Richard C. Willson (willson@simdac.jpl.nasa.govv), Principal Investigator EOS/Active Cavity Radiometer Irradiiiance Monitor (ACRIM)
The science objectives of the ACRIMSAT Mission are in the fields of climatology and solar physics. Sustained changes in the total solar irradiance (TSI) of as little as a few tenths of one percent per century could be primary causal factors for significant climate change on time scales ranging from decades to centuries.1 It is clear from paleoclimate research that periodic solar irradiance-driven climate changes have occurred.2 There is compelling evidence that some of these may have been driven by intrinsic solar variability. 3,4 A precise, long-term record of solar luminosity variation is required to provide empirical evidence of the sun's role in climate change and to separate its effect from other climate drivers. The same record, together with other solar observations, will yield an improved understanding of the physics of the sun, the causes of luminosity variations, and could eventually lead to a predictive capability for solar driven climate change.
The National Research Council recently published its findings regarding research priorities for Solar Influences on Global Change, one of the seven science element's of the U.S. Global Change Research Program.5 Their recommendations include "monitoring total and spectral solar irradiance from an uninterrupted, overlapping series of spacecraft radiometers employing in-flight sensitivity tracking" as this element's highest priority and most urgent activity. The EOS/ACRIM-SAT mission is designed to be a cost-effective, small-satellite approach to meeting that priority.
The sun is a variable star. Its luminosity has been found to vary by 0.1 percent over a solar cycle in phase with the level of solar magnetic activity.6 Photometric observations of many solar-type stars have revealed that brightness variations correlated with magnetic activity like the sun's are a common phenomenon. Many demonstrate higher variability than the sun, leading to speculation that the sun's variability may have been greater in the past and may be again in the future.7, 8 This would have significant implications for climate change.
A precision TSI database with resolution adequate to relate centuries of systematic TSI variation to climate change must be compiled from the results of many flight experiments. With a nominal lifetime of 5 years per experiment, their contiguous results must be related with the maximum precision accessible to current technology, on the order of 10 ppm. This far exceeds the capability of current "ambient temperature" flight instrumentation to define the "absolute uncertainty" of the TSI (>1000 ppm) and even that of cryogenic instrumentation currently under development (>100 ppm). The uncertainty of modeling TSI using ground-based observations of proxy solar emission features is orders of magnitude less precise.
The approach capable of providing the maximum precision for the long-term TSI database with current measurement technology employs an "overlap strategy" in which successive ambient temperature TSI satellite experiments are compared in flight, transferring their operational precision to the database. The current generation of ambient temperature ACRIM flight instrumentation has demonstrated a capability of providing annual precision smaller than 5 ppm of the TSI.6
The EOS/ACRIM experiment was selected to provide the TSI database during the EOS mission. We propose to accomplish the ACRIM science objectives using a cost-effective ACRIMSAT small-satellite sub-mission to implement an overlap measurement strategy and provide the EOS mission segment of the long term, precision, climate TSI database.
ACRIMSAT uses the Active Cavity Radiometer Irradiance Monitor technology flown successfully on NASA's Solar Maximum Mission, Upper Atmosphere Research Satellite, Spacelab 1 and ATLAS missions. A down-sized version of ACRIM instrumentation will be mated with small-satellite technology to construct dedicated ACRIMSAT satellites. ACRIM-SAT's, with a launch volume of less than 0.25 m 3, can be launched two at a time "piggy back" on Pegasus boosters, reducing launch costs to a minimum. The first two ACRIMSAT's can be on orbit within 24 months of mission startup, enhancing the possibility of implementing the overlap strategy with the Upper Atmosphere Research Satellite ACRIM II experiment during its extended mission and the SOHO/VIRGO experiment prior to the end of its two-year minimum mission. A series of ACRIMSAT's is proposed that could provide overlapping satellite total solar irradiance observations throughout the EOS mission.9
Observations of TSI Variability
The first long term solar monitoring utilizing an Electrically Self Calibrating Cavity (ESCC) sensor was the Earth Radiation Budget (ERB) experiment on the NASA Nimbus 7 spacecraft. The ERB database, beginning in late 1978 and continuing to early 1993, is the longest currently available.10 Limitations imposed on ERB solar observations by the absence of solar pointing on the Nimbus platform sustained a noise level in the ERB results that inhibited recognition of intrinsic solar variability until subsequent detection by JPL's Active Cavity Radiometer Irradiance Monitor I (ACRIM I) experiment on the NASA Solar Maximum Mission (SMM) in 1980.11 The mutually corroborative function of the ACRIM I and ERB results has played an important role in verifying intrinsic solar variability on solar activity cycle time scales.
A series of shorter term TSI experiments have been flown on or deployed from the space shuttle to provide comparison experiments for satellite monitors. The Spacelab 1 and ATLAS flights between 1983 and 1993 employed two different TSI experiments, as has the shuttle-deployed EURECA platform that operated for 10 months in 1992-93. 12-14 The shuttle ACRIM experiment has demonstrated a capability of sustaining flight-to-flight precision on the order of 100 parts-per-million (ppm).13 This precision is comparable or superior to the accuracy achievable by radiometers operating even at cryogenic temperatures, but significantly inferior to that accessible using the overlap strategy with ambient temperature satellite experiments. The principal source of uncertainty for the shuttle flights is the potential for contamination of the instrumentation during integration and launch.
The results of modern TSI monitoring are shown in Figure 1. (Contact the author or refer to printed copy of the newsletter) The Nimbus 7/ERB, SMM/ACRIM I and UARS/ACRIM II experiments have documented direct dependence of the TSI on solar activity. Qualitatively similar results have been obtained with the ERBS, NOAA-9 and NOAA-10 solar monitors. The shuttle-based Spacelab 1 and ATLAS ACRIM observations are reference points for the long term satellite solar monitoring experiments.
Results of TSI Variability Observations
The most significant finding from the precision TSI database thus far is on solar cycle time scales: a direct correlation of luminosity and solar activity.6,15,16 With a 0.1% peak-to-peak amplitude during solar cycle 21, it agrees in sense with that predicted from the coincidence of the "Little Ice Age" climate anomaly and the "Maunder Minimum" of solar activity during the 16th and 17th centuries. 3
Solar cycle TSI variation is predicted with varying degrees of success by linear regression models using the precision TSI database and "proxies" of solar activity, such as the Zurich sunspot number, the 10.7 cm microwave flux, He I 1083 nm full disk equivalent width and the øcore-to-wing ratioÓ of the Mg II line at 280 nm. The use of the He I model led to the initial realization of the primary role of faculae and the bright network in the solar cycle TSI variation.6, 7-19 The "proxy models" of TSI have been useful in providing qualitative explanations of solar phenomena, but it is not surprising in view of the fact that they are statistical constructs and not physical models, that significant errors, relative to satellite observations, are found in some model predictions of TSI.
An inverse relationship between sunspot area and total irradiance has been found on the solar rotational time scale (27 days) with deficits in total irradiance of as much as -0.3%.11 There is growing evidence that most of the missing flux is balanced by excess facular radiation on the active region time scale (months) with the rest redistributed through the bright network on the solar cycle time scale.20,21
On the shortest time scales, solar global oscillations of low degree have been detected in the ACRIM I total irradiance data, including pressure modes (time scales of minutes--the so-called 5-minute oscillations or "P-modes") 22 and possible gravity modes (time scales of hours to days). 23 Interpretation of the 5-minute oscillation results from the ACRIM I experiment has placed an upper limit on differential rotation of the outer solar atmosphere as a function of solar radius, and therefore on solar oblateness, providing support for the relativistic interpretation of the perihelion of Mercury observations. 22 P-mode oscillations are constrained to the convection zone or just below; therefore, the depth within the sun to whichûtheir analysis can provide new physical insight is limited. Should gravity mode oscillations be verified in TSI data, their analysis would yield information on physical processes extending to the solar core.
TSI variations on time scales shorter than a year do not appear to be of direct climatological interest but contain information on solar variability that have provided much new insight into the physics of the sun. Continuous TSI monitoring, particularly by satellites with a high solar pointing duty cycle during each orbitûcan provide the observations that will facilitate future solar models that may predict TSI variability with sufficient precision to anticipate corresponding climate variations.
Present and Planned TSI Monitoring
The Nimbus 7/ERB experiment ceased operations in early 1993. The precision TSI climate database is currently being sustained by a single experiment, the UARS/ACRIM II. The UARS has on-board resources and an orbit that could last to the end of the decade. However, early problems with the battery and solar panel drive systems have raised some doubts about the longevity of UARS. Should it fail before the launch of the SOHO/VIRGO experiment, the TSI database would experience a discontinuity that could only be addressed by reflight of one or more of the shuttle-based TSI experiments. The uncertainty of the discontinuity would not be less than the reproduceability accessible to successive shuttle experiment operations which would compromise the extension of the existing TSI database. An additional concern is always the continuity of the mission operations and data analysis (MO&DA) funding which frequently becomes the scarcest resource of all in "extended" missions.
The ERBS and NOAA-9 experiments continue to function. These have provided the required solar insolation observations for their radiative balance science objectives, but because of infrequent and brief solar observation opportunities, they cannot contribute significantly to the precision of the long term TSI database.
The next TSI experiment, to be launched in mid-to-late 1995, will be the European Space Agency's (ESA) Solar Heliospheric Observer (SOHO)/VIRGO, with a minimum mission lifetime of two years. With the SOHO launch less than a year away and the UARS operational problems seemingly under control, the probability of conducting overlapping observations between ACRIM II and VIRGO seems fairly high.
The next planned NASA experiments are a series of ACRIM's included in the Earth Observing System program as flights-of-opportunity currently scheduled to begin in the 1999-2000 timeframe. The major concern in the effort to sustain the TSI database during the late 1990s is the probable cessation of UARS/ACRIM II and possible cessation of SOHO/VIRGO observations prior to the inception of EOS/ACRIM observations in 1999 or 2000. Failure to overlap these experiments could result in a catastrophic loss of relative precision between the first 20 years of the long term, precision TSI database and that to follow.
Sustaining the TSI Database
Monitoring solar luminosity variability with maximum precision demands not only state-of-the-art technology but the use of an optimum research strategy. Following is an evaluation of approaches to sustaining the precision TSI database with the requisite 10 ppm or smaller discontinuities between experiments.
The "Overlap" Strategy with Ambient Temperature Radiometers
A relative precision smaller than 10 ppm should be readily achievable for the data of overlapped solar satellite monitors, assuming a sufficiency of overlapping comparisons and adequate degradation calibrations. The principal source of uncertainty for satellite experiments is degradation of their sensors by extended solar exposure during multiyear missions. The series of ACRIM experiments have employed a three-fold sensor redundancy and phased operational modality that calibrates such degradation with residual an uncertainty of less than 50 ppm per decade.6 The optimum overlap strategy is the intercomparison of successive, high precision satellite solar monitoring experiments at a precision level defined by their operation in the space flight environment. The backup overlap strategy would involve intercomparisons by a "third party" flight experiment, such as another satellite experiment or the shuttle-based TSI experiments, that have made intercomparisons with two successive but non-overlapping satellite solar monitors.
The "overlap strategy" was to have begun with the overlap of the SMM/ACRIM I and UARS/ACRIM II experiments. Unfortunately the SMM mission ended in late 1989, two years before the delayed UARS could be launched. The relationship between the ACRIM I and ACRIM II experiments has instead been established using a "third party" overlap strategy based on the results of mutual comparisons of ACRIM I and ACRIM II with the less precise but long lived Nimbus 7/ERB experiment. The results are shown in Table 1. The ratio of ACRIM I to ACRIM II is 1.002060 with linear detrending of the degrading Nimbus 7/ERB results. The statistical uncertainty of 10 ppm demonstrates the ability of the backup "overlap strategy" to produce high precision.
The Nimbus 7/ERB experiment does not have a degradation calibration capability and linear detrending can only approximate the effects of degradation on the comparison results. The uncertainties of the results in Table 1; therefore, include some systematic error and as such, represent an upper limit for the backup overlap strategy.
Absolute Radiometry
The "absolute" uncertainty (relative to S.I. units) of the current generation of TSI flight instrumentation, which operates at ambient temperatures, is about 1000 ppm in the laboratory and about 3000 ppm in flight experiments. 24,25 Ambient temperature TSI radiometry is a mature technology that reached its inherent design limits nearly 20 years ago. It has been thoroughly flight tested in various configurations on balloon, rocket, shuttle and satellite flight platforms.
The absolute uncertainty of a new generation of TSI sensors operating near the temperature of liquid Helium approaches 100 ppm in the laboratory environment. 26 Cryogenic sensors face some daunting challenges in their transformation into space flight experiments, however. They must use much smaller apertures (0.3 cm diameter) than their laboratory versions to minimize solar heating that would otherwise prevent their Stirling cycle coolers from maintaining temperatures below the required 20 K.
Aperture area determination is the single most limiting source of absolute error with TSI radiometers. The smaller apertures required by cryogenic radiometers are extremely difficult to make and measure accurately.
Contamination is a major source of uncertainty in TSI flight experiments, and this is of particular concern for cryogenic sensors. At low temperatures they would function as "getters" for condensables and particulates. Accumulation of contaminants on the rim of their small areas would cause larger errors than for the larger area apertures of ambient temperature instrumentation. A realistic expectation for their eventual inflight performance would likely be in the several-hundred ppm uncertainty range.
Clearly, the absolute uncertainty of neither the ambient or cryogenic temperature TSI sensor technology is adequate to sustain the contiguous, long term database at the 10 ppm level.
The Use of Solar "Proxy Models"
The use of so-called proxy models of TSI has been advanced by some as an approach for "bridging gaps" between flight observations of the TSI. Proxies are solar line emmision or absorption features that characterize certain processes of the solar atmosphere. The proxy models are statistical constructs based on the regression of the time series of the proxies against the time series of TSI observations. The resulting TSI "models" have provided significant solar physical insight but are qualitative in nature. They are not rigorous physical models in any sense.
The discovery of TSI variability on solar active region timescales stimulated the first simple proxy models. The deficit effect of sunspots on total irradiance was approximated by a simple approximation to a solar atmosphere radiative transfer function called the photometric sunspot index (PSI). It was computed using the projected areas and contrasts of sunspots, taking into account the limb darkening effect. 27 Hoyt and Eddy developed a model using their sunspot blocking function and the Zurich sunspot index to predict the total irradiance variability as far back as 1874. 28 However, irradiance models based only on the sunspots could explain just about half of the total irradiance variation observed by ACRIM I.
The next obvious step was to incorporate faculae into the models. Active region faculae were recognized as significant contributors of excess flux, relative to the undisturbed photosphere, and as a probable mechanism of offsetting the energy deficit of sunspots in active regions. 20,30-32 Similar conclusions were derived from UV observations made by the Solar Mesosphere Explorer (SME) mission. 33,34 More recently, precision ground-based photometry of the solar disk has convincingly demonstrated these effects for faculae.35
As the results of the SMM/ACRIM I and Nimbus 7/ERB followed the solar magnetic activity level from the maximum of solar cycle 21 to the minimum marking the end of cycle 21 and the beginning of cycle 22, the interest of modelers shifted to the solar-cycle timescale. The models of several investigators indicated that the distributed, faculaelike, "active network" provides a significant contribution to the total irradiance variation on solar cycle timescales (~11 years).35,36, 37 The active network is thought to be populated by residual faculae from old, decaying active regions and/or faculaelike features deriving from the distributed solar magnetic field. Major features of the irradiance data during the latter part of solar cycle 21 and the beginning of cycle 22 were qualitatively reproduced by linear regression models using the full disk equivalent width of the He-line at 1083 nm and 10.7 cm radio flux.
The success of proxy irradiance models did not extend to the maxima of solar cycles 21 and 22, however, where they produced estimated fluxes significantly lower than the ACRIM I and ERB TSI observations (see Fig. 2 (Contact the author or refer to printed copy of the newsletter)). While some modelers have chosen to call the experimental data near solar maxima into question,35 it is more likely that these simple regression models, based on chromospheric spectra, cannot adequately describe the complex solar cycle behavior of the photosphere, from which most of the TSI emanates.
Linear regression models should not be expected to provide more than general insight into total irradiance variations. Multi-variate spectral analysis has been shown to be a more effective approach to studying the combined effect of various solar events on the solar irradiance. 37 This technique has found that during the maximum of solar cycle 21 in 1980 most of the power spectral density of ACRIM I's TSI time series was explained by sunspots. During solar minimum (1984-85) more than 80% of the power spectral density at the average solar rotation period (27-day) was caused by faculae and the active network. Multivariate analysis also delineates power spectral peaks not explained by sunspots, faculae or the bright magnetic network near periods of 9 and 27 days, indicating that yet to be discovered solar events are modifying total solar irradiance. This underscores the fact that the underlying physics of solar variability is not well understood and reliance on simple proxy models of TSI for critical links between observations would be a scientifically unjustified approach.
A cross section of viewpoints by experts on the viability of using proxy TSI models for "bridging observational gaps" can be found in the appendix. This must be viewed as a new field of research that may not produce useful results at the level of precision demanded by the long-term TSI database. The concensus is that predicted TSI can be uncertain by as much as 0.1 % near times of solar maximum activity. Unforunately the potential data gap of the late 1990s will occur near the time of solar cycle 23's maximum activity.
Conclusions
The overlap strategy employing flight tested ambient temperature TSI radiometers is the only approach with a high probability of sustaining the long-term climate TSI database with the precision required. A sensibly conservative overlap strategy that can provide a high probability of success requires launch of EOS/ACRIMSAT in 1997, at least several months before the expiration of SOHO's two-year minimum mission.
The Nimbus 7/ERB experiment has ceased. The ERBE experiments cannot provide the precision required by the overlap strategy. Overlap of UARS/ACRIM II and SOHO/VIRGO appears probable, but the outlook for overlap of SOHO/VIRGO and EOS/ACRIM, under current launch plans (1999), must be viewed with considerable pessimism. The design lifetime for SOHO is a two-year minimum mission, with a four-year supply of orbit maintenance resources. Reliance on it for longer than the minimum mission would be unwise due to the many untested features of this libration point satellite, its complement of TSI sensors with limited flight heritage and marginal degradation calibration capabilities.
Appendix
Assessment of Solar Proxy Models as Predictors of TSI
I sought the opinions of three experts in the TSI proxy modeling field to provide a statement of their viewpoints on the viability of the models as TSI predictors. The three, Judith Lean, Jeff Kuhn and Judit Pap, have pursued the proxy issue from different directions and represent a cross-section of current thinking on this topic. The distillation of their opinions is that proxy models are useful for providing physical insight into solar physical processes if highly precise TSI observations are available, but their use as TSI predictors capable of bridging gaps in TSI monitoring with the required precision will not be a realizable capability for a long time, if ever.
My query to the experts was:We are trying to understand the state-of-the-art in the ability of proxy models to estimate the TSI on different timescales and at different parts of the solar cycle. My feeling is that the RISE program will make major progress in this area in the future but that as a science, despite some excellent preliminary work, proxy modeling is immature and that it would be dangerous to rely on it in the near future to sustain TSI observations over flight data gaps.
Response of Judith Lean:
"I agree with you entirely, although I may not be as optimistic as you are that even RISE will provide quantitative models with sufficent accuracy. Processing ground-based data such as CaK images is turning out to be very difficult because of instrumental effects--even Jack Harvey has to remove sufficient instrumental effects from his CaK gong images. In removing the instrumental effects, it is quite possible that any background irradiance component is also being removed.
Since I know you know all the problems with the various
present-day proxy models (need to improve suspot blocking,
facular determination etc, etc.), I am faxing you a figure
that compares some models used for HISTORICAL
reconstructions that assume different long-term backgrounds.
The H&S model uses the length of the solar cycle to
determine the background component, and I have recently
developed a background component based on the GSSN to add to
the FL mode (which lacks a background component entirely!).
The figure shows that even a gap of two years at the
appropriate time could mean an error of about 0.5 Wm
Response of Jeff Kuhn:
"1) It is possible that statistical interpolation using
several proxies (at a minimum Ca K + UV/EUV + CM) could
provide an irradiance signal accurate enough to improve the
climate modeling. But I don't believe this is the most
interesting problem with the modeling efforts, i.e., there
are bigger problems on this front.
2) The most interesting physical problem (to my thinking)
is the problem of the origin of this variability. I do not
believe the proxy data have/will provide much more than some
evidence for the number of statistically independent
components to this variability.
3) Spatially resolved proxy (nonbolometric, e.g., Ca K)
will lead to some new information on the origins of this
variability.
4) Satellite irradiance measurements are critical to
interpreting ground-based spatially resolved observations.
Without the integrated time-dependent signal (proxies here
are USELESS) in combination with resolved high precision
photometry efforts to understand the variability mechanism
we lose much of the reason for such ground-based
photometry."
Response of Judit Pap:
"The detection of total irradiance variations by satellite
based experiments during the last 15 years stimulated
modelling efforts to help identify their causes and to
provide estimates for time intervals when no satellite
observations exist. The most outstanding problem is the lack
of a quantitative physical model for the variations in total
irradiance, therefore, one has to rely on empirical models
based on 'proxy indicators' of solar activity. The current
empirical models of total solar irradiance developed from
the Photometric Sunspot Index (PSI) and proxy data for
bright magnetic features (faculae, plages and the magnetic
network) disagree with the observations at the time of solar
maximum.17 It has also been found that a considerable
remaining variability exists in total solar irradiance after
removing the effect of sunspots and bright magnetic features
over a broad range of periods including 300, 27, 13.5, and 9
days.38 It is not
clear whether these unexplained variations
are caused by additional solar effects, such as large scale
motions39 and surface
temperature changes
40 or they are
related to inaccuracies in the current proxy data. The PSI
model for the effect of sunspots on solar
irradiance,11,20
,27
has been calculated from the area,
position, and contrast of sunspot groups, published in the
NOAA/WDC Solar Geophysical Data (SGD) catalog. However,
these data are not based on photometric measurements and
each observatory has a different method to estimate the area
and the heliographic coordinates of sunspots and as a
consequence, ~25% to 50% noise is introduced in these
sunspot data.
Full disk measurements in the CaII K line, MgII h & k
lines, and the HeI 1083 nm line equivalent width are used
for modelling the effect of bright magnetic features. It has
been shown that the long-term variation of total irradiance
is primarily caused by the changing emission of faculae and
the bright magnetic network.17
However, the observations
used for studying the effect of the bright features are full
disk measurements and, therefore, they are not capable of
distinguishing between the facular and network contribution
to irradiance changes. In addition, these proxy indices
represent chromospheric conditions, while more than 90% of
total solar irradiance originates from the photosphere where
the physical conditions are completely different than in the
chromosphere. In order to clarify the role of faculae and
the network in total irradiance changes, one should use
spatially resolved data instead of full disk proxies; high
resolution and photometrically calibrated images of the
photosphere are required for measuring the network (and
facuale) area and intensity. These measurements are
necessary to better understand our present surrogates and
they are essential for improving irradiance models.
The crucial questions are: (1) to what extent are the
current models capable of expaining the observed irradiance
changes, and (2) what is the precision of these more
adequate models. Kuhn has shown that surface temperature
changes may also cause long-term irradiance variations as a
consequence of temporal changes in differential rotation in
the interior of the Sun, a solar dynamo magnetic field near
the base of the convective zone or large scale convective
cells. If the observed irradiance change over the solar
cycle represents a global effect, proxy models will not be
able to replace the direct observations. Even by analyzing
the highest resolution solar images, the accuracy of the
models will not be better than about 1%. On the other hand,
daily and continuous irradiance observations are necessary:
coupling the studies of irradiance changes related to
surface manifestations of solar activity will lead to a
better understanding of the underlying physical mechanism
causing solar variability. The ultimate goal is to
understand (1) how, (2) why, and (3) in that time scale the
solar energy flux varies in order to reconstruct and predict
the solar induced climate changes."
References
