Annual average total irradiance increases by 1–2% per decade at three monitoring stations in Oregon over the period from 1980 to 2007. Direct normal irradiance measurements increase by 5% per decade over the same time period. The measurements show no sign of a dimming before 1990. The impact of high concentrations of stratospheric aerosols following the volcanic eruptions of El Chichón and Mount Pinatubo are clearly seen in the measurements. Removing these years from the annual average all-sky time series reduces the trends in both total and direct normal irradiance. Clear-sky periods from this long direct normal time series are used in conjunction with radiative transfer calculations to test whether part of the increase could be caused by anthropogenic aerosols. All three sites show relatively low clear-sky measurements before the eruption of El Chichón in 1982, suggesting higher aerosol loads during this period. After removing the periods most strongly impacted by volcanic eruptions, two of the sites show statistically significant increases in clear-sky direct normal irradiance from 1987 to 2007. Radiative transfer calculations of the impact of volcanic aerosols and tropospheric water vapor indicate that only about 20% of that clear-sky increase between background aerosol periods before and after the eruption of Mount Pinatubo can be explained by these two factors. Thus a statistically significant clear-sky trend remains between 1987 and 2007 that is consistent with the hypothesis that at least some of the increase in surface irradiance could be caused by a reduction of anthropogenic aerosols.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Analyses of surface solar irradiance reported a decrease in total irradiance until around 1990 and a following increase at many monitoring stations worldwide [e.g., Wild et al., 2005, and references therein]. One of the hypothesized causes for the “dimming” and following “brightening” is changing concentrations of anthropogenic aerosols. Streets et al.  calculated global emissions of sulfur dioxide and black carbon between 1980 and 2000, and found that global emissions peaked in 1988–1989, and were at a minimum in 2000, consistent with the observed change from dimming to brightening.
 Determining how much of the change seen in surface solar irradiance is due to anthropogenic aerosols would help reach better conclusions about anthropogenic climate forcing. If the observed decrease in surface irradiance caused by anthropogenic aerosols is significant, then the global warming from increased greenhouse gases might be underestimated [e.g., Pittock, 2006].
 Total irradiance measurements are the primary measurement used in global dimming and brightening studies both because they are the most widely available measurement, and because the total surface solar flux is an important energy budget parameter. Measurements of direct normal irradiance are more useful than total irradiance for detecting changes in aerosol concentrations, however, because of their higher sensitivity to scattering and smaller measurement uncertainties. Scattering by aerosols is primarily in the forward direction. Thus, aerosol scattering has only a small diminishing effect on the total irradiance reaching the surface, but a larger effect on how much of the downwelling surface irradiance is direct or diffuse radiation. Changes in absorbing aerosols impact direct and total components of radiation equally, and so could be seen in either measurement component. Pyrheliometers used to measure direct normal irradiance have smaller percentage uncertainties than pyranometers measuring total irradiance because the accuracy of the direct normal measurements is impacted less by the direction of incoming radiation (cosine response) than the total irradiance measurements. But more importantly for trend studies, the calibration of pyrheliometers is more stable over time than pyranometers. This study uses a long time series of direct normal measurements in Oregon to examine possible changes due to aerosol extinction under clear skies.
2. About Data
 The University of Oregon Solar Radiation Monitoring Lab (UO SRML) has been operating measurement sites around the Northwestern United States since 1975. Three stations in Oregon have high-quality total and direct normal shortwave surface irradiance measurements from thermopile-based instruments starting in 1978–1979. The locations of these measurement stations are shown in Figure 1.
 The three measurements stations are located in different climate zones. The Eugene monitoring station is located at the end of the Willamette River Valley at an elevation of 150 m. Eugene has higher precipitation and cloudiness in the winter than the other stations. Two possible local anthropogenic aerosol sources in the area are field burning and paper mills. Forest fires affect irradiance measurements at both the Eugene and Burns areas. Hermiston is located along the Columbia River, also at a low altitude (180 m), but surrounded by a much drier climate. While the population of Hermiston is quite small (15,000), several possible sources of particulate matter exist in the region including a coal power plant, two natural gas burning steam and electricity cogeneration plants, a weapons incinerating facility, and various agricultural sources. Additionally, both the Hermiston and Burns measuring stations are located at agricultural research stations, possible sources of dust. Burns is located in eastern Oregon on a high desert plateau (altitude 1265 m). Dust from agriculture is likely the predominant local source of particulate matter at the Burns station.
 Total irradiance measurements are made with Eppley Precision Spectral Pyranometers (PSPs) and direct normal irradiance measurements with Eppley Normal Incidence Pyrheliometers (NIPs). The data are integrated over 5-min intervals and stored for retrieval. The data have been manually screened for errors. Instruments were calibrated in the field every 1–2 years as funding allowed.
 Special attention was given to calibration values used with the instruments in order to minimize the impact of changing calibration methodology and instrument degradation on the trends as described in a previous study [Riihimaki and Vignola, 2008]. Correcting for instrument degradation is particularly important with pyranometers. For example, the sensitivity of one pyranometer declined by 9% over 20 years of use in the field. One of the advantages of using direct normal irradiance in trend studies is the stability of the NIPs. Because calibration histories of the NIPs showed no detectable decline over 25–30 years within the 2% uncertainty of the instrument calibrations, only one calibration value was used for each NIP over the entirety of its use in the field. The trends in clear-sky direct normal irradiance are positive, indicating that the change is not caused by instrument calibration. Instrument degradation results in a loss of detector sensitivity that would produce a negative trend.
 Total irradiance measurements from PSPs at 5-min resolution have absolute uncertainties of 5% or larger. The absolute uncertainty of annual average total irradiance could be as low as 2–3%, when the zenith angle systematic errors are averaged out. When calculating trends, the relative uncertainty is more important than the absolute accuracy, however. The relative uncertainty of PSPs is on the order of 1.5–2%. Taking into account the relative uncertainty and the corrections for calibration methodology changes, changes in total irradiance of 2–3% or larger are likely detectable in the annual average total irradiance measurements.
 The NIPs have an uncertainty in the field of 2–3%, even for high-resolution 5-min measurements. Under optimal conditions uncertainties in NIP measurements for longer-term averages over monthly or annual time periods would reduce the percentage uncertainties of NIPs to less than 2%. However, under field conditions at these sites, including manual alignment adjustment for declination, a 2–3% absolute uncertainty for both monthly and annual averages is thought to be more realistic. The relative uncertainty of NIPs is 1–2%, however, and irradiance changes on that order or larger should be detectable in the direct normal irradiance time series in this study.
3. All-Sky Trends in Annual Averages
 Annual average total irradiance is plotted for each site in Figure 2. Monthly averages are calculated from 24-hour daily average irradiance. Annual averages are calculated as means of monthly averages. Eugene and Hermiston have total irradiance records with >98% of daily averages available. About 96% of daily averages are available at Burns. In order for a daily average to be calculated, each hour was required to contain at least one 5-min measurement.
 More daily averages of direct normal irradiance are missing than of total irradiance. However, most of the missing direct normal daily averages were filled in with correlations from the total irradiance using the method described by Vignola and McDaniels . The percentage of missing daily averages filled in with this method were 1.4%, 9.9%, and 8.2%, at Eugene, Burns, and Hermiston, respectively. Direct normal measurements from a rotating shadowband pyranometer are also used to fill 4.8% of missing daily averages at Hermiston. The remaining missing direct normal daily averages are <1.5% of all daily averages at each site. Annual averages begin in 1980 at Hermiston and Burns, and in 1976 in Eugene. Data are not available at Hermiston between 1990 and 1992 when the station was closed.
 Quadratic polynomial fits of annual average total irradiance (Figure 2) and direct normal irradiance (Figure 3) are plotted using least squares regression. The measurements from the first decade do not decline as seen in other studies [e.g., Dutton et al., 2006]. Instead, the irradiance exhibits a general tendency to increase over the entire time period for both total and direct normal irradiance.
Table 1 lists the slopes of linear regression fits to annual averages of both total and direct normal irradiance. The uncertainties given in Table 1 represent 1 standard error of the least squares regression slope coefficient including a correction for first-order autocorrelation. Following standard practice, the standard error of the slope is calculated as
where a and b are the regression coefficients, is the average value of x, and df is the number of degrees of freedom. Typically, df is equal to the number of measurements minus two. If measurements from consecutive time steps are autocorrelated, they are not independent, and the number of degrees of freedom is reduced. An adjustment was made to the degrees of freedom here by assuming first-order autocorrelation. Thus, df was multiplied by a factor of (1 − ϕ)/(1 + ϕ), where ϕ is the correlation coefficient between y values at consecutive time steps (from Halstead Harrison, personal communication, 2008, including an unpublished memo at http://www.atmos.washington.edu/∼harrison/reports/df.pdf).
Table 1. Slopes of Linear Trend Lines Fit to Annual Average All-Sky Irradiance Data at Three Sites in Oregona
Percent per Annum
Percent per Annum
Percent per Annum
Slopes are given in both percentage change per annum and in W m−2/a. No vol denotes trend calculations with the years most strongly impacted by volcanic eruptions removed (1982–1985 and 1991–1994). Uncertainties represent 1 standard error of the slope coefficient as described in section 3.
0.20% ± 0.07%
0.31 ± 0.11
0.13% ± 0.05%
0.29 ± 0.11
0.10% ± 0.04%
0.19 ± 0.07
0.19% ± 0.08%
0.29 ± 0.12
0.08% ± 0.06%
0.18 ± 0.14
0.08% ± 0.05%
0.16 ± 0.10
Direct normal irradiance
0.54% ± 0.17%
0.83 ± 0.26
0.49% ± 0.21%
1.10 ± 0.47
0.48% ± 0.18%
0.75 ± 0.28
0.43% ± 0.16%
0.66 ± 0.25
0.22% ± 0.14%
0.50 ± 0.31
0.36% ± 0.15%
0.57 ± 0.23
 The linear regression fits to both annual average total irradiance and direct normal irradiance are statistically significant at a 95% confidence level for all three sites. The 95% confidence level is calculated as twice the standard error given in Table 1. Total irradiance increases by 1–2% per decade (2–3 W m−2 per decade) at all three sites. Annual averages of direct normal irradiance show stronger increases (5% per decade or 8–11 W m−2 per decade) than total irradiance, as expected, because of the higher sensitivity to scattering and larger values since measurements are made normal to the incoming radiation.
 The time series appear to be strongly impacted by the volcanic eruptions of El Chichón (March 1982) and Mount Pinatubo (June 1991). Excluding the years most heavily impacted by volcanic eruptions (1982–1985, 1991–1994) reduces the slope of linear trends calculated from annual average data, particularly for the shorter time series beginning in 1980 (see the rows labeled “No vol” in Table 1). Of the total irradiance measurements, only the slope of the longer time series in Eugene (beginning in 1976) remains statistically significant with a magnitude of approximately 2% per decade. The slopes of fits to total irradiance measurements at Burns and Hermiston drop below 1% per decade when removing the volcanic years and are not statistically significant at a 95% confidence level. The slopes of fits to direct normal irradiance drop to 2–4% per decade, but remain statistically significant at Eugene and Hermiston.
4. Clear-Sky Direct Normal Irradiance
 To more closely examine the impact of aerosol changes on the data, clear (i.e., cloud-free) periods were identified in the 5-min resolution data using the method developed by Long and Ackerman . This clear-sky detection algorithm uses the magnitude and variability of the total and diffuse irradiance to look for smoothly varying periods falling within expected clear-sky values (for details, see Long and Ackerman ). Clear-sky periods were identified at these sites using diffuse irradiance calculated by subtracting the direct component from the total SW irradiance measurements. Once clear-sky periods are detected, daily coefficients for a power law function are fit to the data when possible, and the coefficients are interpolated between values through time for cloudy days in order to produce continuous estimates of clear-sky total, direct, and diffuse SW irradiance. These clear-sky estimates are then used as the expected clear-sky values in the identification algorithm in an iterative process. Since the Long and Ackerman  methodology is intended to detect the presence of clouds, general aerosol effects and changes are also included in the clear-sky estimations. Studies have shown [Dupont et al., 2008] that the Long and Ackerman  clear-sky detection allows haze (primarily subvisual cirrus) up to about 0.15 optical depth in the clear-sky category.
 Monthly averages of direct normal irradiance measurements identified as clear between solar zenith angles of 65° and 75° were calculated for this analysis. These zenith angles were chosen for two reasons. At large zenith angles (low solar elevation angles), solar radiation must travel along a longer path length through the atmosphere than at smaller zenith angles, making the impacts of scattering more visible than at smaller zenith angles. The second advantage of using measurements at these zenith angles is that they are fully represented in all months at all three sites. The minimum solar zenith angle reached in the winter months at the latitudes of these stations is slightly less than 65°.
 The monthly averages were then deseasonalized by subtracting long-term averages for each month. The long-term averages were calculated over all months except those most strongly impacted by volcanic eruptions (March 1982 to December 1985 and June 1991 to December 1994). Figure 4 shows these monthly average anomalies for each site.
 The time series in Figure 4 show four changes in clear-sky irradiance that could be related to aerosol changes: the large dips from the volcanic eruptions of El Chichón and Mount Pinatubo, low clear-sky values before 1982, a positive trend between 1987 and 2007 (neglecting the large dip from 1991 to 1994) at Eugene and Burns, and a step-function-like decrease at Hermiston around June of 2003.
 The effects of the volcanic eruptions of El Chichón and Mount Pinatubo are the most prominent changes in the time series. The eruptions reduce the direct normal irradiance at these zenith angles by as much as 200 W m−2 or 25% in the months after the eruptions. This gives us a reference for the sensitivity of these data to a known large increase in particulate concentrations.
 It is also noteworthy that the clear-sky values in the years directly before the eruption of El Chichón (1980–1981) are on average lower than those before the eruption of Mount Pinatubo (1989–1990) and at the end of the time series. This is particularly visible in the Eugene and Burns time series. The clear-sky irradiance at Hermiston shows low values in the early years, but also during the background periods around 1990 and 2003–2007, as discussed in more detail in the following paragraphs. The early low values are also seen in the all-sky data, and are the reason the time series show only positive trends and not a “dimming” followed by a “brightening.” The fact that the clear-sky measurements from all sites are low before 1982 suggests a regional cause.
 The scatter in the time series at Eugene is greater than that at Burns (Figure 4). This greater variability in the monthly averages of clear-sky data at Eugene may be related to the statistical variability caused by fewer clear periods available in a given month and not just atmospheric changes. The Burns station is located on a high desert plateau where both summers and winters have a high frequency of clear periods. The Eugene station is located in the Willamette Valley, with weather patterns characterized by cloudy winters.
 To better compare the measurements at the three sites for regional events, the time series are smoothed with a lowess filter. The smoothing technique replaces individual measurements with points calculated from a weighted linear regression with a 1-year span. The results are shown in Figure 5. This view of the data clearly shows the low values at the beginning of the time series for Eugene and Burns, and to a lesser extent Hermiston. The smoothed time series from Hermiston deviates more from the other two sites than the Eugene and Burns time series deviate from each other. This is particularly visible at the end of time series when the Hermiston time series is consistently lower than those at Eugene and Burns from June of 2003 to December of 2007.
Streets et al.  modeled anthropogenic aerosol concentrations and found that global average anthropogenic aerosol optical depths decreased between 1988 and 2000. They hypothesized that this might cause the brightening seen in surface irradiance measurements. To more closely examine this period in the surface irradiance data, linear regression lines were fit to the clear-sky anomalies for the years 1987 to 2007, excluding June 1991 to December 1994. These trend lines are plotted in Figure 4, and have slopes of 0.80 ± 0.26, 0.88 ± 0.34, and 0.11 ± 0.45 W m−2/a for Eugene, Burns, and Hermiston, respectively. The uncertainties represent one standard error of the slope calculated as described in section 3. The trends at Eugene and Burns amount to an increase of 16–18 W m−2 in 20 years or a change of 2–2.5%. The increases in clear-sky irradiance over this period could be caused by a number of factors including changes in stratospheric aerosol loads following volcanic eruptions, a change in background or local aerosol concentrations, or changes in water vapor concentrations. A closer investigation of the causes of these trends is described in section 5.
 The change between 1987 and 2007 at Hermiston is smaller and not statistically significant. The cause of the smaller increase at Hermiston than at the other two sites appears to be caused by a dip in irradiance measurements in June of 2003.
 Neither the positive trends in clear-sky irradiance at Burns and Eugene, nor the decrease at Hermiston in 2003 appear to be caused by instrument calibration or replacement. As was noted in section 2, the calibration values used with the NIPs making these measurements were not changed over the time period each was in use. One would expect a calibration error to cause a gradual decrease that is not visible in Figure 4. The change at Hermiston cannot be explained by switching instruments because the same instrument has been used to take measurements at the site since 1999, which does not correspond to the decrease in 2003. New instruments also cannot explain the positive trend between 1987 and 2007 at Burns because the instrument currently in use was installed at the site in July of 1987.
 The cause of the decrease in irradiance in Hermiston in 2003 clear-sky irradiance at Hermiston is not known, however, station maintenance and an increase in local aerosol optical depth were considered as possibilities.
 Some problems have been found in Hermiston measurements in recent years caused by a decreased frequency of station maintenance. Less frequent cleanings or alignment could cause a decrease in irradiance, though no sudden change happened with station maintenance in 2003. To confirm that maintenance issues did not cause the decrease in irradiance, the measurements were compared to direct normal irradiance calculated from a rotating shadowband pyranometer (RSP) at the site. Clear-sky RSP measurements also show lower irradiance values around 2003, and do not appear to deviate systematically from the direct normal NIP measurements in June of 2003. Thus, instrument maintenance does not appear to be the cause of the 2003 decrease at Hermiston.
 A second possible explanation for the decrease in irradiance at Hermiston is an increase in local aerosol concentrations. No changes have yet been found for any known particulate sources that correspond directly to the apparent timing of the step function in the measurements, though two known sources began operating within a year of June 2003. According to the Calpine website, the natural gas powered steam and electricity generation plant in Hermiston began operations in August of 2002 (http://www.calpine.com/power/plant.asp?plant=62). The Department of Environmental Quality's website reports that the Umatilla Chemical Depot (located 8 miles west of Hermiston) conducted trial burn operations in the beginning and end of 2003 and more continuous incineration sometime afterward (http://www.deq.state.or.us/umatilla/cdp.htm). It is not known whether the emissions from these sources would be sufficient to impact the clear-sky direct normal irradiance in the manner observed at Hermiston. A change in agricultural practices in the region or at the agricultural research station where the instruments are located could also impact the irradiance by changing the amount of dust in the air. More specific measurements of particulate matter or aerosol optical depth would be helpful to examine whether the variability in local aerosol sources accounts for the deviation between Hermiston and the other two sites.
5. Modeling Stratospheric Aerosol Changes in Eugene
 Estimates of the impact of volcanic eruptions and tropospheric water vapor concentrations on the clear-sky irradiance were calculated for Eugene using available data sets. Changes in these two factors are assumed to have the largest impact on clear-sky surface irradiance besides anthropogenic aerosols. This calculated irradiance is compared to measured clear-sky irradiance to further examine how much of the clear-sky irradiance increase between 1987 and 2007 (plotted in Figure 4) could be caused by anthropogenic aerosols. Section 5.1 describes the radiative transfer model and input data sets used to make those calculations. Modeled and measured direct normal irradiance at zenith angles of 65–75° are compared in section 5.2.
5.1. Description of Model and Inputs
 The stratospheric aerosol optical depth data set described by Sato et al.  was used to calculate the clear-sky direct normal irradiance changes caused by volcanic eruptions. For the period from 1980 to October of 1997 monthly average 550-nm stratospheric aerosol optical depths (AODs) are obtained primarily from satellite measurements including the stratospheric aerosol and gas experiment (SAGE) instruments [Sato et al., 1993]. The AODs used in this study are for a zonal average of 43 north latitude. From November 1997 to December 1999 the data set estimates stratospheric AOD assuming an exponential decay from the 1997 values with a 1-year time constant (http://data.giss.nasa.gov/modelforce/strataer/), dropping the 550-nm AOD to 0.0005 in December of 1999. Stratospheric AOD from lidar measurements in Germany [Jäger, 2005], however, indicate that the background stratospheric aerosol levels were reached by 1997, and did not decrease further in the following years. Thus, stratospheric aerosol optical depth was assumed to remain constant at 0.0038 (the last measured value) from October 1997 to December 2007 in these calculations. Section 5.2 also gives the impact on direct normal irradiance assuming 550-nm stratospheric AOD decreases to 0.0005 for the sake of comparison.
 Radiative transfer calculations were done using an eight-stream discrete ordinate (DISORT) model [Stamnes et al., 1988]. Monthly average humidity and temperature profiles were taken from NCEP Reanalysis data [Kalnay et al., 1996]. Humidity profiles are only available for the troposphere, so no impact of stratospheric water vapor changes is considered. A standard summer ozone profile was used throughout the calculations. Correlated-k gaseous absorption was handled using the Guo and Coakley  adaptation of the NASA Langley Fu-Liou code. Lacking more specific information about tropospheric aerosol concentrations, a constant continental type tropospheric aerosol with a 550-nm aerosol optical depth of 0.05 was used. Aerosol scattering properties were taken from aerosol mixes described by Hess et al. .
5.2. High Zenith Angle Calculations
 Monthly average clear-sky direct normal irradiance was calculated for zenith angles of 65–75°. The results of the model calculations are shown in Figure 6 along with Eugene monthly average measurements for the same zenith angle range. The annual cycle is much stronger in the measurements than the modeled values, as expected, because modeled tropospheric aerosol loads are not varied seasonally. Neglecting this difference, the modeled values appear to be in a reasonable range and fit the variation in the volcanic aerosols. Figure 7 shows that the monthly anomalies of the modeled data set also fit the measured values well for the eruptions of El Chichón and Mount Pinatubo, but the radiative transfer calculations using the Sato et al.  data set do not explain the low values in 1980–1981.
Figure 8 shows measured (Figure 8a) and modeled, (Figures 8b–8d), monthly anomalies for 1987–2007 (excluding June 1991 to December 1994) and least squares regression fits to each time series. Each of the graphs in Figure 8 shows monthly average anomalies of clear-sky direct normal irradiance for zenith angles of 65–75°. The radiative transfer calculations in Figure 8b use input data from both the stratospheric aerosol data set [Sato et al., 1993] and monthly average tropospheric water vapor profiles [Kalnay et al., 1996]. The model calculations in Figures 8c and 8d only include interannual changes in either water vapor or stratospheric aerosol optical depth respectively, giving an estimate of the relative impact of these two factors on clear-sky direct normal irradiance. The time series in Figure 8c was calculated assuming the stratospheric aerosol optical depth was zero, and that in Figure 8d was calculated assuming a constant seasonal water vapor cycle from NCEP reanalysis long-term mean water vapor profiles.
Table 2 lists the slopes of the trend lines plotted in Figure 8 with uncertainties given as 1 standard error of the slope coefficient, calculated as described in section 3. Changes in water vapor and stratospheric aerosols over this time period impact the direction of the trend oppositely. The slope of the trend in Figure 8d is 0.34 ± 0.14 W m−2/a, representing the irradiance impact of just the volcanic aerosol changes. The water vapor decreases clear-sky direct normal irradiance at these high zenith angles with a slope of −0.17 ± 0.06 W m−2/a (Figure 8c). The combined impact of stratospheric aerosols and water vapor is 0.19 ± 0.08 W m−2/a. It should be noted, however, that the NCEP reanalysis water vapor profiles are identified as a class B product, meaning they are influenced strongly by the climate model and not just observations and thus may not give reliable values for trend studies [Kalnay et al., 1996]. Though the magnitude of the impact of the water vapor change is not known precisely, the direction is consistent with an increase in temperature over the time period.
Table 2. Slopes of Regression Lines Plotted in Figure 8a
Slope (W m−2/a)
Slopes describe change in clear-sky direct normal irradiance at solar zenith angles of 65–75 from 1987 to 2007. Letters a–d correspond to panels in Figure 8.
0.88 ± 0.34
b: both inputs
0.19 ± 0.08
c: only water vapor
−0.17 ± 0.06
d: only stratospheric aerosols
0.34 ± 0.14
 Only about a fifth of the total 0.88 ± 0.34 W m−2 increase per year in the high zenith angle measurements is explained by the combined impact of stratospheric aerosol concentration changes and water vapor variability in the model data. The difference between the measured and modeled trend is 0.69 ± 0.35 W m−2/a, where the uncertainty is found by adding the slope uncertainties in quadrature. This difference is statistically significant at a 90% confidence level, and supports the possibility that a reduction in anthropogenic aerosols could cause some of the increase in irradiance in Oregon over the past two decades. These results are based on the assumption that the stratospheric aerosol optical depth of October 1997, 0.0038, is an adequate description of the background aerosol for the next ten years. As no major volcanic events have occurred in that time period, it is thought to be a good assumption, though the stratospheric AOD could have dropped slightly lower. To calculate an upper limit on the possible contribution of volcanic aerosols to the clear-sky trend, the model calculations were also made assuming that the background stratospheric AOD diminishes as low as 0.0005 by the year 2000 as calculated in the Sato et al.  data set. The combined impact of stratospheric aerosol reduction and water vapor increases from 1987–2007 could then be as large as 0.49 ± 0.09 W m−2/a, leaving a difference between the measured and modeled clear-sky slopes of 0.34 ± 0.35 W m−2/a. This indicates that even were the background stratospheric AOD to go to nearly zero, the irradiance measurements do not exclude the possibility that some of the change in surface irradiance is caused by anthropogenic aerosol concentrations.
 While the modeled results here suggest that volcanic aerosols and water vapor do not explain the positive trend in clear-sky direct normal irradiance, they cannot confirm that anthropogenic aerosol decreases cause the trend. Other factors could also cause increases in clear-sky direct normal irradiance including changes in the level of haze or subvisual cirrus as these conditions are not screened out using the Long and Ackerman  clear-sky detection method. Variability in natural aerosol sources like forest fires might also contribute to the increases in clear-sky irradiance. Measurements of aerosol optical depth would be a valuable addition to this study to more precisely evaluate how much of the clear-sky changes are due to aerosol concentration changes. Spectral irradiance measurements are available at the Eugene site from a MultiFilter Rotating Shadowband Radiometer (MFR-SR) from 1997 until the present, but have not yet been analyzed. These measurements can be used to better separate the impacts of aerosols and water vapor, and to examine any changes in AOD over the last decade.
6. Conclusions and Outlook
 Both total and direct normal irradiance have increased in Oregon over the last 3 decades. The three sites used in this study show no indication of a dimming before 1990, except the low irradiance values following the eruption of El Chichón. In fact, the years 1980 and 1981 in the data set appear to be lower than periods later in the time series that are similarly not strongly impacted by major volcanic eruptions. The causes of these low values are unknown, but are seen at all three sites, suggesting a regional cause. No change in instrument setup has been found that would systematically decrease measurements at all three sites over those years. Calibration drift tends to cause a decrease in the measurements over time rather than an increase. Clear-sky direct normal irradiance is also lower before 1982 than other periods with background levels of AOD, indicating that the cause is not only a matter of cloudiness.
 One possibility is that the low values were caused by the eruption of Mount St. Helens in 1980, though the eruption is not thought to have impacted atmospheric transmission for a long period of time. Michalsky and Stokes  found that on the order of days to a few weeks after eruptions of Mount St. Helens in 1980, the baseline turbidity returned to levels consistent with 1979 turbidity. Their analysis was based on direct normal spectral measurements in Richland, Washington, ∼50 miles north of the Hermiston station. Stratospheric lidar measurements in Germany, however, indicate an increase in backscatter following the eruption of Mount St. Helens in May of 1980 lasting 11 months and the eruption of Alaid in 1981 lasting 8 months [see Jäger, 2005, Table 1]. More careful comparisons between measurements of these volcanic emissions and solar irradiance would be needed to determine whether the low values in the Oregon data could be caused by these volcanic eruptions.
 A second possibility is that the lower irradiance values before 1982 were caused by higher levels of particulate matter from anthropogenic sources. Streets et al.  modeled emissions of SO2 and black carbon in the United States from 1980 to 2000, and found that emissions were highest in 1980 and steadily decreased over the next 20 years. Measurements of aerosol optical depth or particulate matter would be helpful to confirm whether or not the low irradiance values before 1982 come from higher aerosol loading, though it is difficult to find aerosol measurements in the 1980s.
 Direct normal clear-sky irradiance increased significantly at two of the sites between 1987 and 2007 in time series of measurements made at solar zenith angles of 65–75 degrees. Clear-sky measurements from the third site, Hermiston, decrease at the end of the time series. Visual inspection suggests a step-function-like change around June of 2003 in the Hermiston time series. The cause of this change is unknown, though it may be related to a local increase in atmospheric particulates.
 Radiative transfer model calculations were made for the Eugene site to estimate the impact of volcanic aerosols and water vapor on the clear-sky direct normal irradiance. Stratospheric AOD values based on SAGE measurements [Sato et al., 1993] were used until October of 1997 and any further decreases in background stratospheric AOD were assumed to be negligible. The combined effect of volcanic aerosols and water vapor explained about 20% of the measured clear-sky irradiance increase from 1987 to 2007 at solar zenith angles of 65–75 degrees, suggesting that a decrease in anthropogenic or other regional sources of aerosol concentrations could also be a cause of the increase in irradiance.
 Future work is needed to determine whether the increase in clear-sky irradiance is caused by aerosol changes, and if those changes are sufficient to explain the all-sky irradiance trends. Work is currently being done to calculate limits on the possible change in aerosol optical depth consistent with the irradiance measurements. Spectral irradiance measurements are also available in Eugene from 1997 to the present. Analysis of these measurements would give a better idea of the nature of the background aerosol in Eugene and any change that has occurred over the last decade.
 Thanks go to the Eugene Water and Electric Board, the Bonneville Power Administration, and the National Renewable Energy Laboratory for supporting the operation of the University of Oregon Solar Radiation Monitoring Laboratory. NCEP Reanalysis data were provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, from their Web site at http://www.cdc.noaa.gov/. Stratospheric aerosol optical depth data were obtained from the NASA Goddard Institute for Space Studies Web site at http://data.giss.nasa.gov/modelforce/strataer/. Thanks go to J. A. Coakley Jr. for providing radiative transfer code and helpful feedback on model calculations, H. Harrison for help in calculating trend uncertainties, and Martin Wild and two anonymous reviewers for helpful discussion. L. D. Riihimaki wishes to acknowledge support from the Engineering and Technology Industry Council of Oregon. C. N. Long acknowledges the support of the Climate Change Research Division of the U.S. Department of Energy as part of the Atmospheric Radiation Measurement (ARM) Program.