Broadband Albedo Observations in the Southern Great Plains

Claude E. Duchon School of Meteorology, University of Oklahoma, Norman, Oklahoma

Search for other papers by Claude E. Duchon in
Current site
Google Scholar
PubMed
Close
and
Kenneth G. Hamm School of Meteorology, University of Oklahoma, Norman, Oklahoma

Search for other papers by Kenneth G. Hamm in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Time series of daily broadband surface albedo for 1998 and 1999 have been analyzed from six locations in the network of 22 Atmospheric Radiation Measurement Program Solar–Infrared Radiation Stations distributed from central Kansas to central Oklahoma. Two of the stations are in Kansas, and four are in Oklahoma; together they reasonably encompass the variation in geography in the southern Great Plains. Daily precipitation totals locally measured or obtained from nearby Oklahoma Mesonet stations and time series of biweekly maximum normalized difference vegetation index obtained from NOAA’s Advanced Very High Resolution Radiometer were used to determine linkages between surface albedo and amount of precipitation and degree of green vegetation. As part of this determination, daily albedo was categorized according to sky condition, that is, clear, partly cloudy, or overcast, with appropriate boundaries for each category. The more notable results are the following: 1) 2-yr mean annual albedos varied by more than 20% among the six sites, the lowest albedo being 0.18 and the highest albedo being 0.22; 2) the numerical difference was about 4 times the maximum interannual mean difference among the six stations, indicating the importance of geographic location; 3) for sites with a large amount of bare soil, a systematic decrease in albedo in response to rainfall events and a systematic increase in albedo as the soil dried were observed; 4) at the one site with total vegetation cover, that is, no bare soil, albedo response to precipitation events was suppressed; 5) no relation was found between mean annual albedo and annual precipitation; 6) whether days were classified as clear or partly cloudy had little influence on daily albedo, but overcast days typically reduced albedo, sometimes substantially; and 7) the main contributor to low albedos on overcast days with rain was the wet surface; the contribution by the overcast sky was secondary.

Corresponding author address: Claude E. Duchon, School of Meteorology, University of Oklahoma, 100 East Boyd St., Suite 1310, Norman, OK 73019. Email: cduchon@rossby.metr.ou.edu

Abstract

Time series of daily broadband surface albedo for 1998 and 1999 have been analyzed from six locations in the network of 22 Atmospheric Radiation Measurement Program Solar–Infrared Radiation Stations distributed from central Kansas to central Oklahoma. Two of the stations are in Kansas, and four are in Oklahoma; together they reasonably encompass the variation in geography in the southern Great Plains. Daily precipitation totals locally measured or obtained from nearby Oklahoma Mesonet stations and time series of biweekly maximum normalized difference vegetation index obtained from NOAA’s Advanced Very High Resolution Radiometer were used to determine linkages between surface albedo and amount of precipitation and degree of green vegetation. As part of this determination, daily albedo was categorized according to sky condition, that is, clear, partly cloudy, or overcast, with appropriate boundaries for each category. The more notable results are the following: 1) 2-yr mean annual albedos varied by more than 20% among the six sites, the lowest albedo being 0.18 and the highest albedo being 0.22; 2) the numerical difference was about 4 times the maximum interannual mean difference among the six stations, indicating the importance of geographic location; 3) for sites with a large amount of bare soil, a systematic decrease in albedo in response to rainfall events and a systematic increase in albedo as the soil dried were observed; 4) at the one site with total vegetation cover, that is, no bare soil, albedo response to precipitation events was suppressed; 5) no relation was found between mean annual albedo and annual precipitation; 6) whether days were classified as clear or partly cloudy had little influence on daily albedo, but overcast days typically reduced albedo, sometimes substantially; and 7) the main contributor to low albedos on overcast days with rain was the wet surface; the contribution by the overcast sky was secondary.

Corresponding author address: Claude E. Duchon, School of Meteorology, University of Oklahoma, 100 East Boyd St., Suite 1310, Norman, OK 73019. Email: cduchon@rossby.metr.ou.edu

1. Introduction

Solar radiation is the primary energy input to the earth–atmosphere system. A portion of this energy is transmitted through the atmosphere to the earth’s surface, where it is either absorbed or reflected back into the atmosphere. The fraction of solar energy absorbed by the earth’s surface is then partitioned into sensible, latent, and ground heat fluxes. Horizontal gradients in these fluxes contribute to differential heating of the atmospheric surface layer and the consequent development of local-scale to synoptic-scale atmospheric circulations. In turn, these circulations contribute to the weather and climate on all spatial scales. Because surface reflection plays a fundamental role in the surface energy budget, knowledge of its spatial and temporal variability is important for understanding the weather and climate of a region.

The reflective property of a land surface is determined by forming the ratio of the upwelling (reflected) solar radiation Ku to the downwelling solar radiation Kd, both of which are irradiances (energy per unit time per unit horizontal area). This ratio is called the broadband surface albedo a, and the product (1 − a)Kd is the absorbed solar energy available for partitioning at the land surface. The dependencies of reflectance on wavelength and direction are incorporated into the calculation of albedo by integrating the upwelling radiation over a “broad band” of wavelengths determined by the particular radiometer and over the hemisphere below the point of measurement, respectively. Similar integrations apply to the downwelling solar radiation.

The Atmospheric Radiation Measurement Program (ARM) was established by the U.S. Department of Energy (U.S. DOE) to improve understanding and, ultimately, the modeling of the processes and properties that affect atmospheric radiation (Stokes and Schwartz 1994; Ackerman and Stokes 2003). While particular focus has been placed on clouds and their radiative properties, ARM recognizes the importance of surface albedo by identifying it as one of the two most important properties (along with the effective surface radiating temperature) for the specification of the radiation field at the earth’s surface (U.S. DOE 1990, p. 29). To provide observations necessary for its modeling objectives, the Southern Great Plains (SGP) Cloud and Radiation Test Bed (CART) was established in an area extending from central Kansas to central Oklahoma (Fig. 1). This 365 km (north–south) by 300 km (east–west) area roughly corresponds to the size of a general circulation model grid cell of 3° latitude by 3° longitude.

Ground-measured broadband surface albedo (or simply albedo) varies with time of day, day of year, surface type and texture, the presence of liquid or frozen precipitation on the surface, and the optical properties of the radiometer. The variation of albedo with time of day is due to the dependence of reflected radiation on the solar position. Minnis et al. (1997), Barnsley et al. (2000), and Grant et al. (2000) investigated this relation by observing the daily sunrise-to-sunset dependence of clear-sky albedo on solar zenith angle. Knowledge of the diurnal cycle of albedo is essential for evaluating the accuracy of a daily satellite albedo estimate, but analysis of albedo on seasonal or yearly time scales is more easily undertaken using a single value of albedo assigned to a day, week, or month.

With this background in mind, the purpose of this investigation was to determine temporal and spatial variations of daily surface albedo at six selected locations within the SGP CART using two years of observations. The results will contribute to fulfilling the ARM objective of relating “observed radiative fluxes and radiances in the atmosphere as a function of position and time. .. to surface properties” (U.S. DOE 1996, p. 1.1) and serve as guidance for developing a climatological description of surface albedo as additional years of data are analyzed.

The organization of the remainder of this paper is as follows. In section 2 we describe measurement site locations in the SGP CART area and the instrumentation that provided the data used in our analysis. In section 3 we address criteria for site selection, classification of cloud cover, and calculations of daily albedo, normalized difference vegetation index (NDVI), and daily precipitation. In section 4 we examine the structure of the daily albedo at each of the six sites in conjunction with daily precipitation, cloud cover, and NDVI for 1998 and 1999. We also examine differences in daily albedo that result from using two methods of measurement. In section 5 we study the impact of overcast days on albedo. In section 6 we demonstrate the independence of mean annual albedo and mean annual precipitation and comment on the relation between albedo and NDVI. Section 7 is a summary of our investigation.

2. Radiation measurements

The SGP CART comprises a central facility, 24 extended facilities (EF), four boundary facilities, and three intermediate facilities. The locations of the various sites and the associated instrumentation could be found online (http://www.arm.gov/) at the time of writing. The solid and open circles in Fig. 2 show the locations of the extended facilities, two of which (EF-13, EF-14) are at the central facility location near the town of Lamont in northern Oklahoma. The boundary and intermediate facilities are not shown. Albedo analysis was performed at the six stations identified by solid circles. Measurements at the central facility and boundary facilities provide detailed information about the atmospheric column above the CART area, and measurements at the extended facilities are used to measure surface energy fluxes, soil water, and soil temperature (Stokes and Schwartz 1994).

The instrument platform at the extended facilities that measures downwelling and reflected broadband solar radiation used in calculating albedo is the Solar–Infrared Radiation Station (SIRS), a complete description of which could be found at the Web site given above. In addition to downwelling and upwelling longwave irradiance, each SIRS platform measures four shortwave irradiance components over the wavelength range of 0.28–2.8 μm: 1) direct normal (DIR), 2) diffuse (DIF), 3) global (Kd), and 4) reflected (Ku). Each component is sampled every 2 s, from which a 1-min average is computed based on the preceding 30 samples. All albedo estimates used here were calculated using 1-min averages.

The instrument platform includes three Eppley Laboratory, Inc., precision spectral pyranometers [PSP; one each to measure the global (hemispheric), diffuse, and reflected components], an Eppley normal incidence pyrheliometer (NIP) to measure the direct normal component, and two Eppley precision infrared radiometers (PIR) to measure the upwelling and downwelling longwave components. The downwelling components are measured at 1.5 m above ground level (AGL) at the south end of each EF. The diffuse component is measured with a shaded PSP mounted on an automatic sun tracker to which is attached the NIP. The reflected component is measured at the north end of each EF using an inverted PSP mounted at 10 m AGL. Adjacent to the inverted PSP is an inverted PIR to measure upwelling longwave irradiance. At this height, 90% of the reflected radiation measured by the inverted PSP is from a circular surface area with radius of 30 m surrounding the nadir point if isotropic scattering of the reflected irradiance is assumed. A plan view of the SIRS instrument configuration for shortwave observations within a generic EF is shown in Fig. 3. The horizontal dimensions are approximate and may vary slightly among sites. The six-sided boundary in Fig. 3 represents the two-wire electric fence enclosing the EF. As shown in the sketch, the 30-m radius influencing the Ku measurement typically includes most of the fenced area and portions of the surrounding area to the north, east, and west but does not include land to the south of the EF. The surface type inside the fenced boundary is usually some variety of grass, but the surrounding area also could include different surface types such as agricultural or wooded areas. Thus, the assumption of isotropic scattering is typically not true but is useful for estimating the area from which most of the reflected solar irradiance comes. This study focused on the six EFs shown in Fig. 2 whose surface types were determined to be those most representative of the surrounding area, as defined in the next section.

3. Experimental procedures

a. Site selection

In order that albedo measurements at the local scale be useful for comparison with albedo measurements at larger spatial scales, it is important that the land surface characteristics at a given site be as uniform as possible. Surface homogeneity over an area of at least 1 km2 was desired so that local measurements could be compared with satellite measurements using the Advanced Very High Resolution Radiometer (AVHRR) in a future study.

In the original ARM plan, EF locations were chosen so that surface heat, moisture, and momentum fluxes could be monitored across the SGP CART area over the widest range of environmental and surface conditions (Barr and Sisterson 2000, p. 43). As a result, EFs were located adjacent to various surface types such as pasture, winter wheat, and mixed crops. The requirement for surface representativeness at each EF was targeted to be between 0.4 and 0.6 km in every direction (Barr and Sisterson 2000, p. 41). Thus, because it was possible that two or more types of surface could exist within 0.5 km of an EF or within the footprint of an AVHRR pixel, a minimum of 1 km2, each site supporting a SIRS platform was evaluated for this study with respect to representativeness of Ku measurements used in calculating albedo. The representativeness was estimated according to the uniformity of vegetation and topography inside and outside the fenced area extending outward at least 0.5 km in all directions. Between 24 October 1998 and 15 January 1999, each of 21 EFs containing a SIRS platform was visited. Since that period, EF-25 near Seminole, Oklahoma, has been decommissioned and EF-21 near Okmulgee, Oklahoma, has become operational.

Uniformity of vegetation was subjectively evaluated by visually considering its type, color, height, leaf area index (LAI), and fractional vegetative cover (FVC). The importance of each characteristic is dependent on site location. For example, although the surface type inside the fenced area at each EF consists predominantly of some variety of grass, many of the sites were in close proximity to a winter-wheat field. Over the course of a year, the wheat field will typically show a much larger variation in color, height, and LAI than the grass inside the fenced area. In such cases, the Ku measurement was considered not to be representative of uniform vegetation. At other sites located in pasture or rangeland, the grass outside the fenced area was usually shorter and had lower LAI than the grass inside the fence because of the presence of cattle. However, for most sites of this type, the difference in vegetation height and LAI was sufficiently small that the effect on Ku was not believed to be significant.

With regard to uniformity of topography, the presence of surface features such as trees or windbreaks, ravines, hills, ponds or flooded areas, buildings, roads, or rock outcroppings can affect correlation with satellite albedo measurements. One or more of these features was observed at a number of sites.

After completion of the above evaluation for representativeness, each EF was placed into one of three categories with respect to our need for satellite comparison as discussed above. Category-1 sites were considered to have sufficient area of surface homogeneity surrounding the measurement site, category-3 sites had insufficient area, and category-2 sites had questionable area. Full tabular results for each site are given in Hamm (2003). Only category-1 sites were investigated. Table 1 summarizes the main geographical properties for each of these sites, and their locations are shown in Fig. 2. We note that the dominant vegetation cover is native grass (ungrazed) at four sites and is rangeland (grazed) at two sites. FVC is high at the four grassland sites and is low at the two rangeland sites, and all sites have some type of loam soil.

b. Sky-cover classification

To assess cloud-cover effects on ground-measured surface albedo, a daily sky classification was performed for the six category-1 sites for 1998 and 1999. Frequent human observations have traditionally provided good estimates of sky cover because all portions of the sky can be viewed at once. However, sky-cover observations are not taken at the EFs except during biweekly maintenance visits. Meteorological satellites also can be used to evaluate sky cover, but archived geostationary satellite data are not universally available and the temporal resolution of polar-orbiting satellites is often limited to one overpass per day. Therefore, in this study, a subjective method of evaluating sky cover was employed in which diurnal time series of SIRS irradiances were visually inspected at each of the six locations.

As previously stated, a pyranometer detects broadband solar radiation over a hemispheric field of view; however, the irradiance measurement is weighted toward the portion of the sky in which the sun is located (Duchon and O’Malley 1999). If no clouds are present between the sun and the pyranometer, the diurnal time series of global irradiance Kd resembles a smooth parabolic-like curve with a maximum at solar noon as seen in Fig. 4a. The DIR, DIF, and Ku irradiances also exhibit symmetric diurnal distributions (note that Ku is plotted as a negative value in Fig. 4 to distinguish it easily from DIF). Figure 4a, then, provides an example of a day without any local cloud cover. The sum of the diffuse irradiance and the product of the direct normal irradiance with the cosine of the solar zenith angle is equal to the global irradiance.

Effects of clouds passing between the sun and the SIRS instruments appear in the diurnal time series as fluctuations in the magnitude of the irradiances from their clear-sky envelopes. Because of reflection from clouds, global irradiance with a cloudy sky can exceed global irradiance with a clear sky. The responses of direct normal, global, and diffuse irradiances to clouds can be seen in Fig. 4b for a partly cloudy day. As cloud cover increases and becomes a deep, uniform layer, diffuse irradiance essentially becomes equivalent to global irradiance, and direct normal irradiance is reduced to near zero. This situation defines an overcast day, an example of which is shown in Fig. 4c.

Following considerable experimentation (Hamm 2003, 32–35), it became evident that the most practical cloud-cover classification, in the sense of independent repeatability of daily estimates of cloud cover, was three mutually exclusive categories (clear, partly cloudy, and overcast), but with appropriate boundaries. A clear day was defined to be a day during which at least 75% of the time the observed global irradiance Kd approximated the expected clear-sky irradiance. Figure 4a is a case in which the percentage of time the day was clear was 100%. Totally clear days seldom occur. Among all six sites, the average number of completely clear days per month was less than 2 for all months except October, November, and December, when it was between 3 and 4.

Using the definition of overcast sky above, a day was classified as overcast if the direct normal irradiance was approximately zero the entire day, diffuse and global irradiances were essentially the same, and their magnitudes did not exceed 200 W m−2 for more than 1 h. An example is shown in Fig. 4c. Note that overcast days may or may not have precipitation. A partly cloudy day, as shown in Fig. 4b, was defined as the intermediate case between a clear day and an overcast day in which a clear sky was observed on the irradiance plots less than 75% of the time for days that were not classified as being overcast. Some subjectivity is introduced into the classification scheme, because the amount of clear sky during each day is visually estimated from the time series of irradiances. Table 2 shows the number of days in each sky category for each year and, for later use, the mean annual albedos. Except for El Reno, Oklahoma, the primary reason the total number of days in either year is fewer than 365 is missing data. The El Reno site was established in mid-1998. More discussion about Table 2 will be provided in section 4a.

Long and Ackerman (2000) describe an automated method for identifying clear and cloudy skies at 1-min intervals using a series of four tests applied to time series of global and diffuse irradiances. This method should be investigated for application to making daily cloud classification more objective than our method. Moreover, it is probably the only practical way of developing a daily cloud classification when many stations are involved over a long period of time.

c. Calculation of daily albedo

The daily broadband albedo is the ratio of the reflected shortwave irradiance to the downwelling shortwave irradiance, with each variable integrated over 1 day defined by sun zenith angles less than or equal to 80°, or from about 1 h after sunrise to 1 h before sunset. The reduced day length was used to minimize propagation of instrument cosine response error (Eppley Laboratory, Inc. 1976). Because the contribution to downward irradiance is small at large zenith angles, the impact of stopping the integration at 80° should be small. Thus, the daily albedo α is given by
i1558-8432-45-1-210-e1
where Ku and Kd are 1-min averages of the hemispheric irradiances. Zenith angles were calculated using the date and time information in the SIRS data files along with the solar position algorithm given by Michalsky (1988).
The SIRS data archive offers two methods to compute albedo. Common to both is the reflected irradiance Ku measured by the single downward-looking pyranometer at 10 m. The downward irradiance Kd can be measured in two independent ways. The first is using an upward-looking pyranometer that yields Kd directly. We will call the method to calculate albedo using the measurement of downward global irradiance the “global” method. The second way to estimate downward global irradiance is to sum the diffuse sky irradiance with the product of the direct normal irradiance and cosine of the zenith angle Z. That is,
i1558-8432-45-1-210-e2
This approach to calculating albedo will be called the “summation” method. In the ideal situation, the estimates of Kd from the two methods should be very close. In lieu of comparing Kd from the two methods, we will, instead, compare the consequent differences in daily albedo at each site. The global method was used throughout our analysis because of the greater reliability than when using the summation method, as was evident from visual inspection of the 1-min data and reports of misalignment of the sun tracker required in the summation method [see Corrective Maintenance Reports online at http://www.ops.sgp.arm.gov/ and Anderberg (1999)].

All calculations of broadband albedo reported in this paper were completed prior to 2001. As a result, corrections for thermal effects in all-black pyranometers that can be performed now were not included. Bush et al. (2000) found that the bias in global irradiance measured by a PSP ranged from about −1 W m−2 in overcast skies to −12 W m−2 in clear skies and from −16 to −23 W m−2 when it was shaded to measure diffuse sky irradiance. Dutton et al. (2001) developed a correction scheme that substantially reduced thermal offset errors for diffuse solar irradiance for any amount of cloud cover. The advancements in irradiance measurements have resulted in improved pyranometer calibrations and their uncertainty. The work by Myers et al. (2001, 2004) at the National Renewable Energy Laboratory (NREL) is especially noted. They conclude, “the best measurement accuracy for broadband radiation is on the order or 3%” (Myers et al. 2004). Given the improvements in calibration procedures at NREL that include corrections for thermal offset, zenith angle calculation, and zenith angle response, Myers et al. (2001) propose that clear-sky global irradiance was underestimated by 2.5%–3% before March of 2000. If the comparatively smaller error in the downward-facing pyranometer were to be ignored, the corresponding daily albedo would be about 2.5%–3% too high on a completely clear day. For such days, which seldom occur, an albedo of 0.25 calculated using the global method actually would be 0.24. Differences of this magnitude will have a negligible effect in interpreting the temporal structure of time series of daily albedo in section 4.

d. Calculation of NDVI

The NDVI is a satellite-derived data product used to monitor the amount and health of green vegetation. The index makes use of the fact that green vegetation reflects a much higher fraction of incident radiation in the near infrared than in the visible red and is given by
i1558-8432-45-1-210-e3
where ρnir is the near-infrared reflectance and ρred is the visible red reflectance (Tucker 1979).

For green vegetation, NDVI has a positive value that increases between leaf emergence and peak chlorophyll production and decreases as vegetation matures and senesces. Typical values of NDVI for vegetated surfaces range from 0.1 to 0.6, with higher values being associated with greater plant density and greenness (McGuffie and Henderson-Sellers 1987). For bare soil and rock surfaces, the near-infrared and visible reflectances are closer in value, which results in a relatively small NDVI. For snow- or ice-covered surfaces and clouds, the visible reflectance is greater than the near-infrared reflectance (Lillesand et al. 2004, 468–469; Dickinson 1983, p. 309) so that for these surfaces NDVI has a negative value.

The sampling frequency of NDVI is determined by the daytime overpass frequency of the particular satellite. In this study, the reflectances in Eq. (3) were measured approximately once per day by the AVHRR on the National Oceanic and Atmospheric Administration (NOAA)-14 polar-orbiting satellite in which the ρnir band (channel 2) is 0.72–1.10 μm and the ρred band (channel 1) is 0.58–0.68 μm. The dataset used was obtained from the ARM data archive through the External Data Center.

As noted above, the daily temporal resolution can be jeopardized by the occurrence of clouds, which effectively block the satellite’s view of the surface, resulting in an anomalously low value of NDVI. To minimize the cloud effect, the maximum value of NDVI from the measurements available over successive 14-day periods was used to represent the NDVI for that period (Goward et al. 1991). Selecting the maximum value should also increase the probability that the associated pixel will be near the nadir (Holben 1986). Thus, if any one of the measurements was taken over cloud-free skies at a given location, at least one of the NDVI values should be larger than zero, its magnitude depending on the amount of green vegetation. A 2-week period of measurements, as opposed to, say, a daily period, was chosen to increase the likelihood of observing at least one clear sky during times of overpass. In the various figures of NDVI that will be shown below, the maximum value for each 14-day period is plotted on the final day of the period.

There are a host of errors associated with NDVI estimates from AVHRR. They include inexact geolocation of a satellite pixel, uncertainty of occurrence of a cloud-free satellite overpass during a 14-day period, no correction for atmospheric scattering, calibration errors, and the effect of changing view angle and sun angle. Goward et al. (1991) and Cihlar et al. (1994) discuss in detail many of errors associated with estimating NDVI. Here we briefly discuss the first three as they relate to our analysis. With regard to the first source, we assumed a 1-pixel (or 1 km, at best) location uncertainty in the north–south and east–west directions. To reduce the resulting error, the average NDVI of a 3 × 3 array of pixels (thus 9 km2) was used. The error in NDVI was estimated by Hamm (2003) to be typically less than ±0.1 units. With regard to the second error, we cannot explicitly determine whether there were no cloud-free overpasses during a given 14-day interval at a site. However, there was no maximum NDVI less than 0.04 from any site. With regard to the third source of error, atmospheric scattering tends to decrease NDVI relative to the true value (McGuffie and Henderson-Sellers 1987) so that there could be a systematic negative bias in the NDVI estimates we used. However, we are more interested in seasonal trends and relative seasonal means of NDVI than its exact magnitude.

e. Calculation of daily precipitation

To relate changes in albedo to changes in surface wetness, time series of daily precipitation were calculated using the ARM data archive and Oklahoma Mesonet (Brock et al. 1995) data files. At the two Kansas EFs, Plevna (EF-4) and Coldwater (EF-8), daily precipitation was measured by ARM Surface Meteorological Observation System rain gauges. At two of the remaining four sites in Oklahoma, Pawhuska (EF-12) and Cordell (EF-22), on-site rain gauge measurements were available for both 1998 and 1999. Daily precipitation totals for the other two sites, Morris (EF-18) and El Reno (EF-19) were not available. In their place, values were used from Oklahoma Mesonet stations at Okmulgee (OKMU) and El Reno (ELRE), respectively. The separation between Morris and OKMU is about 13 km, and the separation between El Reno and ELRE is approximately 2 km. Considering only separation distance, we expect the precipitation measured at ELRE to be, overall, a good substitute for that at El Reno and that from OKMU to be a fair substitute for Morris. Both pairs of differences can be expected to be greater for convective precipitation than for stratiform precipitation.

4. Analysis of daily albedos

a. Introduction

An albedo study based on ground measurements of solar radiation has an advantage over satellite measurements because the observations are more direct and yield uniformly high accuracy in all cloudiness conditions. To investigate whether ground-measured surface albedo depends on the duration of cloud cover, each daily albedo was placed into one of the three sky-cover categories defined in section 3b.

As discussed earlier, Table 2 shows the number of days for which albedo estimates were calculated in each sky-cover category by location and year. Days during which there was believed to be snow or ice on the ground were not included in subsequent analyses. Such days were determined by examining air temperature, precipitation, abnormally high Ku, and albedo. Among the six stations, the albedo for days that were thought to be influenced by snow or ice ranged from 0.25 to 0.92. With the exception of 1998 at El Reno (which opened in the middle of 1998, as noted earlier), the annual number of daily albedos varied between 208 and 349. The criterion for determining whether a day was acceptable for calculating daily albedo was the availability of Ku and Kd data for at least 80% of the possible minutes between daylight hours defined by the solar zenith angle less than 80°. In fact, for the days that were accepted, the vast majority had 100% of the daylight data.

The format for the analyses to follow is to show time series of daily albedo for 1998 and 1999 along with daily precipitation and biweekly NDVI on sets of four-panel figures, one set for each of the six sites (Figs. 5, 8, 11, 13, 16 and 18, described below). With one exception, the axial limits for each variable are the same for each set of figures. The time axis (abscissa) is the day of the year, and across the top of each figure is an approximate scale according to month of year in which the corresponding letter is located at the middle date of the month. Each of these figures will precede or follow a figure that shows the time series of daily albedo differences between the global method and summation method.

b. Plevna (EF-4)

Figure 5 shows time series of daily albedo, daily precipitation, and biweekly maximum NDVI for 1998 and 1999 for Plevna (see Fig. 2). Each daily albedo in Figs. 5a and 5c, and all similar panels at all six sites, were derived using the global method (section 3c), and each was categorized as being clear, partly cloudy, or overcast as defined earlier. Time series of the differences between albedo using the global and summation methods are shown in Fig. 6. Table 3 summarizes the annual range in albedo differences between these two methods and the annual average difference for the number of days in common for each site for each of the two years. No absolute difference in daily albedos at Plevna exceeded 0.009, which was the best overall comparison between methods among the six sites. Figure 6 shows some semblance of an annual cycle with the largest positive differences in the winter months and largest negative differences in the summer months. Recall that in section 3c it was pointed out that both diffuse and global irradiances in 1998 and 1999 were underestimates. By taking differences of albedo, there is a tendency for cancellation of errors. Given the generally small magnitude of these differences, any conclusion to be drawn from the analysis that follows will be independent of the method used to calculate albedo.

Daily precipitation amount is plotted using the right-hand ordinate in Figs. 5a and 5c. Examination of the maximum biweekly NDVI time series in Figs. 5b and 5d reveals the expected seasonal increase in NDVI from about day 90 (end of March) to an annual maximum around day 200 (middle of July) and a sustained decrease between about day 240 (late August) and day 330 (late November). Thus, the times of vegetation green-up and senescence were similar in both years.

Figures 5a and 5c indicate that the daily albedo at Plevna varied primarily between 0.11 and 0.21 during the two years except for the last 3 months of 1999 when it increased from about 0.18 to 0.23. The increase can be explained as follows. Approximately 285 mm of rain accumulated from 20 different rain days between day 270 and day 330 (late September to late November) in 1998, but only 13 mm accumulated from five rain days for the same period in 1999. The resulting lightened color of the surface in late 1999 as compared with late 1998 produced an increase in albedo of 0.05 over the 60-day period. During December of 1999, the daily albedo varied about an average of 0.22 during which time 17 mm accumulated from 7 rain days.

Periods of increasing and decreasing albedo can be observed after day 240 (late August) of 1998 and between days 90 (end of March) and 210 (end of July) of 1999. Figure 7 contains an expanded view of these periods. In Fig. 7a, increasing albedo appears to be associated with clear (open triangles) or partly cloudy (solid triangles) days and a lack of rain, whereas decreasing albedo is typically marked by rain and either overcast (solid circles) or partly cloudy days. The variations range in amplitude from up to 0.05 from days 240 to 300 (late August to late October) to as much as 0.08 during the remainder of the year.

In Fig. 7b, the most noticeable periods of increasing and decreasing albedo occur between days 90 and 135 (end of March and mid-May—no data were available between days 138 and 145). Periods of increasing albedo are not characterized by frequent clear days, but they again are associated with a lack of rainfall. For example, the 0.04 albedo increase between days 115 and 135 occurred while only 7 mm of rainfall were measured. In contrast to this period, between days 160 and 180, the mean albedo decreased only about 0.02 despite the accumulation of over 100 mm of rain. The reason may be that the increase in grass cover, as indicated by the higher NDVI in Fig. 5d, helped to limit the reduction in albedo during this time.

Figures 5 and 7 show that, as defined in section 3b, albedos for clear days are usually indistinguishable from albedos for partly cloudy days. This finding supports an observation made by Kung et al. (1964) that “the albedo values measured under cloudy conditions seem to be comparable with those taken under clear sky conditions.” Therefore, even if the albedos for clear and partly cloudy days had been combined into a single sky-cover category, the conclusions drawn from the separate designations would remain unchanged. The only dependence on sky cover observed in this study is that albedos for overcast days are commonly lower than albedos for the other two sky categories. This relationship can be seen easily by examining Figs. 5a and 5c and will also be observed in similar figures for the other sites. In section 5 we will look in more detail at albedos on overcast days by comparing their values for days with rain and days without rain during dry periods. That we expect overcast days with rain (or on days following rain) to exhibit a reduced albedo is considered to be a result of the reduced backscatter of light resulting from water in the soil or on vegetation canopy or both [see Dickinson (1983) for further explanation].

In summary, periods of increasing and decreasing daily albedos were observed at Plevna during both years. A period of decreasing albedo appears to be caused by surface wetting in response to rain and is frequently followed by a period of increasing albedo resulting from the subsequent drying of soil and vegetation. Albedo variations of this kind were observed over periods of 2 weeks or longer, and their amplitudes seem to be larger during months before or after the peak in NDVI, presumably when more bare soil is exposed. The lowering of the albedo either on or immediately following a day in which there was measurable precipitation was as large as 0.06. Overcast skies seem to keep these albedos low because of low Kd and the consequent slow drying of the wet surface. The low albedo values (<0.15) associated with overcast days are due to either daytime precipitation or surface wetting by antecedent rainfall or both and are especially noticeable during the autumn, winter, and spring months.

c. Coldwater (EF-8)

Figure 8 shows the variation in daily albedo measured during 1998 and 1999 at Coldwater (see Fig. 2). Table 3 shows that the range in albedo difference between the global and summation methods is similar to that at Plevna for 1998 but that the positive limit is considerably greater in 1999. Time series of albedo differences at Coldwater for the 2-yr period are presented in Fig. 9 and tend to parallel those at Plevna with the largest positive differences (>0.01) occurring during December of 1999. There were reports of sun-tracker misalignment prior to 1 March 1999, but none was reported after this date. Again, the differences in albedo are not large enough to change the conclusions that will be drawn from the analysis that follows.

In Fig. 8, the albedos for clear and partly cloudy days ranged between 0.15 and 0.25, with most values greater than 0.17, and albedos for overcast days ranged from 0.13 to 0.22. Inspection of the NDVI time series in Fig. 8d reveals a peak in NDVI between days 150 (end of May) and 210 (late July) in 1999 and a pronounced decrease between days 210 and 240 (late August). During the former period, the peak in NDVI was between 0.05 and 0.15 higher than in 1998, implying greater vegetative cover, likely because of the greater amount of rainfall prior to day 120 in 1999 than in 1998 and its coincidence with the time of normal spring green-up. It is also possible that winter wheat is present in one or more pixels of the 9-pixel average. The basis for this possibility is the peak in NDVI at the end of November of 1998 and early December of 1999, a consequence of the green-up after autumn planting, followed by a decline in NDVI as the winter wheat becomes dormant until spring. Verification would require coregistration of individual NDVI pixels with the same pixels in a land cover classification. More will be said about the influence of winter wheat on NDVI in the sections for El Reno and Cordell.

Comparing Fig. 8 with Fig. 5 for Plevna (111 km northeast; Fig. 2), a similar pattern of increasing albedo is observed between days 270 (late September) and 330 (late November) for 1999 at both locations. The autumn increase at Coldwater is associated with clear skies and lack of rain in the same way as described for Plevna. During this period, the albedo at Coldwater increased by 0.06, in response to only a 17-mm rainfall accumulation from 2 rain days with 39 clear days. Corresponding values at Plevna were an increase in albedo of 0.05 in response to 13-mm accumulation from 5 rain days with 34 clear days. Thus, both sites in western Kansas were strongly influenced by the same synoptic weather regime.

In the expanded views of late spring and summer months in Fig. 10, multiweek variations in albedo were as large as 0.10 between days 150 (end of May) and 270 (late September) in 1998 and 0.08 in 1999. These variations are larger in amplitude than those at Plevna and may be due to the grazed rangeland surface type and its comparatively high fraction of bare soil (see Table 1). The sharp decreases correspond to wetting of the soil resulting from rain events, and the increases are associated with the drying of soil, which is especially rapid with clear skies. For example, between days 150 and 187 of 1998 (Fig. 10a), during which only 11 mm of rain were measured and no overcast days occurred, there was an attendant increase in albedo from 0.20 to almost 0.25. Over the next 4 weeks (days 188–215), 114 mm of rain were measured and the albedo decreased to 0.15, but without any overcast days. Within that time period, there was an initial drop in albedo from 0.24 to 0.17 between days 188 and 195 in response to 49 mm of rain. The next 10 days were dry and mostly clear, and the albedo increased to only 0.18—an example showing that a major rain event can result in low albedo even on nonovercast days following the event. Then it decreased to 0.15 between days 206 and 212 as 64 mm of rain were measured. A similar cyclic pattern in response to dry and wet periods can be observed between days 220 and 270 as the albedo initially increased from 0.16 to 0.19, then decreased slightly to 0.18 in response to a 37-mm rain event, after which it increased again to 0.20.

Figure 10b shows the albedo and precipitation time series for days 150–270 of 1999. During the 5-week period between days 180 (late June) and 213 (end of July), approximately 25 mm of rain were measured, mostly on 1 day, and the albedo increased from 0.19 to 0.22. The effect of this rainfall event produced a slight decrease in albedo, which lasted for about 5 days. Beginning with day 214, nearly 25 mm of rain fell, and the albedo dropped to about 0.19. On the three partly cloudy or clear days following the rain, the albedo was 0.185 or 0.186—another example of the effect a rain event can have on albedo on nonovercast days after the event. No rain was measured during the 3-week period from day 221 to day 240, and the albedo increased to almost 0.24. A number of rainfall events over the following 4 weeks (days 241–270) resulted in another substantial decrease in albedo.

d. Pawhuska (EF-12)

As indicated in Table 1, the Pawhuska site in extreme northern Oklahoma (Fig. 2) is covered by dense, ungrazed tallgrass, varying in color from brown in winter to green in summer. Figures 11b and 11d show a corresponding large annual cycle in NDVI from about 0.1 to around 0.5, with the latter value being considerably higher than that at either Plevna or Coldwater. Thus, if phenology of vegetation has a measurable effect on broadband albedo, as inferred from NDVI, it should be noticeable at this location.

Table 3 shows there were substantial differences between the albedos calculated from the two methods, and, according to Fig. 12, they occurred roughly between days 120 and 230. Corrective maintenance logs report sun-tracker misalignment on days 113, 211, 223, 229, and 236 during 1998, the limits of which bracket the period of large negative differences between the global and summation methods of calculating daily albedo. Thus, sun-tracker misalignment might account for some or most of these differences.

In general, the daily albedo time series in Figs. 11a and 11c varied smoothly over the two years. The extreme variations observed at Coldwater during both years and at Plevna in 1999 are not seen here. Instead, in 1998, the highest albedos occurred from days 135 to 223 (mid-May to 10 August) during which period one can observe small decreases in albedo in response to rainfall events. The sudden decrease in albedo of almost 0.02 on day 223 occurred after 0.51 mm of rain on day 222, well out of proportion to what we saw at Plevna and Coldwater. Such a large response to such a small rainfall amount is unlikely. Just as unlikely, though, is the increase in albedo in days following the 22 mm of rain accumulated on days 210 and 211. There are also a number of days with missing or bad data during the period from day 211 to day 223, resulting in only 5 days with albedo. Of interest is that site maintenance visits occurred on these same days (211 and 223). In both visits, light dust was reported; however, such reports are very common. Thus, the cause of the sudden drop in albedo on day 223 is not evident.

The remainder of the year showed no extended dry periods. The numerous overcast days are related, mainly, to the many days with rain. The albedos for clear and partly cloudy days between rain days were confined to the range of 0.14–0.18.

In Fig. 11c, the daily albedo during the first 6 months of 1999 (days 1–182) for clear and partly cloudy days showed remarkably little variation, remaining between 0.16 and 0.19 despite a rainfall total in excess of 700 mm and a large increase in NDVI. From days 240 to 270 (late August to late September), the albedo decreased slightly from 0.19 to 0.17 in response to 150 mm of rain. For the next 30 days (270–300), there was negligible rainfall, but, in contrast to an immediate noticeable increase in albedo that likely would be observed at Plevna and Coldwater for a similar situation, the albedo began to increase only after about 15 days. Whether the soil surface is wet or dry apparently matters little because of the dense overlying vegetation. The upward trend in albedo continued for the remainder of 1999 but slowed because of both 30 mm of rainfall on day 300 and rainfall that exceeded 100 mm in early December.

The results from Pawhuska suggest that locations in the southern plains with dense vegetation will yield small variations in daily albedo with respect to rain events in comparison with locations with a significant fraction of exposed soil, such as at Coldwater. The broadband reflective properties of dry and wet soil are well known (Coulson and Reynolds 1971; Graser and Van Bavel 1982; Weidong et al. 2002), but they are less well known for wet and dry vegetation. We are not surprised to observe the low average albedo at Pawhuska because of the “trapping” of radiation by the tall dense vegetation (Dickinson 1983). The same phenomenon likely results in an albedo that is less sensitive to rainfall and vegetation color in comparison with albedos from Plevna and Coldwater, where there is less vegetation and more exposed soil.

e. Morris (EF-18)

The Morris site in eastern Oklahoma (Fig. 2) is characterized by native grassland (Table 1). Examination of Fig. 13a shows that from the beginning of the year through 12 September (day 255) the albedos for clear and partly cloudy days varied between 0.19 and almost 0.26 (note the change in limits of albedo in this figure). Again, the response of daily albedo to rainfall events can be seen, particularly around days 120 and 220, with the latter followed by a steep increase in albedo during a nearly month long period of no rain and then a rapid drop with 90 mm of rain on day 256. After day 300 (late October), the day-to-day fluctuation in albedo was unusually large, perhaps the result of the rapid succession of clear and rainy days.

Albedo values of 0.28 during early January and late November of 1999 (Fig. 13c) were the highest among the six EFs analyzed. Only 6 mm of rain fell from 22 December 1998 to 20 January 1999, which may account for the relatively high albedo in mid-January. Between days 270 and 330 (late September to late November) of 1999, the albedo increased from 0.22 to 0.28 in mainly clear weather. During this 2-month period, 30 days were classified as clear at Morris, with only 76 mm of rain measured on four different days at OKMU, located 13 km away (section 3e). As a whole, 1999 can be characterized by a U-shaped variation in albedo, with the lowest values from days 110 to 180 (20 April–end of June). During this period, approximately 540 mm of rain were measured at OKMU.

Overall, the most noteworthy feature at Morris is the high albedo relative to other sites, particularly in the winter months. The NDVI plots in Figs. 13b and 13d do not provide evidence for high albedos, and their occurrence is in contrast to the comparatively low mean annual albedos at Plevna and Pawhuska (Table 2), which also have high fractional vegetation cover (Table 1). Figure 14 is similar to the previous time series plots of albedo differences. It shows an overall increase in differences beginning in early August of 1998 (around day 210), reaching a maximum in December of 1998–January of 1999, and then decreasing to the end of April (around day 120). The increase in albedo difference beginning in September of 1999 tends to parallel that in 1998. The sun tracker was realigned on day 223 in 1998 and on days 84, 168, 264, 292, and 351 in 1999. A connection between days of realignment and corresponding changes in the magnitudes of the differences is not apparent. The actual role, if any, the sun tracker played in accounting for the albedo differences in Fig. 14 is unknown. If the argument were made that albedo differences are due, in fact, to the pyranometer used in the global method, the time series of daily albedo in Fig. 13 would decrease by the magnitude of the differences in Fig. 14, but its general structure would be similar. In summary, the high albedos at Morris appear to be real, and their cause should be investigated.

f. El Reno (EF-19)

The SIRS platform was installed at the El Reno site in central Oklahoma (Fig. 2) in June of 1998, with data available beginning on 8 July (day 189; R. Peppler 2001, personal communication). The range in albedo differences in Table 3 for the data available in 1998 is approximately the same as at Plevna for the entire year, and both sites have the same range of differences in 1999, with two exceptions (days 303 and 304 in Fig. 15). The pattern of maximum positive differences occurring during the winter months and maximum negative differences occurring during the summer months that was observed at Plevna and Coldwater was repeated here, as can be seen in Fig. 15. This will be true also at the Cordell site. In fact, all six sites seem to have a more or less similar rhythm of daily albedo differences.

Figure 16a shows, overall, a large decrease in albedo from day 189 until day 270 (late September). Over this 80-day period, only 44 mm of rain were measured at nearby ELRE, and the NDVI was at or below 0.2, as seen in Fig. 16b. The dry summer conditions at El Reno were typical of an intense drought that affected most of south central and western Oklahoma from April through September of 1998 (Johnson 1998, p. 2). During the first 25 days of the record, the daily albedo varied from 0.23 to 0.25. This period was followed by a sharp drop in response to rainfall of 6.6 mm on day 215 and 0.8 mm on day 224 and a month-long increase, the 5.6 mm of rain at ELRE on day 244 notwithstanding. The increase in albedo ended abruptly with 3.0 mm of rain on days 255 and 256 that reduced the albedo by 0.05. Such a large drop in albedo with so little rainfall may be a consequence of the overall dry summer wherein dry reflective soil sharply darkened, even with a small rainfall amount. We must be mindful, though, that the rainfall at the El Reno site could have been considerably more than shown, given that the separation between it and the Mesonet station (ELRE) is 2 km. Similarly, there may have been negligible rainfall at the ARM El Reno site on day 244 during the month-long increase in albedo. With the return of substantial rainfall around day 260 (middle of September) at ELRE, the albedo at El Reno decreased considerably, and regular rainfall kept the albedo between 0.12 and 0.21 for the remainder of the year.

In 1999, the albedo for clear and partly cloudy skies varied around the value 0.20 until the end of March and then dropped briefly to 0.17, but without explanation. Perhaps rain fell at the El Reno site but not at ELRE on day 89 or 90. Over the next 60 days, until day 150 (end of May), the albedo increased steadily to about 0.23. This increase is worthy of note because over 200 mm of rain were measured during this period. In addition, NDVI reached 0.5—its peak value for the year.

The general structure of the NDVI plots for 1998 and 1999 is consistent with winter wheat being observed (McPherson 2003, p. 67). That is, the peak in NDVI was reached in mid-April, followed by senescence, harvest from late May to early June, the presence of stubble and bare soil until planting in October followed by green-up in November–December, dormancy from December to February, and green-up again until mid-April and a new peak in NDVI. In fact, the El Reno site is located just inside the eastern border of the Oklahoma winter-wheat belt, an area characterized by either winter wheat or a mix of winter wheat and grassland (McPherson 2003, p. 42). Although no winter wheat was observed within a radius of at least 1 km from the El Reno site, because we used a 3 × 3 array of AVHRR pixels, that is, an area of about 9 km2, to determine NDVI (see section 3d), the NDVI plots strongly suggest winter wheat was planted beyond the 1-km radius. The low NDVI during July, August, and September of 1998 is a consequence of the summer drought mentioned earlier. In contrast, the NDVI for the same period in 1999 was much higher because native grasses in the 9-km2 area surrounding the El Reno site likely grew well in response to much greater rainfall. The conclusion from Fig. 16 is that attempts to relate satellite estimates of NDVI and albedo to surface measurements of albedo will be greatly influenced by the uniformity of the vegetation over the area from which the satellite estimates are taken.

The two large rain events before and after day 180 in 1999 had virtually no impact on albedo. The combination of wet soil and high fractional vegetation cover, similar to conditions observed at Pawhuska, apparently diminished the effects of even large rain events on albedo. However, with a much drier second half of the year, more examples of the response of albedo to rainfall and dry periods can be seen. The increase in albedo between days 270 and 330 parallels that at Coldwater (Fig. 8c) and Morris (Fig. 13c), with numerous clear days—33 at El Reno, 39 at Coldwater, and 30 at Morris—but different times of occurrence and amounts of rainfall. Much of the increase in albedo at Coldwater with its low FVC is likely due to the brightening of soil in response to the lack of rain, whereas at El Reno the increase in albedo may be due to brightening associated with vegetation senescence.

g. Cordell (EF-22)

From Table 3 the range in albedo differences between the two methods for both years was from about −0.012 to 0.006. It is interesting that, not only at Cordell but at all sites, with the exception of Morris 1999, where warm-season data are missing, the absolute value of the annual average differences or bias is less than 0.005, despite the comparatively large range in daily values. It seems there is a tendency for the negative warm-season differences to balance the positive cool-season differences. The time series of daily albedo differences is shown in Fig. 17 where, again, the largest positive differences are in winter and the largest negative differences are in summer. The sun tracker at Cordell was realigned on day 224 in 1998 and days 41, 55, 153, and 167 in 1999. Any effects of these realignments are not apparent in Fig. 18.

Figure 18 shows that at this west-central Oklahoma site (Fig. 2), practically all albedos for clear and partly cloudy days ranged between 0.17 and 0.24 during the two years. Between days 150 and 210 (end of May to end of July) of 1998, there was a steady increase in albedo from 0.20 to 0.24, apparently because of only 37 mm of rain during the 2-month period. The albedo varied about the value 0.22 over the next 2 months (between days 210 and 270), during which time only 23 mm of rain were measured. Nevertheless, albedo response to the small rain events on days 216 and 223 can be seen. After day 270, frequent rainfall reduced the clear and partly cloudy daily albedos to between 0.17 and 0.21.

We observe that the pattern of NDVI at Cordell for both years tends to follow that at El Reno (Fig. 16) and is similarly a signature for winter wheat. Like El Reno, Cordell is located in the winter-wheat belt (McPherson 2003, p. 42). Not only are the times of peak NDVI the same at both sites for both years, so also is the time of the lowest NDVI at both sites in 1998, that is, from day 180 to day 270, which coincides with the 3-month-long dry period. We attribute the lower NDVI at Cordell than at El Reno during this period to its lower FVC (Table 1). Because NDVI is derived from a much larger area than is daily albedo, an area that for El Reno and Cordell includes winter wheat, providing a link between these variables for these sites is problematic.

In 1999, fluctuations in daily albedo at Cordell paralleled those at Coldwater (222 km to the north) between days 180 and 300 (July through late October). Figure 19 shows the similarity of the fluctuations between the two locations. The primary difference is that, between days 240 and 255, Coldwater received 44 mm of rain while Cordell received only 8 mm, and the resulting reduction in albedo was correspondingly larger at Coldwater than at Cordell. After day 270, the albedo at Cordell increased with the frequent occurrence of clear days, just as it did at Coldwater and the other four sites. Missing SIRS irradiance data at Cordell from day 306 through day 335 prevented analysis of the albedo for that time period.

As a possible example of the influence of soil type on albedo, consider the subset of albedos in Fig. 19. Table 1 shows that among the six sites studied, these two have the greatest amount of bare soil or lowest FVC, and Cordell has a clayey soil while Coldwater has a sandy soil. A sandy soil drains water faster than a clay-type soil (Dingman 1994, p. 225). Figure 20 is an expanded version of Fig. 19 and shows the decrease in albedo that occurred between days 215 and 220 of 1999 at both locations and the subsequent increase in albedo between days 220 and 240. Figure 20a shows that over the 21-day period from day 220 to day 240, the albedo at Coldwater increased steadily from 0.18 to 0.23, whereas that at Cordell in Fig. 20b increased more slowly and leveled off at approximately 0.21. This result is in accord with the clayey soil at Cordell drying less quickly than the sandy soil at Coldwater and, therefore, remaining darker for a longer period of time. In addition, the clay-type soil at Cordell is likely naturally darker than the sandy soil at Coldwater (Dingman 1994, p. 315).

5. Daily albedo on overcast days

In the previous section we cited numerous examples in which low daily albedo was coincident with rainfall on the same day, because of, at least in part, the wetness of the soil and vegetation. Because many rain days were also overcast days, the question arose: What was the contribution, if any, of an overcast sky to the observed low albedos? In this section we attempt to answer this question.

The approach we took was to compare daily albedos on overcast days with rain with daily albedos on overcast days during dry periods and, further, to compare the latter albedos with those from surrounding clear and partly cloudy days. If the latter comparison were to show no difference, we would conclude that the albedos for rain days with an overcast sky are entirely a consequence of the wetness of the surface.

We analyzed data from the four sites that had a rain gauge. Table 4 shows a comparison of daily albedo on overcast days at these sites for two conditions. The first condition was the coincidence of overcast days that had rainfall equal to or exceeding 2 mm, an amount that should have been sufficient to wet the surface. For example, in column 3 at Plevna for 1998 there were 15 days that met this condition. Column 4 shows that the average albedo for the 15 days was 0.131 and that the range in daily albedo among the 15 days was from 0.107 to 0.160.

The second condition was the occurrence of an overcast day during a dry period. The results are shown in column 5 where, as examples, the albedo at Plevna during 1998 was 0.185 on day 155, and was 0.143 and 0.156, respectively, on successive days 353 and 354. The procedure for selecting these days was as follows. Periods of no rain comprising at least three successive partly cloudy or clear days followed by an overcast day were identified. If the minimum number of days was present, there had to be no significant rainfall (less than 1 mm) on days prior to the first dry day. The purpose was to be reasonably assured that the soil and vegetation were dry. Of the 22 days (see column 5) that met the second condition, there was only one occurrence of minimal requirements. That was day 353 in 1998 at Plevna; 0.25 mm of rain were recorded on the day prior to first day of the 3-day sequence, which, in turn, was preceded by 7 no-rain days. The typical case was that there were many more than a few no-rain days before the overcast day.

A visual comparison of the albedos in column 4 with those in column 5 indicates that the average daily albedo on overcast days with rain was noticeably less than the daily albedo on overcast dry days. That there is a substantial range in albedo in column 4 for any given year and site is probably a result of whether the rain occurred prior to, during, or after daylight hours, or some combination thereof. The relationship between time of occurrence of precipitation and albedo within the course of a day is a logical future study. Such a study could shed light on the effect of water on the dome, which is ignored here.

To test whether the albedo for an overcast day during a dry period was different than the albedos on surrounding clear and partly cloudy days, we averaged albedos for the three previous days and three following days (when possible). The resulting 3-day averages and ranges are shown in columns 6 and 7. Mainly because of precipitation following the overcast day, many of the potential values in column 7 could not be calculated. The results show that 8 of the 12 days in column 5 for which previous and following albedo calculations could be made had an albedo less than any value in the range of albedos (and thus the average). The average albedo of these 12 days is 0.186, whereas the average albedo of the 60 days before and after the overcast day is 0.199. With one exception, the remaining 10 albedos in column 5 were also less than any value in the range of albedos. Their average is 0.178 as compared with 0.199 for the associated 24 previous days. Although the results are not conclusive, given the limited number of cases, they suggest that the daily albedo for an overcast day is less than the albedo for a clear or partly cloudy day with similar surface conditions in the amount of 0.01–0.02. The reduction in daily albedo is in accord with model simulations of albedo using the Moderate-Resolution Atmospheric Transmittance (MODTRAN) algorithm in which Liang et al. (1999) separated the diffuse albedo from the direct albedo and found the former tends to be lower than the latter. In comparison, over all 99 overcast days with rain (column 3), the average albedo was 0.144, or about 0.03–0.04 less than an overcast day during a dry period. From this comparison we conclude that a wet surface is the primary contributor to low albedos on overcast days with rain; the contribution from the overcast sky is secondary.

6. Seasonal and annual albedos and NDVI

Figure 21 is a plot of the mean annual albedos, taken from Table 2, at each of the six locations versus total measured precipitation for both 1998 and 1999. A mean annual albedo is the simple average of four seasonal arithmetic mean albedos, the seasons defined as January–March, April–June, and so on. This type of average was used because of the disparity in the number of available daily albedos in each season because of missing data; the nontraditional definition of seasons was used so they could be easily adapted to annual means. The seasonal means are given in Hamm (2003). Because there are comparatively few overcast days and they primarily occur in the cooler half of the year, only albedos for clear and partly cloudy days were used in calculating seasonal means.

Figure 21 shows that there is little relation between mean annual albedo and annual precipitation. For example, although Pawhuska and Morris both measured between 1100 and 1300 mm both years, Pawhuska had the lowest annual albedos and Morris had the highest. The low albedos at Pawhuska are consistent with native tallgrass vegetation (Table 1 and section 4d), but the cause for the relatively high albedo at Morris remains to be explained. Figure 21 also shows that there is considerable variation in both albedo and precipitation among stations but there was little change in the mean annual albedo at a given location from 1998 to 1999. From Table 2, the difference in mean annual albedo is 0.001 or less at four of the six sites. On the one hand, this is an unexpectedly small difference. On the other hand, there was only one case in which a seasonal mean changed by more than 0.02 from one year to the next (Hamm 2003, his appendix A). At least for this 2-yr period, the seasonal and annual albedos were remarkably stable at these four sites.

It was shown that for three sites—Plevna and Pawhuska in both years and Morris in 1999—the annual cycle of NDVI tends to follow the pattern of lowest value during winter months, followed by increasing value associated with spring green-up to a maximum in summer months, followed by decreasing NDVI in late summer and autumn as senescence begins, and returning to the low winter value. The annual variations of NDVI at El Reno, Cordell, and, perhaps to some extent, at Coldwater seem to be linked to the presence of winter wheat in some of the pixels in the 9-pixel average. With the exception of Morris, where there is a tendency for a ∪-shaped albedo and a ∩-shaped NDVI, in particular in 1999, the annual variation in albedo for the remaining sites appears to be independent of NDVI and to be driven mainly by the occurrence of rainfall events, wet periods, and dry periods in relation to the degree of vegetation cover (contrast the albedo variation at Coldwater with that at Pawhuska).

7. Summary

Time series of daily albedo were analyzed for 1998 and 1999 at six extended facilities in the ARM SGP CART extending from central Kansas to central Oklahoma. Two methods of calculating daily albedo were applied. In one method, the downwelling solar irradiance was measured by an Eppley precision spectral pyranometer, and in the other method it was calculated by summing the diffuse sky irradiance measured by a sun-shielded PSP and the horizontal projection of the direct normal solar irradiance measured by an Eppley normal incidence pyrheliometer, with all measurements made at a height of 1.5 m. Common to both methods was the reflected solar irradiance measured by a downward-looking PSP at 10 m. Daily integration of the downwelling and reflected solar irradiances was performed for sun elevation angles greater than 10° to reduce cosine response error. Because of the not-infrequent occurrence of misalignment of the sun tracker and more missing data in the summation method, the global method was used exclusively to estimate daily albedo. Nevertheless, at each site we presented time series of daily albedo differences between the two methods that show an annual cycle of differences with maxima in the winter and summer months. Given the improvements in calibration procedures discussed in section 3c, it will be interesting to compare albedo differences in later years with those in 1998 and 1999.

Selection of the six sites was based on observed uniformity of surface properties over a 1-km2 area, with the result being that these six are the best among the 24 EFs with respect to horizontal homogeneity of vegetation, uniformity of elevation, and lack of obstructions.

In addition to our basic goal to begin a spatial climatological description of daily albedo for the SGP CART area, our interest was to examine the dependence of daily albedo on precipitation, vegetation, and sky cover. Vegetation was determined by fractional vegetation cover and maximum 2-week normalized difference vegetation index, which is a measure of the degree of green vegetation. FVC was visually estimated and classified as being either high or low. The six sites are widely distributed across the SGP CART area encompassing 143 000 km2 (Fig. 2) and thus have different types of vegetation, different FVC, and different soil types (Table 1). Three categories of daily sky cover were used: clear day—at least 75% of the daylight time the sky was essentially clear, overcast day—the sky was overcast the entire day, and partly cloudy day—neither a clear day nor an overcast day.

With regard to the results, we provide first some general comments aimed particularly at NDVI. Overall, at Plevna, Coldwater, Pawhuska, and Morris, the NDVI tended to the expected annual cycle of lowest values during the winter months and highest values during May–July. A noteworthy departure from the annual variation occurred at Coldwater in 1999 and at Morris in 1998. The former departure was suggestive of winter wheat, and the latter occurred during an extended rainy period during which there may not have been a clear sky during the times of satellite overpass. At El Reno and Cordell, the highest values of NDVI occurred in April, the lowest values occurred in autumn, and a secondary maximum occurred in November–December. Although the sites themselves are located in native grassland and rangeland, respectively, both are also located in the Oklahoma winter-wheat belt. We propose that the existence of wheat fields outside the approximate 0.5-km radius of approximately uniform grassland or rangeland centered at each site, in significant measure accounts for the observed behavior of NDVI. In this regard, it is important to remember that, because of pixel location uncertainty, the determination of NDVI was based on 9 AVHRR pixels centered on the site, thus encompassing at least 9 km2, which, given the observed annual variation in NDVI, apparently included areas of winter wheat. We conclude that efforts to link satellite estimates of NDVI and albedo successfully to surface measurements of albedo will require uniformity of the vegetation over the area from which the satellite estimates are taken. The reason is that a ground-based measurement of albedo is, effectively, a “point” measurement relative to the 1-km2 area of a single AVHRR pixel.

We found that periods of increasing or decreasing daily albedo are strongly related to the occurrence or nonoccurrence of rainfall and FVC or, equivalently, the amount of exposed soil. Systematic decreases and increases in daily albedo in response to rain events followed by dry periods, respectively, were easily observed at five of the six locations. For those sites with large amounts of bare soil (Coldwater and Cordell), there were often changes in time series of daily albedo resembling a “V shape” in response to rainfall events followed by clear, dry periods. For periods with little or no precipitation followed by days of rain, the sequence of albedos conversely had a “Λ shape” appearance. Pawhuska had unit FVC (no bare soil) with the result that variations in albedo showed weak response to rain events.

Other results are as follows:

  1. The maximum difference between mean albedos over the 2-yr period among the six sites is about 4 times the difference in annual means at any site, that is, the spatial variation is about 4 times the interannual variation among the six sites. The spatial variation appears to be due to the pronounced differences in FVC among the six sites, because the site with the smallest 2-yr mean albedo (Pawhuska) had the densest vegetative cover. Therefore, for a site like Pawhuska that has a high FVC, a relatively “flat” daily albedo time series with little seasonal or interannual variation might be predicted. On the other hand, for a site with a low FVC or large proportion of bare soil, for example, Coldwater, the variation of daily albedo cannot be predicted except to the extent that it is highly sensitive to both daily and subseasonal precipitation totals. Thus, as seen in Fig. 21, there is little relation between mean annual albedo and annual precipitation total.

  2. The degree of cloudiness does not have a noticeable effect on daily albedo for days during which the sky was mostly clear from sunrise to sunset or for days that had variable periods of clear and cloudy sky. On the other hand, albedos are observed to be usually much lower for days that have an overcast sky with rain.

  3. The low albedos (≤0.20) observed on overcast days with rain were attributed primarily to the wet surface and secondarily to sky condition.

  4. The annual variation of NDVI (a 9-pixel average) is not a good predictor of the annual variation in albedo. We attribute this, in part, to the enormous mismatch in spatial measurement scales of the two variables coupled with inhomogeneity of the land surface.

Acknowledgments

The authors thank several ARM scientists who provided valuable assistance for this research. Chad Bahrmann and Jim Teske arranged access to ARM extended facilities so that surface conditions could be documented. Mike Splitt generously provided software that was used as a starting point for the albedo calculations. Dr. Peter Minnett provided technical guidance on the use of AVHRR satellite data. Randy Peppler provided valuable information about data quality and availability. David Groff, Don Bond, and Dan Rusk helped us to extract site-maintenance information from the ARM archive. Dr. Tom Stoffel at the National Renewable Energy Laboratory provided us documentation on radiometer inaccuracies. We especially appreciate the numerous insightful comments from Prof. Peter Lamb at the University of Oklahoma and from three anonymous reviewers who read the initial manuscript with great care and suggested substantive ways to improve its quality.

Oklahoma Mesonet precipitation data were provided through the courtesy of the Oklahoma Mesonet Project, a cooperative venture between Oklahoma State University and the University of Oklahoma. Solar radiation, precipitation, and AVHRR reflectance data were obtained from the Atmospheric Radiation Measurement Program sponsored by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, Environmental Sciences Division. The research was funded by ARM Grant 125-6018 through the Cooperative Institute for Mesoscale Meteorological Studies (CIMMS).

REFERENCES

  • Ackerman, T. P., and G. M. Stokes, 2003: The Atmospheric Radiation Measurement Program. Phys. Today, 56 , 3844.

  • Anderberg, M. H. L., 1999: A review of SIRS data quality at the ARM Southern Great Plains site. Ninth ARM Science Team Meeting Proc., San Antonio, TX, ARM, 8 pp.

  • Barnsley, M. J., P. D. Hobson, A. H. Hyman, W. Lucht, J-P. Muller, and A. H. Strahler, 2000: Characterizing the spatial variability of broadband albedo in a semidesert environment for MODIS validation. Remote Sens. Environ., 74 , 5868.

    • Search Google Scholar
    • Export Citation
  • Barr, S., and D. L. Sisterson, 2000: Locale analysis report for the Southern Great Plains. ARM Rep. ARM-00-001, 66 pp.

  • Brock, F. V., K. C. Crawford, R. L. Elliott, G. W. Cuperus, S. J. Stadler, H. L. Johnson, and M. D. Eilts, 1995: The Oklahoma Mesonet: A technical review. J. Atmos. Oceanic Technol., 12 , 519.

    • Search Google Scholar
    • Export Citation
  • Bush, B. C., F. P. J. Valero, A. S. Simpson, and L. Bignone, 2000: Characterization of thermal effects in pyranometers: A data correction algorithm for improved measurement of surface insolation. J. Atmos. Oceanic Technol., 17 , 165175.

    • Search Google Scholar
    • Export Citation
  • Cihlar, J., D. Manak, and N. Voisin, 1994: AVHRR bidirectional reflectance effects and compositing. Remote Sens. Environ., 48 , 7788.

  • Coulson, K. L., and D. W. Reynolds, 1971: The spectral reflectance of natural surfaces. J. Appl. Meteor., 10 , 12851295.

  • Dickinson, R. E., 1983: Land surface processes and climate—Surface albedos and energy balance. Advances in Geophysics, Vol. 25, Academic Press, 305–353.

    • Search Google Scholar
    • Export Citation
  • Dingman, S. L., 1994: Physical Hydrology. Macmillan, 575 pp.

  • Duchon, C. E., and M. S. O’Malley, 1999: Estimating cloud type from pyranometer observations. J. Appl. Meteor., 38 , 132141.

  • Dutton, E. G., J. J. Michalsky, T. Stoffel, B. W. Forgan, J. Hickey, D. W. Nelson, T. L. Alberta, and I. Reda, 2001: Measurement of broadband diffuse solar irradiance using current commercial instrumentation with a correction for thermal offset errors. J. Atmos. Oceanic Technol., 18 , 297314.

    • Search Google Scholar
    • Export Citation
  • Eppley Laboratory, Inc., 1976: Instrumentation for the measurement of the components of solar and terrestrial radiation. Eppley Laboratory, Inc., 12 pp.

  • Goward, S. N., B. Markham, D. G. Dye, W. Dulaney, and J. Yang, 1991: Normalized difference vegetation index measurements from the Advanced Very High Resolution Radiometer. Remote Sens. Environ., 35 , 257277.

    • Search Google Scholar
    • Export Citation
  • Grant, I. F., A. J. Prata, and R. P. Cechet, 2000: The impact of the diurnal variation of albedo on the remote sensing of the daily mean albedo of grassland. J. Appl. Meteor., 39 , 231244.

    • Search Google Scholar
    • Export Citation
  • Graser, E. A., and C. H. M. Van Bavel, 1982: The effect of soil moisture on soil albedo. Agric. Meteor., 27 , 1726.

  • Hamm, K. G., 2003: Albedo observations in the ARM Southern Great Plains Cloud and Radiation Testbed for 1998 and 1999. M.S. thesis, School of Meteorology, University of Oklahoma, 140 pp.

  • Holben, B. N., 1986: Characteristics of maximum-value composite images from temporal AVHRR data. Int. J. Remote Sens., 7 , 14171434.

  • Johnson, H. L., 1998: 1998 Oklahoma annual summary. Oklahoma Climatological Survey, 24 pp.

  • Kung, E. C., R. A. Bryson, and D. H. Lenschow, 1964: Study of a continental surface albedo on the basis of flight measurements and structure of the earth’s surface cover over North America. Mon. Wea. Rev., 92 , 543564.

    • Search Google Scholar
    • Export Citation
  • Liang, S., A. H. Strahler, and C. Walthall, 1999: Retrieval of land surface albedo from satellite observations: A simulation study. J. Appl. Meteor., 38 , 712725.

    • Search Google Scholar
    • Export Citation
  • Lillesand, T. M., R. W. Kiefer, and J. W. Chipman, 2004: Remote Sensing and Image Interpretation. 5th ed. John Wiley and Sons, 763 pp.

  • Long, C. N., and T. P. Ackerman, 2000: Identification of clear skies from broadband pyranometer measurements and calculation of downwelling shortwave cloud effects. J. Geophys. Res., 105 , 1560915626.

    • Search Google Scholar
    • Export Citation
  • McGuffie, K., and A. Henderson-Sellers, 1987: Climatology from space: Data sets for climate monitoring and climate modelling. Remote Sensing Applications in Meteorology and Climatology, R. A. Vaughan, Ed., D. Reidel, 375–389.

    • Search Google Scholar
    • Export Citation
  • McPherson, R. A., 2003: The impact of Oklahoma’s winter wheat belt on the mesoscale environment. Ph.D. dissertation, University of Oklahoma, 197 pp.

  • Michalsky, J. J., 1988: The Astronomical Almanac’s algorithm for approximate solar position (1950-2050). Sol. Energy, 40 , 227235.

  • Minnis, P., S. Mayor, W. L. Smith Jr., and D. F. Young, 1997: Asymmetry in the diurnal variation of surface albedo. IEEE Trans. Geosci. Remote Sens., 35 , 879891.

    • Search Google Scholar
    • Export Citation
  • Myers, D. R., T. L. Stoffel, I. Reda, S. M. Wilcox, and A. M. Andreas, 2001: Recent progress in reducing the uncertainty in and improving pyranometer calibrations. ASME J. Sol. Energy Eng., 124 , 4450.

    • Search Google Scholar
    • Export Citation
  • Myers, D. R., I. Reda, S. Wilcox, and A. Andreas, 2004: Optical radiation measurements for photovoltaic applications: Instrumentation uncertainty and performance. Organic Photovoltaics V, Z. Kafafi and P. Lane, Eds., International Society for Optical Engineering (SPIE Proceedings Vol. 5230), 142–153.

    • Search Google Scholar
    • Export Citation
  • Peppler, R. A., D. L. Sisterson, and P. Lamb, 2000: Site scientific mission plan for the Southern Great Plains CART site: January–June, 2000. Argonne National Laboratory Information and Publishing Division Rep. ARM-00-006, 34 pp.

  • Stokes, G. M., and S. E. Schwartz, 1994: The Atmospheric Radiation Measurement (ARM) Program: Programmatic background and design of the Cloud and Radiation Test Bed. Bull. Amer. Meteor. Soc., 75 , 12011221.

    • Search Google Scholar
    • Export Citation
  • Tucker, C. J., 1979: Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ., 8 , 127150.

  • U.S. DOE, 1990: Atmospheric Radiation Measurement Program Plan. U.S. Department of Energy Rep. DOE/ER-0441, 121 pp.

  • U.S. DOE, 1996: Science Plan for the Atmospheric Radiation Measurement Program (ARM). U.S. Department of Energy Rep. DOE/ER-0670T UC-402, 71 pp.

  • Weidong, L., F. Baret, G. Xingfa, T. Qingxi, Z. Lanfen, and Z. Bing, 2002: Relating soil surface moisture to reflectance. Remote Sens. Environ., 81 , 238246.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Location of the ARM SGP CART.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 2.
Fig. 2.

Overall view of ARM SGP CART (adapted from Peppler et al. 2000). The solid circles are the locations of the six EFs that were studied and are the category-1 sites referred to in section 3a. The remaining EFs are identified by open circles. County lines are shown along with the county name for each EF.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 3.
Fig. 3.

Location of SIRS shortwave instruments within a generic EF. See text for further explanation.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 4.
Fig. 4.

Diurnal time series of the four SIRS component irradiances for (a) a cloud-free day, (b) a partly cloudy day, and (c) an overcast day at Cordell (EF-22, Fig. 2) during mid-October 1998. Variables are defined in text.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 5.
Fig. 5.

(a) Daily albedo for clear, partly cloudy (ptcldy), and overcast (ovc) days and daily precipitation (precip) at Plevna for 1998. (b) Maximum biweekly NDVI at Plevna for 1998. (c), (d) Same as (a) and (b), respectively, but for 1999.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 6.
Fig. 6.

Daily albedo differences (global − summation) for (left) 1998 and (right) 1999 at Plevna.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 7.
Fig. 7.

Expanded views of two time periods of Figs. 5a and 5c: (a) late summer and autumn months of 1998 and (b) late spring and early summer months of 1999.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 8.
Fig. 8.

Same as Fig. 5, but for Coldwater.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 9.
Fig. 9.

Same as Fig. 6, but at Coldwater.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 10.
Fig. 10.

Expanded views of late spring and summer months in Figs. 8a and 8c: (a) 1998 and (b) 1999.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 11.
Fig. 11.

Same as Fig. 5, but for Pawhuska. Note the three extended periods of missing albedo in 1999 during April–July.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 12.
Fig. 12.

Same as Fig. 6, but at Pawhuska. Note the change of scale of the difference axis.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 13.
Fig. 13.

Same as Fig. 5, but for Morris. Note the two extended periods of missing albedo in 1999 from August to September and almost all of December.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 14.
Fig. 14.

Same as Fig. 6, but at Morris. Note the change of scale of the difference axis.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 15.
Fig. 15.

Same as Fig. 6, but at El Reno.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 16.
Fig. 16.

Same as Fig. 5, but at El Reno.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 17.
Fig. 17.

Same as Fig. 6, but at Cordell.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 18.
Fig. 18.

Same as Fig. 5, but for Cordell. Daily albedo data are missing for days 306–335 in 1999.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 19.
Fig. 19.

Expanded views of days 180–330 in 1999 at (a) Coldwater and (b) Cordell.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 20.
Fig. 20.

Expanded views of Fig. 19 for days 210–240.

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Fig. 21.
Fig. 21.

Mean annual albedo vs annual precipitation at each of the six locations (PLV = Plevna, CLD = Coldwater, PAW = Pawhuska, MOR = Morris, ELR = El Reno, and COR = Cordell).

Citation: Journal of Applied Meteorology and Climatology 45, 1; 10.1175/JAM2317.1

Table 1. Geographical information for the six selected EFs. Native grasses are ungrazed; rangeland is grazed. Soil types provided by D. Bond (2005, personal communication).

i1558-8432-45-1-210-t01
Table 2.

Number of days in each sky-cover category and mean daily albedo for each year at each of the six EFs that was investigated. The El Reno site was established in mid-1998.

Table 2.
Table 3.

Daily albedo differences: global method − summation method. The El Reno site was established in mid-1998.

Table 3.
Table 4.

Daily albedo on overcast days.

Table 4.
Save
  • Ackerman, T. P., and G. M. Stokes, 2003: The Atmospheric Radiation Measurement Program. Phys. Today, 56 , 3844.

  • Anderberg, M. H. L., 1999: A review of SIRS data quality at the ARM Southern Great Plains site. Ninth ARM Science Team Meeting Proc., San Antonio, TX, ARM, 8 pp.

  • Barnsley, M. J., P. D. Hobson, A. H. Hyman, W. Lucht, J-P. Muller, and A. H. Strahler, 2000: Characterizing the spatial variability of broadband albedo in a semidesert environment for MODIS validation. Remote Sens. Environ., 74 , 5868.

    • Search Google Scholar
    • Export Citation
  • Barr, S., and D. L. Sisterson, 2000: Locale analysis report for the Southern Great Plains. ARM Rep. ARM-00-001, 66 pp.

  • Brock, F. V., K. C. Crawford, R. L. Elliott, G. W. Cuperus, S. J. Stadler, H. L. Johnson, and M. D. Eilts, 1995: The Oklahoma Mesonet: A technical review. J. Atmos. Oceanic Technol., 12 , 519.

    • Search Google Scholar
    • Export Citation
  • Bush, B. C., F. P. J. Valero, A. S. Simpson, and L. Bignone, 2000: Characterization of thermal effects in pyranometers: A data correction algorithm for improved measurement of surface insolation. J. Atmos. Oceanic Technol., 17 , 165175.

    • Search Google Scholar
    • Export Citation
  • Cihlar, J., D. Manak, and N. Voisin, 1994: AVHRR bidirectional reflectance effects and compositing. Remote Sens. Environ., 48 , 7788.

  • Coulson, K. L., and D. W. Reynolds, 1971: The spectral reflectance of natural surfaces. J. Appl. Meteor., 10 , 12851295.

  • Dickinson, R. E., 1983: Land surface processes and climate—Surface albedos and energy balance. Advances in Geophysics, Vol. 25, Academic Press, 305–353.

    • Search Google Scholar
    • Export Citation
  • Dingman, S. L., 1994: Physical Hydrology. Macmillan, 575 pp.

  • Duchon, C. E., and M. S. O’Malley, 1999: Estimating cloud type from pyranometer observations. J. Appl. Meteor., 38 , 132141.

  • Dutton, E. G., J. J. Michalsky, T. Stoffel, B. W. Forgan, J. Hickey, D. W. Nelson, T. L. Alberta, and I. Reda, 2001: Measurement of broadband diffuse solar irradiance using current commercial instrumentation with a correction for thermal offset errors. J. Atmos. Oceanic Technol., 18 , 297314.

    • Search Google Scholar
    • Export Citation
  • Eppley Laboratory, Inc., 1976: Instrumentation for the measurement of the components of solar and terrestrial radiation. Eppley Laboratory, Inc., 12 pp.

  • Goward, S. N., B. Markham, D. G. Dye, W. Dulaney, and J. Yang, 1991: Normalized difference vegetation index measurements from the Advanced Very High Resolution Radiometer. Remote Sens. Environ., 35 , 257277.

    • Search Google Scholar
    • Export Citation
  • Grant, I. F., A. J. Prata, and R. P. Cechet, 2000: The impact of the diurnal variation of albedo on the remote sensing of the daily mean albedo of grassland. J. Appl. Meteor., 39 , 231244.

    • Search Google Scholar
    • Export Citation
  • Graser, E. A., and C. H. M. Van Bavel, 1982: The effect of soil moisture on soil albedo. Agric. Meteor., 27 , 1726.

  • Hamm, K. G., 2003: Albedo observations in the ARM Southern Great Plains Cloud and Radiation Testbed for 1998 and 1999. M.S. thesis, School of Meteorology, University of Oklahoma, 140 pp.

  • Holben, B. N., 1986: Characteristics of maximum-value composite images from temporal AVHRR data. Int. J. Remote Sens., 7 , 14171434.

  • Johnson, H. L., 1998: 1998 Oklahoma annual summary. Oklahoma Climatological Survey, 24 pp.

  • Kung, E. C., R. A. Bryson, and D. H. Lenschow, 1964: Study of a continental surface albedo on the basis of flight measurements and structure of the earth’s surface cover over North America. Mon. Wea. Rev., 92 , 543564.

    • Search Google Scholar
    • Export Citation
  • Liang, S., A. H. Strahler, and C. Walthall, 1999: Retrieval of land surface albedo from satellite observations: A simulation study. J. Appl. Meteor., 38 , 712725.

    • Search Google Scholar
    • Export Citation
  • Lillesand, T. M., R. W. Kiefer, and J. W. Chipman, 2004: Remote Sensing and Image Interpretation. 5th ed. John Wiley and Sons, 763 pp.

  • Long, C. N., and T. P. Ackerman, 2000: Identification of clear skies from broadband pyranometer measurements and calculation of downwelling shortwave cloud effects. J. Geophys. Res., 105 , 1560915626.

    • Search Google Scholar
    • Export Citation
  • McGuffie, K., and A. Henderson-Sellers, 1987: Climatology from space: Data sets for climate monitoring and climate modelling. Remote Sensing Applications in Meteorology and Climatology, R. A. Vaughan, Ed., D. Reidel, 375–389.

    • Search Google Scholar
    • Export Citation
  • McPherson, R. A., 2003: The impact of Oklahoma’s winter wheat belt on the mesoscale environment. Ph.D. dissertation, University of Oklahoma, 197 pp.

  • Michalsky, J. J., 1988: The Astronomical Almanac’s algorithm for approximate solar position (1950-2050). Sol. Energy, 40 , 227235.

  • Minnis, P., S. Mayor, W. L. Smith Jr., and D. F. Young, 1997: Asymmetry in the diurnal variation of surface albedo. IEEE Trans. Geosci. Remote Sens., 35 , 879891.

    • Search Google Scholar
    • Export Citation
  • Myers, D. R., T. L. Stoffel, I. Reda, S. M. Wilcox, and A. M. Andreas, 2001: Recent progress in reducing the uncertainty in and improving pyranometer calibrations. ASME J. Sol. Energy Eng., 124 , 4450.

    • Search Google Scholar
    • Export Citation
  • Myers, D. R., I. Reda, S. Wilcox, and A. Andreas, 2004: Optical radiation measurements for photovoltaic applications: Instrumentation uncertainty and performance. Organic Photovoltaics V, Z. Kafafi and P. Lane, Eds., International Society for Optical Engineering (SPIE Proceedings Vol. 5230), 142–153.

    • Search Google Scholar
    • Export Citation
  • Peppler, R. A., D. L. Sisterson, and P. Lamb, 2000: Site scientific mission plan for the Southern Great Plains CART site: January–June, 2000. Argonne National Laboratory Information and Publishing Division Rep. ARM-00-006, 34 pp.

  • Stokes, G. M., and S. E. Schwartz, 1994: The Atmospheric Radiation Measurement (ARM) Program: Programmatic background and design of the Cloud and Radiation Test Bed. Bull. Amer. Meteor. Soc., 75 , 12011221.

    • Search Google Scholar
    • Export Citation
  • Tucker, C. J., 1979: Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ., 8 , 127150.

  • U.S. DOE, 1990: Atmospheric Radiation Measurement Program Plan. U.S. Department of Energy Rep. DOE/ER-0441, 121 pp.

  • U.S. DOE, 1996: Science Plan for the Atmospheric Radiation Measurement Program (ARM). U.S. Department of Energy Rep. DOE/ER-0670T UC-402, 71 pp.

  • Weidong, L., F. Baret, G. Xingfa, T. Qingxi, Z. Lanfen, and Z. Bing, 2002: Relating soil surface moisture to reflectance. Remote Sens. Environ., 81 , 238246.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Location of the ARM SGP CART.

  • Fig. 2.

    Overall view of ARM SGP CART (adapted from Peppler et al. 2000). The solid circles are the locations of the six EFs that were studied and are the category-1 sites referred to in section 3a. The remaining EFs are identified by open circles. County lines are shown along with the county name for each EF.

  • Fig. 3.

    Location of SIRS shortwave instruments within a generic EF. See text for further explanation.

  • Fig. 4.

    Diurnal time series of the four SIRS component irradiances for (a) a cloud-free day, (b) a partly cloudy day, and (c) an overcast day at Cordell (EF-22, Fig. 2) during mid-October 1998. Variables are defined in text.

  • Fig. 5.

    (a) Daily albedo for clear, partly cloudy (ptcldy), and overcast (ovc) days and daily precipitation (precip) at Plevna for 1998. (b) Maximum biweekly NDVI at Plevna for 1998. (c), (d) Same as (a) and (b), respectively, but for 1999.

  • Fig. 6.

    Daily albedo differences (global − summation) for (left) 1998 and (right) 1999 at Plevna.

  • Fig. 7.

    Expanded views of two time periods of Figs. 5a and 5c: (a) late summer and autumn months of 1998 and (b) late spring and early summer months of 1999.

  • Fig. 8.

    Same as Fig. 5, but for Coldwater.

  • Fig. 9.

    Same as Fig. 6, but at Coldwater.

  • Fig. 10.

    Expanded views of late spring and summer months in Figs. 8a and 8c: (a) 1998 and (b) 1999.

  • Fig. 11.

    Same as Fig. 5, but for Pawhuska. Note the three extended periods of missing albedo in 1999 during April–July.

  • Fig. 12.

    Same as Fig. 6, but at Pawhuska. Note the change of scale of the difference axis.

  • Fig. 13.

    Same as Fig. 5, but for Morris. Note the two extended periods of missing albedo in 1999 from August to September and almost all of December.

  • Fig. 14.

    Same as Fig. 6, but at Morris. Note the change of scale of the difference axis.

  • Fig. 15.

    Same as Fig. 6, but at El Reno.

  • Fig. 16.

    Same as Fig. 5, but at El Reno.

  • Fig. 17.

    Same as Fig. 6, but at Cordell.

  • Fig. 18.

    Same as Fig. 5, but for Cordell. Daily albedo data are missing for days 306–335 in 1999.

  • Fig. 19.

    Expanded views of days 180–330 in 1999 at (a) Coldwater and (b) Cordell.

  • Fig. 20.

    Expanded views of Fig. 19 for days 210–240.

  • Fig. 21.

    Mean annual albedo vs annual precipitation at each of the six locations (PLV = Plevna, CLD = Coldwater, PAW = Pawhuska, MOR = Morris, ELR = El Reno, and COR = Cordell).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 297 117 6
PDF Downloads 119 37 2