solar-zenith angle is near 45°.

4 Snow Albedos

[22] Snow is an excellent reflector in the visible wavelengths where spectral solar radiation has its highest irradiance; therefore, it has a major effect on the radiation balance at the Earth's surface. Measuring the albedo of snow and its spectral and solar-zenith angle dependence is difficult in terms of getting consistent results as snow albedo decreases as the snow grains increase in size, and as absorbing particles deposit and mix on the snow. In back-to-back seminal papers on the spectral albedo of snow Wiscombe and Warren [1980] and Warren and Wiscombe [1980] described theoretically the expected wavelength dependence and magnitude of the albedo and its dependence on a number of variables.

[23] Figure 15 is the spectral albedo for a completely clear day with snow on the ground at Table Mountain. The solar-zenith angle at solar noon is about 59°, and the end points are at 80°. Figure 16 is a similar clear-day plot for early December of the next winter with a noon solar-zenith angle at about 62° and end points at 80°. The plots are very similar with similar wavelength dependences and magnitudes. The subtle differences in wavelength dependence are within the uncertainties of the measurements. There is a similar tilt to both plots with higher albedos in the mornings. The snowfall occurred in both instances the previous day or two with below freezing temperatures continuing throughout the days shown. The snow cover in December (Figure 16) was thicker than in January (Figure 15), but both were around 6-7 cm. There is no clear indication of solar-zenith angle dependence in either plot.

Figure 15.

Albedo on a day with below freezing temperatures for a thinly snow covered (few cm) surface. Tilted values do not show solar-zenith angle dependence. Wavelength dependence is highest in the near infrared, which is not expected for snow.

Figure 16.

Very similar albedos and wavelength dependence for a slightly thicker snowpack for a winter day that is similar to the Figure 15 data.

[24] In Figure 17 the clear-day albedo measurements followed a heavy snowfall with 35 cm on the surface for 5 February 2012 compared to 7 cm in Figure 16 and less than that in Figure 15. The albedos are slightly higher, but more importantly the wavelength dependence has changed with peak reflectivity in the 615 and 673-nm wavelengths rather than the 870 and 940-nm wavelengths in Figures 16 and 17. This change in the wavelength dependence is possibly explained by the thickness of the snow as discussed in Wiscombe and Warren [1980]. The thinner snow packs of Figures 15 and 16 allow sensing of the surface beneath the snowpack that has a higher reflectivity in the near infrared than it does in the wavelengths shorter than 700 nm (see Figure 6 for dry vegetation). It is also possible that some of the dry vegetation is protruding through the thin snow cover. Another possibility discussed in the companion paper Warren and Wiscombe [1980] concludes that a small amount of absorber in the snow such as soot or dust that absorbs in the visible, but not in the near infrared, could qualitatively explain the wavelength dependence in Figures 15 and 16. The differences between Figures 15/16 and Figures 17/18, and the freshness of the snow, however, suggest snowpack thickness as the more probable cause.

Figure 17.

Albedos for a clear day with thick snow cover. Note that the albedos are higher and the wavelength dependence is now similar to theoretical predictions with 615 and 673 peak values. Tilt is similar to two previous figures, and there is no notable solar-zenith angle dependence.

[25] Figures 18 and 19 are plots for cloudy days before and after the day plotted in Figure 17. Figure 18 has high, constant values throughout the day as expected since the direction of downwelling radiation is effectively constant throughout the day. However, Figure 19 has high values in the morning and then shows somewhat smaller values in the afternoon even though it is cloudy all day. This figure is the first in this paper to, perhaps, indicate that the formation of larger snow particles may be producing lower reflectivity in the afternoon. Although the temperature stayed well below freezing all day, the solar irradiance was near clear day values in the afternoon.

Figure 18.

Constant snow albedos as expected for a cloudy day with wavelength dependence similar to Figure 17. These data were acquired two days prior to Figure 17. The higher values may be associated with smaller snow grains from the fresh snowfall event.

Figure 19.

Cloudy day snow albedos in this figure may indicate snow grains evolving to larger sizes causing somewhat lower albedo values in the afternoon.

5 Discussion and Future Directions

[26] Grenfell et al. [1994] derived an equation to calculate the effect of surface tilt on the measurement of albedo. The Table Mountain albedo towers are mounted above a surface that has an estimated slope, based on topographic maps, of 1° to 2° with the western edge higher than the eastern edge. The slopes of the clear-day snow albedos in Figures 15 through 17 are consistent with this formula and this estimation of the slope. In Figure 2 the Sioux Falls, South Dakota, site slope is such that east is higher than west leading to the tilt of snow albedo in Figure 2 that again is consistent with the estimates, but has the opposite morning to afternoon behavior seen near Boulder. The most probable explanation for the changes in magnitude of the albedo throughout the day in most of these plots is associated with surface slope and not with snow grain size change. The possible exception is the subtle change noted in Figure 19 that occurred on an overcast day. The same tilt argument explains the asymmetry of albedo for the vegetated surfaces in Figures 8 through 13.

[27] Wiscombe and Warren [1980] theoretically predicted snow surface albedo to increase with solar-zenith angle. Dirmhirn and Eatons [1975] measurements were qualitatively in agreement with these calculations. Lupi et al. [2001] found similar results for snow-covered surfaces. Recently, Dumont et al. [2010] integrated their bidirectional reflectance distribution functions measured over snow for three different incidence angles and drew the conclusion that albedo increases with solar-zenith angle. However, Wuttke et al. [2006] found that snow albedos did not change with solar-zenith angle as we have also noted in Figures 15 – 17. Some of the increases with solar-zenith angle could be the failure to correct offsets and cosine responses for the downwelling measurement as discussed in the first section of this paper and as demonstrated in Figure 2. It is clear that the condition of the snow will influence the solar-zenith angle dependence, for example, the surface will become more specular with melting and freezing and should lead to an increase in albedo with solar-zenith angle. We will look for these cases in future studies. We conclude, however, that the fresh snow observed in Figures 15-17 in below freezing temperatures behaves like a Lambertian surface over the incidence angles measured.

[28] McFarlane et al. [2011] have developed a method to estimate the continuous spectral distribution of albedo with the six wavelengths of the MFRSR. We intend to improve the technique by adding a longer wavelength filter at 1625 nm to the albedo measurement. From their paper's Figure 5 it is clear that this information should improve the spectral agreement throughout the near infrared.

[29] Plans are to add spectral upwelling measurements to all of the SURFRAD [Augustine et al., 2005] sites and to add new MFRSR downwelling measurements with this wavelength addition at 1625 nm. The spectral albedo data will permit progress on retrieving the spectral single scattering albedo on clear days and cloud optical depths and effective cloud particle radii when there are overcast conditions. These sites will provide useful data for improving and validating satellite measurements. In addition to the seven fixed SURFRAD sites, we are developing one or two moveable SURFRAD sites for special field studies. The first moveable SURFRAD site has the 1625-nm channel, and it is already showing interesting results.

[30] In conclusion, the paper has demonstrated that measurements to calculate albedo have to not only be well calibrated, but must be corrected for often neglected issues like offsets and angular responses associated with 2π sr radiometers. Calibrations for instruments making upwelling and downwelling diffuse (ground and sky) measurements should be made at 45°. Albedos of vegetation show both dramatic and subtle changes that can be used to detect the state of the ecosystem. Although hints of these changes were mentioned here, further analysis in needed to fully exploit the potential of the MFRSR NDVI and solar-zenith angle dependency changes. The grade of the surface being measured explains the albedo asymmetry, perhaps, for the first time; this will be a crucial correction for satellite albedo comparisons. Once the asymmetry is corrected, very fresh snow albedos appear close to Lambertian in behavior counter to many other snow albedo measurements. There is more research needed to understand the effect of the depth of the snow and its evolution with time and temperature and the resulting effects on the albedo and solar-zenith angle dependence.


[31] The authors would like to thank Steve Cooper and John Augustine for their help in developing the infrastructure to enable these albedo measurements. Our thanks to Ellsworth Dutton and Peter Kiedron for reviewing the manuscript and suggesting key changes. This research was supported in part by the Office of Biological and Environmental Research of the U.S. Department of Energy as part of the Atmospheric Radiation Measurement Program.