Observed trends in sunshine duration over South America


  • Carlos Raichijk

    Corresponding author
    1. GERSolar, Instituto de Ecología y Desarrollo Sustentable (INEDES), Universidad Nacional de Luján, Rutas 5 y 7, 6700 Luján, Buenos Aires, Argentina
    • GERSolar, Instituto de Ecología y Desarrollo Sustentable (INEDES), Universidad Nacional de Luján, Rutas 5 y 7, 6700 Luján, Buenos Aires, Argentina.
    Search for more papers by this author


In this work, temporal series of monthly mean values of sunshine duration were studied in five climatic regions of South America. For this purpose, data from 237 meteorological stations throughout South America were taken into account in: Argentina (20), Bolivia (2), Brazil (190), Paraguay (11), Peru (3), and Uruguay (11). The stations have been grouped into the following five climatic regions according to the classification of the Brazilian Institute of Geography and Statistics (IBGE): Equatorial, Tropical Equatorial, Tropical Central Brazil—Warm, Tropical Central Brazil—Mesothermal, and Humid Temperate. The non-parametric test of Mann-Kendall was used to evaluate possible trends for both the period of maximum extension of available data, 1961–2004, as well as for the sub-periods 1961–1990 and 1991–2004. From this analysis, significant trends could be observed, first decreasing between 1961 and 1990, and then increasing from 1990 onwards. The trends found for the regions studied are in agreement with those observed in other regions of the planet, thus providing further evidence of the phenomenon that has been referred to as global dimming and brightening. Copyright © 2011 Royal Meteorological Society

1. Introduction

The study of solar radiation incidence has assumed fundamental importance in the context of the present climate changes. Changes in the levels of radiation over the years could have a strong impact on the Earth's entire ecosystem, climate, and economical activity. In recent years, a decrease in solar radiation incidence has been observed from the 1950s to the 1980s (Abakumova et al., 1996; Stanhill and Cohen, 2001; Liepert, 2002), a phenomenon known as global dimming. Later studies (Wild et al., 2005; Pinker et al., 2005) have shown a change in the decreasing trend from the second half of the 1980s decade onwards, referred to as global brightening. Although the causes of these variations are still not clear, they can be attributed to changes in atmospheric transmittance, both due to variations in cloudiness and anthropogenic aerosol concentrations (Wild, 2009). Liepert (2002) has found that the decrease in surface solar radiation observed in the USA between 1961 and 1991 was larger for cloudy days (−18 W/m2) than for clear sky days (−8 W/m2), where the variation may be due to direct effects of aerosols. Cutforth and Judiesch (2007), by analysing correlations of solar radiation with other meteorological variables measured for 7 locations of the Canadian Prairies between 1951 and 2005, have suggested that the main cause for the decrease in radiation incidence was the increase detected in atmospheric water content and cloudiness. Stjern et al. (2009), when analysing records of 11 stations in northwest and Arctic Europe, have found a good degree of correlation between incident solar radiation and cloud cover which would allow explaining the trends mentioned earlier. On the other hand, Wild et al. (2005), Streets et al. (2006) and Ruckstuhl et al. (2008) have suggested that the recent increase in incident radiation could be due to the decrease in anthropogenic aerosol emissions originated by more effective regulation policies in industrialised countries, and to the economic decline of most eastern European countries in the late 1980s.

However, Alpert and Kischa (2008) question the global character of these phenomena. They have observed that solar radiation decreases only in areas with a population density higher than 10 per/km2, according to the analysis of year-to-year variations of annual radiation fluxes between 1964 and 1989 for 317 stations from the GEBA pyranometric world net (Gilgen et al., 1998). In addition to this, Gueymard and Myers (2009) have noticed that in many cases the observed variation is smaller than the uncertainties introduced by the measurement process, mainly due to instrumental degradation, loss of data, change of sensors, or calibration methods. To illustrate this, they have analysed trends observed at the station of NREL (Golden, USA) using simultaneous global radiation data from 5 different sensors during the period 2001–2007. For the month of December, for instance, they have detected negative trends with all instruments but with magnitudes that differ significantly.

Different studies in South America (Righini et al., 2005; Tiba et al., 2005; Grossi Gallegos et al., 2006; Raichijk et al., 2006) have focused on rescuing sunshine duration data since this is one of the parameters most linked to solar radiation with historical records spanning large time intervals and ample spatial coverage in the whole region. These characteristics, also available on a global scale, highlight the great importance of sunshine series for temporal evolution studies of solar radiation incidence over long periods of time. Hence, the temporal behaviour of sunshine series has been analysed in different regions of the planet. Angell (1990) and Stanhill and Cohen (2005) in the USA, Pallé and Butler (2001) in Ireland, Kaiser and Qian (2002), and Yang et al. (2009) in China, have reported decreases in mean values of the sunshine from the 1950s. Stanhill and Cohen (2008) have observed a change of trend from the mid 1980s in Japan. Sanchez-Lorenzo et al. (2007), interpolating sunshine records of 72 stations of the Iberian Peninsula, have generated grids of 1° × 1° resolution. The study was conducted over the entire region and for four sub-regions defined by principal components analysis. The regions thus obtained are consistent geographically and can be identified with different climate types. The temporal evolution of the mean annual series for the whole region showed a decrease in sunshine duration from the 1950s to the early 1980s, followed by a positive trend up to the end of the 20th century. A similar trend in temporal evolution in sunshine—especially during spring time—was reported from studying 79 locations throughout Western Europe (Sanchez-Lorenzo et al., 2008). You et al. (2009) have found in the eastern and central Tibetan Plateau a good degree of correlation between regional annual and seasonal means of sunshine duration obtained from data of 71 ground-measurement stations, and satellite-derived values of solar radiation downward fluxes and different climatic variables from ERA-40 reanalysis dataset. The correlations are consistent with the temporal evolution observed for the series of these parameters: a decreasing trend in sunshine duration after the 1980s (in opposition to the results found for most areas of the world) corresponds to a decreasing trend in solar radiation downwards fluxes as well as an increasing trend in low cloud amount, water vapor pressure, precipitation, and relative humidity during the same period.

In the present work, monthly means records of sunshine duration obtained employing Campbell-Stoke heliographs for 237 meteorological stations throughout South America from 1961 to 2004 were analysed. The stations have been grouped into five climatic regions. In order to detect possible trends, the temporal behaviour of the annual and seasonal series of sunshine duration has been studied in each region, together with shortwave radiative fluxes, for all skies (SWAll) and clear skies (SWClear), and cloud fraction (CF) obtained from NASA's Surface Radiation Budget (SRB) Release 3.0 dataset.

2. Data and climatic regions

Monthly mean sunshine data was analysed for six countries in South America (Table VII). In Table I are indicated the number of meteorological stations considered for each country as well as the monitoring time corresponding to each dataset used.

Table I. Number of stations considered for each country and time period of each dataset used
CountryPeriodNumber of stations

The records for each station were made consistent, eliminating those values that were off by more than ± 1.96σ from the average value for the period considered (compatible with the level of significance for the test to be used). Owing to the fact that the series of data for nearly all the stations analysed are incomplete, the stations were grouped by climatic regions in order to facilitate the interpolation of the available data for each month over these areas. The mean errors associated with this estimation depend on the density of stations, the variability in the sunshine field, and the accuracy of the measured values.

In previous work, for the study of the spatial variability of sunshine in the flat region with Humid Temperate climate of South America (Raichijk et al., 2006), the so-called spatial variability coefficient (Hay and Suckling, 1979) could be defined as a function of the distance between all pairs of recording stations with good degree of fitting. The spatial variability coefficient defined in this way can be used for the determination of interpolation error or spatial coverage by stations for a given extrapolation error. In Figure 1, the spatial variability coefficient, CV(%), defined with a confidence level of 90%, is plotted as a function of the distance between stations in the Humid Temperate region.

Figure 1.

Spatial variability coefficient, CV(%), as a function of the distance between stations in the Humid Temperate Region (Raichijk et al., 2006)

The Equation for the obtained fitted curve is the following:

equation image(1)

where d is the distance in km between pairs of stations.

It can be seen from Equation (1) that it is possible to extrapolate data for a measuring station up to a distance of 240 km without getting an error larger than 15%. On the other hand, as Gandin (1970) proposes, extrapolating (1) to distance zero and dividing by the root of 2, the mean measurement error of sunshine can be estimated, which in this region is equal to 7.7% with a 90% confidence level. Taking the case studied earlier as an example, it was considered that for a sufficiently flat given climatic region the conditions of homogeneity and isotropy are fulfilled, so that the values obtained in those stations can be interpolated over the whole region considered without introducing significant errors.

Monthly mean values of shortwave radiative fluxes and cloud fraction at 1°× 1° degree resolution, provided by NASA's Surface Radiation Budget (SRB) Release 3.0 dataset (http://eosweb.larc.nasa.gov/PRODOCS/srb/table_srb.html) and available for the 1984–2005 period, were also used for this study.

The climatic and topographic characteristics for each region are detailed below, indicating within brackets the number of stations used for each country. The classification and terminology introduced by the Brazilian Institute of Geography and Statistics (IBGE) was taken into account.

Equatorial Region: Af/Am according to the Köeppen-Geiger classification (Geiger and Pohl, 1953). This region comprises the Amazonia region of Brazil (33), Peru (3), and Bolivia (2). Hot with average annual temperatures between 24 and 27 °C, and average annual precipitations between 1500 and 2500 mm, without a dry season, or with a short dry season, under 3 months. Flat region, maximum elevation: 415 m, Gleba Celeste station, Mato Grosso, Brazil.

Tropical Equatorial Region: Aw/Bsh according to the Köeppen-Geiger classification. This region comprises the northeast Brazilian Sertão (42), going south until approximately 10°S latitude. Hot, semi-arid with a prolonged dry season. Flat region, maximum elevation: 680 m, Arcoverde station, Pernambuco, Brazil.

Tropical Central Brazil—Warm Region: Aw according to the Köeppen-Geiger classification. This region comprises the Brazilian Central Plain (34). Semi-humid with a dry winter. Flat region, with altitude between 400 and 1000 m, maximum elevation: 1097 m, Itamarandiba station, Minas Gerais, Brazil.

Tropical Central Brazil—Mesothermal Region: Cw according to the Köeppen-Geiger classification. This region comprises the centre and south of Goias, south of Minas Gerais, north of Rio Janeiro, San Pablo and Mato Grosso do Sul states, and north of Paraná in Brazil (61) plus Eastern region of Paraguay (9). Average annual temperatures between 15 and 18 °C with a dry winter. This region comprises the Sierras of Mantiqueira, with altitude between 1000 and 2800 m, maximum elevation: 1642 m, Campos do Jordao station, Sao Pablo, Brazil.

Humid Temperate Region: Cfa according to the Köeppen-Geiger classification. Centre and south of Paraná, Santa Catarina and Rio Grande do Sul states in Brazil (20), Mesopotamic and Humid Pampa regions in Argentina (20), Uruguay Republic (11), plus Encarnación and Capitan Miranda stations in Paraguay (2). Average annual temperatures between 10 and 15 °C without a dry season. Although mainly a flat region, it comprises Sierras Geral and do Mar in South Brazil, with altitude up to 1000 m, maximum elevation: 1047 m, Bom Jesus station, Santa Catarina, Brazil.

3. Methods

For each region, kriging was used to interpolate (with a spatial resolution of 1°× 1°) the monthly mean values of sunshine from their respective stations for each month of the period 1961–2004. Interpolation could not be performed when the number of stations was less than 5, since in these cases the mean coverage radius by measurement station could be up to 1000 km, which according to Equation (1) would imply interpolation errors greater than 25%. The excluded cases are: Equatorial Region, years between 1961 and 1969; Tropical Equatorial Region, year 1961 and December 1989; Tropical Central Brazil—Warm Region, years 1966, 1967, and 1968; Humid Temperate Region, September 1963. Figure 2 indicates the locations of stations and grid points of interpolated values for the different climatic regions analysed.

Figure 2.

Location of the sunshine duration stations (black dots) and grid points of interpolated values (gray dots) for the different climatic regions: (a) Equatorial, (b) Tropical Equatorial, (c) Tropical Central Brazil—Warm, (d) Tropical Central Brazil—Mesothermal and (e) Humid Temperate

For each grid point, the annual and seasonal sums of the interpolated values were determined for every year of the corresponding series. The following seasonal sums were considered: summer, addition of January, February, November, and December values; winter, addition of May, June, July, and August values; spring-autumn, addition of March, April, September, and October values. Finally, the regional mean values were determined by averaging the values obtained for all grid points for each year of the series.

In Figure 3, the temporal evolution of the regional means for the annual sums for each region is shown. The curve that best fits the series, with its respective determination coefficient, R2, is indicated as well.

Figure 3.

Temporal evolution of the regional means for the annual sums of sunshine duration for each region studied: (a) Equatorial, (b) Tropical Equatorial, (c) Tropical Central Brazil—Warm, (d) Tropical Central Brazil—Mesothermal and (e) Humid Temperate

In the three climatic regions located north of 15°S latitude, a parabolic type fit can be observed, with an inflexion in the tendencies from the second half of the 1980s onwards. On the other hand, in the two regions that stand south of this latitude, Tropical Central Brazil—Mesothermal (with a very low determination coefficient R2) and Humid Temperate, the fitting obtained is linear with decreasing trends in both cases.

In order to unify criteria of analysis for the different regions, it was decided to evaluate possible trends in both the maximum available period, 1961–2004, as well as in the sub-periods 1961–1990 and 1991–2004. The method originally set up by Mann (1945), which was then re-formulated by Kendall in 1948 (Kendall and Stuart, 1979), was used. This type of non-parametric models inform about the sign of change and its significance. For the estimation of the magnitude of change, linear fitting was employed assuming the normality and homogeneity of the variance throughout the series.

It was also decided to express the magnitude of change observed, i.e. the slope obtained from the linear regression, in the cases when a significance level higher than 90% is found (p-level ⩽ 0.1 when performing the non-parametric test of Mann-Kendall) and a Pearson linear correlation coefficient, r, with a magnitude equal or higher than 0.5. In case of having a significance higher than 90% but with equation image (expressed in cursive writing) only the sign of the found change will be indicated. In Tables II, III, IV, V and VI, are indicated the results obtained from studying the trends of the regional means for the annual, summer, winter, and spring-autumn sums, in each of the 5 regions studied.

Table II. Trends in regional means of sunshine duration for annual, winter and spring-autumn sums for Equatorial Region
Equatorialslope (h/year)r-Pearsonp-levelslope (h/year)r-Pearsonp-levelslope (h/year)r-Pearsonp-level
Annual sums..........0.76− 0.45 ± 0.10− 0.730.00060.83 ± 0.110.910.0003
Summer sums..........0.64− 0.16 ± 0.03− 0.740.00080.23 ± 0.050.810.001
Winter sums..........0.66− 0.15 ± 0.04− 0.650.0010.32 ± 0.060.830.0003
Spring-autumn sums..........0.90− 0.15 ± 0.04− 0.640.0030.27 ± 0.050.840.002
Table III. As in Table II, but for Tropical Equatorial Region
Tropical Equatorialslope (h/year)r-Pearsonp-levelslope (h/year)r-Pearsonp-levelslope (h/year)r-Pearsonp-level
Annual sums..........0.12..........0.280.38 ± 0.130.660.02
Summer sums..........0.140.350.07..........0.25
Winter sums+0.340.02..........0.33..........0.61
Spring-autumn sums+0.310.05..........0.200.26 ± 0.080.690.007
Table IV. As in Table II, but for Tropical Central Brazil—Warm Region
Tropical Central Brazil—Warmslope (h/year)r-Pearsonp-levelslope (h/year)r-Pearsonp-levelslope (h/year)r-Pearsonp-level
Annual sums..........0.79− 0.35 ± 0.12− 0.520.01..........0.58
Summer sums..........0.65..........0.12..........0.66
Winter sums..........0.23− 0.12 ± 0.03− 0.590.002..........0.37
Spring-autumn sums..........0.670.480.0050.23 ± 0.110.530.03
Table V. As in Table II, but for Tropical Central Brazil—Mesothermal Region
Tropical Central Brazil—Mesothermalslope (h/year)r-Pearsonp-levelslope (h/year)r-Pearsonp-levelslope (h/year)r-Pearsonp-level
Annual sums0.320.03− 0.26 ± 0.07− 0.570.00070.56 ± 0.180.670.01
Summer sums..........0.190.370.06..........0.34
Winter sums0.380.01− 0.12 ± 0.02− 0.700.0001..........0.47
Spring-autumn sums..........0.600.370.070.37 ± 0.090.770.001
Table VI. As in Table II, but for Humid Temperate Region
Humid Temperateslope (h/year)r-Pearsonp-levelslope (h/year)r-Pearsonp-levelslope (h/year)r-Pearsonp-level
Annual sums0.490.0010.400.020.29 ± 0.140.500.06
Summer sums0.330.010.350.02..........0.32
Winter sums..........0.25..........0.56+0.370.02
Spring-autumn sums0.480.003..........0.21..........0.27
Table VII. Coordinates and annual means values of sunshine duration of stations utilised in this work (the stations are ordered by latitude; country is indicated as: Argentina (Ar), Bolivia (Bo), Brazil (Br), Paraguay (Pa), Peru and Uruguay (Ur) and Climatic Region as: Equatorial (E), Tropical Equatorial (TE), Tropical Central Brazil—Warm (TCB-warm), Tropical Central Brazil—Mesothermal (TCB-meso) and Humid Temperate (HT)
RegionCountryStationLat. (°)Long. (a)Alt. (m)Annual mean(hours)RegionCountryStationLat. (°)Long. (a)Alt. (m)Annual mean(hours)
EBrIauaretê0.62− 69.201203.9EBrCoari− 4.08− 63.13465.2
EBrMacapa− 0.03− 51.05146.5TEBrBacabal− 4.25− 44.78256.3
EBrS.G.da Cachoeira− 0.12− 67.00904.7EBrItaituba− 4.27− 55.58455.6
EBrSoure− 0.73− 48.52106.5EBrBenjamin Constant− 4.38− 70.03654.2
EBrBarcelos− 0.98− 62.92405.1TEBrJaguaruana− 4.78− 37.77128.4
EBrTracuateua− 1.08− 46.93366.0TEBrCaxias− 4.87− 43.351037.2
EBrBelem− 1.45− 48.47106.1EPeruGenaro Herrera− 4.9− 73.631264.5
EBrBreves− 1.67− 50.48155.9TEBrMacau− 5.12− 36.7737.5
EBrTuriacu− 1.72− 45.40446.1TEBrMorada Nova− 5.12− 38.37438.3
EBrPorto de Moz− 1.73− 52.23165.7TEBrCrateus− 5.17− 40.672967.3
EBrObidos− 1.92− 55.52376.1TEBrQuixeramobim− 5.17− 39.28797.8
EBrMonte Alegre− 2.00− 54.081456.4TEBrMossoro− 5.20− 37.30388.0
EBrCameta− 2.25− 49.50246.9EBrMaraba− 5.35− 49.15955.6
EBrFonte Boa− 2.53− 66.17554.5TEBrBarra do Corda− 5.50− 45.271536.2
TEBrSao Luis− 2.53− 44.30506.3TEBrImperatriz− 5.53− 47.481236.1
EBrBelterra− 2.63− 54.951756.1TEBrApodi− 5.62− 37.821508.5
EBrParintins− 2.63− 56.73295.9TEBrCeara Mirim− 5.65− 35.65617.6
TEBrAcarau− 2.88− 40.13167.9TEBrGrajau− 5.80− 46.451636.1
TEBrParnaiba− 3.08− 41.77798.0EBrManicore− 5.82− 61.30504.7
EBrManaus− 3.12− 59.95674.9EPeruSan Ramón− 5.93− 76.081204.8
EBrItacoatiara− 3.13− 58.43404.3TEBrTaua− 6.00− 40.423986.9
EBrAltamira− 3.20− 51.20744.9TEBrColinas− 6.05− 44.251796.4
EBrTucurui− 3.72− 49.72405.6TEBrFlorania− 6.12− 36.823247.6
TEBrZe Doca− 3.72− 45.53456.5TEBrIguatu− 6.37− 39.302178.1
TEBrChapadinha− 3.73− 43.351037.5TEBrCruzeta− 6.43− 36.582268.1
TEBrSobral− 3.73− 40.331097.2EPeruEl Porvenir− 6.58− 76.312304.9
TEBrFortaleza− 3.77− 38.55267.9EBrSao Felix do Xingu− 6.63− 51.972064.0
EBrCodajas− 3.83− 62.08484.8TEBrSão Gonçalo− 6.75− 38.222338.9
EBrTefe− 3.83− 64.70475.0TEBrFloriano− 6.77− 43.021237.6
TEBrCampos Sales− 7.00− 40.385837.6TCB-warmBrDiamantino− 14.40− 56.452864.5
TEBrAraguaina− 7.20− 48.202285.5EBoAngosto del Bala− 14.55− 67.553005.1
EBrLabrea− 7.25− 64.83614.3TCB-warmBrVitoria da Conquista− 14.88− 40.808746.2
TEBrBarbalha− 7.32− 39.304097.9TCB-warmBrEspinosa− 14.92− 42.855697.8
TEBrCarolina− 7.33− 47.471926.5TCB-warmBrFormoso− 14.93− 46.258406.0
TEBrBalsas− 7.53− 46.032596.5TCB-warmBrMocambinho− 15.08− 44.024528.0
EBrCruzeiro do Sul− 7.63− 72.671703.7TCB-warmBrMonte Azul− 15.08− 42.756037.7
TEBrTriunfo− 7.82− 38.1211057.7TCB-warmBrGoianesia− 15.22− 49.006506.8
TEBrMonteiro− 7.88− 37.076047.6TCB-warmBrJanuaria− 15.45− 44.374747.8
EBrTarauaca− 8.17− 70.771904.4TCB-mesoBrFormosa− 15.53− 47.339356.7
TEBrC. D Araguaia− 8.25− 49.281576.0TCB-warmBrCuiaba− 15.55− 56.121516.6
TEBrArcoverde-8.43− 37.056817.8TCB–mesoBrBrasilia− 15.78− 47.9311596.5
TEBrCabrobo− 8.52− 39.333417.7TCB-mesoBrPirenopolis− 15.85− 48.977406.4
TEBrPedro Afonso− 8.97− 48.181876.6TCB-warmBrAragarcas− 15.90− 52.233456.6
TEBrAlto Parnaiba− 9.10− 45.932857.0TCB-warmBrArinos− 15.90− 46.055197.0
TEBrBom Jesus do Piaui− 9.10− 44.123317.6TCB-mesoBrGoias− 15.92− 50.135126.5
TEBrPaulo Afonso− 9.37− 38.222537.6TCB-warmBrPedra Azul− 16.00− 41.286486.3
TEBrPetrolina− 9.38− 40.483708.0TCB-warmBrCaceres− 16.05− 57.681185.5
EBrRio Branco− 9.97− 67.801604.6TCB-warmBrSalinas− 16.17− 42.304715.7
TCB-warmBrMonte Santo− 10.43− 39.304656.5TCB-warmBrUnai− 16.37− 46.554606.7
TCB-warmBrPorto Nacional− 10.72− 48.422396.7TCB-warmBrMontes Claros− 16.68− 43.836467.2
TCB-warmBrBarra− 11.08− 43.174018.4TCB-warmBrAracuai− 16.83− 42.052896.2
TCB-warmBrIrece− 11.30− 41.877478.1TCB-warmBrParacatu− 17.23− 46.877126.4
TCB-warmBrPeixe− 12.02− 48.532426.9TCB-warmBrPirapora− 17.35− 44.925057.5
EBrGleba Celeste− 12.20− 56.504155.7TCB–warmBrJoao Pinheiro− 17.70− 46.177607.8
TCB-warmBrTaguatinga− 12.40− 46.436036.7TCB-mesoBrIpameri− 17.72− 48.177726.9
TCB-warmBrItaberaba− 12.52− 40.282506.0TCB-mesoBrRio Verde− 17.80− 50.927746.2
TCB-warmBrLencois− 12.57− 41.384385.5TCB-warmBrItamarandiba− 17.85− 42.8510975.7
TCB-warmBrBom Jesus da Lapa− 13.27− 43.424408.2TCB-warmBrTeofilo Otoni− 17.85− 41.503565.6
TCB-warmBrItiruçu− 13.35− 40.127555.2TCB-mesoBrJatai− 17.88− 51.726626.2
EBoIxiamas− 13.76− 68.135005.5TCB-mesoBrCatalao− 18.18− 47.958406.9
TCB-warmBrCaetite− 14.07− 42.488827.3TCB-mesoBrDiamantina− 18.25− 43.6012966.3
TCB-warmBrPosse− 14.10− 46.378256.8TCB-mesoBrPatos de Minas− 18.52− 46.439406.8
TCB-mesoBrCapinopolis− 18.72− 49.556207.4TCB-mesoBrPresidente Prudente− 22.12− 51.384357.1
TCB-mesoBrG. Valadares− 18.85− 41.931485.5TCB-mesoBrIvinhema− 22.32− 53.933697.0
TCB-mesoBrC. Do Mato Dentro− 19.02− 43.436525.5TCB-mesoBrResende− 22.45− 44.434405.3
TCB-mesoBrPompeu− 19.22− 45.006906.6TCB-mesoPaPedro Juan Caballero− 22.58− 55.736527.2
TCB-mesoBrSete Lagoas− 19.47− 44.257327.4TCB-mesoBrCampos Do Jordao− 22.75− 45.6016424.8
TCB-mesoBrAimores− 19.48− 41.07826.7TCB-mesoBrEcologia Agricola− 22.80− 43.68336.0
TCB-mesoBrUsiminas− 19.48− 42.532985.0TCB-mesoBrTaubate− 23.03− 45.555775.5
TCB-mesoBrAraxa− 19.60− 46.9310236.6TCB-mesoBrJacarezinho− 23.15− 49.974706.9
TCB-mesoBrParanaiba− 19.70− 51.1847.3TCB-mesoBrLondrina− 23.32− 51.135666.9
TCB-mesoBrUberaba− 19.73− 47.957377.5TCB-mesoBrMaringa− 23.40− 51.925426.9
TCB-mesoBrCaratinga− 19.80− 42.156096.1TCB-mesoPaConcepción− 23.41− 57.3746.9
TCB-mesoBrJoao Monlevade− 19.83− 43.128595.4TCB-mesoBrUbatuba− 23.45− 45.0784.6
TCB-mesoBrFlorestal− 19.87− 44.427486.4TCB-mesoBrSao Paulo− 23.50− 46.627925.4
TCB-mesoBrBelo Horizonte− 19.93− 43.939156.9TCB-mesoBrSantos− 23.93− 46.33134.3
TCB-mesoBrSanta teresa− 19.93− 40.586485.2TCB-mesoBrItapeva− 23.95− 48.887076.2
TCB-mesoBrBambui− 20.03− 46.006616.0TCB-mesoBrCampo Mourao− 24.05− 52.376166.5
TCB-mesoBrVenda Nova− 20.38− 41.187105.2TCB-mesoPaSalto del Guairá− 24.05− 54.312656.7
TCB-mesoBrVotuporanga− 20.42− 49.985027.3TCB-mesoBrGuaira− 24.08− 54.252306.4
TCB-mesoBrCampo Grande− 20.43− 54.725307.1TCB-mesoPaSan Estanislao− 24.66− 56.431926.5
TCB-mesoBrCaparaó− 20.52− 41.908436.2TCB-mesoBrIguape− 24.72− 47.552.664.1
TCB-mesoBrFranca− 20.55− 47.4310266.8HTBrCastro− 24.78− 50.0010095.0
TCB-mesoBrViçosa− 20.75− 42.856896.0TCB-mesoPaAsunción− 25.25− 57.511017.1
TCB-mesoBrItaperuna− 21.20− 41.901236.3HTBrCuritiba− 25.43− 49.279245.4
TCB-mesoBrBarbacena− 21.25− 43.7711265.4HTBrIrati− 25.47− 50.638375.3
TCB-mesoBrSao Joao del Rei− 21.30− 44.279915.7TCB-mesoPaCiudad del Este− 25.53− 54.61966.7
TCB-mesoBrSao Simao− 21.48− 47.556176.9TCB-mesoBrParanagua− 25.53− 48.524.54.2
TCB-mesoBrCoronel Pacheco− 21.58− 43.254355.6TCB-mesoPaVillarrica− 25.76− 56.431617.0
TCB-mesoBrMachado− 21.67− 45.928735.6TCB-mesoPaCaazapá− 26.18− 56.361406.7
TCB-mesoBrCampos− 21.75− 41.33116.2TCB-mesoPaSan Juan Bautista− 26.66− 57.151266.8
TCB-mesoBrLavras− 21.75− 45.009186.7HTBrChapeco− 27.12− 52.626796.5
TCB-mesoBrJuiz de fora− 21.77− 43.359395.0HTBrIrai− 27.18− 53.232476.3
TCB-mesoBrSao Carlos− 21.97− 47.878566.7HTPaCapitan Miranda− 27.28− 55.832236.7
TCB-mesoBrCordeiro− 22.02− 42.355055.6HTPaEncarnación− 27.33− 55.83916.8
TCB-mesoBrSao Lourenco− 22.10− 45.029536.4HTBrCampos Novos− 27.38− 51.209476.3
HTArCorrientes− 27.45− 58.76627.4HTUrPaysandú− 32.35− 58.04617.2
HTArResistencia− 27.45− 59.05525.8HTUrMelo− 32.37− 54.191006.8
HTBrLages− 27.82− 50.339375.7HTArMarcos Juarez− 32.70− 62.151156.9
HTBrPasso Fundo− 28.22− 52.406846.5HTUrPaso de los Toros− 32.80− 56.53757.0
HTBrSão Luiz Gonzaga− 28.40− 55.022456.3HTArRosario− 32.92− 60.78257.3
HTBrCruz Alta− 28.63− 53.604726.7HTUrTreinta y Tres− 33.22− 54.39466.4
HTBrBom Jesus− 28.67− 50.4310485.5HTUrMercedes− 33.25− 58.07177.0
HTBrCaxias do Sul− 29.17− 51.207595.8HTBrS. Vitória do Palmar− 33.52− 53.35246.5
HTArReconquista− 29.18− 59.70537.1HTArSan Pedro− 33.68− 59.68287.2
HTBrTorres− 29.35− 49.7256.0HTArLaboulaye− 34.13− 63.371377.3
HTBrSanta maría− 29.70− 53.70956.0HTUrColonia− 34.46− 57.84237.0
HTBrUruguaiana− 29.75− 57.08626.9HTUrRocha− 34.49− 54.31186.5
HTArCeres− 29.88− 61.95887.2HTArJunin− 34.55− 60.92816.9
HTBrPorto alegre− 30.05− 51.17476.1HTArSan Miguel− 34.55− 58.73267.2
HTArMonte Caseros− 30.26− 57.65547.2HTArBuenos Aires− 34.58− 58.4866.7
HTUrArtigas− 30.40− 56.511216.9HTUrCarrasco− 34.83− 56.01336.6
HTBrEncruzilhada do Sul− 30.53− 52.524275.9HTArPehuajó− 35.87− 61.90877.0
HTUrRivera− 30.90− 55.542426.6HTArAnguil− 36.50− 63.981657.0
HTArRafaela− 31.18− 61.551007.4HTArSanta Rosa− 36.57− 64.271917.0
HTBrBagé− 31.33− 54.102426.1HTArCoronel Suarez− 37.43− 61.882336.9
HTUrSalto− 31.43− 57.98446.9HTArPigüé− 37.60− 62.383047.0
HTArParaná− 31.78− 60.48787.4HTArTres Arroyos− 38.00− 60.251156.7
HTBrRio Grande− 32.03− 52.1036.3       

4. Results and discussion

The possible existence of trends in sunshine duration was evaluated for five climatic regions in South America. For this purpose, in each of the five regions, the monthly mean values were interpolated from their respective stations and the regional mean of the annual and seasonal sums were determined for the periods 1961–2004 and the sub-periods 1961–1990 and 1991–2004.

The error of estimation for the interpolated values must be taken into account in order to evaluate the variations obtained in each region for each period. Basing on a previous study of sunshine spatial variability in the Humid Temperate Region (Raichijk et al., 2006), the interpolated values were estimated with a mean relative error ranging between a minimum value of 7.7% (measurement error) and a maximum value which depends on the density of stations. The number of stations with available data varies from month to month for each region. The extreme cases correspond to the months of August and September 1991 in the Equatorial Region, which had only five stations and, therefore, a mean coverage radius by station of approximately 800 km. This, according to Equation (1), implies interpolation errors of up to 23–24%.

The monthly mean values of shortwave radiative fluxes taken from NASA's SRB dataset were compared with corresponding ground-measured fluxes over the period 1992–2005 from stations of the Baseline Surface Radiation Network (BSRN). For all stations and years, the mean bias was determined to be − 3.1 W/m2, and the root mean square difference, 20.5 W/m2 (http://eosweb.larc.nasa.gov/PRODOCS/srb/readme/readme_srb_rel3.0_shortwave_monthly.txt). When the comparisons were made with ground measured values from different stations of Argentina located in the Humid Temperate Region (Raichijk, 2009), it was found that the relative root mean square differences varied between 5 and 18%, depending on the degree of climatic homogeneity of the area covered by the respective pixels.

The results observed for each studied climatic region are detailed below (the variation obtained for a given period is expressed in cursive writing if it is less than the estimation error of the parameter).

Equatorial Region: In this region, the largest variations and more pronounced change of trends in sunshine duration was observed. During the 1970–1990 period the relative variations in the annual sums was − 15.8 ± 3.4% with a maximum in summer of − 19.9 ± 4.2% and minimum in winter of − 12.7 ± 3.4%. For the 1991–2004 period, a change in sign of the trends was detected, with a positive variation in the annual sums of 19.3 ± 2.6% and similar behaviour in different seasons. Consistent with these results, for the 1984–2005 period, the SWAll fluxes showed positive significant trends with variations in the annual and seasonal means varying between 4.4 and 5.8% and the CF, negative trends with variations between − 8.2 and − 9.7%. The SWClear fluxes only showed a negative significant trend in summer with a small variation equal to − 1.9%.

Tropical Equatorial Region: Significant trends for the 1991–2004 period were found for annual and spring-autumn sums of sunshine, with positive variations equal to 6.2 ± 2% and 12.6 ± 3.8%, respectively. The SWAll fluxes and CF also showed significant trends for the annual and spring-autumn means during the 1984–2005 period, with positive variations of 4.2 and 4.7% in the SWAll fluxes and negative variations of − 11.7 and − 14.6% in the CF. For the SWClear fluxes, increasing trends for the annual and winter means were detected, with variations of 2.1 and 4.8%, respectively.

Tropical Central Brazil—Warm Region: Decreasing trends for the sunshine were determined during the 1961–1990 period for the annual sums, with a variation of − 12.8 ± 4.3%, and for winter sums, − 11.7 ± 3.2%. A positive trend was found for 1991–2004 only for spring-autumn sums, with a variation of 11.8 ± 5.4%. The SWAll fluxes for the 1984–2005 period showed positive trends with variations of 2.5% in the annual means, 3.3% in winter, and 3.9% in spring-autumn. Equals variations in the SWClear fluxes are observed. The CF showed negative trends with variations of − 7.8 and − 14.7% in annual and spring-autumn means, respectively.

Tropical Central Brazil—Mesothermal Region: A change in the sign of trends for annual sums of sunshine was observed, with a negative variation of − 9.8 ± 2.7% for 1961–1990 period and a positive variation of 10 ± 3.2% for 1991–2004. Variations only were detected in winter for the first period, − 12.8 ± 2.4%, and in spring-autumn for the next period, 20.3 ± 4.8%. The SWAll fluxes during the 1984–2005 period showed positive significant trends for annual, winter, and spring-autumn means, but with values of the Pearson linear correlation coefficient lower than 0.5 in all cases. The SWClear fluxes and CF showed almost the same variations in sign and magnitude, as for the Tropical Central Brazil—Warm Region.

Humid Temperate Region: A significant trend was observed only for the 1991–2004 period for the annual sums of sunshine, with a variation of 5.6 ± 2.8%. From the study of the time series of monthly mean values of sunshine for the 1956–2000 period in 28 stations of Argentina located in this region, Grossi Gallegos and Spreafichi (2006) found significant trends in only 13 of them, with negative variations in most cases below the measurement error. Also, significant trends could not be observed for 11 stations of Uruguay for the 1986–2004 period (Grossi Gallegos and Spreafichi, 2007). The SWAll fluxes during the 1984–2005 period did not show significant trends, the SWClear only for the annual means, with a small variation of 1.7% and the CF in summer with a negative variation of − 8.2%.

5. Conclusions

The results found for all the regions studied prove that whenever significant trends were found, these were in agreement with those observed in other regions of the planet. That is, negative trends for regional means values of sunshine duration for periods up to the end of the 1980s, global dimming, and a change in sign of the trends observed from the 1990s onwards, global brightening. The temporal behaviour during 1984–2005 period of satellite-derived values of shortwave radiative fluxes, for all and clear skies, and cloud fraction from the SRB dataset are consistent in all climatic regions, with the results obtained for the sunshine duration, i. e. increasing trends in sunshine are linked to increases in the solar radiative fluxes and these increases, to declines in cloudiness.


We thank the directors of the following organisations for their invaluable contribution to this work: Servicio Meteorológico Nacional of Argentina (SMN), Instituto Nacional de Meteorología of Brazil (INMET), Instituto Nacional de Tecnología y Normalización of Paraguay (INTN), Servicios Nacionales de Meteorología e Hidrología of Peru and Bolivia (SENAMHI), and Dirección Nacional de Meteorología of Uruguay (DNM).