New Zealand dimming and brightening

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 114,, doi:10.1029/2008jd011401, 2009 New Zealand dimming and brightening J. B. Liley 1 Received 31 October 2008; revised 18 February 2009; accepted 2 March 2009; published 28 April 2009. [1] Recent papers in the international literature on global dimming found a mixed pattern in data from New Zealand. Closer analysis of pyranometer data from four long-term sites shows a downward trend up to 1990, with a reversal at three of them after that time. How much can be attributed to the direct aerosol effect is uncertain from the pyranometer data, but aerosol optical depth data from Lauder show too little aerosol for this to be a substantial component. A comparison with much longer records of sunshine hours shows that there was a trend of increasing cloudiness to around 1990, and a decline since then, consistent with the global pattern. Citation: Liley, J. B. (2009), New Zealand dimming and brightening, J. Geophys. Res., 114,, doi:10.1029/2008jd011401. 1. Introduction [2] Several studies, mostly based on the Global Energy Balance Archive (GEBA), have shown that solar irradiance at Earth s surface declined from 1960 to 1990 [Liepert, 2002; Stanhill and Cohen, 2001; Gilgen et al., 1998]. The trend was especially strong in the Americas, parts of the USSR, Israel, sub-saharan Africa, and some of Asia. The global average trend was about 3 Wm 2 per decade, for a total of 6 to 9 Wm 2 up to 1990. Clear sky data for the United States show a decline in 2 decades of 8 W m 2, more than 1/3 of the decline for all sky conditions [Liepert, 2002], suggesting that part of the trend can be attributed to increasing aerosol optical depth (AOD), with the rest due to increased cloudiness, some of which may be an indirect aerosol effect. Attribution of dimming to aerosol change suggests anthropogenic origins, and certainly this view is supported by Alpert et al. [2005], whose work seems to have strongly influenced the conclusions about global dimming in chapter 3 of the IPCC Fourth Assessment Report [Trenberth et al., 2007]. [3] Regardless of its origins, the decline in solar surface irradiance has been shown [Roderick and Farquhar, 2002] to be consistent with decreased pan evaporation, and reduced diurnal temperature range (DTR). [4] Australia and New Zealand show both increases and decreases in GEBA data, but only two New Zealand sites were used by Liepert [2002]. New Zealand data are consistent with the global trends in pan evaporation [Roderick and Farquhar, 2005] and DTR [Salinger and Griffiths, 2001]. [5] Subsequent analysis, based on high-quality data series such as those from BSRN, shows a reversal around 1990, with increasing radiation thereafter at a majority of sites [Wild et al., 2005]. High precision and reproducibility in 1 National Institute of Water and Atmospheric Research, Lauder, New Zealand. Copyright 2009 by the American Geophysical Union. 0148-0227/09/2008JD011401 measurements is needed to determine any trends from these much shorter series. 2. New Zealand Irradiance [6] There are few New Zealand climate stations with pyranometer data before 1970. The four with the earliest data are at Auckland (37.0 S, 174.8 E, WMO 931190, Nov 1969), Wellington (41.4 S, 174.9 E, WMO 934370, Jan 1954), Christchurch (43.5 S, 172.5 E, WMO 937800, Jan 1960), and Invercargill (46.4 S, 168.3 E, WMO 938440, Jan 1954). Monthly means of solar radiation at these sites are shown in Figure 1, expressed as a fraction of model clear-sky values. [7] The model here is a functional fit, as described later, to hourly values in the latter part of the record that show a smooth diurnal curve of near maximum height for the time of year. These fits give a clear-sky total for each day, and thence for each month, of an average year. It is used just to remove the seasonal cycle in solar elevation, day length, Earth-Sun distance, and average aerosol. For comparison with other sites and studies, note that the annual average hourly downwelling short-wave fluxes for the four sites here are: Auckland, 172 W m 2 ; Wellington, 161 W m 2, Christchurch, 156 W m 2 ; Invercargill, 142 W m 2. Figure 1 also shows trend lines fitted to the data for the periods before and after 1990. These trends seem to concur with the predominant global trends; declining irradiance for the earlier period, and an increase in global irradiance since 1990. [8] The data in Figure 1 need to be treated with caution, as there is uncertainty about the stability of pyranometer calibration over decades. Instruments in the National Institute of Water and Atmospheric Research (NIWA) climate network, which now includes over 100 pyranometers, are cycled between stations every 1 2 years as they are removed for calibration and redeployed, but for much of the period used here calibration was much less frequent. The measurements at these sites were with Eppley 10-J bulb-type pyranometers up to the early 1970s, Eppley 8 48a BWP instruments for much of that decade, and Eppley PSP instruments thereafter. After 2000, some data are from Licor LI200 silicon sensors. 1of9

Figure 1. Monthly mean irradiance as a fraction of clear-sky model values, for four long-term New Zealand records. Gaps in the data in recent times have been filled with data from adjacent stations, provided the overlap period of the two stations shows negligible differences. [9] Other climate stations in the network have similar issues, and shorter records. The difficulty in maintaining stable calibration and obtaining reliable solar radiation data was the motivation for the Baseline Surface Radiation Network (BSRN), established from 1992 at the initial sites. The only New Zealand BSRN site was established at Lauder (45.0 S, 169.7 E, WMO 938510) in 1999, and that data record is as yet too short to be of use for this study. [10] Linear trends as calculated and shown in Figure 1 seem to represent the data reasonably well, but it is important to test the robustness of the conclusions against alternative analyses. The data from before 1969 are derived from daily totals of irradiance, recorded manually. Subsequent automation of data collection means that hourly data were recorded after that time. Both data sets were examined for further insight. [11] The hourly data can be used to distinguish days that are nearly cloud-free, and not surprisingly these correspond to the envelope of highest daily totals. The ratio of measurements to model values for clear days also provides a 2of9

Table 1. Trends in Daily Irradiance as a Fraction of Clear-Sky Model for Four New Zealand Sites a All Days Brightest 10% of Days Pre-1990 Post-1990 Pre-1990 Post-1990 Auckland 6.7 ± 0.6 0.8 ± 0.6 5.6 ± 0.2 0.9 ± 0.2 Wellington 1.2 ± 0.3 1.8 ± 0.6 1.2 ± 0.1 0.5 ± 0.2 Christchurch 2.0 ± 0.3 0.1 ± 0.6 2.4 ± 0.1 0.6 ± 0.2 Invercargill 3.0 ± 0.3 0.5 ± 0.6 2.4 ± 0.1 0.7 ± 0.3 a Unit is %/decade. measure of the intercomparability of different instruments, provided the dates of changes are known and aerosol effects are small or slowly varying. [12] There are, in some instances, discernible steps in the ratios for clear-sky data at times of instrument change. The steps are small and more-or-less uniformly distributed in time, so they do not significantly affect the trends, but they do increase the residuals and thus the uncertainties. One period in the Christchurch data, with a large downward step (25%) in May 1987 and equal upward step in August 1988, was rescaled to give a smooth profile in the clear-sky measured:model ratio. Because of its proximity to the 1990 knot in the spline fits, adjusting these data reduced both the downward and subsequent upward trends for Christchurch, which were otherwise similar to the Invercargill values. [13] Clear-sky comparisons of daily values also show that the seasonal cycle, relative to the model, differs somewhat between instruments. Again, this does not change the trends, but increases the residuals. [14] Table 1 shows the trends calculated for daily total radiation divided by model clear-sky values. The first two numeric columns are for all data, as in Figure 1, with small differences between the trends arising from the different order of averaging and dividing by model values. The calculated uncertainty for daily values is smaller because of the much larger number of data points fitted, albeit with extra sources of noise. In both analyses the uncertainties are only those from statistical fitting with the usual assumption of independently identically distributed additive random error. Any bias over extended periods, as may arise from the calibration cycle, can skew the trends with little evidence in the computed uncertainties. [15] Possible causes of the irradiance trends, if they are not artifacts of measurement, include the direct effect of aerosols, and changes in the prevalence or optical properties of clouds. To distinguish these effects, we can consider the trend for a subset of days with highest recorded radiation. Such analysis, for the 10% with highest irradiance fraction in each year, is shown in the third and fourth numeric columns of Table 1. The choice of 10% within each year gives similar results to selecting nearly clear-sky days from hourly data, but is applicable also to the early part of the record for which only daily totals were recorded. [16] These numbers would suggest that most of the dimming up to 1990 is caused by increasing aerosol, or by calibration trends. After 1990, the trend at all sites except Wellington is effectively the same for all days as it is for the brightest, again suggesting that the trends are in aerosol or calibration. At Wellington, the large difference between trends after 1990 suggests a 2.3%/decade decrease in the effect of cloud on downwelling short-wave irradiance. Table 2. Trends in Daily Irradiance a All Days Brightest 10% of Days Pre-1990 Post-1990 Pre-1990 Post-1990 Auckland 6.4 ± 0.6 0.4 ± 0.6 2.3 ± 0.2 0.2 ± 0.2 Wellington 1.0 ± 0.3 1.3 ± 0.7 0.6 ± 0.1 0.6 ± 0.2 Christchurch 1.9 ± 0.4 0.6 ± 0.6 1.1 ± 0.1 0.5 ± 0.2 Invercargill 2.6 ± 0.3 0.1 ± 0.6 1.0 ± 0.1 0.0 ± 0.2 a Unit is %/decade. [17] Again we must be wary of how trends are calculated. Dividing by model values gives more weight to any trends in winter measurements, if they differ from those for summer. An alternative is to fit both decadal trend and seasonal cycle simultaneously, by multilinear regression. Using three Fourier terms (annual, semiannual, and 4-monthly cycles) as predictors gives trends as in Table 2, for all days, and for the brightest 10%. [18] This analysis gives similar results for the full data set, though the trends are smaller, and the Christchurch post-1990 trend is reversed. In contrast, the trends for the brightest days change markedly from those calculated in Table 1 for irradiance fraction. The numbers in Table 2 should be more relevant to any changes in energy balance, as they treat summer and winter irradiance equally. [19] The difference between trends for all days and for the brightest days is a measure of changes in cloud prevalence or optical properties, resulting from changes in the water cycle or perhaps from indirect aerosol effects. These differences are shown in Table 3, where they are also related to the associated change in downwelling short-wave irradiance averaged over 24 h. [20] Though trends are not uniform, nor as pronounced as Figure 1 suggests, they are still broadly consistent with global dimming and brightening observed elsewhere. It is also important to note that their magnitude, in terms of radiative energy, is large relative to other components like changes in greenhouse gases. [21] Restricting the data to just clear skies gives the component due to aerosol extinction or to a drift in calibration of the instruments, but cannot distinguish the two. Other studies [e.g., Liepert, 2002] have considered the calibration of instruments over decades to be sufficiently trustworthy that the clear-sky changes can be attributed to aerosols, especially as there has been substantial decline and subsequent improvement in air quality throughout much of the northern hemisphere. For New Zealand, the magnitude of the change seems unlikely. 3. Effect of Pinatubo [22] In June 1991, the eruption of Mount Pinatubo in the Philippines injected about 20 Mt of SO 2 into the stratosphere, generating around 50 Mt of sulfate aerosol that caused a major perturbation to the global climate for the next few Table 3. Irradiance Trends Attributable to Cloud Percent per Decade W m 2 per Decade Pre-1990 Post-1990 Pre-1990 Post-1990 Auckland 4.1 ± 0.6 0.6 ± 0.7 6.8 ± 1.0 1.0 ± 1.1 Wellington 0.4 ± 0.3 1.9 ± 0.7 0.6 ± 0.5 3.1 ± 1.1 Christchurch 0.7 ± 0.4 0.1 ± 0.7 1.1 ± 0.6 0.1 ± 1.0 Invercargill 1.6 ± 0.3 0.1 ± 0.7 2.2 ± 0.4 0.1 ± 0.9 3of9

Table 4. Trends in Daily Irradiance, Excluding Pinatubo Years a All Days Brightest 10% of Days Pre-1990 Post-1990 Pre-1990 Post-1990 Auckland 5.4 ± 0.6 0.4 ± 0.6 2.1 ± 0.2 0.4 ± 0.2 Wellington 0.6 ± 0.3 0.8 ± 0.7 0.5 ± 0.1 0.8 ± 0.2 Christchurch 1.6 ± 0.4 0.8 ± 0.6 1.1 ± 0.1 0.5 ± 0.2 Invercargill 2.6 ± 0.3 0.2 ± 0.6 0.9 ± 0.1 0.1 ± 0.2 a Unit is %/decade. years, with a peak radiative forcing (at the tropopause) in 1992 of 3 to 4 Wm 2 [Intergovernmental Panel on Climate Change, 1996]. As this approximates the time when dimming changed to brightening, it may have had an effect on the trends. [23] The use of a first-order spline fit, continuous but not differentiable, means that any change at the knot, the point where the two lines intersect, will affect the calculated slopes of both lines. Masking the data from July 1991 to June 1993 (inclusive) does indeed change the fitted trends, as shown in Table 4, which should be compared with Table 2. The change is greatest for Auckland, with its shorter record, but also more northerly location. Pinatubo aerosol was especially concentrated in a tropical reservoir, spreading out from the equator, mostly to the winter hemisphere, over the next 2 to 3 years, so its effect may have been greater for Auckland. [24] Although the aerosol cloud from Pinatubo can be used to account for some of the dimming and brightening trends, even this does not seem to be a simple direct aerosol effect; its effect is much smaller for the brightest days. Separate measurements of global (G), diffuse (F) and direct (R) radiation have been made at Kaitaia (35.1 S, 173.3 E, WMO 930120), Paraparaumu ( 40.9 S, 175.0 E, WMO 934200), and Invercargill since 1987, and at the Lauder BSRN site since 1999. The instruments at the first three sites are prone to various problems, mostly affecting the alignment, so only those data that pass several checks for range and consistency are considered valid. The Paraparaumu and Invercargill data sets are more complete, and both correspond to sites used above (Paraparaumu is 46 km from Wellington). Hourly data for G and R at both sites, over the period of Pinatubo influence, are considered here. [25] To remove seasonal variation, both G and R are modeled as constants times the cosine of solar zenith angle raised to a power, G ¼ G 0 ðcos ZÞ g R ¼ R 0 ðcos ZÞ r ð1þ : The fit for G on a cloudless day is very close for all hours of the day, but R falls off more sharply at the ends of the day which are not of concern to the present analysis, so the fit was only applied for the period when R > 700 W m 2. Fits as above gave a set of values for both g and r; these were averaged to find fixed values of g = 1.25 and r =0.3 for a second series of fits that allow comparability of G 0 and R 0 over time. Both values correspond to the (somewhat artificial) extrapolation of equations (1) to overhead Sun. Days that were nearly cloud-free were determined by a choice of criteria on the mean sum of squares for the fits, c 2 G < 0:01G 0 c 2 R < 0:02R 0: ð2þ [26] The results are shown in Figure 2, with asterisks denoting R values and circles for G. Many more clear days were found in the (more complete) Paraparaumu data set, and the Invercargill data unfortunately contained few winter data; the latter was a result of more frequent cloud, rather than the 700 W m 2 threshold for fitting R. [27] Figure 2 shows that direct radiation was reduced by about 15% at both sites in the 2 years after the eruption, but global radiation was minimally reduced. Much of the light scattered out of the direct beam still reached the surface as diffuse radiation. Although there was certainly some loss of downwelling solar radiation through backscatter to space by Pinatubo aerosol, this may have been compensated, in the measurement of G, by the angular ( cosine ) response of the instruments. For lower solar elevation, forward scattering can increase the measured irradiance by increasing radiation around the zenith, where the instrument response is greatest. [28] Whatever the reason, the observation that Pinatubo aerosol did not reduce global radiation under clear skies by more than a few percent suggests that some of the Pinatubo effect, as discerned above, was a result of changes in cloud brought about by the radiative and photochemical changes induced by the enhanced stratospheric aerosol. 4. Aerosol Optical Depth [29] A 5% reduction in global irradiance would require an increase of about 0.05 in purely absorptive AOD (across the visible spectrum, due to soot for example). For the predominant aerosols, which mostly scatter rather than absorb, energy removed from the direct beam, as measured by AOD, largely reappears as diffuse radiation in global irradiance. For aerosols with high single-scatter albedo, any increase in AOD would have to be much larger (e.g., 0.2) to account for a 5% reduction in irradiance. There may have been some change around Auckland, New Zealand s largest city, but the city straddles a narrow isthmus, and its air is generally well-flushed by coastal breezes. Certainly there has been no long-term AOD increase of this size for the country. Measurements at Lauder, for 15 months in 1996/ 1997 and again from 1999 to the present [Liley and Forgan, 2009], show monthly average AOD at 500 nm to range seasonally from 0.02 to 0.05 (Figure 3). These are well below the values needed to account for the measured decline in clear-sky irradiance, accumulated over 3 decades, of 4% to 10% at the four sites above. Those sites are coastal, with higher AOD than Lauder because of marine aerosol, but absorbing aerosol sufficient to cause such clear-sky dimming would be detectable across New Zealand as a whole. The Lauder AOD data, starting within a decade of the end of the dimming period, suggest that the clear-sky change is largely a calibration artifact. [30] The decadal trend in AOD, if indeed there is any, is only 0.5% (of its mean value, 0.035, at 500 nm), about 3 orders of magnitude less than would be visible in the pyranometer data. 5. Sunshine Hours [31] Although pyranometer measurements in New Zealand only started in 1954, at two stations, the New Zealand climate network includes an alternative and independent solar radi- 4of9

Figure 2. Relative effects of Pinatubo aerosol on global (circles) and direct (asterisks) radiation at (top) Paraparaumu and (bottom) Invercargill. Data are deseasonalized by scaling to overhead sun with a fitted function as described in the text. ation data set. This is the record of sunshine hours, at 207 sites around New Zealand and the South Pacific, starting at some sites from 1905, and providing over 62,000 monthly totals as shown in Figure 4. The monthly totals, typically 50 300, are plotted as a fraction of their potential maximum, the total daylight hours. [32] The measurements were all made with Campbell- Stokes recorders up until the late 1990s, when Kipp and Zonen sunshine duration recorders (CSD1) were progressively introduced in automated weather stations. [33] It appears from Figure 4 that the records of sunshine hours show the familiar trend of declining radiation until 1990, and increasing thereafter. This is suggested particularly by the decile lines (10th and 30th percentiles, median, 70th and 90th percentiles), but again caution is needed. The number of sites varies greatly over the decades, and if a larger number of cloudy sites were in operation between 1950 and 1990 that could affect the apparent trend. A good example is Campbell Island (52.5 S, 169 E), a meteorological station in the Southern Ocean where sunshine hours Figure 3. Aerosol optical depth (AOD) at 500 nm over Lauder, Central Otago. Dark line represents a model fit of annual cycle with decadal trend as shown. 5of9

Figure 4. Monthly mean sunshine hour fraction for all New Zealand sites in the NIWA Climate Database. Lines show deciles 1, 3, 5, 7, and 9 for 5-year intervals. were recorded from August 1941 until August 1995. The data with sunshine hour fraction less than 0.15 in Figure 4, which at first sight appear to be erroneous, are almost all from Campbell Island. [34] Thus, it is important to determine trends for each location, and then combine them as appropriate. Though measurements at a few sites started as early as 1905, many stations have been repositioned or replaced by a nearby station over the century. Linking all sites within 8 km of each other, and within 125 m altitude, gave groups with smooth transitions between the time series for each station in the group. To compare with global trends, and the above analyses of pyranometer data, a first-order spline fit is used again. This time it allows one linear trend up to 1950, another to 1990, and a third thereafter, with continuity at the joins. [35] Fitting the site groups with the largest data series gave results as shown in Table 5 for locations with 400 or more months of data (overlaps in several instances give more months of data than the total for the period). They show mostly negative trends to 1990, and positive trends thereafter, including some remarkably large values. [36] Some of the largest trends can be a result of short data series, as reflected in small N values in the table. Trends were not calculated for sites and periods with less than N i = 60 months (5 years) of data. Instrument or procedural changes may also be a factor, especially as the move from paper recorders to automated sensors occurs during the post-1990 period. Nevertheless, there were no consistent differences between the data from the two sensors, and omitting all Electronic Weather Sensor (EWS) sites from the record does not substantially change the results, except to the extent that consequently shorter records give greater uncertainties in the post-1990 trends. [37] The main New Zealand landmass is narrow from west to east, though it extends over 12 degrees of longitude. In latitude it ranges from 34 to 47 degrees, from the subtropics to the edge of the Southern Ocean, and its weather and climate are strongly affected by both the El Niño Southern Oscillation (ENSO) and the Interdecadal Pacific Oscillation (IPO). If major changes in cloud cover over decades are connected to such cycles, they might be expected to show some geographical pattern. This question is addressed in Figure 5, which shows the trends in sunshine hour fraction for the 1950 1990 and post-1990 periods for those site groups with valid data out of the top 100 by total length of record. This gives 86 locations for the earlier period, and 54 for the later. [38] Figure 5 shows no consistent geographical pattern in trends, such as a latitudinal or longitudinal gradient. A cluster of sites in the southwest of the South Island show small positive trends in the 1950 1990 period, when most areas have a negative trend, but the cluster is surprising. It includes a wide range of climates, as exemplified by annual precipitation, which varies from 6 m in the rain forest of western Fiordland, up to 12 m on the Southern Alps, and down to less than 0.5 m (500 mm) in South Canterbury and Central Otago east of the Alps. It is possible that a change in strength or frequency of the prevailing westerly winds could have caused this spatially consistent trend in cloud cover, but differences in the predominant cloud types over the three climate regions [Uddstrom et al., 2001] make the connection seem less likely. Some effect of altitude is also plausible, as the interiors of both the North and South Islands extend to over 2500 m in altitude, but there are no sunshine-hour records from above 1100 m. The 17 sites over 300 m in altitude show no significant difference in average trends from the sites at lower altitude. [39] The mean and its standard error for the trends at the locations and in the periods of Figure 5 are shown in Table 6. The trends are also expressed in W m 2, using a regression of irradiance fraction on sunshine hour fraction for 22 sites with enough simultaneous data (cf. A. Tait and B. Liley, Interpolation of daily solar radiation for New Zealand using satellite data-derived cloud cover surface, submitted to Weather and Climate, 2009). [40] These numbers are consistent in sign with GEBAand BSRN-derived global trends as described earlier. Within 6of9

Table 5. Trends in Sunshine Hour Fraction of Daylight Hours for New Zealand Sites a Location Latitude Longitude N 1 Pre-1950 N 2 1950 1990 N 3 Post-1990 Ruakura, Hamilton 37.8 175.3 201 3.8 ± 2.2 893 0.2 ± 0.5 159 1.1 ± 1.5 Wellington 41.3 174.8 515 0.6 ± 0.5 479 0.3 ± 0.5 222 4.9 ± 1.6 Nelson 41.3 173.3 441 0.3 ± 0.5 500 1.1 ± 0.4 218 6.9 ± 1.3 Hokitika 42.7 171.0 446 0.8 ± 0.7 480 1.1 ± 0.6 223 7.9 ± 1.8 Invercargill 46.4 168.4 353 1.0 ± 0.8 479 1.6 ± 0.6 279 10.1 ± 1.5 Auckland, Albert Park 36.9 174.8 483 2.3 ± 0.5 545 1.2 ± 0.5 11 Christchurch Gardens 43.5 172.6 251 0.2 ± 1.2 517 1.7 ± 0.5 222 1.1 ± 1.4 Gisborne Aero 38.7 178.0 281 1.6 ± 0.5 479 0.8 ± 0.5 221 3.4 ± 1.6 Palmerston North 40.4 175.6 238 2.1 ± 1.8 508 2.1 ± 0.7 223 6.4 ± 2.0 Lincoln 43.6 172.5 505 0.6 ± 0.5 454 0.2 ± 0.6 0 Rotorua, Whakarewarewa 38.2 176.3 239 4.9 ± 1.4 496 1.3 ± 0.5 219 4.7 ± 1.5 Queenstown 45.0 168.7 239 3.8 ± 1.9 480 0.0 ± 0.7 221 1.7 ± 2.1 Timaru Gardens 44.4 171.2 239 0.1 ± 1.5 479 0.3 ± 0.6 221 3.5 ± 1.6 Napier 39.5 176.9 238 3.4 ± 1.3 480 1.0 ± 0.5 212 5.7 ± 1.5 Mount Cook, The Hermitage 43.7 170.1 232 3.7 ± 1.8 478 0.5 ± 0.7 201 4.0 ± 2.0 Tauranga Aero 37.7 176.2 204 3.6 ± 1.5 480 0.9 ± 0.5 217 7.0 ± 1.4 Ashburton 43.9 171.7 215 4.6 ± 1.4 478 0.8 ± 0.5 200 3.8 ± 1.7 Dunedin Gardens 45.9 170.5 247 1.6 ± 0.8 478 3.4 ± 0.6 167 15.6 ± 1.7 Wallaceville 41.1 175.1 121 8.8 ± 4.2 497 0.2 ± 0.7 236 1.7 ± 1.8 Alexandra 45.3 169.4 239 4.5 ± 1.5 478 1.0 ± 0.6 71 7.6 ± 2.4 Kaitaia Aero 35.1 173.3 0 467 0.2 ± 0.6 310 0.1 ± 1.2 Hanmer Forest 42.5 172.9 239 0.1 ± 1.7 480 2.6 ± 0.7 57 Auckland, Mangere 37.0 174.8 0 564 1.8 ± 0.9 213 3.6 ± 1.6 Westport Aero 41.7 171.6 153 0.8 ± 3.1 479 3.0 ± 0.7 96 2.0 ± 6.1 Greymouth Aero 42.5 171.2 32 474 2.7 ± 0.8 221 12.9 ± 2.1 Waingawa 41.0 175.6 239 5.4 ± 1.7 474 1.6 ± 0.8 14 Taumarunui 38.9 175.3 32 479 2.6 ± 0.9 208 5.6 ± 2.5 Waihi 37.4 175.8 239 0.3 ± 1.7 474 3.5 ± 0.7 0 Blenheim 41.5 174.0 273 0.3 ± 1.1 437 1.0 ± 0.6 0 Waimate 44.7 171.0 383 7.5 ± 0.9 315 3.0 ± 1.1 0 Lake Tekapo 44.0 170.4 249 7.1 ± 1.3 384 1.1 ± 0.8 60 5.9 ± 2.0 New Plymouth 39.1 174.1 398 3.4 ± 0.8 285 0.3 ± 1.2 0 Wanganui 39.9 175.0 151 8.1 ± 2.7 475 0.2 ± 0.6 56 Paraparaumu Aero 40.9 175.0 0 443 0.7 ± 0.7 222 3.2 ± 1.7 Waipukurau Aero 40.0 176.5 58 477 1.5 ± 0.7 65 0.5 ± 7.8 Levin 40.7 175.3 134 2.1 ± 1.2 415 2.6 ± 1.0 13 Thames 37.2 175.6 47 304 3.1 ± 0.9 203 10.7 ± 2.1 Gore 46.1 168.9 107 5.4 ± 4.3 444 1.9 ± 0.7 0 Dannevirke 40.2 176.1 0 315 0.7 ± 1.3 235 8.6 ± 1.9 Stratford Demo Farm 39.3 174.3 0 318 2.3 ± 1.2 220 2.8 ± 1.9 Balclutha, Finegand 46.3 169.7 0 290 4.2 ± 1.3 240 10.6 ± 1.7 Whakatane Mill 38.0 177.0 0 380 1.2 ± 0.8 146 1.7 ± 1.8 Taupo 38.7 176.1 0 472 1.5 ± 0.7 49 Dargaville 35.9 173.8 74 10.0 ± 8.5 285 2.8 ± 1.0 155 2.9 ± 2.0 Grassmere Salt Works 41.7 174.1 35 459 1.6 ± 0.6 0 Winton 46.2 168.3 0 310 2.6 ± 1.2 184 4.0 ± 1.8 Te Kuiti 38.3 175.2 0 332 5.5 ± 1.4 147 12.2 ± 2.9 Whangarei Hospital 35.7 174.3 0 425 1.9 ± 0.8 47 Palmerston 45.5 170.7 0 249 3.3 ± 1.6 222 3.6 ± 1.8 Motueka, Riwaka 41.1 173.0 0 294 2.5 ± 1.1 170 1.7 ± 1.5 Te Paki Station, Te Hapua 34.5 172.8 173 1.4 ± 2.4 287 3.3 ± 1.3 0 Ohakea Aero 40.2 175.4 0 431 0.3 ± 0.9 22 Vernon Lagoon 41.5 174.0 0 220 4.1 ± 1.3 221 4.2 ± 1.4 Whatawhata 37.8 175.1 0 424 0.8 ± 0.9 0 New Plymouth Aero 39.0 174.2 0 205 0.8 ± 2.0 219 3.8 ± 1.8 Omarama, Tara Hills 44.5 169.9 0 409 1.3 ± 0.8 0 Waipoua Forest 35.7 173.6 239 0.1 ± 2.3 170 5.9 ± 3.5 0 Foxton 40.5 175.3 0 347 2.1 ± 1.0 62 11.6 ± 3.4 Haast 43.9 169.0 83 3.1 ± 6.9 320 2.1 ± 1.3 0 Highbank Power Station 43.6 171.7 0 402 2.1 ± 0.9 0 a Unit is %/decade relative to 1990. N 1,N 2, and N 3 are the number of monthly means of daily sunshine in the three periods. They exceed the total number of months in the period in several instances where two or more adjacent instruments measured simultaneously for an extended period. the respective uncertainties, they also agree with the 1950 1990 trends in pyranometer data, but the mean trend in sunshine hours is much larger for the period since 1990 than is suggested by the pyranometer data. There are of course many (100) pyranometer sites with data from 1990 onward, but the uncertainty in calibration stability compromises their value for studies of this kind. Despite their simplicity, or perhaps because of it, the sunshine recorders of 19th century design still provide a valuable measurement record. As noted, omitting the EWS data has little effect; the trends in Table 6 change to 1.60 ± 0.38 for 1950 1990, and 5.86 ± 1.17%/decade for 1990 2008. Stanhill and Cohen [2005] discuss some interaction of sunshine records with other climate variables, but conclude that such data are 7of9

and largely anthropogenic. Although that study was clearly very influential in the conclusions about global dimming in chapter 3 of the IPCC Fourth Assessment Report [Trenberth et al., 2007], there is no mention there or elsewhere of the error in equation (3) of Alpert et al. [2005] for their cumulative mean slope, a n ¼ a n 1 þ a n N ; instead of the correct alternatives, in their notation, ð3þ Figure 5. Trends in sunshine hours, expressed as the decadal change in percentage of maximum possible hours. sufficiently consistent for trend analysis. Interestingly, the trends that they discern (to 1987) in the conterminous United States are neither as consistent nor as large as the trends in pyranometer data. The latter are some of the largest observed globally, and may well be largely anthropogenic. For New Zealand trends, this is unlikely. [41] Once again, it is important to note how large the trends are in short-wave radiant energy flux. Compared with numbers on the order of 1.7 W m 2 radiative forcing (at the tropopause) for the increase in atmospheric CO 2 from preindustrial levels, trends as above and in the global dimming and brightening literature are undeniably important. Of course, they are likely to be balanced in part by countervailing effects on long-wave outgoing flux, if cloudy days correlate with cloudy nights. This concurs with the reduced diurnal temperature range reported by Salinger and Griffiths [2001]. Thus they are not obviously inconsistent with global warming due to greenhouse gas increases [Wild et al., 2007]. What is clear, however, is that the known large radiative forcings of clouds are not randomly distributed in space or time, so their effects cannot just be assumed to cancel over regional scales or long time periods. Climate records show that this is an area that models need to address better. 6. Pacific Islands [42] The NIWA Climate Database also holds sunshine records for many islands in the South Pacific. Table 7 shows these data, and New Zealand s Raoul, Chatham, and Campbell Islands, analyzed as above. [43] The observed changes are very large for some islands, though mostly for short records, and the pattern is not as consistent as for the New Zealand landmass. They are shown here mainly for interest, and because the tropical location of many makes any secular change in cloud cover over the regions that they represent even more important for global energy balance. 7. Global or Local? [44] The New Zealand and Pacific Island data clearly bear directly on the issue addressed by Alpert et al. [2005]; whether the dimming phenomenon is global or just local a n ¼ 1 n X n i¼1 a i ¼ a n 1 þ a n a n 1 n : ð4þ It is not clear that this was only a typographical error, and did not mar their calculations, as it could produce the type of hysteresis shown in their Figure 2. Either way, a large anthropogenic effect in the vicinity of dense population does not negate the finding of long-term change in sparsely populated regions, and indeed Alpert et al. [2005] find a downward trend of 0.16 W m 2 a 1 that is comparable to the trends found elsewhere in the global dimming literature. [45] In contrast, Alpert and Kishcha [2008] conclude that there was no dimming over the 70% of global land area with human population density Pd < 10 km 2, because the trend (from 1964 to 1989) for 44 such sites was not significantly different from zero. As shown by their Figure 1, the lack of significance results from wide scatter rather than a lack of downward trend, so the more relevant question for their thesis is whether the trend is significantly different from that found for the 273 more populous sites. They do not test the hypothesis that there is no difference. [46] Whether or not population density affects the trends in radiation, it is clear from the analysis of Alpert and Kishcha [2008] (e.g., their Figure 1) that latitude-adjusted average radiation is lower in densely populated areas. Though it is suggestive of an anthropogenic aerosol effect, this relationship could also arise if cloudier areas, such as harbors, river deltas, and higher-rainfall regions, have supported the greatest population density. [47] The average human population density of New Zealand is now about 16 km 2, but in 1950 it was 8km 2, and much lower if averaged over its predominantly maritime region. With 1500 km to the nearest large landmass, any exogenous anthropogenic pollution or other aerosol (e.g., dust storms, volcanic eruptions, or natural fires) must also affect a large area of the Southern Ocean. From the pyranometer data analyzed here, it is unclear that there is any substantial contribution to New Zealand dimming or brightening from the direct aerosol effect, but there is certainly a significant effect from decadal changes in cloud prevalence or optical properties. Because it is consistent with average trends in many other parts of the globe, this dimming and subsequent brightening supports the conclusion that there has indeed been global change, on a Table 6. Mean Decadal Trends in Sunshine Hour Fraction for the Site Groups With the Longest Records Percent per Decade Period Number of Sites Mean ± SEM a SD Wm 2 per Decade 1950 1990 86 1.40 ± 0.34 3.13 1.93 ± 0.47 1990 2008 54 3.93 ± 0.69 5.09 5.42 ± 0.96 a Standard error of mean. 8of9

Table 7. Trends in Sunshine Hour Fraction of Daylight Hours for Pacific Island Sites a Location Latitude Longitude N 1 Pre-1950 N 2 1950 1990 N 3 Post-1990 Kiribati, Tarawa 1.4 172.9 0 144 7.1 ± 5.6 115 24.6 ± 6.6 Kiribati, Banaba 0.9 169.6 0 137 5.4 ± 6.5 80 27.8 ± 11.8 Tuvalu, Funafuti Aero 8.5 179.2 133 26.6 ± 5.5 150 9.1 ± 1.9 142 0.1 ± 3.4 Samoa, Apia 13.8 171.8 303 2.0 ± 1.4 463 5.1 ± 0.9 29 Vanuatu,Vila 17.8 168.3 0 119 3.7 ± 6.6 160 1.3 ± 5.7 Fiji, Nadi Aero 17.8 177.4 34 526 1.0 ± 0.7 26 Niue, Vaipapahi 19.0 169.9 0 60 9.0 ± 10.4 74 7.2 ± 8.3 Niue, Kaimiti 19.1 169.9 0 15 115 5.2 ± 5.3 Vanuatu, Burtonfield 19.5 169.2 0 79 13.5 ± 7.1 77 9.0 ± 7.4 Ci Mauke 20.1 157.3 0 73 19.5 ± 9.0 76 11.7 ± 8.9 Vanuatu, Aneityum 20.2 169.8 0 69 21.0 ± 8.3 68 1.3 ± 10.1 Tonga, Fua amotu Aero 21.2 175.1 0 118 4.8 ± 4.4 154 9.8 ± 3.4 Ci Rarotonga Aero 21.2 159.8 0 417 1.2 ± 0.9 130 10.7 ± 3.7 Ci Mangaia 21.9 158.0 0 71 12.4 ± 10.5 74 4.1 ± 10.5 Pitcairn Island 25.0 130.0 0 43 210 7.5 ± 3.0 Raoul Island 29.2 177.9 116 8.0 ± 3.7 473 1.4 ± 0.6 207 0.7 ± 1.6 Chatham Island 44.0 176.6 0 388 1.0 ± 1.2 50 Campbell Island 52.5 169.1 94 9.6 ± 14.2 469 3.5 ± 1.9 62 11.9 ± 24.2 a Unit is %/decade relative to 1990. time scale of decades, of sufficient magnitude to warrant its consideration in models of global radiative balance. [48] In this context, the Pacific Island data, especially in concert with temperature and radiation records from those sites, are worthy of more detailed scrutiny and analysis. 8. Conclusions [49] Analysis of New Zealand pyranometer data for the few sites with a long record shows general agreement with trends described in the global dimming and brightening literature. The performance and calibration stability of standard pyranometers is barely adequate for observing small changes in mean radiation over decades, and this makes it very difficult to reliably distinguish the direct effect of aerosols in clear sky data. The clear skies over New Zealand, even at a time when global dimming was most pronounced, suggest that observed changes in pyranometer data for clear skies are indeed a calibration artifact. [50] Nevertheless, 100 years of sunshine hour records tell a story of changing cloudiness on decadal time scales, over a region larger than that represented by New Zealand. Explaining, or better, predicting such changes is an important challenge for models of Earth s radiative balance. [51] Acknowledgments. All New Zealand data used herein are available from the NIWA Climate Database. The Pacific Island data are held on behalf of the respective meteorological authorities, whose permission is required for access. Their continued work to collect and archive the data is gratefully acknowledged. References Alpert, P., and P. Kishcha (2008), Quantification of the effect of urbanization on solar dimming, Geophys. Res. Lett., 35, L08801, doi:10.1029/ 2007GL033012. Alpert, P., P. Kishcha, Y. J. Kaufman, and R. Schwarzbard (2005), Global dimming or local dimming?: Effect of urbanization on sunlight availability, Geophys. Res. Lett., 32, L17802, doi:10.1029/2005gl023320. Gilgen, H., M. Wild, and A. Ohmura (1998), Means and trends of short wave irradiance at the surface estimated from Global Energy Balance Archive data, J. Clim., 11, 2042 2061. Intergovernmental Panel on Climate Change (1996), Climate Change 1995: The Science of Climate Change, 572 pp., Cambridge Univ. Press, Cambridge, U. K. Liepert, B. G. (2002), Observed reductions of surface solar radiation at sites in the United States and worldwide from 1961 to 1990, Geophys. Res. Lett., 29(10), 1421, doi:10.1029/2002gl014910. Liley, J. B., and B. F. Forgan (2009), Aerosol optical depth over Lauder, New Zealand, Geophys. Res. Lett., doi:10.1029/2008gl037141, in press. Roderick, M. L., and G. D. Farquhar (2002), The cause of decreased pan evaporation over the past 50 years, Science, 298, 1410 1411. Roderick, M. L., and G. D. Farquhar (2005), Changes in New Zealand pan evaporation since the 1970s, Int. J. Climatol., 25, 2031 2039, doi:10.1002/joc.1262. Salinger, M. J., and G. M. Griffiths (2001), Trends in New Zealand daily temperature and rainfall extremes, Int. J. Climatol., 21, 1437 1452, doi:10.1002/joc.694. Stanhill, G., and S. Cohen (2001), Global dimming: A review of the evidence for a widespread and significant reduction in global radiation with discussion of its probable causes and possible agricultural consequences, Agric. For. Meteorol., 107, 255 278, doi:10.1016/s0168-1923(00)00241-0. Stanhill, G., and S. Cohen (2005), Solar radiation changes in the United States during the twentieth century: Evidence from sunshine duration measurements, J. Clim., 18, 1503 1512, doi:10.1175/jcli3354.1. Trenberth, K. E., et al. (2007), Observations: Surface and atmospheric climate change, in Climate Change 2007: The Physical Science Basis, Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, edited by S. Solomon et al., pp. 235 336, Cambridge Univ. Press, Cambridge, U. K. Uddstrom, M. J., J. A. McGregor, W. R. Gray, and J. W. Kidson (2001), A high-resolution analysis of cloud amount and type over complex topography, J. Appl. Meteorol., 40(1), 16 33, doi:10.1175/1520-0450(2001)040<0016:ahraoc>2.0.co;2. Wild, M., H. Gilgen, A. Roesch, A. Ohmura, C. N. Long, E. G. Dutton, B. Forgan, A. Kallis, V. Russak, and A. Tsvetkov (2005), From dimming to brightening: Decadal changes in solar radiation at Earth s surface, Science, 308, 847 850, doi:10.1126/science.1103215. 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. J. B. Liley, National Institute of Water and Atmospheric Research, Private Bag 50061, Lauder, Central Otago 9352, New Zealand. (b.liley@niwa.co.nz) 9of9