Decadal variability of surface incident solar ... - Wiley Online Library

6 downloads 0 Views 921KB Size Report
Jul 14, 2015 - Intercomparisons between observation-based and model-based Rs performed in this research show that sunshine duration (SunDu)-derived ...
PUBLICATIONS Journal of Geophysical Research: Atmospheres RESEARCH ARTICLE 10.1002/2015JD023420 Key Points: • Pyranometers had a sensitivity problem from 1958 to 1990 in China • Observations of diffuse and total solar radiation were impaired • Solar radiation decreased before 1990 and remained stable afterwards over China

Correspondence to: K. Wang, [email protected]

Citation: Wang, K., Q. Ma, Z. Li, and J. Wang (2015), Decadal variability of surface incident solar radiation over China: Observations, satellite retrievals, and reanalyses, J. Geophys. Res. Atmos., 120, 6500–6514, doi:10.1002/2015JD023420. Received 23 MAR 2015 Accepted 19 JUN 2015 Accepted article online 23 JUN 2015 Published online 14 JUL 2015

Decadal variability of surface incident solar radiation over China: Observations, satellite retrievals, and reanalyses Kaicun Wang1,2, Qian Ma1,2, Zhijun Li1,2, and Jiankai Wang3 1

College of Global Change and Earth System Science, Beijing Normal University, Beijing, China, 2Joint Center for Global Change Studies, Beijing, China, 3Chinese Meteorological Administration, Beijing, China

Abstract Existing studies have shown that observed surface incident solar radiation (Rs) over China may have important inhomogeneity issues. This study provides metadata and reference data to homogenize observed Rs, from which the decadal variability of Rs over China can be accurately derived. From 1958 to 1990, diffuse solar radiation (Rsdif) and direct solar radiation (Rsdir) were measured separately, and Rs was calculated as their sum. The pyranometers used to measure Rsdif had a strong sensitivity drift problem, which introduced a spurious decreasing trend into the observed Rsdif and Rs data, whereas the observed Rsdir did not suffer from this sensitivity drift problem. From 1990 to 1993, instruments and measurement methods were replaced and measuring stations were restructured in China, which introduced an abrupt increase in the observed Rs. Intercomparisons between observation-based and model-based Rs performed in this research show that sunshine duration (SunDu)-derived Rs is of high quality and can be used as reference data to homogenize observed Rs data. The homogenized and adjusted data of observed Rs combines the advantages of observed Rs in quantifying hourly to monthly variability and SunDu-derived Rs in depicting decadal variability and trend. Rs averaged over 105 stations in China decreased at 2.9 W m2 per decade from 1961 to 1990 and remained stable afterward. This decadal variability is confirmed by the observed Rsdir and diurnal temperature ranges, and can be reproduced by high-quality Earth System Models. However, neither satellite retrievals nor reanalyses can accurately reproduce such decadal variability over China. 1. Introduction The amount of solar radiation incident at the surface (Rs) is determined mainly by atmospheric clouds and aerosols as well as solar elevation. Surface absorbed Rs transfers into latent and sensible heat fluxes [Wang et al., 2010a, 2010b], which heat the air above the surface and provide energy and moisture for cloud and precipitation processes. The visible part of Rs can also be absorbed by photosynthetic organisms and used in photosynthesis. Rs has been linked to observed variable decadal warming rates [Wang and Dickinson, 2013b; Wild et al., 2007]. Existing studies have shown that observed Rs data may have major inhomogeneity problems because of sensitivity drift and instrument replacement [Shi et al., 2008; Tang et al., 2011; Wang, 2014]. In particular, it has been shown that measurement biases can explain discrepancies between observed and simulated Rs decadal variability [Wang, 2014]. This study provides metadata and reference data to homogenize observed Rs data over China. The metadata of Rs observations, including instruments, measurement methods, instrument calibrations, and changes in stations, are essential information to examine the homogeneity of the observed Rs. However, such information was unavailable when the observed Rs data were released. Existing studies on Rs either lack such important information or include inconsistent information [Liang and Xia, 2005; Qian et al., 2006; Shi et al., 2008; Tang et al., 2011; Xia, 2010a; Xia et al., 2006; Yang et al., 2006].

©2015. The Authors. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

WANG ET AL.

In this research, such metadata were collected from Chinese journals that are inaccessible to the international community. These publications were authored by instrument developers and meteorological observers who documented first hand information about the instruments used in China. Such information provides evidence to understand the inhomogeneity of observed Rs. With this information, it can be explicitly claimed that the sensitivity drift from 1957 to 1989 can be attributed to the pyranometers used to measure diffuse and total solar radiation, whereas measurements of direct solar radiation (Rsdir) were not impaired. The observed decadal variability of Rsdir, Rsdif, and Rs can confirm this inference based on instrument metadata (see section 8 for detailed information).

SURFACE SOLAR RADIATION OVER CHINA

6500

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

Another great challenge in homogenizing observed Rs data is finding homogenous reference data. Rs exhibits strong spatial variation, and stations measuring Rs are very sparsely distributed. Therefore, neighboring stations cannot be used as reference stations to homogenize observed Rs data. As a result, few studies have been performed on homogenizing observed Rs data, although homogenization has been widely accepted for air temperature and precipitation [Li et al., 2009; Peterson et al., Figure 1. Number of stations where data for observed surface incident solar radiation 1998]. Furthermore, Rs was (Rs) were available each month. China has performed Rs measurements since 1957. measured with the same methThe number of stations measuring Rs exceeded 60 in 1961 and generally increased ods and instruments in China, afterward. The number of stations changed substantially during two periods: from 1966 to 1972 and from 1990 to 1993. which suffered from similar sensitivity drift problem. This research proposes to homogenize observed Rs data using sunshine duration (SunDu) derived Rs as reference data, which are available at each Rs station in China. Previous studies have argued that SunDu is almost free from sensor sensitivity drift and can be used to calculate decadal Rs variability [Tang et al., 2011; Wang et al., 2012b]. However, some studies have claimed that SunDu is only a proxy for Rs and is of low quality. SunDu measurements also depend on SunDu recorder types and geolocation of their measurement stations. A recent study by the authors addressed this issue by calibrating the equation to calculate Rs from SunDu at each station with observed Rs data [Wang, 2014]. In this study, intercomparisons of observation-based (observed Rs and SunDu-derived Rs) and model-based (climate models, reanalyses, and satellite retrievals) estimates have been carried out. These comparisons indicated that SunDu-derived Rs has the best agreement with model-based Rs estimates, including satellite retrievals, reanalyses, and the Coupled Model Intercomparison Project phase 5 (CMIP5) Earth System Models (ESMs) at time scales longer than 1 month. These intercomparisons show that it is reasonable to homogenize observed Rs data using SunDu-derived Rs as reference data. The homogenized Rs combines the advantages of observed and SunDu-derived Rs and can accurately depict Rs variability from diurnal to decadal time scales. From this viewpoint, homogenized Rs provides a better data set for its scientific applications.

2. Rs Observations Over China Coordinated Rs measurements started in 1957 over China. By 1961, there were more than 60 measuring stations (Figures 1 and 2), and the number generally increased afterward, except for two periods: 1966–1972 and 1989–1993. China replaced its instruments during 1990–1993. The measurement history can be roughly divided into two stages: 1957–1989 and 1990–present (Tables 1 and 2). During the first stage, there were 88 stations, of which 75 were first-class stations and 13 were second-class stations. In first-class stations, Rsdir and Rsdif were measured every hour during the daytime by pyrheliometers and shaded pyranometers, respectively (Tables 1 and 2). Initially, China imported Yanishevsky thermoelectric pyrheliometers and pyranometers from the Union of Soviet Socialist Republics (USSR) to perform these measurements. The Yanishevsky thermoelectric pyrheliometer is a first-class instrument which is recommended by the World Meteorological Organization (WMO). The Yanishevsky black-white pyranometer is a WMO second-class instrument [Garg and Garg, 1993], with a spectral range of 0.25–4.0 μm [Garg and Garg, 1993]. WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6501

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

Figure 2. Station types and data durations of surface-incident solar radiation (Rs)-measuring weather stations, which are identified by various markers and colors. A square without an accompanying cross indicates that the station was abandoned after 1990, and a cross without an accompanying square indicates that the station was added to the network after 1990. Before 1990, the stations were divided into first class and second class. At first-class stations (approximately 85% of the total stations, shown as red squares), Rs was calculated as a sum of direct solar radiation (Rsdir) observed by a pyrheliometer and diffuse solar radiation (Rsdif) observed by a shaded pyranometer (see Tables 1 and 2). The second-class stations measured Rs using pyranometers. After 1993, only the first-class stations (approximately 17% of total stations, shown as red crosses) measured Rs, Rsdir, and Rsdif at the same time; other stations merely measured Rs with pyranometers.

The measurements were performed by manually reading a microampere meter connected to the radiometers to record the electric current of their thermopiles, and the resulting value was used to calculate the radiation value using the known sensitivity of the radiometers. Rs was calculated as a sum of observed Rsdir and Rsdif at first-class stations [Zhang and Lu, 1988] (see also Table 2). At the second-class stations, Rs was observed with black-white thermopile pyranometers every 30 minutes. It is well known that black-white type pyranometers are not the ideal instrument for measuring Rs due to their long response time and high directional response errors. However, they have been recommended for measurements of Rsdif because of their small thermal offset error [Augustine et al., 2005; Wang et al., 2013]. Table 1. Instruments Used for Surface Solar Radiation Measurements (Rs) in China, Modified From Zhang and Lu [1988 and 1990] Pyrheliometer Specifications Instrument type Thermopile type Thermopile coating Dome Spectral range (μm) Response time (95%) 1 2 Sensitivity range (μV W m ) Stability (sensitivity change per year) Sun tracking Temperature response error Thermal offset (nighttime) Cosine response error Azimuth response error

WANG ET AL.

1957–1989

1990 to Present

DFY1 Solid black General-purpose lacquer No 0.3–4.0 μm ≤13 s 5.7–14.2 Unknown

TBS2 or DFY3 Solid black Optical lacquer

Manual

Automatic, failed frequently ±2% ≤20 μV

±10% Unknown

Quartz glass 0.3–4.0 μm ≤25 s 7–14 ±1%

SURFACE SOLAR RADIATION OVER CHINA

Pyranometer 1957–1989

1990 to Present

DFY2 TBQ2 or DFY4 Black-white Solid black General-purpose Optical lacquer lacquer General-purpose glass Double quartz glass 0.3–2.4 μm 0.3–4.0 μm ≤60 s ≤25 s 7–14 7–14 Unknown ±2%

±10% Unknown ±30% Unknown

±2% ≤100 μV ±15% ≤8.6%

6502

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

Table 2. Measurement Method and Calibration Process for Surface Solar Radiation (Rs) Instruments in China, Modified From Zhang and Lu [1988 and 1990] Period Type and number of stations Measuring variables Recording method Sampling frequency National reference groups Calibration process

1957–1989

1990 to Present

First class: 75; second class: 13 (88 in total)

First class: 17; second class: 33; third class: 48 (98 in total) First class: direct, diffuse and Rs; second class: Rs and net radiation; third class: Rs Measuring voltage automatically by RYJ-2 or DRG-C data logger 60 (RYJ-2) or 360 (DRB-C) per hour

First class: direct and diffuse, total (Rs) was calculated from these; second class: total only Measuring electric current manually using a microampere meter First-class stations: hourly; second-class stations: half-hourly Two cavity pyrheliometers (H-F) (after 1981) After 1981, instruments were calibrated using second standard references at regional centers every other year.

Two cavity pyrheliometer (H-F) and one absolute radiometer (PMO6) Instruments were directly calibrated against national reference groups every other year, and sensitivity was monitored at regional centers each year.

With the inevitable issues of instrument aging, imitations of the USSR radiometers (DFY1 and DFY2) were used later at most stations (Table 1). Due to technical limitations, the quartz glass required for the dome of the pyranometers was replaced by general-purpose glass for Chinese imitations (DFY2), which had a lower transmittance than quartz glass and did not cover the full solar spectral regime [Zhang and Lu, 1988, 1990]. The thermopiles of a pyranometer should be coated by optical lacquer. However, thermopiles of Chinese-made pyranometers (DFY2) were coated by general-purpose lacquer, which has a significant directional reference, resulting in a high cosine response error of up to 35% to the pyranometers [Mo et al., 2008] (Table 1). Moreover, the general-purpose lacquer coating of the thermopile is easy to peel off, particularly the black coating part. This introduced different degradation rates of the black and white coatings of the thermopile, resulting in an important sensitivity drift in the pyranometers [Yang et al., 2010]. The actinometer carrier scale used before 1981 in China was the International Pyrheliometer Scale (IPS). In 1981, the WMO recommended the World Radiometric Reference (WRR), and China then selected this reference for use. The daily solar radiation data before 1981 were multiplied by 1.022 to convert to the WRR before the data were released in China [Ma et al., 1998]. China also followed the recommendations of the USSR in calibrating its radiometers. All the radiative receivers were calibrated using a multistep method [Ma et al., 1998; Shi et al., 2008]: calibration once a month at the stations against the reference instruments, calibration of the reference radiometers against regional reference instruments, and calibration of the regional reference instruments every 2 years against Chinese reference instruments, which were regularly calibrated against reference pyrheliometers in Tokyo, Japan, or Pune, India [Shi et al., 2008]. In 1981, China imported two cavity pyrheliometers (H-F No.19743 and No.20294) as its national reference groups to calibrate its first class references at the regional centers [Yang et al., 2007], which in turn were used to calibrate the working instruments at each station every other year (Table 2). Before 1990, the electric current was measured, and the instrument sensitivity from the calibration process was used to calculate the radiation values. However, the instrument sensitivity depends on ambient temperature because the resistance of the thermopiles changes with temperature. The Rs observations therefore had strong temperature response errors [Lu and Shi, 1987; Zhou et al., 1978]. To address the problem of radiometer aging, China replaced its instruments from 1990 to 1993 [Bai, 1987; Zhang and Lu, 1988] (Tables 1 and 2). New automated instruments replaced the old manual instruments. The data logger connected to a new instrument sampled the voltage rather than the electric current every minute. The measurement scheme was also replaced. After 1993, the number of Rs stations was increased to 98, although some stations were abandoned (Figure 2). These stations were classified into three classes. At the 17 first-class stations, Rs, Rsdir, and Rsdif were measured in parallel by pyranometers, pyrheliometers, and shaded solid black thermopile pyranometers, respectively (Figure 3). However, their solar trackers failed frequently and introduced a high missing data rate for Rsdir [Lu and Bian, 2012; Mo et al., 2008]. In the second- and third-class stations (81 stations in total), Rs was directly measured by a solid black thermopile pyranometer (Table 1). Briefly, since 1993, most Rs measurements over China have been WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6503

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

made by solid black thermopile pyranometers, which have strong directional errors [Lu and Bian, 2012; Mo et al., 2008] (see also Tables 1 and 2). To address the sensitivity drift problem, calibration processes were substantially improved after 1991. China imported an absolute pyrheliometer (PMO-6 No.850406) in 1991 [Yang et al., 2007], which is a selfcontained system developed at the Physikalisch-Meteorologisches Observatorium Davos (PMOD)/WRC (Table 2). This PMO-6 absolute pyrheliometer and the two H-F type cavity pyrheliometers make up the national reference group of China, which has been calibrated by referFigure 3. Current instruments used to measure (a) surface incident solar ences at the WRC every 5 years. To radiation (Rs), (b) direct solar radiation (Rsdir), and (c) diffuse solar radiation improve calibration accuracy, all (Rsdif) at the Beijing weather station. the working instruments in China are now calibrated directly against the national references at Beijing rather than at the regional centers [Yang et al., 2007]. After 1993, 80% of the 98 Rs stations used TBQ-2 pyranometers, whereas the other 20% used DFY4 pyranometers (Table 1). Although the instruments were substantially improved, the uncertainty of the new instruments was still higher than that of the first-class instruments recommended by the WMO. The Chinese-developed solid black thermopile pyranometers have high thermal offset and directional response errors, including cosine and azimuthal response errors [Lu et al., 2002; Mo et al., 2008; Yang et al., 2010] (see Table 1), partly because Chinese manufacturers are not required to specify these errors [Yang et al., 2012]. Although Rsdir has been measured at the 17 first-class stations since 1993, their solar trackers failed often, and the rate of missing data was very high. The performance of solar trackers in China has been substantially improved in 2007 [Yang et al., 2012]. The stability of the Chinese-developed instruments is also worse than that of the WMO recommended first-class pyranometers [WMO, 2008] (see also Table 1). In the 1990s, these Chinese-developed instruments were deployed at the Chinese Ecosystem Research Networks (CERN). However, all these instruments were replaced several years later because they could not meet CERN accuracy requirements [Mo et al., 2008]. A new generation of higher-accuracy pyranometers and pyrheliometers has been developed, but they have not been deployed for operational measurements [Mo et al., 2008].

3. Sunshine Duration (SunDu)-Derived Rs Over China SunDu is a standard observation at weather stations in China. Most weather stations in China have used Jordan sunshine recorders to measure SunDu since the 1950s (Figure 4). Only 18 stations (~2% of all weather stations) in high-latitude regions in northeastern China used Campbell-Stokes recorders, which were replaced by Jordan sunshine recorders in January 2012 [Lu et al., 2012]. The Jordan sunshine recorder has a copper cylinder with two small holes in its sides. The direct solar beam penetrates the small holes, illuminating the light-sensitive paper inside the copper cylinder [Xu et al., 2011]. The length of the band of changed color on the light-sensitive paper is a measure of SunDu [Che et al., 2005; Wang et al., 2012b; Zhao et al., 2010]. Because the recording material (light sensitive paper) in a SunDu recorder is replaced each day, it does not suffer from a sensitivity drift problem [Sanchez-Lorenzo and Wild, 2012].

WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6504

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

Since the early twentieth century, SunDu has been used to estimate Rs using the Ångström formula [Angström, 1924]. In 2006, Yang et al. [2006] proposed the following formula to estimate Rs, which involved parameterizing the radiative extinctions of air and cloud separately: Rs =Rc ¼ a0 þ a1  n=N þ a2 ðn=NÞ2 (1) where n is the measured SunDu, N is the theoretical value of SunDu, and Rc is the daily total solar radiation at the surface under clear-sky (or cloud-free) conditions. The effects of Rayleigh scattering, water vapor absorption, and ozone absorption can be accounted for in Rc using meteorological observations [Yang et al., 2006]. In Yang et al. [2006], winter- and summer-averaged aerosols based on Hess et al. [1998] were included to calculate Rc, which is out-of-date, and Hess et al. [1998] can only provides a geolocation dependent long-term average of atmospheric aerosol loading. It works similarly to the constant parameter a0 in equation (1), which is calibrated at each station. It Figure 4. Current sunshine duration recorder used in the Beijing weather has been shown that Rs calculated station. Most weather stations in China have used this Jordan sunshine recorder to measure SunDu since the 1950s. Only 18 stations (~2% of all from equation (1) can accurately weather stations) in high-latitude regions in Heilongjiang Province in reflect impact of aerosols on Rs [Tang northeastern China used the Campbell-Stokes recorders, which were et al., 2011; Wang et al., 2012b]. This is replaced by Jordan sunshine recorders in January 2012 [Lu et al., 2012]. because the observed SunDu can accurately reflect long-term trend of atmospheric aerosols [Magee et al., 2014; Sanchez-Lorenzo et al., 2008; Sanchez-Romero et al., 2014]. In particular, the observed reduction in SunDu has been attributed to the increase in atmospheric aerosols [Che et al., 2005; Qian et al., 2006; Wang et al., 2012c; Zhao et al., 2010] and SunDu-derived Rs calculated using equation (1) has been demonstrated to capture accurately the impacts of both clouds and aerosols on Rs in China during recent decades [Tang et al., 2011; Wang et al., 2012b]. In 1962, the WMO defined SunDu as the duration for which direct solar beam irradiance is greater than 120 W m2 during 1 day and recommended the Campbell-Stokes recorder as the standard instrument. However, the threshold of the Jordan recorder is generally 10% higher than 120 W m2 and varies over a broad range depending on station location (i.e., latitude) and weather conditions [Xia, 2010b; Zhang and Tan, 2000]. To address this issue, equation (1) was calibrated at each station in China. This compensated for the disadvantages of SunDu: (1) SunDu does not directly provide an estimate of Rs, and (2) the threshold of a SunDu recorder changes with the SunDu recorder type and the site environment. To obtain a reliable equation, only stations where Rs data were available over 10 years or more were used. There were 105 stations meeting this requirement. SunDu was then used to calculate Rs from 1950 to 2012 at each station on a daily scale and averaged to generate monthly averages. Detailed validation information for the SunDu-derived Rs can be found in Wang [2014], which shows that Rs can be calculated accurately from SunDu data from daily to decadal time scales.

WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6505

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

4. Satellite Rs Retrievals Satellite Rs retrievals were calculated from cloud observations. Two recently released satellite Rs products were compared and evaluated in this research. The first product was the GEWEX SRB Rs product from 1983 to 2007 [Zhang et al., 2004]. Its cloud inputs are the visible and infrared radiances from the International Satellite Cloud Climatology Project (ISCCP) [Rossow and Schiffer, 1999; Stubenrauch et al., 2013]. The ISCCP cloud products combine cloud observations from polar orbit and geostationary satellites directly. The different amounts of data from the polar orbit and geostationary satellites and their different capabilities for detecting low-level clouds introduced inhomogeneities into the ISCCP cloud data [Dai et al., 2006; Evan et al., 2007]. Because clouds determine Rs, this inhomogeneity of the ISCCP cloud products can be expected to introduce significant inhomogeneity to the Rs values calculated from the cloud products. The second satellite product, Clouds and Earth’s Radiant Energy System (CERES) Energy Balanced And Filled (EBAF), addresses these issues by retaining the amount of clouds as detected by the polar orbit satellites and multiplying this amount by a variable factor representing the diurnal cloud cycle derived from the geostationary satellite observations [Doelling et al., 2013]. Furthermore, Cloudsat radar (CPR) and the CALIPSO Lidar profile [Winker et al., 2010] can help to detect low-level and cirrus clouds [Kato et al., 2011; Minnis et al., 2008]. The CERES EBAF algorithms use the CERES observations at the top of the atmosphere [Wielicki et al., 1996] to constrain their calculations using satellite-retrieved surface, cloud, and aerosol properties as input, primarily from the Moderate Resolution Imaging Spectroradiometer (MODIS). Adjustments were made if substantial biases were found in the calculations. This process guaranteed a consistent radiation product and substantially improved the accuracy of their Rs products [Kato et al., 2013]. CERES EBAF Rs data have been available since 2000.

5. Rs From Reanalyses Rs from the latest two reanalyses, ERA-Interim [Dee et al., 2011] and Modern-Era Retrospective analysis for Research and Application (MERRA) [Rienecker et al., 2011], were studied in this research. The reanalyses simulated cloud parameters to calculate Rs. The reanalyses produced reasonable estimates of atmospheric humidity by assimilating radiosonde and satellite observations, from which the cloud parameters are derived. Both reanalyses used the maximum-random overlapping scheme for clouds. Neither reanalysis allowed aerosol loadings to change annually.

6. CMIP5 ESM Simulations of Rs CMIP5 ESMs calculate Rs using the radiative transfer model with input from modeled cloud and aerosol parameters. Monthly Rs from historical runs of 48 CMIP5 ESMs were downloaded for this research (Table 4). The historical runs covered the period from 1850 to 2005, a choice which was forced by atmospheric composition changes from both anthropogenic and natural sources [Taylor et al., 2012].

7. Comparisons of Monthly Rs From Different Estimates In this research, intercomparisons were performed between observation-based (observed and SunDu-derived) and model-based (satellite retrievals, reanalyses, and model simulations) estimates of Rs. Monthly means and anomalies were compared at the 105 stations in China. The medians of the statistical parameters at the 105 stations are shown in Table 3. The observed Rs were compared with model-based Rs. The same comparisons were also conducted for SunDu-derived Rs. The results are summarized in Table 3, which indicate that the agreement between SunDu-derived Rs and model-based Rs estimates is better than the agreement between observations and model-based Rs estimates. This result is surprising because SunDu is only an indirect observation of Rs and is likely due to the significant inhomogeneity of Rs observations over China (see section 8 for detailed information). Furthermore, the pyranometers used to measure Rs have significant thermal offset and directional response errors. Because the observations were used to calibrate equation (1), the biases of the observations are the same as those of the SunDu-derived Rs (Table 3). Satellite retrievals were the best among the model-based Rs estimates in terms of bias, standard deviation, and correlation coefficient (Table 3), followed by reanalyses, and CMIP5 ESM simulations were the worst. The CERES EBAF Rs was slightly better than the GEWEX SRB (Table 3). MERRA had a stronger positive bias WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6506

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

a

Table 3. Statistical Summary of Comparisons Between Observation-Based Estimates and Model-Based Rs Estimates CMIP5 Models

Observed Rs SunDu-Derived Rs Homogenized Rs 1 Homogenized Rs 2

ERA-Interim

MERRA

CERES-EBAF

GEWEX SRB

Bias

STD

R

Bias

STD

R

Bias

STD

R

Bias

STD

R

Bias

STD

R

23.8 23.9 21.5 21.5

33.7 31.4 32.2 32.4

0.80(0.01) 0.83(0.02) 0.82(0.01) 0.81(0.01)

21.0 21.8 19.9 19.8

19.1 15.0 16.5 17.6

0.94(0.67) 0.96(0.74) 0.96(0.70) 0.96(0.70)

29.8 30.5 30.7 30.5

23.2 18.7 20.0 20.4

0.93(0.56) 0.96(0.64) 0.94(0.60) 0.94(0.61)

8.1 8.6 6.2 6.1

13.1 12.5 12.5 13.0

0.97(0.72) 0.97(0.74) 0.98(0.73) 0.97(0.73)

9.9 10.2 7.1 6.0

17.0 13.7 15.4 15.4

0.95(0.69) 0.97(0.79) 0.96(0.73) 0.96(0.73)

a

The homogenized Rs 1 and 2 were output from two homogenization methods: (1) mean adjusted and (2) quantile-matching (Q-M) adjusted. The statistical 2 2 parameters (bias (W m ), standard deviation (STD) (W m ), and correlation coefficient) were calculated from monthly Rs averages at each station in Figure 2, and their median values are shown here. The medians of bias, STD, and R for the 48 CMIP5 ESMs are shown here. The numbers in brackets for column R were calculated from the monthly Rs anomalies (with the seasonal cycle removed). Among the observations, SunDu-derived Rs had the best agreement with modelbased Rs estimates, whereas CERES EBAF satellite Rs retrievals best agreed with observations.

and a higher standard deviation than ERA-Interim, which is consistent with previous studies [Bosilovich et al., 2011; Naud et al., 2014; Wang and Zeng, 2012; Wang and Dickinson, 2013a]. The monthly Rs anomalies from CMIP5 ESMs were almost unrelated to the observation-based estimates due to their poor performance in simulating seasonal cloud variation [Zhang et al., 2005]. CMIP5 ESMs substantially overestimated Rs over China by 15–20 W m2 because of their underestimation of clouds, in particular of stratus clouds in southern China [Zhang and Li, 2013; Zhang et al., 2014]. Compared to observed Rs, the CERES EBAF Rs had a positive bias of 8 W m2. Rs was measured by solid black thermopile pyranometers (types: TBQ2 or DFY4) after 1995 in China. No results for the thermal offset of Chinese-developed pyranometers during the daytime have been reported. However, many studies on other solid black pyranometers have been conducted [Philipona, 2002], which show that Rs may be underestimated by 1%–2%. Therefore, the positive bias of CERES EABF Rs should be less than 5 W m2.

8. Decadal Rs Variability Over China In the work described in this section, the impact of the sensitivity drift problem on decadal Rs variability over China was investigated. From 1958 to 1993, Rsdir and Rsdif were measured separately by pyrheliometers and shaded pyranometers in China, and Rs was calculated as the sum of the two. There are 65 stations where measurements of both Rsdir and Rsdif were available for more than 120 months, from which regional anomalies over China were calculated and are shown in Figure 5. From 1961 to 1991, the decreasing trend in observed Rs (8.97 W m2 per decade, p = 0.00) is much stronger than that of SunDu-derived Rs (3.60 W m2 per decade, p = 0.00). At the same time, Rsdir decreased by 7.38 W m2 per decade (p = 0.00). Rs is primarily determined by clouds and aerosols, which exert primarily a scattering effect. In general, a reduction in Rsdir indicates a lesser increase in Rsdif. It is reasonable to assume that approximately half the reduction in Rsdir transfers to Rsdif. If so, Rsdif should increase by 3.69 W m2 per decade, and Rs should decrease by 3.69 W m2, which is consistent with results for SunDu-derived Rs. However, the observed Rsdif decreased by 1.24 W m2 per decade (p = 0.00) rather than increasing at 3.69 W m2 per decade. These results confirm that the overestimation of observed Rs from 1961 to 1991 was primarily caused by the sensitivity drift of the pyranometers which were used to measure Rsdif. The decadal variability of Rs from different estimates over China was then compared from 1960 to 2012. Except for special claims, the regional average of Rs was calculated at the 105 stations where both SunDu and observations of Rs were available. The satellite retrievals, reanalyses, and CMIP5 ESM simulations of Rs at the nearest pixel (grid cell) to the stations were averaged into regional averages. Then regional anomalies were calculated with a reference period of 1960–2012 by removing the seasonal cycle of Rs. Figure 6 shows that Rs observations have significant inhomogeneity over China. The dimming trend from 1961 to 1990 was 8.0 W m2 per decade averaged over the 105 stations in China (see also Figure 1 and Table 4). Observed Rs data over China contain an important inhomogeneity from 1990 to 1993, i.e., an abrupt increase in Rs (Figure 6). Two factors explain this inhomogeneity. First, the old instruments had a sensitivity drift problem and underestimated Rs before 1990. Second, the stations measuring Rs substantially changed both in numbers (Figure 1) and in geolocations from 1989 to 1993 (Figure 2). From

WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6507

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

Figure 5. Regional average annual anomalies of observed total solar radiation (Rs, red), sunshine duration (SunDu)-derived Rs (black), observed direct solar radiation (Rsdir, green), and observed diffuse solar radiation (Rsdif, blue) over China. From 1958 to 1993, Rsdir and Rsdif were measured separately by pyrheliometers and shaded pyranometers in China, and Rs was calculated as their sum. There are 65 stations where measurements of both Rsdir and Rsdif were available for more than 120 months. Regional anomalies were calculated from these stations. From 1961 to 1991, the decreasing 2 2 trend of observed Rs (8.97 W m per decade, p = 0.00) is much stronger than that of SunDu-derived Rs (3.60 W m 2 per decade, p = 0.00). At the same time, Rsdir decreased by 7.38 W m per decade (p = 0.00). It is reasonable to assume 2 that approximately half the reduction in Rsdir transfers to Rsdif. If so, Rsdif should increase by 3.69 W m per decade, and 2 Rs should decrease by 3.69 W m , which is consistent with the trend of SunDu-derived Rs. However, the observed Rsdif 2 2 decreased by 1.24 W m per decade (p = 0.00) rather than increasing at 3.69 W m per decade. These results confirm that the overestimation of observed Rs from 1961 to 1991 was primarily caused by the sensitivity drift of the pyranometers which were used to measure Rsdif.

1990 to 1993, the network of Rs measuring stations was restructured. Figure 2 show that seven stations were abandoned and 24 stations were added to the network. The impact of these changes on the time series of regional average Rs was investigated. It was found that the changes in number and location of measuring stations introduced a spurious increase in the regional average Rs over China from 1990 to 1993 (Figure 6). After 1990, Rs should have remained stable during rather than significantly brightening over China, which was confirmed by studies based on diurnal air temperature range [Qian, 2014; Wang et al., 2012a; Wang and Dickinson, 2013b]. The atmospheric aerosol loading from elevated urbanization and economic development canceled out the impact of decreasing cloud fraction on Rs [Qian, 2014; Rienecker et al., 2011; Xia, 2010a]. SunDu-derived Rs is almost free from the sensitivity drift problem and provides an Rs trend of 2.9 W m2 per decade from 1961 to 1990 (Table 4). A similar result was obtained by Tang et al. [2011]. The decreasing trend before 1990 and the nearly constant SunDu-derived Rs afterward were well reproduced by the GISS ESMs (Figure 7). This agreement was due to the inclusion of a near real emission inventory of atmospheric aerosols by CMIP5 ESMs [Taylor et al., 2012]. Although all the CMIP5 ESMs used the same emission inventory, atmospheric aerosol loading also depends on many other processes, such as dry and wet deposition, long-distance transport, and atmospheric chemical processes. Different ESMs performed substantially differently on these processes and provided different trends of Rs (Table 4). The parameterization of the aerosol characteristics and the aerosol-cloud interaction in the GISS ESMs have been extensively investigated and evaluated [de Boer et al., 2013; Lee and Adams, 2010; Li et al., 2010; Menon et al., 2008]. The decreasing trend of Rs over China from 1983 to 2007 from GEWEX SRB was primarily caused by inhomogeneity due to inaccurate use of the polar orbit and of geostationary satellite cloud observations [Dai et al., 2006; Evan et al., 2007; Pinker et al., 2005]. Furthermore, GEWEX SRB did not include interannual variability of tropospheric aerosols. This issue has been corrected in the CERES EABF Rs products. However, the CERES EBAF data were gathered over too short a period to derive a reliable trend due to high annual variability (Figure 7).

WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6508

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

ERA-Interim and MERRA similarly predicted an increasing trend in Rs after 1990. This trend was also invalid and occurred because these products ignored the canceling effect of increased aerosol loading [Streets et al., 2006; Streets et al., 2008; Wang et al., 2009]. Similar results have been reported for other reanalysis products [Xia et al., 2006].

9. Homogenization of Observed Rs Data Figure 6. Annual anomalies of surface incident solar radiation (Rs) averaged from 105 stations over China: raw data for observed Rs (red line) and their homogenization (blue line), sunshine duration (SunDu)-derived Rs averaged from the data when observed Rs was available at the 105 stations (green line, see also Figure 1), and SunDu-derived Rs averaged using the full time series from 1960 to 2012 at the 105 stations (black line). The substantial differences between the black and green lines in 1966–1972 and 1990–1993 show the impact of changing data availability on regional average Rs over China (Figure 1). The annual variability of homogenized Rs followed that of SunDu-derived Rs because SunDu-derived Rs was used as a reference for the homogenization. Rs inhomogeneity was significant before 1994 due to instrument replacement, changes in measurement methods from 1990 to 1993, and degradation of instrument sensitivity before 1989.

A homogeneous climate time series is defined as one whose variations are caused only by variations in climate [Aguilar et al., 2003]. Nonclimatic factors may hide the true climatic signals and patterns and thus potentially bias the conclusions of climate studies [Costa and Soares, 2009]. It is important to assess the homogeneity of long-term climate records before they can be reliably used. Homogenization is a process of nonclimatic inhomogeneity detection and adjustment, which includes four basic steps [Aguilar et al., 2003]: (1) metadata analysis and basic quality control, (2) creation of reference time series, (3) breakpoint detection, and (4) data adjustment. Previous sections have explained that Rs observations over China suffered from a serious inhomogeneity problem due to a sensitivity problem with the pyranometers used, whereas SunDu did not have such a sensitivity drift problem. Because SunDu observations were available at each Rs station, it is reasonable to use SunDu-derived Rs as a reference to homogenize the observed Rs data. Inhomogeneities in the observed Rs data were detected by examining the difference series between the observed and SunDu-derived Rs series. It was assumed that the SunDu-derived Rs series accurately reflected the climate of the region, so that any significant departures from this reference could be directly associated with discontinuities in the observed Rs data [Li et al., 2009]. An open-source software package, RHtest4, was used to perform the homogenization. The current (RHtest4) and previous versions of this software have been used to homogenize daily and monthly air temperature and precipitation data [Vincent et al., 2012; Xu et al., 2013]. Several comparison studies have recommended the use of this software [Venema et al., 2012]. In this study, the monthly Rs were homogenized using two methods: mean adjusted and quantile-matching (Q-M) adjusted [Wang, 2008a, 2008b; Wang, 2008a, 2008b]. The data in Table 3 indicate that both methods work well but that the mean adjusted method has slightly better performance. Therefore, only the results of the mean adjusted method are shown in Figure 6. The data in Table 3 indicate that the agreement between the homogenized Rs and the model-based estimates of Rs was better than the agreements between the raw data and the model-based estimates, because the inhomogeneity of the raw data had been removed (Figure 6). In addition, the homogenized Rs was 2 W m2 higher than the raw data. This difference was attributed to underestimation of Rs before 1990 due to reduced instrument sensitivity (Figures 6 and 7). The decadal variability of the homogenized Rs was similar to that of the SunDu-derived Rs.

WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6509

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

Table 4. Surface-Incident Solar Radiation (Rs) Trends From 1961 to 1990 as Simulated by the Coupled Model Intercomparison Project Phase 5 (CMIP5) 48 Earth a 2 System Models (ESMs), Compared With Those From Raw Rs Observational Data and Sunshine Duration (SunDu) (Units: W m per decade) Trend

Trend

ESM

Trend

Low

High

p Value

ESM

Trend

Low

High

p Value

CMCC-CESM HadCM3 FGOALS-g2 BCC-CSM1-1 BNU-ESM CanCM4 CanESM2 FIO-ESM MIROC-ESM-CHEM MIROC-ESM IPSL-CM5A-LR IPSL-CM5B-LR GFDL-CM2p1 GFDL-CM3 GFDL-ESM2G GFDL-ESM2M GISS-E2-H-CC GISS-E2-H GISS-E2-R-CC GISS-E2-R CESM1-CAM5-1-FV2 CESM1-WACCM NorESM1-ME NorESM1-M Raw Rs observations

0.49 0.60 1.01 0.18 1.27 1.46 1.41 0.64 1.88 1.52 0.05 0.98 1.61 1.66 0.18 1.09 3.11 2.57 2.49 2.90 2.36 1.79 0.81 0.16 8.00

2.22 1.55 2.04 1.32 2.32 2.54 2.62 1.56 3.05 2.71 0.80 1.75 3.24 2.7 1.67 2.21 4.0 3.52 3.31 3.91 3.54 2.96 1.67 0.86 9.22

1.25 0.35 0.02 0.96 0.22 0.39 0.20 0.27 0.70 0.33 0.70 0.21 0.01 0.63 1.31 0.02 2.24 1.62 1.67 1.91 1.17 0.62 0.04 1.18 6.76

0.586 0.223 0.064 0.756 0.025 0.013 0.029 0.176 0.004 0.018 0.895 0.018 0.061 0.004 0.82 0.064 0.000 0.000 0.000 0.000 0.001 0.006 0.072 0.762 0.000

CMCC-CMS CSIRO-Mk3-6-0 MPI-ESM-LR MPI-ESM-MR MPI-ESM-P IPSL-CM5A-MR INMCM4 ACCESS1-0 ACCESS1-3 HadGEM2-AO HadGEM2-CC HadGEM2-ES CNRM-CM5-2 CNRM-CM5 MIROC5 BCC-CSM1-1-M MRI-CGCM3 MRI-ESM1 CCSM4 CESM1-BGC CESM1-CAM5 CESM1-FASTCHEM CMCC-CM MIROC4H SunDu-derived Rs

0.61 0.85 0.15 0.25 1.47 0.43 0.39 1.66 0.19 1.23 0.67 0.95 2.09 1.54 1.69 0.75 0.71 0.56 1.63 1.2 2.03 0.45 0.58 1.22 2.90

1.66 2.07 0.99 1.20 0.55 1.10 1.28 2.73 1.25 2.18 1.70 2.03 2.93 2.52 2.67 2.15 1.44 1.24 2.6 2.32 3.26 1.50 1.72 2.33 3.94

0.43 0.37 1.29 0.70 2.40 0.23 0.50 0.59 0.88 0.27 0.35 0.13 1.25 0.55 0.71 0.65 0.02 0.12 0.67 0.08 0.80 0.59 0.56 0.10 1.86

0.255 0.182 0.795 0.605 0.004 0.210 0.393 0.005 0.734 0.018 0.205 0.095 0.000 0.005 0.002 0.297 0.065 0.118 0.002 0.044 0.003 0.397 0.325 0.040 0.000

a

The p values from Student’s t confidence test are also shown here. If the p value is less than 0.05, this indicates that the trend is statistically significant, i.e., it passes the α = 0.05 confidence test. The low and high values represent the low and high 95% confidence interval of the trend.

Figure 7. Annual anomalies of surface-incident solar radiation (Rs) averaged over 105 stations over China: sunshine duration (SunDu)-derived Rs (thick black line), CMIP5 GISS four-model averages (red line), ERA-Interim (green dotted line), MERRA (green line with plus symbols), CERES SYN (blue dotted line), and GEWEX SRB (blue line with plus symbols). CMIP5 GISS accurately simulated the decreasing trend before 1990 and the near-constant trend afterward. Forty-eight CMIP5 ESM averages simulated similar decadal variability, but with lower rates of decrease. The reanalyses (ERA-Interim and MERRA) exhibited increasing Rs values after 1990 because the reanalysis systems did not include the canceling effect of increased aerosol loading. GEWEX SRB predicted a spurious decreasing trend in Rs from 1983 to 2007 due to inhomogeneity caused by imperfect use of the polar orbit and of geostationary satellite cloud observations.

WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6510

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

10. Conclusions and Discussion In this study, a thorough literature review of the history of Chinese Rs measurements was performed, and intercomparisons between observation-based estimates (direct observations and their homogenizations, as well as SunDu-derived Rs) and model-based estimates of Rs (satellite retrievals, reanalyses, and CMIP5 ESM) were carried out. A preliminary effort was made to homogenize observed Rs over China using SunDu-derived Rs as a reference. The following conclusions can be derived: 1. Rs observations have significant inhomogeneity over China. Before 1989, Rs was calculated as a sum of Rsdir and Rsdif observations measured by pyrheliometers and shaded pyranometers separately. Due to technical limitations and irregular calibration, pyranometers before 1990 had a strong sensitivity drift problem. This introduced crucial spurious decreasing trends into Rsdif and Rs data. The sensitivity drift problem of the pyrheliometers was less critical, and therefore, Rsdir data were more reliable [Luo et al., 2001]. From 1990 to 1993, these instruments were replaced, which introduced a spurious brightening of Rs during this period. This spurious brightening was exacerbated by changes in the network of stations measuring Rs, including their number and their geolocation. After 1993, Rs was measured directly using solid black thermopile pyranometers, which may have an important negative thermal offset during the daytime. 2. Previously reported rates of decreasing and increasing Rs over China were substantially overestimated because of inhomogeneity of Rs observations. The dimming trend from 1961 to 1990 should have been 2.9 W m2 per decade, as obtained from SunDu-derived Rs, rather than the 8.0 W m2 per decade determined from observed raw data. The trend was nearly zero after 1990 (see Table 4). The trend derived from SunDu-derived Rs was confirmed by independent estimates of aerosol optical depth [Luo et al., 2001; Wang, 2014], observed changes in diurnal temperature range [Wang et al., 2012a; Wang and Dickinson, 2013b], and the observed pan evaporation [Yang et al., 2015]. 3. Monthly SunDu-derived Rs had the best agreement with model-based estimates of Rs over China. SunDu-derived Rs can be used as reference data to homogenize observed Rs data. Homogenization was performed on monthly mean Rs data. The adjustments derived from the homogenization process can be applied to both monthly and daily observed data Rs. The homogenized Rs combines the advantages of the observed Rs in depicting diurnal and daily Rs variation and the advantages of SunDu-derived Rs in depicting monthly to decadal Rs variability. On average, the homogenized Rs was ~ 2 W m2 higher than the raw data. This difference was due to underestimation of Rs before 1990 because of the degradation in instrument sensitivity. The thermal offset of the solid black thermopile pyranometers was negative and may have introduced an underestimation of Rs by several W m2 after 1995. Therefore, the biases presented in Table 3 should be several W m2 less. 4. On average, current CMIP5 ESMs overestimated Rs over China by 15–20 W m2. CMIP5 ESMs cannot accurately simulate monthly to annual Rs variability due to their poor performance in simulating seasonal cloud variability. Decadal variability from SunDu-derived Rs was well reproduced by some ESMs. This level of agreement is due to inclusion of a near-real emission inventory of atmospheric aerosols by CMIP5 ESMs. Although all the CMIP5 ESMs used the same emission inventory, atmospheric aerosol loading also depends on many other processes, such as dry and wet deposition, long-distance transport, and atmospheric chemical processes. Different ESMs performed substantially differently in these processes and provided different trends of Rs. 5. Among the model-based Rs estimates, satellite Rs retrievals had the best agreements with the observed and SunDu derived Rs due to their better estimates of cloud parameters. The CERES EABF Rs was slightly better than the GEWEX SRB estimate. However, the time duration of CERES EABF products is too short to derive a reliable trend. The decreasing trend of Rs over China from 1983 to 2007 was not an actual decrease but instead was caused by inhomogeneity due to imperfect usage of the polar orbit and of geostationary satellite cloud observations. The GEWEX SRB did not contain interannual variability tropospheric aerosols either. These issues had been corrected in the CERES EABF Rs products. The biases of the CERES EABF and GEWEX SRB Rs products should have been less than 5 W m2. 6. The accuracy of Rs from reanalyses was better than that of CMIP5 ESMs and worse than that of satellite retrievals. ERA-Interim and MERRA similarly predicted an increasing Rs trend after 1990. This trend was also invalid and occurred because these products ignored the canceling effect of increased aerosol loading. The MERRA Rs products had a strong positive bias of 25–30 W m2 over China. WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6511

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

This study demonstrates the benefits of using SunDu as an indirect observation of Rs. SunDu is only a proxy of Rs and relies on Rs observations to calibrate equation (1), which converts it to Rs. Available daily SunDu observations are known to have difficulty in estimating Rs at time scales shorter than 1 day. However, SunDu has the advantage of being almost free of inhomogeneity. SunDu-derived Rs can be used to detect inhomogeneity in observed Rs data. However, SunDu has been regarded as an outdated observation and has been abandoned by many countries, including the United States. However, this paper argues that observations of SunDu, as an inexpensive alternative or backup for Rs observations, should be retained and continued to provide Rs data for climatic studies. The comparisons performed here enabled better estimation of decadal Rs variability over China. Although this study was performed using Chinese data, it can be performed anywhere that both SunDu and Rs observations are available. Such studies will permit a critical revision of global Rs dimming and brightening data that depend on direct Rs observations [Wild, 2009] and that may be subject to inhomogeneity. This study focused on decadal Rs variability over China. It was found that SunDu-derived Rs depicts decadal variability of Rs well from the 1960s to the 1990s over China. However, we cannot find ground truth to quantify uncertainty of the SunDu-derived Rs. Furthermore, one should look at Rs variability at different time scales. At hourly and daily time scales, Rs can vary from 0 to 1000 W m2. At monthly time scale, Rs can vary from 0 to 500 W m2. At annual time scale, Rs variability is in the tens of Wm2. At decadal time scale, Rs variability is less than 10 W m2. The impacts of the sensitivity drift problem (several W m2), thermal offset (several W m2), and directional response error (tens of W m2) are much less than Rs variability at hourly, daily, and monthly times scales. Under these conditions, it is good to use the observed Rs to validate satellite retrievals and model simulations. In such studies, hourly, daily, or monthly averages of Rs are generally used. In this study, observed daily Rs were also used to calibrate equation (1) which relates Rs to sunshine duration. However, the decadal variability (several W m2) is of similar magnitude to the measurement errors caused by sensitivity drift, thermal offset, and directional response errors. In particular, the systematic sensitivity drift of the pyranometers could have introduced important spurious trends into observed diffuse radiation and Rs. Because the recording material of the sunshine recorder is replaced each day, sunshine duration does not have a sensitivity drift problem, although it has important random errors, and sunshine duration-derived Rs is of much lower accuracy than observed Rs at hourly and daily time scales. In summary, both observed and SunDu-derived Rs are imperfect, but they have their advantages at different time scales. The observed Rs is much better at hourly and daily time scales, whereas SunDu-derived Rs has an advantage in depicting decadal variability. This study represents a preliminary effort to synthesize the advantages of both estimates by homogenizing the observed Rs data set with SunDu-derived Rs as a reference data set. This method combines the advantages of Rs at shorter time scales and the advantages of SunDu-derived Rs at longer time scales. Acknowledgments This study was funded by the National Basic Research Program of China (2012CB955302), the National Natural Science Foundation of China (41175126, 41205036 and 91337111), and the Fundamental Research Funds for the Central Universities (2013YB37). The sunshine duration and surface incident solar radiation data used in this study were released by the Chinese Meteorological Administration (http://cdc.cma.gov.cn/home.do).

WANG ET AL.

References Aguilar, E., I. Auer, M. Brunet, T. C. Peterson, and J. Wieringa (2003), Guidelines on climate metadata and homogenization Rep., 55 pp., World Meteorological Organization, Geneva, Switzerland. Angström, A. (1924), Solar and terrestrial radiation. Report to the international commission for solar research on actinometric investigations of solar and atmospheric radiation, Q. J. R. Meteorol. Soc., 50(210), 121–126. Augustine, J. A., G. B. Hodges, C. R. Cornwall, J. J. Michalsky, and C. I. Medina (2005), An update on SURFRAD—The GCOS Surface Radiation budget network for the continental United States, J. Atmos. Oceanic Technol., 22(10), 1460–1472. Bai, S. L. (1987), Development of the pyranometer model TBQ-2, Acta Energ. Sol. Sin., 8(2), 188–195. Bosilovich, M. G., F. R. Robertson, and J. Y. Chen (2011), Global energy and water budgets in MERRA, J. Clim., 24(22), 5721–5739. Che, H. Z., G. Y. Shi, X. Y. Zhang, R. Arimoto, J. Q. Zhao, L. Xu, B. Wang, and Z. H. Chen (2005), Analysis of 40 years of solar radiation data from China, 1961–2000, Geophys. Res. Lett., 32, L06803, doi:10.1029/2004GL022322. Costa, A. C., and A. Soares (2009), Homogenization of climate data: Review and new perspectives using geostatistics, Math. Geosci., 41(3), 291–305. Dai, A., T. R. Karl, B. Sun, and K. E. Trenberth (2006), Recent trends in cloudiness over the United States—A tale of monitoring inadequacies, Bull. Am. Meteorol. Soc., 87(5), 597–606. de Boer, G., S. E. Bauer, T. Toto, S. Menon, and A. M. Vogelmann (2013), Evaluation of aerosol-cloud interaction in the GISS ModelE using ARM observations, J. Geophys. Res. Atmos., 118, 6383–6395, doi:10.1002/jgrd.50460. Dee, D. P., et al. (2011), The ERA-Interim reanalysis: Configuration and performance of the data assimilation system, Q. J. R. Meteorol. Soc., 137(656), 553–597. Doelling, D. R., N. G. Loeb, D. F. Keyes, M. L. Nordeen, D. Morstad, C. Nguyen, B. A. Wielicki, D. F. Young, and M. Sun (2013), Geostationary enhanced temporal interpolation for CERES flux products, J. Atmos. Oceanic Technol., 30(6), 1072–1090.

SURFACE SOLAR RADIATION OVER CHINA

6512

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

Evan, A. T., A. K. Heidinger, and D. J. Vimont (2007), Arguments against a physical long-term trend in global ISCCP cloud amounts, Geophys. Res. Lett., 34, L04701, doi:10.1029/2006GL028083. Garg, H. P., and S. N. Garg (1993), Measurement of solar radiation—I. Radiation instruments, Renewable Energy, 3(4–5), 321–333. Hess, M., P. Koepke, and I. Schult (1998), Optical properties of aerosols and clouds: The software package OPAC, Bull. Am. Meteorol. Soc., 79(5), 831–844. Kato, S., et al. (2011), Improvements of top-of-atmosphere and surface irradiance computations with CALIPSO-, CloudSat-, and MODIS-derived cloud and aerosol properties, J. Geophys. Res., 116, D19209, doi:10.1029/2011JD016050. Kato, S., N. G. Loeb, F. G. Rose, D. R. Doelling, D. A. Rutan, T. E. Caldwell, L. Yu, and R. A. Weller (2013), Surface irradiances consistent with CERES-derived top-of-atmosphere shortwave and longwave irradiances, J. Clim., 26(9), 2719–2740. Lee, Y. H., and P. J. Adams (2010), Evaluation of aerosol distributions in the GISS-TOMAS global aerosol microphysics model with remote sensing observations, Atmos. Chem. Phys., 10(5), 2129–2144. Li, J., L. Liu, A. A. Lacis, and B. E. Carlson (2010), An optimal fitting approach to improve the GISS ModelE aerosol optical property parameterization using AERONET data, J. Geophys. Res., 115, D16211, doi:10.1029/2010JD013909. Li, Q., H. Zhang, X. Liu, J. Chen, W. Li, and P. Jones (2009), A mainland china homogenized historical temperature dataset of 1951–2004, Bull. Am. Meteorol. Soc., 90(8), 1062–1065. Liang, F., and X. A. Xia (2005), Long-term trends in solar radiation and the associated climatic factors over China for 1961–2000, Ann. Geophys., 23(7), 2425–2432. Lu, W. H., and Z. Q. Bian (2012), Field experiment and data analysis of high precision solar radiation measuring system, Meteorol. Hydrol. Mar. Instrum., 3, 1–5. Lu, W. H., and H. Z. Shi (1987), Temperatuer characteristics of solar radiation instruments, Meteorol. Mon., 13(2), 26–28. Lu, W. H., Y. Q. Mo, and D. Wang (2002), Characteristics investigation for pyranometers, Acta Energ. Sol. Sin., 23(3), 313–316. Lu, Z. T., S. J. Yu, and S. B. Hou (2012), Discussions on measurement uncertainty of sunshine duration by Jordan and Campbell-Stokes sunshine recorders, Heilongjiang Meteorol., 29(2), 31–32. Luo, Y., D. Lu, X. Zhou, W. Li, and Q. He (2001), Characteristics of the spatial distribution and yearly variation of aerosol optical depth over China in last 30 years, J. Geophys. Res., 106(D13), 14,501–14,513, doi:10.1029/2001JD900030. Ma, Y. Z., X. N. Liu, and S. Xu (1998), The description of Chinese radiation data and their quality control procedures, Meteorol. Sci. Tech., 2, 53–56. Magee, N. B., E. Melaas, P. M. Finocchio, M. Jardel, A. Noonan, and M. J. Iacono (2014), Blue hill observatory sunshine: Assessment of climate signals in the longest continuous meteorological record in North America, Bull. Am. Meteorol. Soc., 95(11), 1741–1751. Menon, S., A. D. Del Genio, Y. Kaufman, R. Bennartz, D. Koch, N. Loeb, and D. Orlikowski (2008), Analyzing signatures of aerosol-cloud interactions from satellite retrievals and the GISS GCM to constrain the aerosol indirect effect, J. Geophys. Res., 113, D14S22, doi:10.1029/ 2007JD009442. Minnis, P., et al. (2008), Cloud detection in nonpolar regions for CERES using TRMM VIRS and Terra and Aqua MODIS Data, IEEE Trans. Geosci. Remote Sens., 46(11), 3857–3884. Mo, Y. Q., Y. Yang, H. L. Liang, and D. Wang (2008), Investigation report on technology of status and development of meteorological radiation observation in China, Chin. J. Sci. Instrum., 29(8), 518–522. Naud, C. M., J. F. Booth, and A. D. Del Genio (2014), Evaluation of ERA-Interim and MERRA Cloudiness in the Southern Ocean, J. Clim., 27(5), 2109–2124. Peterson, T. C., et al. (1998), Homogeneity adjustments of in situ atmospheric climate data: A review, Int. J. Climatol., 18(13), 1493–1517. Philipona, R. (2002), Underestimation of solar global and diffuse radiation measured at Earth′s surface, J. Geophys. Res., 107(D22), 4654, doi:10.1029/2002JD002396. Pinker, R. T., B. Zhang, and E. G. Dutton (2005), Do satellites detect trends in surface solar radiation?, Science, 308(5723), 850–854. Qian, C. (2014), Impact of land use/land cover change on changes in surface solar radiation in eastern China since the reform and opening up, Theor. Appl. Climatol., 1–9, doi:10.1007/s00704-014-1334-5. Qian, Y., D. P. Kaiser, L. R. Leung, and M. Xu (2006), More frequent cloud-free sky and less surface solar radiation in China from 1955 to 2000, Geophys. Res. Lett., 33, L01812, doi:10.1029/2005GL024586. Rienecker, M. M., et al. (2011), MERRA: NASA′s Modern-Era Retrospective Analysis for Research and Applications, J. Clim., 24(14), 3624–3648. Rossow, W. B., and R. A. Schiffer (1999), Advances in understanding clouds from ISCCP, Bull. Am. Meteorol. Soc., 80(11), 2261–2287. Sanchez-Lorenzo, A., and M. Wild (2012), Decadal variations in estimated surface solar radiation over Switzerland since the late 19th century, Atmos. Chem. Phys., 12(18), 8635–8644. Sanchez-Lorenzo, A., J. Calbo, and J. Martin-Vide (2008), Spatial and temporal trends in sunshine duration over western Europe (1938–2004), J. Clim., 21(22), 6089–6098. Sanchez-Romero, A., A. Sanchez-Lorenzo, J. Calbó, J. A. González, and C. Azorin-Molina (2014), The signal of aerosol-induced changes in sunshine duration records: A review of the evidence, J. Geophys. Res. Atmos., 119, 4657–4673, doi:10.1002/2013JD021393. Shi, G. Y., T. Hayasaka, A. Ohmura, Z. H. Chen, B. Wang, J. Q. Zhao, H. Z. Che, and L. Xu (2008), Data quality assessment and the long-term trend of ground solar radiation in China, J. Appl. Meteorol. Climatol., 47(4), 1006–1016. Streets, D. G., Y. Wu, and M. Chin (2006), Two-decadal aerosol trends as a likely explanation of the global dimming/brightening transition, Geophys. Res. Lett., 33, L15806, doi:10.1029/2006GL026471. Streets, D. G., C. Yu, Y. Wu, M. Chin, Z. Zhao, T. Hayasaka, and G. Shi (2008), Aerosol trends over China, 1980–2000, Atmos. Res., 88(2), 174–182. Stubenrauch, C. J., et al. (2013), Assessment of global cloud datasets from satellites: Project and database initiated by the GEWEX radiation panel, Bull. Am. Meteorol. Soc., 94(7), 1031–1049. Tang, W. J., K. Yang, J. Qin, C. C. K. Cheng, and J. He (2011), Solar radiation trend across China in recent decades: A revisit with quality-controlled data, Atmos. Chem. Phys., 11(1), 393–406. Taylor, K. E., R. J. Stouffer, and G. A. Meehl (2012), An overview of CMIP5 and the experiment design, Bull. Am. Meteorol. Soc., 93(4), 485–498. Venema, V. K. C., et al. (2012), Benchmarking homogenization algorithms for monthly data, Clim. Past, 8(1), 89–115. Vincent, L. A., X. L. Wang, E. J. Milewska, H. Wan, F. Yang, and V. Swail (2012), A second generation of homogenized Canadian monthly surface air temperature for climate trend analysis, J. Geophys. Res., 117, D18110, doi:10.1029/2012JD017859. Wang, A., and X. Zeng (2012), Evaluation of multireanalysis products with in situ observations over the Tibetan Plateau, J. Geophys. Res., 117, D05102, doi:10.1029/2011JD016553. Wang, K. C. (2014), Measurement biases explain discrepancies between the observed and simulated decadal variability of surface incident solar radiation, Sci. Rep., 4, 6144. Wang, K. C., and R. E. Dickinson (2013a), Global atmospheric downward longwave radiation at the surface from ground-based observations, satellite retrievals and reanalyses, Rev. Geophys., 51, 150–185, doi:10.1002/rog.20009.

WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6513

Journal of Geophysical Research: Atmospheres

10.1002/2015JD023420

Wang, K. C., and R. E. Dickinson (2013b), Contribution of solar radiation to decadal temperature variability over land, Proc. Natl. Acad. Sci. U.S.A., 110(37), 14,877–14,882. Wang, K. C., R. E. Dickinson, and S. Liang (2009), Clear sky visibility has decreased over land globally from 1973 to 2007, Science, 323(5920), 1468–1470. Wang, K. C., R. E. Dickinson, M. Wild, and S. Liang (2010a), Evidence for decadal variation in global terrestrial evapotranspiration between 1982 and 2002: 1. Model development, J. Geophys. Res., 115, D20112, doi:10.1029/2009JD013671. Wang, K. C., R. E. Dickinson, M. Wild, and S. Liang (2012b), Atmospheric impacts on climatic variability of surface incident solar radiation, Atmos. Chem. Phys., 12(20), 9581–9592. Wang, K. C., K. E. Dickinson, Q. Ma, J. A. Augustine, and M. Wild (2013), Measurement methods affect the observed global dimming and brightening, J. Clim., 26(12), 4112–4120. Wang, K., H. Ye, F. Chen, Y. Xiong, and C. Wang (2012a), Urbanization effect on the diurnal temperature range: Different roles under solar dimming and brightening, J. Clim., 25(3), 1022–1027. Wang, X. L. (2008a), Accounting for autocorrelation in detecting mean shifts in climate data series using the penalized maximal t or F Test, J. Appl. Meteorol. Climatol., 47(9), 2423–2444. Wang, X. L. (2008b), Penalized maximal F test for detecting undocumented mean shift without trend change, J. Atmos. Oceanic Technol., 25(3), 368–384. Wang, X. L., H. Chen, Y. Wu, Y. Feng, and Q. Pu (2010b), New techniques for the detection and adjustment of shifts in daily precipitation data series, J. Appl. Meteorol. Climatol., 49(12), 2416–2436. Wang, Y., Y. Yang, N. Zhao, C. Liu, and Q. Wang (2012c), The magnitude of the effect of air pollution on sunshine hours in China, J. Geophys. Res., 117, D00V14, doi:10.1029/2011JD016753. Wielicki, B. A., B. R. Barkstrom, E. F. Harrison, R. B. Lee, G. Louis Smith, and J. E. Cooper (1996), Clouds and the Earth’s Radiant Energy System (CERES): An earth observing system experiment, Bull. Am. Meteorol. Soc., 77(5), 853–868. Wild, M. (2009), Global dimming and brightening: A review, J. Geophys. Res., 114, D00D16, doi:10.1029/2008JD011470. Wild, M., A. Ohmura, and K. Makowski (2007), Impact of global dimming and brightening on global warming, Geophys. Res. Lett., 34, L04702, doi:10.1029/2006GL028031. Winker, D. M., et al. (2010), The CALIPSO mission: A global 3D view of aerosols and clouds, Bull. Am. Meteorol. Soc., 91(9), 1211–1229. WMO (2008), Measurement of radiation, in Guide to Meteorological Instruments and Methods of Observation, pp. 1–41, WMO. [Available at http://www.wmo.int/pages/prog/www/IMOP/CIMO-Guide.html.] Xia, X. (2010a), A closer looking at dimming and brightening in China during 1961–2005, Ann. Geophys., 28(5), 1121–1132. Xia, X. (2010b), Spatiotemporal changes in sunshine duration and cloud amount as well as their relationship in China during 1954–2005, J. Geophys. Res., 115, D00K06, doi:10.1029/2009JD012879. Xia, X. A., P. C. Wang, H. B. Chen, and F. Liang (2006), Analysis of downwelling surface solar radiation in China from National Centers for Environmental Prediction reanalysis, satellite estimates, and surface observations, J. Geophys. Res., 111, D09103, doi:10.1029/ 2005JD006405. Xu, J. Q., K. Masuda, Y. Ishigooka, T. Kuwagata, S. Haginoya, T. Hayasaka, and T. Yasunari (2011), Estimation and verification of daily surface shortwave flux over China, J. Meteorol. Soc. Jpn., 89A, 225–238. Xu, W., Q. Li, X. L. Wang, S. Yang, L. Cao, and Y. Feng (2013), Homogenization of Chinese daily surface air temperatures and analysis of trends in the extreme temperature indices, J. Geophys. Res. Atmos., 118, 9708–9720, doi:10.1002/jgrd.50791. Yang, H., Z. Li, M. Li, and D. Yang (2015), Inconsistency in Chinese solar radiation data caused by instrument replacement: Quantification based on pan evaporation observations, J. Geophys. Res. Atmos., 120, 3191–3198, doi:10.1002/2014JD023015. Yang, K., T. Koike, and B. S. Ye (2006), Improving estimation of hourly, daily, and monthly solar radiation by importing global data sets, Agric. For. Meteorol., 137(1–2), 43–55. Yang, Y., D. Wang, W. H. Lu, Y. Q. Mo, and L. Ding (2007), National references of solar radiation of China and its quantity dissemination, Chin. J. Sci. Instrum., 28(SV), 429–435. Yang, Y., L. Ding, and D. Wang (2010), Experiments and analysis of pyranometer on nighttime zero offset, Meteorol. Mon., 36(11), 100–103. Yang, Y., L. Ding, J. H. Quan, and X. H. Cheng (2012), Uncertainty analysis of calbiration results of a homemade pyranometer, Meteorol. Sci. Tech., 40(1), 20–24. Zhang, M. H., et al. (2005), Comparing clouds and their seasonal variations in 10 atmospheric general circulation models with satellite measurements, J. Geophys. Res., 110, D15S02, doi:10.1029/2004JD005021. Zhang, W. M., and Z. H. Lu (1988), Technical report on telemering radiometer, Meteorol. Mon., 14(7), 20–23. Zhang, W. M., and Z. H. Lu (1990), Introdction to telemetry radiation instruments and methods of observation, Meteorol. Mon., 16(8), 53–55. Zhang, W. M., and Y. X. Tan (2000), Approach error of sunshine recorder, Meteorol. Mon., 26(5), 17–19. Zhang, Y., and J. Li (2013), Shortwave cloud radiative forcing on major stratus cloud regions in AMIP-type simulations of CMIP3 and CMIP5 models, Adv. Atmos. Sci., 30(3), 884–907. Zhang, Y., H. Chen, and R. Yu (2014), Simulations of stratus clouds over eastern China in CAM5: Sensitivity to horizontal resolution, J. Clim., 27(18), 7033–7052. Zhang, Y. C., W. B. Rossow, A. A. Lacis, V. Oinas, and M. I. Mishchenko (2004), Calculation of radiative fluxes from the surface to top of atmosphere based on ISCCP and other global data sets: Refinements of the radiative transfer model and the input data, J. Geophys. Res., 109, D19105, doi:10.1029/2003JD004457. Zhao, D., Y. Luo, G. Gao, C. Zhu, and Y. Shen (2010), Long-term changes and essential climatic characteristics of sunshine duration over China during 1961–2007, Resour. Sci., 32(4), 701–711. Zhou, Y. H., S. Z. Bao, and F. Z. San (1978), Measurement of temeprature effect of relative solar radiation instruments, Meteorol. Mon., 4(2), 32–33.

WANG ET AL.

SURFACE SOLAR RADIATION OVER CHINA

6514