Abstract

Several changes in U.S. observational practice [in particular, the introduction of the Automated Surface Observing System (ASOS) in the early 1990s] have led to a challenging heterogeneity of time series of most ground-based cloud observations. In this article, an attempt is made to preserve/restore the time series of average low cloud cover (LCC) over the country up to the year 2001 using cloud sky condition and cloud-base height information collected in the national archive data and to describe its spatial and temporal variability. The switch from human observations to ASOS can be bridged through the use of frequency of overcast/broken cloudiness. During the past 52 yr, the nationwide LCC appears to exhibit a significant increase but all of this increase occurred prior to the early 1980s and thereafter tends to decrease. This finding is consistent with similar changes in the frequency of days with precipitation. When the cloud-type information was still available (i.e., during the pre-ASOS period), it was found that the overall LCC increase was due to the increase in stratiform and cumulonimbus cloud occurrences while cumulus cloud frequency decreased.

1. Introduction

Visual observations of cloudiness are important baseline datasets in climate studies. They have been used in analyses of climate change and meteorological processes, to evaluate cloud–climate interactions in climate models, and to validate radiance-derived satellite cloud characteristics. Multidecadal time series of total cloud cover (TCC) and cloud-type frequency over the United States and many other countries from national meteorological station networks have been documented, along with their association with other climatic variables (e.g., Henderson-Sellers 1989; Angell 1990; Karl and Steurer 1990; Sun et al. 2001). This paper addresses the spatial and temporal changes in low cloud cover (LCC) since the late 1940s over the United States, which are derived from sky condition and cloud height information collected in the national archive [the Surface Airways Hourly Data (TD-3280), Steurer and Bodosky 2000].

2. Low cloudiness observations in the United States

LCC is defined as the portion of sky covered by clouds whose base heights are below 2 km. They include cumulus (Cu), cumulonimbus (Cb), stratus (St), stratocumulus (Sc), fractocumulus (Fc), fractostratus (Fs), and nimbostratus (Ns). Obscuring phenomena, such as fog and/or precipitation, are considered as a special low cloud type. The United States did not directly report LCC in its routine weather observation. Instead, the LCC information is in Surface Airways Hourly Data, as sky condition corresponding to a particular cloud layer, the total amount of sky covered by the first two and first three cloud layers (C2C3), low cloud types and/ or cloud-layer-base heights. C2C3 records are available for approximately 40%–50% of all cloud observations from the late 1940s up to the introduction of the Automated Surface Observing System (ASOS) in the early 1990s. Sky condition information (clear, scattered, broken, or overcast skies) is present in almost all cloud observations. Several changes were made in U.S. historical cloudiness observing practices (Karl and Steurer 1990; Sun et al. 2001). The most critical in the past 50 yr was the introduction of ASOS in the early 1990s. Starting in September 1992, ASOS has gradually replaced human observers, and has become the major nationwide provider of cloud observations after 1995. The implementation of ASOS in the U.S. meteorological network resulted in homogeneity of total cloud-cover time series no longer being preserved and cloud-type and cloud-opacity reports no longer being made. As described in section 4, ASOS has also systematically underestimated LCC by misreporting scattered clouds as clear skies. ASOS has changed the principle of traditional cloud observations. Human observers report weather conditions in their area at a fixed time to produce a “snapshot” of cloud condition, while ASOS uses laser beam ceilometers to sample sky conditions in a small vertical column of air (up to ∼3500 m) at 30-s intervals, and then averages these data over the recent 30 min to get the time-averaged cloud observations. According to the ASOS manual [the (National Weather Service) NWS 1998], a portion of “few scattered” cloud conditions, which are usually reported by observers as fair weather Cu, may be reported as “clear” sky by ASOS ceilometers. This bias, as expected, leads to the overestimation of both low cloud amount when present (LAWP) and “no-low-clouds” frequency.

Another significant change in cloud observing practice occurred in July 1996 when the NWS changed units of cloud observations to eighths (octas), instead of tenths, and that has also affected sky condition reports. Before ASOS, sky condition was reported as clear, “thin scattered” and “scattered” (for sky coverage from 0.1 to 0.5), “thin broken” and “broken” (for sky coverage from 0.6 to 0.9), “thin overcast,” “overcast,” and “obscuration.” According to Steurer and Bodosky (2000), these categories are converted to sky coverage in 0%, 30%, 30%, 75%, 75%, 100%, 100%, and 100%, respectively. ASOS does not distinguish between thin and opaque clouds. Between the 1992 ASOS implementation and 1996, sky condition was therefore reported as clear, scattered, broken, overcast, and obscuration. After July 1996, definitions of sky conditions have been changed by the introduction of a few scattered category (that describes skies with less than two-eighths). Thus, for the period after July 1996, categories clear, few scattered, broken, overcast, and obscuration, are converted to sky coverage in 0%, 19%, 44%, 44%, 75%, 75%, 100%, and 100%, respectively.

3. Methods

In order to keep LCC time series as homogeneous as possible and updated, in this note sky condition information is used to evaluate LCC variations. Cloud cover for particular cloud types is calculated using cloud-layer type and amount information in Surface Airways Hourly Data but only for the pre-ASOS period. The latter approach is used to construct the stratiform (St, Sc, Fs, Fc, Ns, and fog) and convective (Cu and Cb) cloud- cover time series (shown in Fig. 3).

Fig. 3.

Daytime annual time series of low cloud cover, stratiform cloud cover, and convective cloud cover area averaged over the contiguous United States for the pre-ASOS period. Annual number of days with precipitation above 0.5 mm during 1949–2002 is depicted. Linear trends during 1949–94 for low, stratiform, and convective cloud cover and precipitation days are 2.3% per 10 yr, 4.1% per 10 yr, −3.6% per 10 yr, and 3.6% per 10 yr, respectively, and for precipitation days during 1949–2002 is 2.4% per 10 yr. All the trend estimates are presented in percent of climatological mean (1961–90) and are significant at the 0.05 or better levels

Fig. 3.

Daytime annual time series of low cloud cover, stratiform cloud cover, and convective cloud cover area averaged over the contiguous United States for the pre-ASOS period. Annual number of days with precipitation above 0.5 mm during 1949–2002 is depicted. Linear trends during 1949–94 for low, stratiform, and convective cloud cover and precipitation days are 2.3% per 10 yr, 4.1% per 10 yr, −3.6% per 10 yr, and 3.6% per 10 yr, respectively, and for precipitation days during 1949–2002 is 2.4% per 10 yr. All the trend estimates are presented in percent of climatological mean (1961–90) and are significant at the 0.05 or better levels

Due to the significant bias in reports of scattered clouds during the ASOS period, only the LCC time series for the pre-ASOS period (1949–94) is used. However, the ASOS period is used in the low cloud overcast and broken frequency time series, making them the only up-to-date cloudiness characteristic that can be tracked nationwide since the post-WWII period. The assumption that ASOS reliably reported the broken/overcast sky conditions appears to be confirmed by a comparison with independently derived time series of precipitation frequency.

Practically the entire first-order station network was in place over the United States in the late 1940s. This study uses 193 stations well distributed across the contiguous United States (cf. Fig. 2 of Sun et al. 2001), 13 stations in Alaska, and 4 stations in Hawaii, all of which have at least 25 yr of data within the reference period of 1961–90. National and regional (nine regions delimited in Fig. 1, Alaska, and Hawaii) averaging is based on the daytime low cloud characteristics in order to avoid the surface nighttime cloud detection bias due to insufficient illumination (Hahn et al. 1995). The daytime monthly LCC is estimated by averaging all 3-hourly values between 0900 and 1800 LT in each month. Daily precipitation observations at the same stations were used to derive the frequency of days with notable precipitation (above 0.5 mm) and these are compared to the LCC time series.

Fig. 1.

Regional partition of the contiguous United States used in this study

Fig. 1.

Regional partition of the contiguous United States used in this study

The following procedures were performed to calculate the area-averaged LCC and precipitation frequency time series over the contiguous United States. First, the station data within 2.5° × 2.5° grid cells are averaged then the mean monthly value for the reference period (1961–90) is calculated for each grid cell. Next, the monthly anomaly is calculated by subtracting the 1961– 90 mean value at each grid. Finally, the anomalies and means at all grids are area averaged to create the monthly time series. These monthly time series are used to calculate seasonal and annual time series. For regional averages over Alaska and Hawaii, a similar procedure is used, but instead of averaging within grid cells direct arithmetic averaging is employed.

4. Results

Daytime LAWP, area averaged over the contiguous United States for the ASOS period (1996–2001) (not shown), is about 67%, which exceeds four standard deviations from mean LAWP calculated from the human- observed period (1949–94), which is 62%. The abrupt increase in LAWP occurring around 1995–96 is also “reasonably” coincident with a jump of no-low-clouds frequency. The no-low-clouds frequency during 1996– 2001 is 55%, 10% more than that in the early 1950s, the period of most severe droughts in the past 50 yr. Apparently a portion of scattered clouds could also be misreported as clear skies and the statement in the ASOS user's guide (i.e., up to 20% misreports of few scattered clouds as clear skies) is an underestimation. This finding eliminated the ASOS period from the analyses of the time series of scattered cloud conditions.

The daytime LCC time series during 1949–94 area averaged over the contiguous United States (Figs. 2 and 3) show significant upward trends in winter, summer, autumn, and annual means. In this article, linear trends are estimated using the method of median of pairwise slope and its statistical significance is assessed using the Spearman rank-order correction technique. The advantages of these nonparametric tools include that no assumptions regarding the underlying statistical distribution is required and the results are also less affected by the presence of outliers (Lanzante 1996). The low LCC values of the early 1950s, which occurred during severe droughts (cf. Fig. 3), contribute much to these trends. Figure 2 also indicates that a disproportionally large fraction of the upward trend in annual LCC is the consequence of the appreciable autumn increase. An increase in LCC is seen in eight of nine regions of the contiguous United States, particularly in the central and eastern United States (Table 1; Fig. 1).

Fig. 2.

Daytime seasonal low cloud cover time series area averaged over the contiguous United States for the pre-ASOS period. Linear trends for winter, summer, and autumn during 1949–94 are 1.8% per 10 yr, 2.1% per 10 yr, and 3.6% per 10 yr, respectively, and are statistically significant at the 0.05 or better levels. Trends are presented in percent of climatological mean (1961–90). For better visualization, y-axis for winter and summer series is on the left and for spring and autumn series on the right. In this figure and Figs. 3 and 4 and Tables 1 and 2, the linear trends and their significance are assessed using techniques of median of pairwise slope and Spearman rank-order correlation (Lanzante 1996), respectively. As in Fig. 4, DJF, MAM, JJA, and SON are defined as winter, spring, summer, and autumn, respectively

Fig. 2.

Daytime seasonal low cloud cover time series area averaged over the contiguous United States for the pre-ASOS period. Linear trends for winter, summer, and autumn during 1949–94 are 1.8% per 10 yr, 2.1% per 10 yr, and 3.6% per 10 yr, respectively, and are statistically significant at the 0.05 or better levels. Trends are presented in percent of climatological mean (1961–90). For better visualization, y-axis for winter and summer series is on the left and for spring and autumn series on the right. In this figure and Figs. 3 and 4 and Tables 1 and 2, the linear trends and their significance are assessed using techniques of median of pairwise slope and Spearman rank-order correlation (Lanzante 1996), respectively. As in Fig. 4, DJF, MAM, JJA, and SON are defined as winter, spring, summer, and autumn, respectively

Table 1.

Climatology and linear trends of the annual daytime low cloud cover. Trend estimates are based on the time series during 1949– 94 (except 1952–94 for Hawaii) and are presented in precent of the climatological mean of 1961–90. Trend estimates in boldface are statistically significant at the 0.05 or better levels. Variances ascribed by the trend (R2 ) are also shown

Climatology and linear trends of the annual daytime low cloud cover. Trend estimates are based on the time series during 1949– 94 (except 1952–94 for Hawaii) and are presented in precent of the climatological mean of 1961–90. Trend estimates in boldface are statistically significant at the 0.05 or better levels. Variances ascribed by the trend (R2 ) are also shown
Climatology and linear trends of the annual daytime low cloud cover. Trend estimates are based on the time series during 1949– 94 (except 1952–94 for Hawaii) and are presented in precent of the climatological mean of 1961–90. Trend estimates in boldface are statistically significant at the 0.05 or better levels. Variances ascribed by the trend (R2 ) are also shown

The increase in LCC over the contiguous United States is mainly due to a steady increase in stratiform cloud, as convective cloud cover was decreasing (Fig. 3). Sun et al. (2001) found similar changes in their respective frequencies along with increased Cb frequency (particularly in the intermediate seasons), and decreased convective cloudiness that was completely dominated by a decreasing Cu frequency.

The presence of clouds does not always correspond to precipitation. But, over the contiguous United States, changes in cloudiness, including TCC (Angell 1990) and low cloud frequency (Sun et al. 2001), agreed with precipitation frequency on both interannual and multidecadal time scales. Figure 3 demonstrates a close relationship between nationwide annual LCC and precipitation days, attesting to the robustness of the LCC time series. Furthermore, the overall decrease in precipitation days during the past 20 yr supports the inference regarding the nationwide LCC changes during the second half of the twentieth century: the LCC appears to exhibit a significant increase in the past 52 yr but this increase occurred prior to the early 1980s, after which LCC tends to decrease (Table 1; Fig. 3).

Comparable (human observed) information for clear sky and scattered cloudiness from the ASOS period after the mid-1990s was not available. Nevertheless, information about broken and overcast sky covers, which account for about 70% of the mean LCC (and about 90% of interannual variability), provides some indication of LCC changes during the entire 1949–2001 period. Changes in broken/overcast sky frequency are generally similar to those in overcast sky frequency, both spatially and temporally (Table 2) and Fig. 4 presents only the time series for the overcast sky frequency [(OSF) accounts for about 53% of the mean LCC). Annual OSF explained 89% of the variance in the annual LCC time series during 1949–94, suggesting the similarity of low-frequency changes between them. The OSF values in three recent dry years (1999–2001) are below normal, thus making the upward trends over the 1949– 2001 period smaller than those for 1949–94 (except winter). Significant OSF increases are still exhibited in winter, autumn and annually, and regionally in the eastern part of the country, including the upper Mississippi, the Northeast, the South, the Southwest, and the southeast region (Table 2). In California and Nevada a strong decrease in the OSF time series is mostly caused by a shift from higher OSF values before 1975–76 to lower values afterward.

Table 2.

Same as Table 1 but for overcast and broken/overcast sky frequencies (in italics) time series during 1949–2001 (except 1952–2001 for Hawaii)

Same as Table 1 but for overcast and broken/overcast sky frequencies (in italics) time series during 1949–2001 (except 1952–2001 for Hawaii)
Same as Table 1 but for overcast and broken/overcast sky frequencies (in italics) time series during 1949–2001 (except 1952–2001 for Hawaii)
Fig. 4.

Same as Fig. 2 but for daytime overcast sky frequency during 1949–2001. Linear trends for winter and autumn are 2.3% per 10 yr and 3.5% per 10 yr, respectively, and are statistically significant at the 0.05 or better levels. The averaging after 1994 was calculated using the ASOS data

Fig. 4.

Same as Fig. 2 but for daytime overcast sky frequency during 1949–2001. Linear trends for winter and autumn are 2.3% per 10 yr and 3.5% per 10 yr, respectively, and are statistically significant at the 0.05 or better levels. The averaging after 1994 was calculated using the ASOS data

Figure 5 shows OSF annual time series for three large regions over the contiguous United States and nationwide. It illustrates the close relationship between OSF and precipitation event frequency over these regions. The OSF matches well with precipitation days before and after the ASOS introduction, suggesting that ASOS reporting of overcast LCC is consistent with the pre- ASOS period. The observed similarity of the interdecadal variability of OSF and LCC time series and the up-to-date information about nonmonotonic OSF changes (Fig. 5) tend to support the conclusion that nonmonotonic LCC behavior during the second half of twentieth century is as described above.

Fig. 5.

Daytime annual overcast sky frequency and days with precipitation above 0.5 mm over the contiguous United States and three regions: south, Missouri River basin, and northeast. R1 and R2 are the correlation coefficients between overcast frequency and precipitation days during 1949–2001 based on unfiltered data and data after the linear trends were removed, respectively. The cloud averaging after 1994 was calculated using the ASOS data

Fig. 5.

Daytime annual overcast sky frequency and days with precipitation above 0.5 mm over the contiguous United States and three regions: south, Missouri River basin, and northeast. R1 and R2 are the correlation coefficients between overcast frequency and precipitation days during 1949–2001 based on unfiltered data and data after the linear trends were removed, respectively. The cloud averaging after 1994 was calculated using the ASOS data

No significant LCC trends are found in Alaska during 1949–94. Instead, an abrupt increase in LCC around 1976 is apparent. Similar changes are also present in TCC and in the overcast frequency time series. These changes may be related to the widespread decadal climate shift of 1976 over the North Pacific basin (Miller et al. 1994). There is a significant decrease in LCC over the Hawaiian Islands. This phenomenon is also shown in overcast and broken/overcast sky frequency for the entire 1949–2001 period (Table 2), supporting the notion that a systematic change in the radiation energy budget might have occurred in the Tropics (Wielicki et al. 2002).

Sun (2003) showed the possibility of reconstruction of TCC for the ASOS period with the help of the International Satellite Cloud Climatology Project (ISCCP) D2 data (Rossow and Schiffer 1999) using their high temporal correlation with human-observed TCC. A similar comparison between the human-observed LCC and LCC derived from the (infrared) IR-only analysis in the ISCCP D2 archive, found no meaningful relationships. However, the LCC estimates derived by the sum of the ISCCP D2 cloud-type amounts detected from the visible/infrared (VIS/IR) analysis (which includes amounts of Cu, Sc, St, Ns, and deep convection clouds) shows a significant correlation with the human-observed LCC. This is mostly due to the high correlation of ground- and satellite-based estimates of stratiform clouds (Sun 2003). The above analysis clearly indicates that the combined VIS/IR analysis is much better than the IR-only analysis in detecting low clouds. But, unlike TCC, the ISCCP-based LCC estimates do not show below-average values for the recent dry years (1999–2001), indicating that ISCCP might not be able to accurately distinguish precipitating cloud types. Thus, no attempt was made in this study to use the ISCCP cloud information to reconstruct ground-based estimates of LCC for the ASOS period.

5. Summary

This study, combined with Sun et al. (2001), shows the pattern of changes in cloudiness over the United States. During the pre-ASOS period (1949–94), low cloud cover increased, and stratiform and cumulonimbus cloud cover increased, while the opposite appeared to be true for cumulus clouds. When extending low cloudiness information to the past 8 yr, with the help of precipitation days and ASOS broken/overcast information, the upward trend in LCC became weaker and nonmonotonic changes can be qualitatively described as follows: LCC increased from the late 1940s to the early 1980s and then decreased to 2002. Similar behavior of nationwide TCC was also noticed by Sun (2003).

Over most of the globe, daytime low clouds cool the surface by their albedo effect and warm the cloudy atmosphere by their absorption of shortwave radiation (Cess et al. 1995), thus reducing the temperature difference between the surface and lower troposphere. Interdecadal changes in LCC revealed in this study may contribute to interdecadal changes in this difference, as suggested by radiosonde and satellite data (Gaffen et al. 2000; Brown et al. 2000). Conversely, due to close coupling of the LCC variability to the thermodynamic characteristics of the atmospheric boundary layer (Klein 1997; Norris 1998), changes in large-scale atmospheric circulation and thermodynamic characteristics could also play a role in the observed LCC changes. Multidecadal variations in LCC and its composition over a region as large as the United States (summarized in this note) are considerable and their climatic causes and consequences are worthy of further study.

Acknowledgments

We thank Sharon Leduc, John Bates, Wallis Trevor, Russell S. Vose, and Anne Waple for helpful suggestions, and Devoyd Ezell for technique support. Insightful comments made by three anonymous reviewers are greatly appreciated. NASA Grant GWEC- 0000-0052 and the NOAA Climate and Global Change Program (Climate Change and Detection Element) provided support for this study.

REFERENCES

REFERENCES
Angell
,
J. K.
,
1990
:
Variation in United States cloudiness and sunshine duration between 1950 and the drought year of 1989.
J. Climate
,
3
,
296
308
.
Brown
,
S. J.
,
D. E.
Parker
,
C. K.
Folland
, and
I.
Macadam
,
2000
:
Decadal variability in the lower-tropospheric lapse rate.
Geophys. Res. Lett
,
27
,
997
1000
.
Cess
,
R. D.
, and
Coauthors
,
1995
:
Absorption of solar radiation by clouds: Observations versus models.
Science
,
267
,
496
499
.
Gaffen
,
D. J.
,
B. D.
Santer
,
J. S.
Boyle
,
J. R.
Christy
,
N. E.
Graham
, and
R. J.
Ross
,
2000
:
Multidecadal changes in the vertical temperature structure of the troposphere.
Science
,
287
,
1242
1245
.
Hahn
,
C. J.
,
S. G.
Warren
, and
J.
London
,
1995
:
The effect of moonlight on observation of cloud cover at night, and application to cloud climatology.
J. Climate
,
8
,
1429
1446
.
Henderson-Sellers
,
A.
,
1989
:
North American total cloud amount variations this century.
Global Planet. Change
,
75
,
175
194
.
Karl
,
T. R.
, and
P. M.
Steurer
,
1990
:
Increased cloudiness in the United States during the first half of the twentieth century: Fact or fiction?
Geophys. Res. Lett
,
17
,
1925
1928
.
Klein
,
S.
,
1997
:
Synoptic variability of low-cloud properties and meteorological parameters in the subtropical trade wind boundary layer.
J. Climate
,
10
,
2018
2039
.
Lanzante
,
J. R.
,
1996
:
Resistant, robust and nonparametric techniques for the analysis of climate data: Theory and examples, including applications to historical radiosonde station data.
Int. J. Climatol
,
16
,
1197
1226
.
Miller
,
A. J.
,
T. P.
Barnett
,
D. R.
Cayan
,
N. E.
Graham
, and
J. M.
Oberhuber
,
1994
:
The 1976–77 climate shift of the Pacific Ocean.
Oceanography
,
7
,
21
26
.
Norris
,
J. R.
,
1998
:
Low cloud type over the ocean from surface observations. Part I: Relationship to surface meteorology and the vertical distribution of temperature and moisture.
J. Climate
,
11
,
369
381
.
NWS
,
1998
:
Automated Surface Observing System (ASOS) user's guide.
National Oceanic and Atmospheric Administration, Washington, DC, 61 pp
.
Rossow
,
W. B.
, and
R. A.
Schiffer
,
1999
:
Advances in understanding clouds from ISCCP.
Bull. Amer. Meteor. Soc
,
80
,
2261
2287
.
Steurer
,
P. M.
, and
M.
Bodosky
,
2000
:
Surface airways hourly (Tech. Doc. 3280) and airways solar radiation (Tech. Doc. 3281).
National Climatic Data Center, Asheville, NC, 50 pp
.
Sun
,
B.
,
2003
:
Cloudiness over the contiguous United States: Contemporary changes observed using ground-based and ISCCP D2 data.
Geophys. Res. Lett., 30, 1053, doi:10.1029/2002GL015887
.
Sun
,
B.
,
P. Ya
Groisman
, and
I. I.
Mokhov
,
2001
:
Recent changes in cloud-type frequency and inferred increases in convection over the United States and the former USSR.
J. Climate
,
14
,
1864
1880
.
Wielicki
,
B. A.
, and
Coauthors
,
2002
:
Evidence for large decadal variability in the tropical mean radiative energy budget.
Science
,
295
,
841
844
.

Footnotes

Corresponding author address: Dr. Bomin Sun, National Climatic Data Center, Federal Building, 151 Patton Avenue, Asheville, NC 28801. Bomin.Sun@noaa.gov