Recent observations made by the Spectral Irradiance Monitor (SIM) on the Solar Radiation and Climate Experiment (SORCE) spacecraft suggest that the Sun’s visible and infrared spectral irradiance increased from 2004 to 2008, even as the total solar irradiance measured simultaneously by SORCE’s Total Irradiance Monitor (TIM) decreased. At the same time, solar ultraviolet (UV) irradiance decreased 3–10 times more than expected from prior observations and model calculations of the known effects of sunspot and facular solar features. Analysis of the SIM spectral irradiance observations during the solar minimum epoch of 2008, when solar activity was essentially invariant, exposes trends in the SIM observations relative to both total solar irradiance and solar activity that are unlikely to be solar in origin. The authors suggest that the SIM’s radically different solar variability characterization is a consequence of undetected instrument sensitivity drifts, not true solar spectrum changes. It is thus doubtful that simulations of climate and atmospheric change using SIM measurements are indicative of real terrestrial behavior.
Since incoming solar radiation supplies the energy that powers Earth’s climate and atmosphere, reliable knowledge of how this radiation varies at different wavelengths is critical for assessing solar-induced climate change. Multiple space-based measurements made in the past 30 years show that from the minimum to maximum of the Sun’s 11-yr cycle, total solar irradiance (TSI), the radiant energy integrated over all wavelengths, increased ~0.1% (Kopp and Lean 2011) and middle ultraviolet spectral energy, at wavelengths from 200 to 300 nm, increased by an order of magnitude more (Lean and Woods 2010). But only the Spectral Irradiance Monitor (SIM) on the Solar Radiation and Climate Experiment (SORCE) spacecraft has observed spectral irradiance simultaneously at the ultraviolet, visible, and infrared wavelengths that compose the bulk of TSI changes. From 2004 to 2008, during the descending phase of solar cycle 23, SIM measurements indicate a 0.3 W m−2 (2%) decrease in UV energy between 250 and 300 nm and a corresponding 0.3 W m−2 (0.2%) increase in visible energy between 600 and 700 nm (Harder et al. 2009). Figure 1 shows these changes in the context of spectral irradiance databases.
In lieu of direct, continuous solar spectral irradiance observations on climatological time scales, models such as the Naval Research Laboratory Solar Spectral Irradiance (NRLSSI) model have been developed to calculate the variations arising from the wavelength-dependent effects of sunspots and faculae, the two primary known sources of solar irradiance variations (Lean 2000; Lean and Woods 2010). NRLSSI reproduces the short-term spectral irradiance changes associated with the Sun’s nominal 27-day rotation, but suggests much smaller solar cycle changes than SIM reports. In the visible spectrum the NRLSSI solar cycle changes are an order of magnitude smaller in amplitude and in phase (rather than out of phase) with solar activity. Analogous models that utilize bolometric (spectrally weighted and integrated) sunspot and facular parameterizations account for a high fraction (92%) of total solar irradiance variance that TIM measures simultaneously with SIM’s spectral irradiance measurements, on both solar rotation and solar cycle time scales (Kopp and Lean 2011).
A recent simulation of climate and atmospheric responses to solar spectrum changes using the SIM measurements reported that “the effects of solar variability on temperature throughout the atmosphere may be contrary to current expectations” (Haigh et al. 2010), concluding that higher solar activity cools, rather than warms, Earth. This statement contradicts numerous analyses that empirically relate global surface warming (not cooling) with higher solar activity. Indeed, the combination of global surface cooling produced by decreasing solar irradiance from 2002 to 2008 and greenhouse gas–induced warming during the same period is an explanation for the minimal upward trend in global surface temperature between 2002 and 2009 (Lean 2010).
During the 2008 minimum between solar cycles 23 and 24, trends in solar activity and total solar irradiance were negligible. We analyze SIM measurements during this period with the goal of clarifying causes for the severe discrepancies between the spectral irradiance changes that SIM measures and NRLSSI models.
2. Measured spectral irradiance variations
Multiple measurements and analyses indicate that high solar activity produces high solar UV irradiance, with shorter UV wavelengths increasing more (Lean and Woods 2010; DeLand and Cebula 2012). The databases of the hydrogen Lyman α emission line at 121.6 nm and the irradiance in a wavelength band from 250 to 300 nm in Figs. 1a,b illustrate this; during times of recent solar activity maxima (1980, 1990, 2000) the UV irradiance increased ~50% at Lyman α and ~1% in the 250–300-nm band. In the visible and near-infrared spectrum, the percentage irradiance changes during the solar cycle are an order of magnitude smaller than in the UV spectrum and, according to the SIM observations shown in Fig. 1c, opposite in phase (at least from 2004 to 2007). Figure 2, which shows monthly averaged percentage spectral irradiance changes from April 2004 to December 2008, indicates that SIM’s measurements are in phase with solar activity at ultraviolet wavelengths and out of phase at visible wavelengths.
The significant differences evident among overlapping measurements in UV irradiance at 250 to 300 nm expose instrumental effects in the database. Because the stability of measurements made prior to the SIM is of order 2% (Woods et al. 1996), separating solar cycle changes from instrumental drifts in the middle ultraviolet spectral irradiance database is a challenging task. Solar cycle amplitudes in spectral irradiance at wavelengths between 120 and 200 nm, exemplified by the Lyman α data in Fig. 1a, are an order of magnitude larger, and therefore measured with smaller uncertainty, than at longer wavelengths. The good agreement among Lyman α irradiance measurements in Fig. 1a compared with the disagreement among middle ultraviolet irradiance measurements in Fig. 1b, confirms this.
Harder et al. (2009) report that the ~2% decrease from 2004 to 2008 in the SIM 250–300-nm band in Fig. 1b and the UV spectral irradiance changes over the same period, shown in Fig. 2, characterize real solar variability because the magnitudes of the variations exceed the long-term relative accuracy of the SIM instrument, reported to be 0.5%–0.1% at wavelengths from 200 to 300 nm and 0.2%–0.05% at wavelengths from 300 to 400 nm (Merkel et al. 2011). Since the decrease in SIM’s UV irradiance from 2004 (annual sunspot number of 40) to 2008 (annual sunspot number of 3) corresponds to about one-third of cycle 23’s full activity range (cycle 23 maximum annual sunspot number of 120 in 2000), the inferred solar cycle amplitude is therefore 2–3 times larger than estimates obtained from previous observations, following careful scrutiny of in-flight instrumental behavior (DeLand and Cebula 2012).
The 0.2% increase from 2004 to 2007 in the 600–700-nm band in Fig. 1c, and the spectral irradiance changes at visible and infrared wavelengths in Fig. 2, are likewise claimed to be real solar increases concurrent with the large UV irradiance decreases because the magnitudes of these variations exceed the presumed long-term relative accuracy of the SIM instrument of better than 0.05% at wavelengths from 400 to 1600 nm. Only the increase in irradiance at 900–950 nm is acknowledged to be an instrumental artifact (J. Harder 2010, personal communication).
3. Comparison of measured and modeled spectral irradiance variations
The NRLSSI model calculates solar spectral irradiance variations by accounting for wavelength-dependent sunspot and facular influences. Because sunspots and faculae are respectively darker and, except in a narrow spectral region near 1625 nm, brighter than the surrounding background “quiet” Sun, their greater coverage of the Sun’s disk during epochs of high solar activity alter spectral irradiance at all wavelengths, relative to solar minimum levels. As the Sun rotates, active regions pass across the hemisphere of the Sun projected to Earth, producing additional short-term spectral irradiance variations. This rotational modulation of spectral irradiance has larger magnitude during high solar activity, when sunspots and faculae are prevalent, than during solar minimum, when they may disappear entirely.
The NRLSSI model closely tracks the short-term solar spectral irradiance changes that SORCE measures. Evident in Fig. 3a is the excellent agreement of the Lyman alpha irradiance variations measured by the Solar Stellar Irradiance Comparison Experiment (SOLSTICE) on SORCE, and modeled by NRLSSI during approximately six solar rotations. The changes are of order 15% and unequivocally solar in origin. Even though the variability at middle UV wavelengths is an order of magnitude smaller, there is nevertheless good agreement in Fig. 3b between rotational modulation of the 250–300-nm band that SIM measures and NRLSSI models. And aside from a longer period trend in the SIM observations not evident in NRLSSI, the measured and modeled irradiance short-term changes in the 600–700-nm band, of order 0.1%, also agree well. This good agreement of the SIM observations and the NRLSSI model of rotationally modulated solar spectral irradiance is evident across the entire spectral range from 200 to 1600 nm (Lean and Woods 2010) and confirms that the sunspot and facular parameterizations in the NRLSSI model are robust.
The fundamental differences between the SIM measurements and the NRLSSI model during the descending phase of solar cycle 23, evident in Figs. 1b,c and 2, must therefore arise from a mechanism that dominates on time scales longer than solar rotation, but differs from (and is of greater strength than) the direct effects of sunspots and faculae. During the quiescent period from May to December 2008, when sunspots were absent from the disk for extended periods and other indicators of solar activity were essentially invariant, solar irradiance is expected to neither increase nor decrease (unless there is some unknown modulator of solar spectral irradiance that is unrelated to solar activity and has no net effect on the spectral integral).
The NRLSSI model indicates negligible irradiance trends at all wavelengths during this quiet period in 2008 and Fig. 4a shows that the trend in the Lyman α observations is also minimal (−0.017 mW m−2 yr−1, less than 1% of the cycle amplitude), and consistent with the NRLSSI model trend (−0.004 mW m−2 yr−1) to within the 2% measurement uncertainty (±0.14 mW m−2). However, Fig. 4b illustrates that the SIM’s UV irradiance continues to decline throughout the 2008 solar minimum period whereas the NRLSSI irradiance does not. The most likely explanation for the SIM changes is not solar activity, which was essentially absent, but uncorrected on-orbit instrument changes.
Assuming that trends in the SIM measurements from May to December 2008 are instrumental, and that they have persisted for the duration of the SORCE mission, permits independent estimates of (lower limit) uncertainties associated with SIM’s measurements from April 2004 to December 2008 (4.66 yr). Figure 5 shows the wavelength-dependent uncertainties obtained by multiplying by 4.66 the trends (yr−1) obtained from linear fits to SIM’s 1-nm binned spectral irradiance observations during solar minimum; the uncertainties estimated in this way are typically a factor of 2–5 larger than Merkel et al. (2011) report for the SIM observations. Furthermore, the uncertainty estimates in Fig. 5 obtained from the trends during 2008 (the sixth year of the SORCE mission) likely underestimate SIM’s instrumental trends in prior years because instrument sensitivities typically change more during the beginning of a space mission. SIM’s UV spectral irradiance measurement uncertainty reaches 2% at some wavelengths, which is typical of the uncertainties reported for prior UV irradiance measurements. Because of these larger uncertainties, it is not possible to attribute the large UV spectral irradiance changes that SIM measures to real solar variations rather than uncorrected instrumental changes.
4. Total irradiance constraints on spectral irradiance variations
An immutable constraint on solar spectral irradiance is that the integral over wavelength equals the total solar irradiance. Yet there are notable differences between SIM’s summed energy between 240 and 1630 nm (scaled by a factor of 1.11) and the total solar irradiance measured independently by TIM, shown together in Fig. 6. Most pronounced are differences in 2004, near the beginning of the SORCE mission, and in 2008, during solar minimum, which suggest that SIM spectral irradiance measurements are not adequately constrained. For example, from May to December 2008, the trend in SIM’s integrated (scaled) energy is −0.25 W m−2 (183 ppm) yr−1, whereas TIM measures a much smaller trend in total solar irradiance of −0.01 W m−2 yr−1 (effectively zero because it is less than TIM’s 10 ppm yr−1 stability). If, as we suggest, the uncertainties in SIM’s UV irradiance measurements are larger, and the UV irradiance changes from 2004 to 2007 smaller, than Harder et al. (2009) report, then uncertainties in SIM’s visible and infrared spectral irradiance are also larger. For example, reducing SIM’s reported UV irradiance changes from April 2004 to December 2008 by a factor of 2 requires a decrease, rather than an increase, in longer wavelength irradiance so that the integrated energy matches the independently measured change in total solar irradiance.
A probable explanation for SIM’s anomalous measurements is that instrument sensitivity decreased at UV wavelengths and increased at visible wavelengths by a greater amount than currently assumed, with the (fortuitous) result of net energy change comparable to that measured in total solar irradiance by TIM. Instrument sensitivity at visible wavelengths may have stabilized during the latter years of the SORCE mission since, as Fig. 1c shows, after about 2007, the SIM spectral irradiance in the 600–700-nm wavelength band varies consistently with the NRLSSI sunspot-facular model. These latter SIM measurements therefore contradict the indication from the earlier measurements that the visible spectrum varies out of phase with solar activity.
SIM’s solar spectral irradiance measurements from April 2004 to December 2008 and inferences of their climatic implications are incompatible with the historical solar UV irradiance database, coincident solar proxy data, current understanding of the sources of solar irradiance changes, and empirical climate change attribution results, but are consistent with known effects of instrument sensitivity drifts.
Trends in SIM observations independent of solar activity (in 2008) suggest SIM’s UV measurements are less stabile by a factor of 2–5 than Merkel et al. (2011) report and that SIM’s UV energy decrease from 2004 to 2007 is likely a result of uncorrected instrument changes. The anomalously large UV decrease compensates an equally anomalous visible energy increase in the visible spectrum, also likely of instrumental origin.
Rather than revising current understanding of solar physics and solar variability effects on climate, what is needed is improved characterization of the SORCE SIM observations. Work is under way to reassess the spectral irradiance variations measured by SIM and a new, advanced SIM has been designed and constructed for flight on the Joint Polar Satellite System (P. Pilewskie 2011, personal communication). To prevent future research following a path of unrealistic solar-terrestrial behavior, the SORCE SIM observations should be used with extreme caution in studies of climate and atmospheric change until additional validation and uncertainty estimates are available.
NASA and NOAA funded this work. Jerry Harder provided SIM spectral irradiance data in the file SIM_daily.sav on 13 May 2011. In preparing this article we appreciate the encouragement of Edouard Bard and Gavin Schmidt, and discussions with Peter Pilewskie.