Volume 45, Issue 9 p. 4281-4289
Research Letter
Open Access

A Revisit of Global Dimming and Brightening Based on the Sunshine Duration

Yanyi He

Yanyi He

State Key Laboratory of Earth Surface Processes and Resource Ecology, College of Global Change and Earth System Science, Beijing Normal University, Beijing, China

Search for more papers by this author
Kaicun Wang

Corresponding Author

Kaicun Wang

State Key Laboratory of Earth Surface Processes and Resource Ecology, College of Global Change and Earth System Science, Beijing Normal University, Beijing, China

Correspondence to: K. Wang,

[email protected]

Search for more papers by this author
Chunlüe Zhou

Chunlüe Zhou

State Key Laboratory of Earth Surface Processes and Resource Ecology, College of Global Change and Earth System Science, Beijing Normal University, Beijing, China

Search for more papers by this author
Martin Wild

Martin Wild

Institute for Atmospheric and Climate Science, ETH Zürich, Zürich, Switzerland

Search for more papers by this author
First published: 19 April 2018
Citations: 70

Abstract

Observations show that the surface incident solar radiation (Rs) decreased over land from the 1950s to the 1980s and increased thereafter, known as global dimming and brightening. This claim has been questioned due to the inhomogeneity and low spatial-temporal coverage of Rs observations. Based on direct comparisons of ~200 observed and sunshine duration (SunDu) derived Rs station pairs, meeting data record lengths exceeding 60 months and spatial distances less than 110 km, we show that meteorological observations of SunDu can be used as a proxy for measured Rs. Our revised results from ~2,600 stations show global dimming from the 1950s to the 1980s over China (−1.90 W/m2 per decade), Europe (−1.36 W/m2 per decade), and the United States (−1.10 W/m2 per decade), brightening from 1980 to 2009 in Europe (1.66 W/m2 per decade) and a decline from 1994 to 2010 in China (−1.06 W/m2 per decade). Even if 1994–2010 is well known as a period of global brightening, the observed and SunDu-derived Rs over China still exhibit declining trends. Trends in Rs from 1923 to 1950 are also found over Europe (1.91 W/m2 per decade) and the United States (−1.31 W/m2 per decade), but the results in Europe may not well represent the actual trend for the European continent due to poor spatial sampling.

Key Points

  • Spatial-temporal comparison of the observed and sunshine duration-derived surface incident solar radiation over China, Europe, and the United States
  • Surface incident solar radiation derived by sunshine duration is proven to depict the observed long-term variability
  • A revisit of global dimming and brightening in China, Europe, and the United States since 1901

Plain Language Summary

Ground-based observations of the surface incident solar radiation (Rs) reveal the phenomena known as global dimming and brightening, that is, a downtrend over land from the 1950s to the 1980s and an uptrend thereafter. However, Rs observations suffer from inhomogeneity issues and low spatial-temporal coverage. Sunshine duration-derived Rs is not present above problems and was utilized here to compare with observed Rs from China, Europe, and the United States over the 1950–2010 common period. Results show a good agreement between two data sets except for the dimming period in China, mainly due to instrument sensitivity drift of Rs observations. Therefore, using more extensive sunshine duration-derived Rs data set at approximately 2,600 stations over China, Europe, and the United States since 1901, a revisit of global dimming and brightening has been reasonably conducted, including the early period prior to the 1950s.

1 Introduction

As the Earth's primary energy source, the surface incident solar radiation (Rs) plays an important role in climate change (Dorno, 1920; Mercado et al., 2009; Sellers et al., 1990). However, Rs is not constant; rather, it undergoes significant decadal variations (Wild, 2009). Ground-based observations of Rs have indicated a continuous decrease from the 1950s to the 1980s and an increase thereafter, a trend that is known as “global dimming and brightening” (Augustine & Dutton, 2013; Dutton et al., 2006; Ohring et al., 2008; Stanhill & Cohen, 2001; K. Wang, Dickinson, et al., 2012; K. Wang et al., 2013; Wild, 2012; Wild et al., 2005). These phenomena have become an important consideration for research on the Earth's energy and water cycles (Long et al., 2009; Roderick & Farquhar, 2002).

Ground-based observations of Rs collected at weather stations since the 1950s, such as those from the Global Energy Balance Archive (GEBA), provide key evidence for global dimming and brightening (Ohmura et al., 1989; Wild et al., 2017). However, the uncertainties in such Rs data are difficult to quantify (Wild, 2009). For example, Rs observations in China suffer from substantial inhomogeneity issues due to instrument replacement and instrument sensitivity drift (K. Wang, 2014; K. Wang et al., 2015). Accordingly, to improve the instrumental accuracy, the Baseline Surface Radiation Network (BSRN) was established during the early 1990s (Ohmura et al., 1998) to independently measure Rs and its diffuse and direct components according to the specifications of the World Climate Research Programme. High-accuracy instruments have been deployed at the BSRN stations, with good maintenance protocols and frequent calibration (Augustine & Dutton, 2013; K. Wang, Augustine, & Dickinson, 2012). Furthermore, parallel observations of Rs at BSRN stations can compensate for each other to improve the data continuity (Augustine & Dutton, 2013). However, the BSRN stations are too sparse (with approximately 50 sites worldwide; Ohmura et al., 1998) to well estimate Rs at the global scale (Ma et al., 2015).

The observed Rs data have been criticized for their poor spatial representativeness, temporal discontinuity, and urbanization effects (Alpert & Kishcha, 2008; K. Wang et al., 2014). The Rs data from reanalyses data sets also have substantial uncertainties arising from the use of cloud fields sourced from model simulations rather than instrumental observations in reanalysis systems (Naud et al., 2014; Zhou et al., 2018). Moreover, most existing reanalyses, such as ERA-Interim and Modern-Era Retrospective Analysis for Research and Applications, do not include interannual variations in atmospheric aerosols, leading to significant biases in the long-term trends of Rs after 1990 in China (K. Wang et al., 2015).

Multiple Rs products based on satellite retrievals, including the Global Energy and Water Exchanges Surface Radiation Budget project, have been employed to investigate the decadal variations of Rs dating back to the 1980s (Pinker et al., 2005). The Global Energy and Water Exchanges Surface Radiation Budget Rs product is derived from cloud observations obtained from the International Satellite Cloud Climatology Project (Rossow & Schiffer, 1999), which suffers from inhomogeneity issues due to different amounts and capabilities of data acquired from geostationary and polar-orbit satellites (Dai et al., 2006; Evan et al., 2007). Furthermore, data from satellite retrievals only cover the brightening period.

The sunshine duration (SunDu) collected at globally distributed weather stations provides an independent proxy of Rs. SunDu records the time duration of a direct solar beam over a given threshold (i.e., 120 W/m2) during a day (Tang et al., 2011; Yang et al., 2006). Observations of SunDu date back to the nineteenth century at some of the sites and SunDu measurement sites are distributed worldwide. The material used to measure SunDu (light sensitive paper) is replaced each day. Therefore, it hardly suffers from the problem of instrument sensitivity drift (Sanchez-Lorenzo & Wild, 2012). Simultaneously, SunDu is almost free from influences of instrument replacement (Stanhill & Cohen, 2005). Even though, SunDu data do not provide a direct estimate of Rs and have the different sensitivity of atmospheric turbidity changes, compared with Rs observations, they are still a good proxy for variations of Rs (Manara et al., 2017). Moreover, existing studies have shown that SunDu-derived Rs estimates roughly depict long-term variability in Rs almost without the problems associated with early radiometry mentioned above (K. Wang, 2014; K. Wang et al., 2015) and reflect to some extent impacts of both aerosols and clouds (Tang et al., 2011; K. Wang, Dickinson, et al., 2012).

In this study, we compared the observed and SunDu-derived Rs at approximately 200 observed and SunDu-derived Rs station pairs over China, Europe, and the United States, meeting record lengths exceeding 60 months and spatial distances less than 110 km, and analyzed the spatial patterns of the trends in the two data sets by establishing 2.5° × 2.5° grids. Finally, we provided a SunDu-derived Rs data set with high spatial-temporal coverage from 1951 to 2010 at approximately 2,600 stations over China, Europe, and the United States, thereby permitting a revisit of global dimming and brightening. In addition, we also attempted to estimate earlier trends in Rs (i.e., prior to 1950) over Europe and the United States.

2 Data and Method

Observed Rs data sets from GEBA (Wild et al., 2017) and the China Meteorological Administration were used in this study (118 sites from 1959 to 2010 over China, 44 sites from 1961 to 2009 over Europe, and 46 sites from 1952 to 1980 over the United States). SunDu and other meteorological data (including the relative humidity, air temperature, and surface pressure) employed to derive Rs at approximately 2,600 meteorological stations over China (2,318 sites from 1951 to 2010), Europe (116 sites from 1901 to 2009), and the United States (137 sites from 1901 to 1987) were selected from the Carbon Dioxide Information Analysis Center, the Global Summary of the Day, the European Climate Assessment & Dataset (ECA&D), and the China Meteorological Administration.

The SunDu-derived Rs data used herein were calculated using the model of the revised Ångström-Prescott equation 1 by Yang et al. (2006):
urn:x-wiley:00948276:media:grl57325:grl57325-math-0001(1)
where a0, a1, and a2 are regression parameters of the observed Rs and against SunDu in equation 1, which have solved at regional and global scales further using monthly data in this study, instead of hourly and daily data in previous research; n is the observed SunDu and N is the theoretical value of SunDu; Rc is the surface solar radiation under clear-sky conditions and is calculated from surface meteorological observations (including the relative humidity, air temperature, and surface pressure), ozone thickness, and turbidity coefficient. Yang et al. (2006) derived turbidity coefficients from aerosol optical depth (AOD) based on Hess et al. (1998). Furthermore, the observed SunDu can accurately reflect long-term trends of atmospheric aerosols (Sanchez-Romero et al., 2014), so that SunDu-derived Rs from this model can capture accurately the impacts of both clouds and aerosols on Rs (Tang et al., 2011; K. Wang, Dickinson, et al., 2012; K. Wang et al., 2015) and reproduce long-term variability and trends in Rs.

To verify the reliability of monthly SunDu-derived Rs data, we selected approximately 200 observed-derived pairs of stations over China, Europe, and the United States; these station pairs must have data record lengths exceeding 60 months and spatial distances must be less than 110 km. It is worth noting that Rs data from GEBA in the United States were utilized only to confirm the agreement between the monthly mean observed and SunDu-derived Rs data, because the problem of incomplete long-term series exists in most data there.

As shown in Figure 1, the highest correlation coefficient between the monthly observed Rs and SunDu-derived Rs occurs in China (r = 0.98, p = 0.00), followed by the United States (r = 0.95, p = 0.00) and Europe (r = 0.92, p = 0.00). Rs data from GEBA and SunDu-derived Rs show the smallest relative standard deviation in the United States (10.77%). In brief, a highly significant correlation coefficient (r = 0.95, p = 0.00) and a small bias (−1.18 W/m2, −0.73%) between both monthly data sets suggest a good performance of monthly SunDu-derived Rs.

Details are in the caption following the image
Comparison of sunshine duration (SunDu)-derived and observed monthly Rs (units: W/m2). Each point represents a monthly value from China (b), Europe (c), and the United States (d) during the period of 1950–2010. (a) Shows the comparison of all data from China, Europe, and the United States. The color bar denotes the scatter density, defined as the number of points in each 200 × 200 axis grid. “N” denotes the number of stations. A highly significant correlation coefficient and a small bias between both data sets suggest a good performance of SunDu-derived monthly Rs.

To minimize the impact of heterogeneous site distribution, we regridded the Rs data sets into 2.5° × 2.5° grid boxes. In addition, we excluded data from either data set during the period of 1990–1993 over China; during this time, Rs observations are affected by an abrupt increase due to instrument replacement issues that might cause a spurious brightening and impact a comparison of the two data sets during the brightening period (K. Wang, 2014).

3 Results

3.1 Comparison of the Trends in the Observed and SunDu-Derived Rs

A direct comparison of observed-derived Rs (with 44 pairs of stations) reveals consistent trends during the periods of dimming (−1.11 W/m2 per decade versus −1.38 W/m2 per decade from 1961 to 1980) and brightening (1.51 W/m2 per decade versus 1.47 W/m2 per decade from 1980 to 2009) in Europe (Figure 2b and Table 1), further suggesting that SunDu-derived Rs data can accurately reproduce the observed trends in Europe from 1961 to 2009. This seems to be inconsistent with Manara et al. (2017), who showed substantial differences in the trends between the SunDu and Rs in Italy during the dimming period but similar trends during the brightening period. Simultaneously, they considered that the observed SunDu and Rs data have different sensitivities for different climatic conditions, especially in the dimming period. However, the spatial and seasonal variations of the solar zenith angle, ozone thickness, and turbidity coefficients that are used to calculate SunDu-derived Rs should effectively account for effects of different climatic conditions and local factors, such as elevation, turbidity, and meteorological conditions, and thus, they are more physically based (Yang et al., 2006) and predisposed to depict long-term variability in Rs.

Details are in the caption following the image
Time series of annual anomalies of the observed and sunshine duration (SunDu)-derived Rs data sets in different regions. (a and b) Represent pairs of the observed Rs (in red line) and SunDu-derived Rs (in black line) stations in China (at 118 sites in 1959–2010) and Europe (at 44 sites in 1961–2009) with trends listed inside. (c) Represents annual and 10-year “LOWESS” (derived from the term “locally weighted scatter plot smooth”) smoothed anomalies (in thick lines) in all reliable SunDu-derived Rs stations over China (in red line), Europe (in blue line), and the United States (in black line) with the number of stations as “N”.
Table 1. Trends (Units: W/m2 per Decade) in the Observed and SunDu-Derived Rs During Three Periods Over China, Europe, and the United States
W/m2 per decade Source Early brightening Dimming Brightening
China Observeda −7.67** −0.53
SunDu-deriveda −2.66** −0.85
SunDu-derivedb −1.90** −1.06
Europe Observeda −1.11 1.51**
SunDu-deriveda −1.38 1.47*
SunDu-derivedb 1.91** −1.36* 1.66*
The United States SunDu-derivedb −1.31* −1.10*
  • Note. “Early brightening” denotes the period of 1923–1950 in Europe and the United States. “Dimming” denotes the periods of 1959–1989, 1961–1980, and 1952–1980 in China, Europe, and the United States, respectively. “Brightening” denotes the periods of 1994–2010 in China and 1980–2009 in Europe. Note that the bold value was calculated during the period from 1950 to 1980.
  • a Represents paired observed-derived Rs stations and
  • b Represents all reliable sunshine duration (SunDu)-derived Rs stations.
  • * The 90% confidence levels.
  • ** The 95% confidence levels.

For China, even though the observed Rs and SunDu-derived Rs data show declining tendencies during the dimming period (1959–1989), the observed data exhibit a more intense trend (−7.67 W/m2 per decade versus −2.66 W/m2 per decade at 118 sites; Figure 2a and Table 1). The observed trend in this period may be spurious and likely due to the impacts of instrument aging and inaccurate calibrations during the period (K. Wang et al., 2015), that is, the degradation of the instrument sensitivity. For example, the general-purpose lacquer coating on the pyranometers constructed in China tends to peel off, which could result in an overestimation of the dimming rate. Our findings are analogous with the previous result of K. Wang et al. (2015) based on 105 observed-derived station pairs in China (−8.9 W/m2 per decade versus −2.9 W/m2 per decade from 1961 to 1990).

The sharp increase in observed Rs values during the period of 1990–1993 in China is potentially attributable to instrument replacement issues and substantially impacts the observed magnitude of the overall trend (Cong et al., 2010; Y. Wang & Wild, 2016), resulting in spurious brightening; however, this issue is not present in the SunDu-derived Rs data set (K. Wang, 2014). After 1993, with further instrument replacement activity and improved calibrations, the observed Rs trends in China are revealed to be closer to those in SunDu-derived Rs (by −0.53 W/m2 per decade versus −0.85 W/m2 per decade from 1994 to 2010; Figure 2a and Table 1).

3.2 Spatial Patterns of the Trends in the Observed and SunDu-Derived Rs

Figure 3 illustrates spatial patterns of the observed and derived Rs trends during the dimming and brightening periods over China (the respective periods of 1959–1989 and 1994–2010) and Europe (the respective periods of 1961–1980 and 1980–2009). SunDu-derived and observed Rs show a same sign of negative trends in most grids during the dimming period over China (from 1959 to 1989) and Europe (from 1961 to 1980), whereas the observed Rs data show larger trend magnitudes and likely due to the impacts of instrument sensitivity drift (Figures 3a–3c). During the dimming period, increasing Rs trends are still estimated in Southwest China and parts of Northeast China (Figure 3a) and in some grid boxes in Southern Europe (Figure 3b).

Details are in the caption following the image
Maps of the decadal trends (units: W/m2 per decade) in 2.5° × 2.5° grids of sunshine duration (SunDu)-derived Rs (a and d), the observed Rs (b and e), and differences between the two data sets (c and f) over China and Europe during two periods of dimming and brightening. “Dimming” denotes the periods of 1959–1989 in China and 1961–1980 in Europe. “Brightening” denotes the periods of 1994–2010 in China and 1980–2009 in Europe.

In the period of 1994–2010, regional differences are exhibited in the two data sets over China, with negative Rs trends in the North China Plain but positive trends in southern China (Figures 3d and 3e). In Europe, the Rs values increase from 1980 to 2009 for both data sets in most grid boxes (Figures 3d and 3e), which is in agreement with the findings of K. Wang et al. (2009) and K. Wang and Dickinson (2013). The increasing Rs trends in parts of Europe during this period are consistent with the substantial decline in the AOD (K. Wang et al., 2009), because the increasing aerosol loadings can reduce Rs both directly (e.g., through scattering effects and absorption) and indirectly (e.g., by modifying the cloud cover and other cloud properties).

3.3 A Revisit of Global Dimming and Brightening

Table 1 provides the trends of the observed and SunDu-derived Rs data sets from approximately 2,600 stations in different regions during three periods. SunDu-derived Rs data have the advantage of longer records and a wider distribution in comparison with the direct observations in Europe and the United States. SunDu-derived Rs data exhibit an uptrend of 1.91 W/m2 per decade (p = 0.06) in Europe (averaged over 10 sites) and a significant downtrend of −1.31 W/m2 per decade (p < 0.05) in the United States (at approximately 90 sites) from 1923 to 1950 (Table 1 and Figure 2c), the period named “early brightening” in previous studies (Ohmura, 2006; Wild, 2009). During this period, a downtrend is present in most grid boxes of southeastern the United States (Figure 4a), but the results for Europe may not be representative of the continent due to poor spatial sampling.

Details are in the caption following the image
Maps of the decadal trends (units: W/m2 per decade) of all reliable SunDu-derived Rs stations over China, Europe, and the United States in 2.5° × 2.5° grids during three periods. “Early brightening” denotes the period of 1923–1950 in Europe and the United States; “dimming” denotes the periods of 1959–1989, 1950–1980, and 1952–1980 in China, Europe, and the United States, respectively. “Brightening” denotes the periods of 1994–2010 in China and 1980–2009 in Europe.

In the period of dimming, the strongest decline in SunDu-derived Rs data occurs in China (−1.90 W/m2 per decade, p < 0.05, from 1959 to 1989), followed by Europe (−1.36 W/m2 per decade, p = 0.07, from 1950 to 1980) and the United States (−1.10 W/m2 per decade, p < 0.05, from 1952 to 1980; Table 1). In particular, significant declines are detected in the southeast China and northeastern the United States (Figure 4b).

During the period of 1994–2010, SunDu-derived Rs data depict a declining trend (−1.06 W/m2 per decade) in China (Table 1), which is in agreement with previous research involving satellite retrievals of Rs (Wu & Fu, 2011). SunDu-derived Rs trends during this period have different spatial patterns, with significant declining trends over the North China Plain (Figure 4c) that is likely due to increased anthropogenic aerosol emissions following the rapid development of the economy and industry (Qian, 2016; Smith et al., 2011). However, the trends of SunDu-derived Rs data in Europe have little spatial variations (Figure 4c), with a significant increase of 1.66 W/m2 per decade (Table 1).

4 Conclusions

The Rs is a key parameter for research on the Earth's energy and water cycles. Based on ground-based observations of Rs, a continuous decrease from the 1950s to the 1980s and an increase thereafter are found to be periods of as “global dimming and brightening.” However, the Rs observations suffer from the inhomogeneity and low spatial-temporal coverage. Here we therefore utilized SunDu and meteorological variables at ~2,600 stations over China, Europe, and the United States from 1901 to 2010, to derive the Rs data for revisiting the global dimming and brightening, which provide high spatial-temporal coverage and are less sensitive to calibration and replacement issues that plague traditional Rs measurements, especially before 1980s.

A direct comparison of Rs observations, only available at a limited number of stations with SunDu-derived Rs shows reasonably consistent trends in Europe, during the periods of dimming (−1.11 W/m2 per decade versus −1.38 W/m2 per decade at 44 sites, from 1961 to 1980) and brightening (1.51 W/m2 per decade versus 1.47 W/m2 per decade at 44 sites, from 1980 to 2009), further suggesting that SunDu-derived Rs data can accurately reproduce the observed trends in Europe from 1961 to 2009.

But for China, even though the observed Rs data show a decreasing trend with SunDu-derived Rs during the dimming period (from 1959 to 1989), the observed data exhibit a larger (and likely erroneous) trend than SunDu-derived Rs (−7.67 W/m2 per decade versus −2.66 W/m2 per decade), due to the negative impacts of instrument sensitivity drift. An internationally coordinated calibration of radiometers was not established until 1979 (Gueymard & Myers, 2008); therefore, it is necessary to reconstruct the Rs data before that time. After 1993, with further instrument replacement activity and an improved calibration, the observed Rs trends in China are revealed to be closer to those in SunDu-derived Rs (by −0.53 W/m2 per decade versus −0.85 W/m2 per decade from 1994 to 2010).

During the dimming period, SunDu-derived and observed Rs show a same sign of negative trends over China and Europe, whereas the observed Rs data show larger trend magnitudes and likely due to the impacts of instrument problems. Negative trends of Rs are agreement with increasing trends of clouds and aerosols over China and Europe in this period (Norris & Wild, 2007, 2009). During the brightening period, the two data sets display spatial variations in Rs trends over China but not over Europe (with increasing Rs). Even if the period is well known for global brightening, the observed Rs and SunDu-derived Rs over most regions of China still exhibit declining trends, especially in North China Plain; that is likely due to increased anthropogenic aerosol emissions following the rapid development of the economy and industry there (Qian, 2016; Smith et al., 2011). In contrast, the increasing Rs trends in parts of Europe during the brightening period maybe caused by with the substantial decline in the AOD (K. Wang et al., 2009).

Our results suggest that SunDu-derived Rs can be useful because of its demonstrated suitability as a proxy for Rs and great spatial-temporal coverage. SunDu-derived Rs estimated from approximately 2600 stations can permit a revisit of global dimming from the 1950s to the 1980s over China (−1.90 W/m2 per decade, p < 0.05), Europe (−1.36 W/m2 per decade, p = 0.07), and the United States (−1.10 W/m2 per decade, p < 0.05), brightening from 1980 to 2009 in Europe (1.66 W/m2 per decade, p = 0.06), and a decline in Rs from 1994 to 2010 for China (−1.06 W/m2 per decade). Even if 1994–2010 is well known for global brightening, the observed and SunDu-derived Rs over China still exhibit dimming. Earlier trends in Rs from 1923 to 1950 are further estimated (1.91 W/m2 per decade for Europe and −1.31 W/m2 per decade for the United States), but the result in Europe may not well represent the actual trend for the European continent during that period due to poor spatial sampling.

Acknowledgments

This study was funded by the National Key R&D Program of China (2017YFA0603601) and the National Natural Science Foundation of China (41525018). The observed Rs data were obtained from GEBA (www.geba.ethz.ch) and the China Meteorological Administration (CMA, http://www.cma.gov.cn). The sunshine duration and other meteorological data (including the relative hunidity, air temperature, and surface pressure) were obtained from the Carbon Dioxide Information Analysis Center (CDIAC, http://cdiac.ess-dive.lbl.gov/ndps/ndp021.html), the Global Summary of the Day (GSOD, https://www.ncdc.noaa.gov), the European Climate Assessment & Dataset (http://www.ecad.eu), and the China Meteorological Administration (CMA, http://www.cma.gov.cn).