© M. Snow et al., Published by EDP Sciences 2025

1 Introduction

1.1 Motivation

Solar irradiance is the most important external input to the Earth’s climate system. It is a factor of 10⁴ larger than all other energy inputs combined (Kren et al., 2017). Wavelengths shorter than 350 nm are absorbed in the atmosphere before reaching the surface, while longer wavelengths deposit only a portion of their energy in the atmosphere, and the remainder is absorbed by the ground. The balance of energy incident on the top of the atmosphere (TOA) to the energy re-radiated to space is a critical dataset for understanding the Earth’s climate system (L’Ecuyer et al., 2015).

Variation of the solar spectral irradiance (SSI) occurs on all timescales from seconds to centuries. The TOA solar spectrum has been measured in the ultraviolet on a near-daily cadence since the early 1980s. The measurement record of SSI in the visible and infra-red began with the Spectral Irradiance Monitor (SIM; Harder et al., 2005) on the Solar Radiation and Climate Experiment (SORCE; Rottman, 2005). The SORCE/SIM dataset spans from 2003 to 2020. The visible wavelength SSI data record continues with the SIM on the International Space Station as part of the Total and Spectral Irradiance Sensor (TSIS-1; Richard et al., 2020). TSIS-1/SIM has been making daily measurements of ∼200–2400 nm, starting in 2018 and continuing to the present. It will be described in more detail in Section 2.1.1.

The integral of SSI over all wavelengths is known as Total Solar Irradiance (TSI). It is typically measured by a bolometer and reported with low cadence and high latency, such as the Total Irradiance Monitor (TIM) on SORCE or TSIS-1 (Kopp and Lawrence, 2005). The purpose of this project is to answer the following question: Can we produce a proxy for TSI suitable for space weather that has high cadence and low latency? Atmospheric models that use TSI or SSI as an input will be able to accurately recreate a solar signal that varies on timescales of minutes and not merely a daily variation.

Observed variation in solar irradiance at a high cadence is not available operationally, so there are few studies of the sensitivity of the atmosphere to rapid irradiance changes. One example of a model run at a high cadence is McInerney et al. (2018), who simulated the solar eclipse of 21 August 2017. Their study using WACCM-X applied a time-varying loss of irradiance localized to the eclipse path across the Earth. In addition to temperature changes, they calculated rapid changes in NmF2, atomic oxygen, and ozone. These effects can last for hours in their model run. The availability of operational intra-day irradiance variation would allow more investigation of the atmosphere’s response to intra-day variations.

In this study, we use observations from the Solar Position Sensor (SPS) on the Extreme ultraviolet and X-ray Irradiance Sensor (EXIS; Eparvier et al., 2009; Machol et al., 2020) aboard the Geostationary Operational Environmental Satellites R-series (GOES-R; Sullivan, 2020) to create a proxy for TSI. Our dataset is from the GOES-16 satellite. In order to understand the solar sources that contribute to the variation seen in the SPS data, we will create a model of the SPS signal using the entire spectrum measured by TSIS-1/SIM (Richard et al., 2020). We can apply the SPS bandpass (Sect. 3) to the TSIS-1/SIM spectrum and then integrate over wavelength. Harder et al. (2022a) have shown that the integrated full SORCE/SIM spectrum (iSSI) includes 96% of the power that contributes to TSI. The remaining 4% is from the long-wavelength infra-red part of the spectrum that shows no significant variability. Coddington et al. (2023) show that this long-wavelength tail of the SIM spectrum can be extended to 200 μm with theoretical spectral models and then integrated to produce the full TSI within measurement uncertainties. Harder et al. (2022a) and Richard et al. (2024) show that the variability of the integrated TSIS-1/SIM full spectrum matches the variation of the TSIS-1/TIM TSI measurement. The wavelengths outside the SIM bandpass add a constant offset to the integrated spectrum but do not contribute to the variation. Section 3 will show that this is also true for the integrated SPS bandpass.

This article will be followed by an article describing the in-flight corrections and calibration of the SPS data. It will also show a comparison of the calibrated SPS data to TSI on timescales shorter than one day. A third article will demonstrate the use of the calibrated SPS measurements to create a high-cadence empirical SSI model. The basic description of the model algorithm will be described in Section 1.2.

1.2 Spectral model overview

We will ultimately use the algorithms that are used to create the NOAA Climate Data Record (CDR) for solar irradiance (Coddington et al., 2016) to estimate a sunspot darkening proxy from SPS measurements. The CDR model uses two inputs: a facular brightening proxy from the Mg II index (Heath and Schlesinger, 1986; Snow et al., 2018, 2019) and a daily sunspot darkening proxy (Brandt et al., 1994) using data collected from the US Air Force Solar Observing Optical Network (SOON; AFWAMAN15-1, 2014). The SOON daily average is collected from sites around the globe, each taking measurements near local noon. Our modification to the Coddington et al. (2016) algorithm will be to use the observed TSI proxy, T_GOES to estimate the sunspot darkening rather than use direct sunspot measurements. The reason is that sunspot observations are only available on a daily cadence, while the SPS data is measured operationally at 4 Hz.

In the CDR model, the sunspot darkening parameter includes all physical processes that decrease irradiance, such as darkening intergranular lanes as well as sunspots, pores, etc. The model is empirical, so the relative contributions of these physical effects are not separable. The model uses linear correlation to the two input parameters to estimate the irradiance as a function of time.

The Coddington et al. (2016) model, in its simplest form, is two equations:

$$T\left(t\right)=T_Q+\Delta T_F\left(t\right)+\Delta T_S\left(t\right)$$(1)

and

$$I(\lambda ,t)=I_Q\left(\lambda \right)+\Delta I_F(\lambda ,t)+\Delta I_S(\lambda ,t),$$(2)

where T_Q and I_Q(λ) are solar minimum TSI and SSI, respectively. I_Q(λ) is the standard solar minimum reference spectrum from the Whole Heliospheric Interval (WHI; Woods et al., 2009). T_Q in Coddington et al. (2016) is the TSI from SORCE/TIM during the WHI. ΔT_F and ΔI_F are variations caused by facular brightening, while ΔT_S and ΔI_S are the variations caused by sunspot darkening. These four quantities are derived from measured inputs, F(t) and S(t), the facular brightening, and sunspot darkening as functions of time. If we can produce a proxy for T(t) from SPS measurements, T_GOES, we can rearrange equation (1) to solve for the variation in TSI due to sunspot blocking, ΔT_S(t).

In this project, our measurement is a proxy for TSI, not S(t), so we need to do some simple algebra to solve for this input to the spectral model. The other input, F(t) will come from the Mg ii measurement from the Extreme Ultraviolet Sensors (EUVS) component of EXIS (Snow et al., 2009). The spectrum at each wavelength is a linear function of a constant plus terms proportional to the variation in time of the facular brightening and sunspot darkening relative to the spectrum at solar minimum.

TSI is also a linear combination of variations due to facular brightening and sunspot darkening. Instead of deriving a model TSI value, we use the measured SPS TSI proxy to determine the sunspot darkening factor and then use that quantity to calculate the SSI at each wavelength.

A fundamental assumption of the Coddington et al. (2016) model is that the variation in time is not a function of wavelength. Different wavelengths have differing variation magnitudes, but they all vary on the same timescales.

2 Data sources

2.1 Total and spectral irradiance sensors (TSIS-1)

Two instruments on TSIS-1, SIM (Richard et al., 2020) and TIM (Kopp and Lawrence, 2005) will be used as calibration standards in the analysis presented here. The TSIS-1 mission began taking data in March 2018 and continues through the present.

2.1.1 Spectral irradiance monitor (SIM)

Richard et al. (2020) describe the TSIS-1/SIM instrument and calibration. It has heritage from the SIM on SORCE, with several critical improvements. The design of the instrument is a Féry prism spectrometer with several detectors that each report a separate wavelength range. Each day, the full spectrum from ∼200–2400 nm is observed and made available through the LASP Interactive Solar Irradiance Datacenter.¹

TSIS-1/SIM was calibrated extensively before flight, and its calibration is maintained on orbit by comparison to redundant observing channels. The SORCE/SIM instrument had only two channels for degradation analysis (Harder et al., 2005), while TSIS-1/SIM has three. The A and B channels are compared weekly, and the comparison to channel C is done semi-annually on the equinoxes (Richard et al., 2024). The degradation correction for TSIS-1/SIM is much more accurate than the SORCE/SIM algorithm. Uncertainty in the integrated spectrum will be discussed in Section 4. Our analysis used version 11 of the SIM data product. Version 12 has been released, but it is an interim release that primarily deals with updating the degradation correction for wavelengths outside the SPS bandpass.²

2.1.2 TSIS-1/TIM

The TIM instrument on TSIS-1 is the same design as SORCE/TIM (Kopp and Lawrence, 2005). Both instruments measure TSI with an electrical substitution radiometer. In simple terms, sunlight heats a blackened cone, while another cone is heated by electrical power. The amount of power required to bring the two cones into thermal equilibrium is the radiant power of the Sun in Wm⁻². This thermal process is relatively slow, requiring stabilization periods of almost 1 min (Kopp and Lawrence, 2005).

The observing pattern for TSIS-1/TIM is to expose the active cavity for a 50 s integration, then close its shutter for 50 s to equilibrate to the reference cavity. The 50 s on, 50 s off pattern repeats during the solar observing period of each spacecraft orbit. TSIS-1 is mounted on the International Space Station in Low Earth Orbit, so there are eclipse periods during each 92-min orbit. This intermittent data is then averaged over 6-h or 24-h intervals to produce the public TSI data product.^3,4 Version 4 was released in 2023 and is the most current public data release.

2.2 EXIS and SPS

There is an EXIS instrument onboard each of the Geostationary Operational Environmental Satellite R-series (GOES-R; Sullivan, 2020). The first EXIS became operational in early 2017 on GOES-16. Two additional satellites in the GOES-R series are currently in orbit, and the fourth member of the series was launched in June 2024. All data shown in this article is from GOES-16/EXIS. One of the instruments included in EXIS is the SPS. SPS is designed to monitor the position of the Sun in the field of view of EXIS, and its primary data product is the angular position of the Sun.

The SPS is a simple instrument with no moving parts. It consists of a limiting aperture and a series of filters in front of a silicon four-quadrant diode. The filter stack is one F2-G12 radiation-hard glass followed by two Schott neutral density filters: NG-2 and NG-3. The transmittance of the two filters has the same dependence on wavelength since they are simply two slabs of NG-3 glass of different thicknesses. These filters reduce the signal reaching the diode and also limit the short wavelength SSI that would damage the detector over time. The wavelength bandpass of the filters and the responsivity curve for the silicon diode are shown in Figure 1. The transmittance of the F2 G12 radiation hard glass is shown in green. Responsivity of the silicon diode and the transmittance of the NG-3 neutral density filter are shown in orange and purple respectively. The modeled SPS is produced by applying these functions to the TSIS-1/SIM spectrum and then integrating over wavelength. No end-to-end preflight calibration of the SPS was performed because it was not one of the required science channels of EXIS. These curves are from the manufacturer’s datasheets, not from preflight laboratory measurements. Therefore, the curves should be taken as qualitatively correct but not quantitatively accurate.

Figure 1 Modeled GOES-R/EXIS SPS signal. (left) The transmittance of the F2 G12 radiation hard glass is shown in green. Responsivity of the silicon diode and the transmittance of the NG-3 neutral density filter are shown in orange and purple, respectively. The modeled SPS is produced by applying these functions to the TSIS-1/SIM spectrum and then integrating over wavelength. (right) The blue curve is a typical full spectrum from TSIS-1/SIM, in units of spectral irradiance [Wm−2 nm−1]. The modeled SPS response is shown as the red curve. Filtered SSI, integrated over wavelength, is the modeled SPS signal (Eq. 3).

Figure 1 also includes the typical responsivity of the AXUV100 silicon diode detector.⁵ The curve labeled “Filtered SSI” is the TSIS-1/SIM spectrum multiplied by the filter transmittances and the responsivity of the diode detectors. The short wavelengths of the solar spectrum are blocked by the F2-G12 glass, and the long wavelengths are outside the silicon diode sensitivity range. The signal produced by the SPS is due entirely to the near-UV, visible, and near-IR bands plus a very small contribution from detector dark current (about 1 part in 30,000).

Equation (3) shows the analytical components of the SPS signal: A_slit is the area of the entrance aperture, I_solar is the incident solar irradiance, T_F2-G12 and T_NG-5 are the relative transmission of the two filters. QE is the quantum efficiency of the sensor. These quantities are strong functions of wavelength, and the product must be integrated over wavelength to produce the net signal. Figure 1 shows the wavelength dependence of these quantities.

$$\mathrm{Signal}\left(t\right)=A_{\mathrm{slit}}{\int }_{{\lambda }_1}^{{\lambda }_2}\mathrm{}{I}_{\mathrm{Solar}}(\lambda ,t)T_{\mathrm{F}2-\mathrm{G}12}\left(\lambda \right)T_{\mathrm{NG}-5}\left(\lambda \right)\mathrm{QE}\left(\lambda \right)\mathrm{d\lambda }.$$(3)

Before flight, the SPS was roughly calibrated to ensure that it would be sensitive enough to produce a high signal-to-noise measurement of the Sun’s position on the four quadrants. A quantitative calibration over wavelength was not required for the SPS data product; therefore, a full preflight calibration was not performed. Laboratory calibration of the QE of a diode is typically measured as output current, Amps, as a function of input optical power at a given wavelength, i.e., A/W. A follow-up article will describe the in-flight corrections for solar distance, temperature variations, and cross-calibration of the SPS to TSI.

3 Integrated spectrum comparison to TSI

The goal of this article is to verify that the integrated subset of SSI measured by the SPS captures the variability of the full spectrum observed by TSI radiometers. Validation will occur in two steps. The first step is to show that the integrated TSIS-1/SIM spectrum reproduces TSI. Harder et al. (2022a,b) has shown that the integrated SORCE/SIM spectrum tracks SORCE/TIM TSI with a 1σ uncertainty of 0.218 Wm⁻², or 166 parts per million (ppm). Richard et al. (2024) have shown that the integrated TSIS-1/SIM spectrum agrees with TSIS-1/TIM TSI to within 0.068 Wm⁻² over the full mission. They found a∼52 Wm⁻² offset between their integrated SSI and TSI.

The second step is to show that the integrated SSI model for the SPS (iSPS) signal reproduces the variation of the full integrated spectrum from TSIS-1/SIM (iSSI). Figure 2 shows the TSIS-1/TIM TSI measurement, iSSI, and the modeled iSPS signal as a function of time for selected time ranges. iSSI and iSPS have been scaled to match the TSIS-1/TIM calibration. The four panels of Figure 2 were chosen to display the agreement between TSI, iSSI, and iSPS over solar rotation timescales. All three quantities are daily averages. Comparisons on shorter timescales will be shown in Section 5.

Figure 2 Time series of TSIS-1 observations: TSI (⋆), integrated SIM spectrum (iSSI, ), and the integrated model SPS spectrum (iSPS, ) highlighting the agreement on solar rotational timescales throughout the TSIS-1 mission.

The TSIS-1/SIM spectrum includes all wavelengths from ∼200 to 2400 nm, which is 96% of the TSI (Richard et al., 2024). The ∼52 Wm⁻² that falls outside this wavelength range is primarily from the long-wavelength end of the spectrum (i.e., the infrared). That part of the spectrum has very little variability (Fontenla et al., 2011; Coddington et al., 2023). The short wavelength end of the spectrum has strong variability but contributes very little power to TSI (Woods et al., 2009). The contribution to the iSPS uncertainty by neglecting the UV part of the spectrum will be discussed in Section 4.

TSIS-1/SIM (V11) spectrum integrates to 1309.35 Wm⁻², leaving ∼52 Wm⁻² unmeasured. Applying the filter transmittances and silicon diode detector responsivity shown in Figure 1 to SSI produces 43.31 Wm⁻² nm⁻¹. This calibration factor converts the measured diode current (in Amps) to SSI (in Wm⁻² nm⁻¹).

4 Error estimates

Figure 3 shows the time series of the daily differences for iSSI () and iSPS (). The irradiance difference for iSSI is systematically higher than iSPS in the early (2018–2020) and late (2021.5–2023) epochs, but is lower in 2020–2021.5. Richard et al. (2024) describe some of the operational challenges of TSIS-1 that explain why the trends vary by ±0.2 Wm⁻² over the mission. It should be noted that these artifacts are in version 11 of the SIM data and are likely to be corrected in future data releases. The V12 release notes⁶ discuss ongoing improvements to the SIM degradation model, which are particularly important for long wavelengths that fall outside the SPS bandpass.

Figure 3 Time series of difference in irradiance between TSIS-1/TSI and the two integrated spectral quantities in Wm−2: iSSI () and iSPS ().

The histogram of differences in integrated irradiance from iSSI and iSPS to TSI is shown in Figure 4. Gaussian fits to the histograms are shown in Figure 4 and were used to determine the standard deviations. The standard deviation of TSI—iSSI for the TSIS-1 instruments is 0.080 Wm⁻² (1−σ). This is less than a third of the full-mission SORCE/SIM integrated difference of 0.22 Wm⁻² (Harder et al., 2022b). The TSIS-1/SIM has lower noise overall, and the correction for long-term trends has the benefit of a third spectral channel (Richard et al., 2020). So, it is reasonable that the deviation of TSIS-1/SIM’s iSSI from TSI would be smaller. The SORCE/SIM deviation is based on the full SORCE mission, including the time period in the early mission where the degradation was changing rapidly and instrument operations were also being modified. Both of those factors can contribute to a larger uncertainty in the SORCE/SIM measurement (Béland et al., 2022).

Figure 4 Distribution of irradiance difference between TSIS-1/TSI and the integrated spectra: iSSI and iSPS.

The deviation of iSPS from TSI is slightly smaller than the deviation for the iSSI. Since iSPS is a subset of iSSI, it is reasonable to ascribe the larger deviation to wavelengths in the daily SSI data product outside the iSPS band. There are occasional data dropouts in the full spectrum, and these will lead to greater uncertainty in iSSI on those days. Wavelengths outside the SPS bandpass produce the double-peaked shape in the irradiance difference histogram which is not present in the iSPS histogram.

The linear correlation between iSPS and iSSI is shown in Figure 5. The line of best fit is shown along with the 95% confidence interval. It is difficult to see the confidence interval because it has an average value of 0.0105 Wm⁻², so it is barely larger than the width of the line. The χ² for this model is 0.0954, also indicating that the linear model is acceptable.

Figure 5 Linear correlation between iSSI and iSPS. The line of best fit is shown with an R2 of 0.942. The 95% confidence interval has an average value of 0.0105 Wm−2, so it is difficult to distinguish from the width of the line. The line with slope = 1 is shown as a dashed curve. The best-fit line has a slope less than one due to the decrease in the difference after 2021 shown in Figure 3.

TSI measurements integrate the solar spectrum at all wavelengths from the extreme ultraviolet to the infrared. iSSI derived from the TSIS-1/SIM spectrum includes only ∼200–2400 nm. The short wavelength part of the Sun’s spectrum certainly does show daily and solar cycle variability (Woods et al., 2022), but how much does UV variability contribute to TSI variability? Integrating the Whole Heliosphere Interval reference spectrum (Woods et al., 2009), from 0.05 to 200 nm yields 0.103 Wm⁻². Figure 7 of Woods et al. (2022) shows that the integrated variability of the 0–200 nm range contributes less than 0.1% of the total variability.

Figure 4 indicates that the 1−σ uncertainty of a daily-averaged iSPS measurement is 0.072 Wm⁻², i.e., 53 ppm. Kopp (2014) shows the standard deviation of SORCE/TIM daily TSI values during the solar minimum between solar cycles 23/24 to be 17 ppm. If the solar variation during that time period were negligible, 17 ppm would be the upper limit to the measurement uncertainty for SORCE/TIM. During that time period, the Variability of solar IRradiance and Gravity Oscillations (VIRGO) aboard the SOlar Heliospheric Observatory (SOHO) and the Advanced Cavity Radiometer Irradiance Monitor (ACRIM) on ACRIMSat TSI measurements had standard deviations of 32 and 39 ppm respectively (Kopp, 2014).

Table 1 shows the 1−σ standard deviations for TSIS-1/TIM TSI, iSSI, and iSPS during solar minimum 24/25 (1 October 2019 to 1 February 2020) in Wm⁻² and ppm. The variation in the TSIS-1/TIM TSI during the 24/25 solar minimum is similar to the SORCE/TIM observations during the previous minimum. SORCE/TIM measurements⁷ during the 24/25 minimum time range have a standard deviation of 0.05 Wm⁻² . The standard deviations of the integrated spectra during the 24/25 minimum are 0.039 Wm⁻² for iSSI and 0.050 Wm⁻² for iSPS. The uncertainty in the iSPS during this time period is the same as the SORCE/TIM uncertainty.

5 Comparison of SPS to other TSI measurements

If the model of the SPS measurement described in Section 3 is correct, then variations in the GOES/EXIS SPS data should agree with variations observed by a TSI radiometer. The long-term trends in the SPS data are dominated by instrument degradation, so no meaningful comparison can be made on annual or solar-rotation timescales. Short-term variations, i.e., less than 24 h, can be relevant for space weather applications. There are few published TSI datasets that have a cadence of less than one sample per 6-h interval. We present comparisons of three.

VIRGO (Fröhlich et al., 1997) has been measuring TSI since 1996. One of the public data products is a 1-min cadence TSI. The full dataset is available from the SOHO archive⁸ as well as a ftp archive from the Physikalisch-Meteorologisches Observatorium Davos (PMOD). While this relatively high-cadence dataset is useful for validation of the SPS TSI proxy, it has extremely high latency, so it would not be suitable as a replacement for SPS in this project.

The Digital Absolute RAdiometer (DARA) aboard the FY3E satellite takes TSI observations at a 1-min cadence (Montillet et al., 2024) during the solar period of the satellite’s orbit. The standard data product is a 6-h average of these integrations. The DARA team has made a preliminary version of the 1-min data taken during July 2022 available to us for comparison.

The TSIS-1/TIM data also takes roughly 1-min integrations during solar observing windows. We include these “level two” observations as well as the public 6-h averages. The data used in this article will be made available to the public.

The latency of the DARA and TSIS-1/TIM data is currently several days, but in principle, it could be made available more promptly. This initial comparison shows data from July 2022. The left panel of Figure 6 shows the data from all instruments for the entire month. It is useful to see the magnitude of the rotational variability tracked by all instruments.

Figure 6 In both panels, the data from SPS is shown as gray dots, and a 5-min median filter of the SPS data is shown as a blue curve. DARA 1-min observations are purple triangles, the orange curve is a 5-min median of the VIRGO observations, and the TSIS-1/TIM data is shown as red diamonds (1-min observations) or a black diamond (6-h average). (left) Datasets for the entire month of July 2022. (right) Datasets for a single day (21 July 2022).

Figure 6 includes offsets to the DARA and VIRGO TSI values to put them on the SI scale that is used by TSIS-1/TIM. The DARA data is calibrated to the WRR scale (Montillet et al., 2024) used by European instruments and differs from the SI scale by a constant offset.

21 July was chosen as a day that shows larger than-average variations in TSI over a few hours in DARA. The TIM and VIRGO data show a steady decline over the day but no significant variations on timescales less than an hour. Late in the day, the DARA samples show a statistically significant deviation from the TIM and VIRGO datasets. The SPS data follows the deviation of the DARA measurements. Unfortunately, no DARA measurements were taken in the middle of the day, so the downward excursion by SPS cannot be validated. Not every day in the July 2022 dataset shows such clear agreement between DARA and SPS, so this should be taken as a best-case scenario.

6 Discussion: sources of SPS instrument degradation

The iSPS model described in this article does not take account of any instrumental artifacts that will be present in the GOES-16/EXIS SPS data. We will briefly mention several sources of signal degradation.

One assumption is that the spectral shapes of the filter transmittances and the diode responsivity do not vary with time. The integrated signal of the SPS does show degradation at a rate of about 3% over five years, but it is not possible to track spectrally-dependent changes at the component level since only the total throughput is recorded.

The initial design of the SPS used a different wavelength limiting filter (Hoya CM500). Preflight radiation testing on the CM500 showed some darkening, but a quantitative calibration was not available. The change to the F2-G12 radiation-hard glass was intended to improve the protection for the other components. Wirtenson and White (1992) discuss the degradation of radiation hard glass in orbit and indicate that the spectral shape of the transmittance does not change significantly. White and Wirtenson (1993) tested samples of radiation-hard glass and found very little darkening up to 1 Mrad of exposure.

Additional sources of SPS degradation could be the deposition of contaminants due to spacecraft thruster firings and outgassing from thermal blankets. The harsh radiation environment in geostationary orbit could also be degrading the detector surface or the electronics. There is no onboard correction for degradation of the SPS signal. The long-term calibration is maintained by cross-calibration with the TSIS-1/TIM TSI record.

7 Conclusions

We have verified that an integrated model spectrum that applies the transmittances of the filters and the responsivity of the AXUV100 silicon diode detector used in the GOES-R/EXIS SPS to the integrated solar spectrum measured by TSIS-1/SIM creates a good proxy for TSI. Previous studies have shown that the integrated full spectrum from SORCE or TSIS-1 SIM is also a good proxy for TSI. Each integrated spectrum can serve as a proxy for TSI with an uncertainty of less than 60 ppm. The wavelength range captured by SPS includes the part of the solar spectrum, which varies with a high correlation to TSI.

The benefit of using the SPS as a proxy for TSI is that GOES-R SPS is an operational measurement with a high cadence. TSI is typically reported as a daily or 6-h average. Short term TSI variation (Woods et al., 2004) can easily be missed by non-operational measurements from low Earth orbit.

The ultimate goal of producing a high-cadence proxy for TSI is to create a high-cadence empirical SSI model based on the Coddington et al. (2016) model that is the NOAA CDR for solar irradiance. That model uses sunspot darkening and facular brightening to model TSI and SSI. Our reformulation uses the measured TSI proxy to infer the sunspot darkening. The facular brightening input to the model will use the operational Mg II index measured by the EUVS-C, which is also on GOES-R/EXIS.

Acknowledgments

This work was supported by the South African National Research Foundation SARChI grant to SANSA and the GOES High cadence Operational Total Irradiance (GHOTI) grant NA20NES4400006 from NOAA to the University of Colorado Boulder. An undergraduate Honours student, Wesley Sephai, was supported by the South African National Astrophysics and Space Science Programme at North-West University. The authors thank Andrew Jones (LASP) for insightful discussions about the AXUV100 characteristics. The editor thanks Serena Criscuoli and two anonymous reviewers for their assistance in evaluating this paper.

References

All Tables

All Figures

Figure 1 Modeled GOES-R/EXIS SPS signal. (left) The transmittance of the F2 G12 radiation hard glass is shown in green. Responsivity of the silicon diode and the transmittance of the NG-3 neutral density filter are shown in orange and purple, respectively. The modeled SPS is produced by applying these functions to the TSIS-1/SIM spectrum and then integrating over wavelength. (right) The blue curve is a typical full spectrum from TSIS-1/SIM, in units of spectral irradiance [Wm−2 nm−1]. The modeled SPS response is shown as the red curve. Filtered SSI, integrated over wavelength, is the modeled SPS signal (Eq. 3).

In the text

	Figure 2 Time series of TSIS-1 observations: TSI (⋆), integrated SIM spectrum (iSSI, ), and the integrated model SPS spectrum (iSPS, ) highlighting the agreement on solar rotational timescales throughout the TSIS-1 mission.
In the text

	Figure 3 Time series of difference in irradiance between TSIS-1/TSI and the two integrated spectral quantities in Wm−2: iSSI () and iSPS ().
In the text

	Figure 4 Distribution of irradiance difference between TSIS-1/TSI and the integrated spectra: iSSI and iSPS.
In the text

	Figure 5 Linear correlation between iSSI and iSPS. The line of best fit is shown with an R2 of 0.942. The 95% confidence interval has an average value of 0.0105 Wm−2, so it is difficult to distinguish from the width of the line. The line with slope = 1 is shown as a dashed curve. The best-fit line has a slope less than one due to the decrease in the difference after 2021 shown in Figure 3.
In the text

Figure 6 In both panels, the data from SPS is shown as gray dots, and a 5-min median filter of the SPS data is shown as a blue curve. DARA 1-min observations are purple triangles, the orange curve is a 5-min median of the VIRGO observations, and the TSIS-1/TIM data is shown as red diamonds (1-min observations) or a black diamond (6-h average). (left) Datasets for the entire month of July 2022. (right) Datasets for a single day (21 July 2022).

In the text