© W.E. McClintock et al., Published by EDP Sciences 2025

1 Introduction

EUVS-C (see the companion paper by McClintock et al., 2025) is one component of the Extreme Ultraviolet Irradiance Sensor (EUVS) instrument (Eparvier et al., 2009). EUVS, together with the X-ray sensor (XRS, Woods et al., 2024), comprise the Extreme Ultraviolet and X-ray Irradiances Sensors (EXIS) investigation (Machol et al., 2020) aboard the GOES-R satellite series, which includes GOES-16, -17, -18, and -19, located in geostationary orbit. Two satellites are operational at one time, and from their vantage points at 75.2° west longitude (GOES-16 and GOES-19) and 137.2° west longitude (GOES-17 and GOES-18), EUVS-C channels observe a narrow range of middle ultraviolet (MUV) wavelengths, which is centered around the MgII k and h emission lines (vacuum values of 279.64 nm and 280.35 nm), with Δλ ~ 0.1 nm spectral resolution, with better than 0.1% precision, and with a 3-second cadence. Whereas emission in the centers (cores) of the k and h lines originates in the chromosphere and exhibits significant short-term (diurnal and 27-day solar rotation time scales) and long-term variability (11-year solar cycle time scales) associated with solar activity, emission from nearby wavelengths originates in the photosphere and is much less variable. Comparing solar irradiances from these two different spectral regions provides a method for removing instrument responsivity changes with time that often corrupt the interpretation of the h and k core emission variability.

Heath & Schlesinger (1986) pointed out that emission at wavelengths a few nm distant from line centers originates in the sun’s photosphere and is, therefore, much less variable than that from the chromosphere. Using 1.1 nm spectral resolution data from the Solar Backscattered Ultraviolet (SBUV) experiment on Nimbus 7, they defined a “MgII Index” as the ratio of solar irradiance in the cores of the lines to that of two wavelengths on either side of line center,

$${\mathrm{ }\mathrm{MgII}}_{\mathrm{SBUV}}=\frac{4·[I_{279.8}+I_{280.8}+I_{280.2}]}{3\cdot [I_{276.6}+I_{276.8}+I_{283.2}+I_{283.4}]}.$$(1)

This index, after appropriate scaling, successfully represented both the short-term variability and long-term variability for far ultraviolet (<204 nm) solar spectral irradiance. Because the MgII index is a ratio measurement, it is largely, but not completely, insensitive to long-term changes in instrument sensitivity (Snow et al., 2019). This is an important advantage over direct measurement of solar spectral irradiance variability, particularly at ultraviolet wavelengths.

The 1.1 nm spectral resolution available in the SBUV instruments was not sufficient to resolve the individual k and h emission cores. Additionally, it has relatively low contrast (I_MAX − I_MIN)/I_Mean, where I_MAX, I_MIN, and I_Mean are maximum, minimum, and mean values, respectively) that arises because the numerator in equation (1) contains both the chromospheric core emission and a significant contribution from the wings of the photospheric line. Using 0.24 nm resolution data from the SOLar-Stellar Irradiance Comparison Experiment (SOLSTICE) investigation aboard the Upper Atmosphere Research Satellite (UARS, Rottman et al., 1993), De Toma et al. (1997) and White et al. (1998) explored the effects of increased resolution on the properties of the MgII index. They defined a new index where the numerator contained the total emission in two 0.36 nm wide wavelength bands centered on h and k, and the denominator was the average of the maxima of parabolic fits to 4 0.5 nm-wide bins in the wings centered on 276.1 nm, 276.6 nm, 283.2 nm, and 283.9 nm. This index, which contained smaller contributions from the photosphere, resulted in a factor of 2.2 increase in contrast and captured more solar activity during the 1996 solar minimum. White et al. (1998) concluded that both indices showed a reduction in rotational variability near solar minimum. Following UARS, instruments with increasingly high spectral resolution were developed. These include the SCanning Imaging Absorption spectroMeter for Atmospheric CartograpHY (SCIAMACHY) with Δλ = 0.21 nm (Bovensmann et al., 1999), the Global Ozone Monitoring Experiment (GOME) with Δλ = 0.17 nm (Burrows et al., 1999) and the SOLar-STellar Irradiance Comparison Experiment II (SOLSTICE II) with Δλ = 0.1 nm (McClintock et al., 2005a) aboard the Solar Radiation and Climate Experiment (SORCE) spacecraft (Rottman, 2005). McClintock et al. (2025) estimated that the index calculated from 0.1 nm SOLSTICE II data has a 10–20% contribution from the photosphere, depending on solar activity.

The index defined by Snow et al. (2005) for SOLSTICE II, which formed the basis for the EXIS investigation measurement requirements, is

$${\mathrm{MgII}}_{\mathrm{SORCE}/\mathrm{SOLSTICE}}=\frac{2·[I_{\mathrm{k}}+I_{\mathrm{h}}]}{[I_{276.6}+I_{276.8}+I_{283.2}+I_{283.4}]}.$$(2)

Wing irradiances were calculated by convolving the full-resolution spectrum with a triangular, 1.1 nm full-width half maximum (FWHM) response function sampled at 0.2 nm. These were summed at the 4 SBUV wavelengths. Core radiances were computed by fitting Gaussian functions to h and k, respectively, and integrating them. The effective integration range was approximately ±0.08 nm about each line center, and the result contained any residual emissions from the photosphere in those bands.

Snow & McClintock (2005) used SOLSTICE II data to show that MgII_{SORCE/SOLSTICE} fluctuates ~1%, 1 − σ, in a typical 24-hour period. They also found index variability of ~0.2% on time scales of 5–10 min, increasing to ~0.3% and ~0.55% for 30 and 80 min, respectively. They suggested that daily variation might be attributed to active regions rotating in and out of view but that those seen on shorter timescales likely represent true variability within the active regions and chromospheric network that give rise to the observed emission. Thus, the ideal instrument for measuring the full intrinsic variability of MgII chromospheric variability on timescales of tens of minutes and longer should be designed to achieve Δλ ~ 0.1 nm spectral resolution with ~0.1% precision with a 1-minute cadence or better. These constraints set the primary measurement objectives for EUVS-C.

McClintock et al. (2025) discuss the EUVS-C definition, design, ground calibration, and predicted flight performance. Here, we provide a summary of the algorithms described by McClintock et al. (2025) that are used to calculate the MgII_EXIS index and their operational implementation. This is followed by an analysis of flight calibration experiments, which demonstrate that EUVS-C meets or exceeds its measurement objectives for spectral resolution, precision, and cadence. Systematic errors, which affect index accuracy on timescales of hours to years, arise from wavelength scale shifts that are not corrected by the operational algorithms used to calculate the index and from spectrally dependent sensitivity loss caused by solar exposure. These can be mitigated using a wavelength shifting technique coupled with the operational algorithm (Sects. 4.2.1 and 4.2.2) and by monitoring instrument radiometric sensitivity loss with quarterly calibration observations (Sect. 4.2.4). Comparing index measurements for GOES-16, -17, and -18 validates the precision and accuracy performance of all three instruments.

2 MgII_EXIS index definition and algorithm implementation

2.1 EXIS index definition

EUVS-C observations consist of 512-element (pixel) spectra acquired at a 3-second cadence with an approximate wavelength range and spectral dispersion of 274–285 nm and ~0.022 nm/pixel, respectively (see Table 1). S_n(j) is defined as the raw signal from a single 3-second observation in pixel j of the nth spectrum reported by the detector in data numbers (DNs) that results from photon events, dark current, and hits from energetic particles, particle-induced gamma rays, and cosmic rays that penetrate the detector shields.

The hits usually appear as spikes in one or two adjacent pixels. These are removed by comparing the signal in the jth spectral element from the nth spectrum, S_n(j), with that from the previous spectrum, S_n−1(j). The logical sequence is

$$\begin{array}{c}\mathrm{If} S_n\left(j\right)-S_{n-1}\left(j\right) \ge \mathrm{ParticleThreshold},\\ \mathrm{then} S_n^{\text{'}}\left(j\right)=S_{n-1}\left(j\right),\\ \mathrm{else} S_n^{\text{'}}\left(j\right)=S_n\left(j\right).\end{array}$$(3)

Based on routine detector dark current monitoring (Sect. 3.2.1), there is no evidence for the persistence of particle hit effects in succeeding spectra. This is partially the result of the readout scheme for the detector that includes two “flushes” at the end of each integration period.

At EUVS-C startup, there is no previous integration, and the particle filtering step is skipped. This simple differencing technique ignores particle hits with values less than the threshold. Additionally, it fails during pointing slews that occur during calibration maneuvers and during eclipse entry and exit. Also, this will not remove successive particle hits in the same detector pixel from the second integration, which is expected only during large particle storms.

S′(j) values are further processed to produce D(j), the raw signal corrected for hits, electrical offset, nonlinearity, and flat-field, which is defined as

$$D\left(j\right)=\left(S^{\mathrm{\prime }}\left(j\right)-D{\left(j\right)}_{\mathrm{Offset}}\right)·\mathrm{FF}\left(j\right)·\mathrm{LIN}.$$(4)

D_Offset is an electrical offset applied to the detector charge-to-voltage amplifier, FF is a flat-field correction for pixel-to-pixel photoresponse nonuniformity, which is typically less than 0.5% across the entire array, and LIN is a correction that accounts for nonlinear effects in the detector photon transfer function. LIN = 1 for S′(j) − D(j)_Offset < 9.0 × 10⁴ DN (Note that the detector output counter in the external electronics saturates at 2¹⁶ DN; therefore, in practice LIN = 1 S′(j) − D(j)_Offset < 6.5535 × 10⁴ DN, McClintock et al., 2025). Figure 1 is a plot of a typical spectrum of D obtained on February 14, 2017, by EUVS-C aboard the GOES-16 satellite.

Figure 1 GOES-16 raw counts (D(j) in equation (4) observed on February 14, 2017, and corrected for particle backgrounds, electrical offset, linearity, and flat field. Dark counts and scattered light have not been subtracted from this spectrum. Pixels 0–60 are masked and used for real-time dark and background correction.

Finally, the detector dark signal and scattered light signal are subtracted to produce the corrected data numbers (D′(j)) that form the basis for MgII_EXIS calculations, which are defined as

$$D^{\mathrm{\prime }}\left(j\right)=\left(S^{\mathrm{\prime }}\left(j\right)-D{\left(j\right)}_{\mathrm{Offset}}\right)·\mathrm{FF}\left(j\right)·\mathrm{LIN}-{D\left(j\right)}_{\mathrm{Dark}}-{D\left(j\right)}_{\mathrm{Scatter}}.$$(5)

Here, D_Dark is the data number count due to dark current within the detector, and D_Scatter is the output resulting from stray and scattered light within the optical system that reaches the detector. Determination of values for the subtracted terms in equations (4) and (5) is discussed in Section 3.

Paralleling the definition of the MgII_{SORCE/SOLSTICE} index, Snow et al. (2009) proposed a MgII_EXIS index definition in which the wing intensities are the weighted sums of pixel-corrected signal data numbers, D′(j). The weighting functions consisted of pairs of trapezoidal functions with FWHM = 1.1 nm at 276.7 nm and 283.3 nm for the blue and red wings, respectively. Core intensities were calculated by fitting Gaussian functions to the k and h line signals and integrating them. This approach was later abandoned to reduce computational complexity.

The current MgII_EXIS operational definition is a further modification of MgII_{SORCE/SOLSTICE} in which the wing weighting functions are trapezoids with FWHM = 110 pixels and full width at base (FWAB) = 150 pixels, centered at 277.4 nm and 282.4 nm, respectively (Snow et al., 2019). Core intensities are calculated by summing signal data numbers in 9 columns centered on the k line and 8 columns centered on the h line (rectangular weighting functions with all values equal to one).

Whereas the SBUV and SOLSTICE blue wing and red wing weighting functions have equal FWHM = 1.1 nm, the blue and red wings for EUVS-C, which are defined in pixel space, are 2.33 nm and 2.25 nm FWHM, respectively in wavelength space due to the nonlinear dispersion of the EUVS-C wavelength scale (Sect. 3.2.2). The EUVS-C weighting functions are plotted in Figure 2 along with a GOES-18 EUVS-C observation of fully corrected counts (Eq. (5)). Also, in contrast to SOLSTICE, EUVS-C corrected data numbers (Eq. (5)) are not radiometrically calibrated before the index calculation. This changes the relative weightings of the red and blue wings to the denominator.

Figure 2 GOES-18 EUVS-C observation made on June 17, 2022. Solid blue and red lines mark the GOES trapezoidal weighting functions (right ordinate). Rectangles, which are 9 pixels wide and 8 pixels wide, respectively, represent the weighting functions for the k and h emission lines used in the current operational algorithm implementation.

2.2 Algorithm implementation

In the current algorithm implementation, signal data numbers in the jth pixel are calculated using modifications of equations (4) and (5) in which FF = 1 and LIN = 1. These assumptions are based on the ground calibration and characterization and were validated during the post-launch instrument checkout as described in Section 3. These assumptions lead to the definition

$$D^{\text{'}}\left(j\right)=S^{\text{'}}\left(j\right)-{D\left(j\right)}_{\mathrm{Offset}}-{D\left(j\right)}_{\mathrm{Dark}}-{D\left(j\right)}_{\mathrm{Scatter}}.$$(6)

The corrected signal values are used along with the weighting functions to calculate MgII_EXIS as the ratio of the sums of weighted average counts in the h and k lines to the sums of weighted average counts in the blue and red wings as follows,

$${\mathrm{MgII}}_{\mathrm{EXIS}}=\frac{{D^{″}}_{\mathrm{h}}+{D^{″}}_{\mathrm{k}}}{{D^{″}}_{\mathrm{BlueWing}}+ {D^{″}}_{\mathrm{RedWing}}}.$$(7)

$$\begin{array}{cc}{D^{″}}_x =\frac{\sum _j{D}^{\text{'}}\left(j\right)·{W\left(j\right)}_x}{\sum _j{W\left(j\right)}_x},& x=\mathrm{h}, \mathrm{k}, \mathrm{Blue}, \mathrm{and} \mathrm{Red}.\end{array}$$(8)

The precise pixel locations of the weighting functions (also referred to as masks) were determined from solar observations made during the instrument commissioning period, which occurred before the beginning of nominal operations. During the operational implementation of the MgII_EXIS algorithm, there is no adjustment in mask location in pixel space on either diurnal or seasonal time scales to account for changes in the wavelength scale caused by pointing errors, Doppler shifts resulting from spacecraft orbital velocity (approximately ±3 km/s resulting in a ±0.13-pixel diurnal shift), and optical element displacements caused by changing thermal gradients within the instrument. The impacts of these simplifying assumptions are discussed in Section 4.2.

3 Flight operation and calibration

3.1 Flight operations

GOES-16, -17, -18, and -19 were launched on November 19, 2016, March 1, 2018, March 1, 2022, and June 26, 2024. Their current operational orbit longitudes are 75.2° west (GOES-16 and GOES-19) and 137.2° west (GOES-17 and GOES-18). GOES-16 and GOES-18 are operational. GOES-17 is in storage after operating from February 12, 2019, through March 13, 2023. GOES-19 has recently been commissioned, and initial checkout and calibration observations are described here.

EXIS is mounted on the Solar Pointing Platform (SPP), which is located on the yoke of the GOES satellite solar array. This arrangement enables it to continuously view the sun except for two 6-week long periods each year when the Earth eclipses reduce observing time by up to 72 min per day. During nominal operations, EUVS-C employs a fixed-length observing cycle that is synchronized with the EXIS instrument clock. It begins with a detector science data readout at a rate of 40 μs per pixel (0.02048 s for the entire array). This is followed by an additional pair of readouts that “flush” the detector to minimize detector lag, which is an incomplete pixel reset that is a characteristic of CMOS detectors. Once the triad of read cycles is complete, a new integration cycle begins. The total cycle length is programmable in increments of 0.25 s with a current value of 3 s (integration period of 2.934 s) for all instruments.

3.2 Flight calibration and characterization and ground calibration validation

3.2.1 Integration time, detector dark current, read noise, and electrical offset

Before operations began, solar spectra were collected at a series of integration times ranging from 0.934 s to 6.934 s in quarter-second increments to determine the nominal operating value of 2.934 s, which provided near maximum detected counts (approximately 5.5 × 10⁴ DN/2.934 s/pixel) that avoids detector counter digitization saturation (2¹⁶ – 1 DN). The charge collected for this signal level is approximately 13 picocoulombs (pC), well below the 24 pC pixel capacity (McClintock et al., 2025); therefore, the detector linearity term (e.g., Eq. (4) is LIN = 1).

Detector dark current (D_Dark) and offset (D_Offset) were measured just after launch by varying the integration time with the EUVS filter wheel (McClintock et al., 2025) set to a blocked position for EUVS-C. Additionally, each EUVS-C is equipped with a flat-field lamp that provides smoothly varying illumination of the detector (McClintock et al., 2025). It was switched on while the filter wheel was in the blocked position. Lamp measurements indicated that each detector exhibited pixel-to-pixel variations <0.4% everywhere and <0.1% over 90% of its area; therefore, similar to LIN = 1, FF = 1 is assumed for radiometric calculations. The filter wheel-lamp sequence is performed weekly to measure the detector flat field response and to track changes in D_Dark, D_Offset, and flat-field that arise from energetic particle exposure in geostationary orbit. These measurements indicate that for all detectors the initial dark current is approximately 2.75 DN/s and that it increases by approximately 0.1 DN/s/year. Initial values of D_Offset were also consistent with ground measurements. On approximately May 15, 2017, GOES-16 exhibited a sudden 2 DN jump in its D_Offset baseline, followed by an increase with time at a rate of approximately 0.5 DN/s/year. A 0.5 DN/s/year baseline increase is also observed in the GOES-17 and GOES-18 offsets but without the jump. Effects of changing D_Offset levels on D′(j) are removed using real-time measurements of masked pixels (Sect. 4.1.1). Read noise values, extrapolated from the dark current measurements were, approximately 2.33 DN (3500 e-). These are consistent with ground measurements (McClintock et al., 2025) and have not changed throughout the missions.

The difference between the filter wheel open and blocked, which is plotted in Figure 3, provides a measurement of light from the Sun that reaches 60 masked pixels on the short-wavelength side of the detector (McClintock et al., 2025) as it is reflected multiple times between the front surface of the detector and the surfaces of the detector window in a “walk” pattern. This declines to a constant value within detection limits by pixel 11. The non-zero values for pixels 0–10 are assumed to be a measure of stray light inside the detector cavity due to sunlight entering the instrument.

Figure 3 GOES-16 EUVS-C signal rate in the masked pixels (0–59) for the average of 1000 standard 2.934 s integrations on February 14, 2017. Counts in pixels 11–59 arise from sunlight multiply reflected between the front surface of the detector and the surfaces of its window in a “walk” pattern. The dashed line is the mean of the first 10 pixels (1.30 DN/s).

3.2.2 Wavelength scale

Flight wavelength scales were determined by comparing radiometrically calibrated spectra (Sect. 3.2.5) taken during standard operations with a single SOLSTICE II spectrum. Assuming they can be described using a second-order polynomial (McClintock et al., 2025) that approximates the sine terms in the standard grating equation leads to the expression

$${\lambda }_N={\lambda }_0+A_1·N+A_2·N^2.$$(9)

where N is the detector pixel number (0–511). Coefficients for equation (9) are summarized in Table 1.

The agreement between EUVS-C and SOLSTICE is better than ±0.008 nm (±0.35 pixels) for GOES-16 and GOES-17 and better than ±0.12 nm (±0.58 pixels) for GOES-18. SOLSTICE accuracy is ±0.0025 nm (McClintock et al., 2005a).

3.2.3 Alignment and pointing effects on wavelength scale

The GOES Sun Pointing Platform is continuously pointed at the Sun by a pair of mechanisms that are controlled by feedback from the Guide Telescope Assembly (GTA) subsystem of the Solar Ultraviolet Imager (Vasudevan et al., 2018), which is coaligned with the EXIS Solar Position Sensor (SPS, McClintock et al., 2025). Solar position within the EXIS FOV is independently measured by its SPS with a precision of 3.3 millidegrees and reported in the EUVS-C Level 1 science data for each 3-second integration period. During instrument commissioning, the alignments of the EUVS-C FOVs relative to the GTA centers were measured for each instrument in the EUVS-C dispersion direction (α) and cross-dispersion direction (β) by sequentially rotating the SPP twice in each axis over a ±2° angular range during a 20-minute period. The results are plotted in Figure 4. Whereas GOES-16 and GOES-17 FOVs are centered within 0.01° relative to their respective GPAs, GOES-18 is offset by Δα = +0.065°, Δβ = +0.07° and GOES-19 is offset by Δα = +0.03°, Δβ = +0.09°. Additionally, the GOES-19 FOV in the cross-dispersion direction is narrow relative to the other instruments by 0.04°. Based on SPS measurements, pointing for EXIS is stable within 0.0033° throughout each of the missions.

Figure 4 EUVS-C field of view relative to the SPP Guide Telescope Assembly as measured after launch. GOES-16 and -17 EUVS-C FOVs are centered within approximately 0.01°. The GOES-18 and 19 EUVS-C FOVs are displaced relative to their GTAs by Δα = +0.065°, Δβ = +0.07° and Δα = +0.03°, Δβ = +0.09°, respectively. Their plots in this figure have been offset by those amounts relative to 16 and 17.

EUVS-C wavelength scale exhibits both stretch and shift as a function of viewing direction (McClintock et al., 2025). These were characterized during the alignment experiments and exhibit essentially identical behavior for all instruments. Spectral displacements can be represented by shift only for β displacements: Δpix = −2.72 · β + 0.81 · β² pixels per degree. Both shift and stretch are required to correct the wavelength for α displacements. For shift, Δpix = −0.45 · α + 1.12 · α² pixels per degree. For a stretch, the wavelength scale expands/contracts by a factor (1 + 0.0106 · α). The effect of wavelength scale displacement on the determination of the MgII index is discussed in Section 4.2.1.

3.2.4 Particle backgrounds

The GOES instruments exhibit nearly identical responses to particle backgrounds, which are removed from the data using the simple sequential replacement algorithm described by equation (3) (Sect. 2.1). The current threshold value used for operational data processing is ParticleThreshold = 17 DN (25,500 e-), which is 2.5 times the maximum noise level (>2.5 σ, see Sect. 4.1.1) in an individual spectrum (Fig. 8). The frequency distribution for events with values >17 DN is approximately exponential, with an e-folding value of approximately 4.3 DN.

On a quiet day (e.g., September 5, 2017), there are approximately 2.5 events per 2.934 s integration when the difference exceeds 17 DN. This number rises to 6.2 events per integration during the major solar storm on September 10, 2017. These events are randomly distributed in wavelength.

Dark events (CMOS array dark current and particle backgrounds) are continuously monitored in real-time every 3 s using masked pixels (1–60). Daily measurements ofthe entire array are performed when the filter wheel moves to the blocked position for calibration. Additionally, a weekly flatfield lamp and filter wheel calibration monitor changes in detector offset, dark current, and flat field (Sect. 3.2.1).

MgII_EXIS values (Eq. (8)) are very weakly dependent on a threshold value. When limits are varied 12 DN < ParticleThreshold < 34 DN, the fractional change in index values is less than ±0.04% for greater than 99% of the typical 28,800 spectra on either quiet or stormy days.

3.2.5 Radiometric performance

Spectra are converted to irradiance using equation (10), and the responsivities plotted in Figure 10 of McClintock et al. (2025) with D_Dark and D_Offset are determined as described in Section 3.2.1,

$${I\left({\lambda }_j\right)}_{\mathrm{SUN}}=\frac{D^{\text{'}}\left({\lambda }_j\right)}{R({\lambda }_j,\theta ,\varphi )·G·\Delta t},$$(10)

$$R\left({\lambda }_j,\theta ,\varphi \right)=R\left({\lambda }_j\right)·F\left(\lambda .\theta ,\varphi \right).$$(11)

I(λ_j) is the solar spectral irradiance in pixel j associated with wavelength (λ_j), reported in photons/cm²/nm/s and R(λ_j) is the instrument responsivity, reported in electrons/(photons/cm²/nm/s) (Fig. 10 of McClintock et al., 2025), G is the detector electron to data number conversion constant (1 DN/1500 e-) and Δt is the integration time (2.934 s). F(λ, θ, φ) is a correction factor whose value is close to 1 that accounts for small variations in R(λ_j) when the Sun is viewed from non-zero azimuth and elevation angles θ and φ, respectively. Based on SPS measurements, which indicate that pointing varies by less than 0.0033° (Sect. 3.2.3), F(λ, θ, φ) is set to 1 in equation (11).

To validate radiometric performance, EXIS-C daily average irradiance values for checkout days (February 14, 2017, May 20, 2018, July 4, 2022, and September 18, 2024, for GOES-16, -17, -18, and -19, respectively, were compared to SOLSTICE daily high-resolution observations, which were obtained from the Laboratory for Atmospheric and Space Physics Interactive Solar Irradiance Datacenter (LISIRD) (https://lasp.colorado.edu/lisirdJ/). The results for GOES-16 are plotted in Figure 5 using SOLSTICE spectra that have ~0.09 nm FWHM spectral resolution (McClintock et al., 2005a) and have been convolved with a triangular profile with a FWHM = 0.025 nm to match that of EUVS-C measurements. GOES spectra were radiometrically calibrated (McClintock et al., 2025) at the National Institute for Standards and Technology’s (NIST’s) Synchrotron User’s Radiation Facility III (SURF III, Arp et al., 2002). The SOLSTICE convolutions improve the agreement with the EXIS data, which have a slightly larger ~0.1 nm FWHM (McClintock et al., 2025).

Figure 5 Comparison of SOLSTICE daily-averaged irradiance on February 14, 2017, with the EUVS-C GOES-16 daily average, shown in blue; the SOLSTICE spectrum has been convolved with a triangular profile with a FWHM = 0.025 nm to match the EXIS-C spectral resolution. SOLSTICE comparisons with GOES-17, -18, and -19 exhibit nearly identical agreement as that with GOES-16.

SOLSTICE observations ended in February 2020, and the comparison with the GOES-18 and GOES-19 observations on July 4, 2022, and September 18, 2024, respectively, used spectra from the same calendar days in 2019. They were scaled in absolute value to 2022 and 2024 using a comparison of observations from the Total and Spectral Irradiance Sensor’s (TSIS) Spectral Irradiance Monitor (SIM) on the days in 2022 and 2024 (Richard et al., 2020) to those from the Solar Radiation and Climate Experiment’s (SORCE) SIM (Harder et al., 2022) on those days in 2019. The accuracy of this approximation is estimated to be better than 1% for the wings of EUVS-C spectra (Woods et al., 2018). It is less accurate for the MgII h and k emission cores.

Figure 6 shows EUVS-C/SOLSTICE spectral ratios, which were computed after linearly interpolating the EXIS-C spectra to the fixed, 0.025 nm-spaced SOLSTICE wavelength scale and binning by 5–0.1 nm sampling. Linear fits to the ratios are also included and have mean values of 0.97, 1.00, and 1.01, respectively. These agreements are consistent with those anticipated from the typical systematic errors of ~2% encountered when intercomparing simultaneous SOLAR irradiance measurements by LASP-built instruments calibrated at SURF III (McClintock et al., 2005b). The significant disagreement in the MgII h and k emissions observed for GOES-18 and GOES-19 is consistent with increased solar activity that is not captured in extrapolation from July 4, 2019 to 2022 and 2024 (Woods et al., 2018).

Figure 6 GOES-16/SOLSTICE (black), GOES-17/SOLSTICE (red), GOES-18/SOLSTICE (blue), and GOES-19/SOLSTICE (orange) sampled at 0.1 nm per spectral bin. Also shown are linear fits to the ratios and a dashed line for Ratio = 1. The ~10% and 25% larger MgII h and k emissions observed for GOES-18 and GOES-19 are consistent with the uncertainty in extrapolating MgII emissions in the SOLSTICE spectra from July 4, 2019 and September 18, 2019 to 2022 and 2024 (Woods et al., 2018).

The standard errors (SEs) for differences between the linear fits and the ratios, defined as standard deviation differences divided by the square root of the number of samples in the fits, are SE = 0.020, SE = 0.021, and SE = 0.029, respectively. These are an order of magnitude larger than values calculated for random errors arising from photon shot noise and most likely arise from pixel-to-pixel registration differences between the SOLSTICE wavelength scale and the interpolated GOES wavelength scales. It is unlikely that there are significant contributions to the SE from uncorrected flat field deviations (Sect. 3.2.1).

The SOLSTICE-to-EUVS-C irradiance comparisons are only valid immediately after the launch of each GOES satellite. EUVS-C radiometric performance monitoring indicates that the sensitivities of all three instruments are degrading at an average rate of ~5%/year (Sect. 4.2.4). The degradation impacts on signal-to-noise and the MgII_EXIS index are discussed in Section 4.2.4.

4 Flight MgII_EXIS index measurement performance

Examples of the GOES-16 and GOES-18 EXIS-C components that comprise the MgII_EXIS index (Eq. (8)) are plotted in Figure 7 for a single arbitrarily chosen day (August 9, 2022). The top panel is D″ values for the red and blue wings, which were calculated from the operational algorithm defined by equation (8) after correcting D′ values (Eq. (6)) for spacecraft-Sun distance and normalizing the median of the resulting D″ values to 1. Bottom panel plots are for corresponding values of the h and k lines and of MgII_EXIS. There is short-term variability (time scales of minutes) in wings that is highly correlated between red and blue, indicating that much of it is solar in origin. This is also true for correlation observed in the h and k line emissions. The instrumental contribution to these short-time scale changes arises primarily from random uncertainties in the individual 3-second observations and are insignificant compared to those from the Sun (Sect. 4.1).

Figure 7 Diurnal relative variability in the GOES-16 and GOES-18 EUVS-C MgIIEXIS index components (Eq. (8)) for August 9, 2022. Diurnal variability in the blue and red wing irradiances arises from Doppler shifts caused by the spacecraft orbital motion that is not corrected in the operational data reduction algorithm (Sect. 4.2.1). Differing satellite longitudes (GOES-16 at 75.2° W and GOES-18 at 137.0° W) cause the profiles of the two instruments to peak at different UTC positions.

The wings also display diurnal variations that are in the order of 0.1% and are out of phase with each other. These primarily arise from Doppler shifts due to spacecraft orbital motion (Sect. 4.2.1) that are not corrected in the operational algorithm, which uses fixed pixel masks to calculate the index components. Because the wings are out of phase, the diurnal variation in the MgII index introduced by their sum (denominator in Eq. (7)) is significantly less than 0.1% (Sect. 4.2.2). Similar variations are also present in the h and k lines, but the strong solar variability in these emissions overwhelms the effect.

GOES pointing performance is sufficiently accurate so that pointing error contributions to both component and index variations are negligible (Sect. 4.2.3).

4.1 Random uncertainty in an individual EUVS-C spectrum and in the MgII_EXIS index

4.1.1 Single spectrum

The analytic estimate for the random uncertainties in the raw counts of a single EUVS-C spectrum after correction for particle backgrounds and electrical offset is

$${\sigma }_{D\left(j\right)}^2={\sigma }_{S\text{'}\left(j\right)}^2=\left[D\left(j\right)·G+5.53\right]\mathrm{ }{\mathrm{DN}}^2.$$(12)

D(j) is the raw data number for pixel j corrected for particle background, linearity, and flat field (Eq. (4)), G is the electron to data number conversion constant (G = 1/1500), and 5.53 = read noise² + digitization noise² reported in data numbers (McClintock et al., 2025).

D′, which is used for the index calculation, is computed by subtracting D_Dark + D_Scatter from D. D_Dark + D_Scatter is estimated from the average of 20 masked pixels (Sect. 3.2.1), D(j) for 5 ≤ j ≤ 24. This approach eliminates any systematic error introduced by baseline changes that have occurred in D_Offset since the beginning of the mission when its initial values were measured. If D(j)_Offset = D(j)_Offset−0 + Δ, then Δ would remain after subtracting D(j)_Offset−0 and the average of pixels 5–24 would then be a measure of Δ + D_Dark + D_Scatter instead of simply D_Dark + D_Scatter. The uncertainty in D_Dark + D_Scatter (and the implicit value of Δ) is

$${\sigma }_{D_{\mathrm{Dark}+\mathrm{Scatter}}}^2={\sum }_{j=5}^{j=24}\left[D\right(j)·G+5.53]/400.$$(13)

Since D(j) ≤ 10 DN for pixels 5–24 (Fig. 3), ${\sigma }_{D_{\mathrm{Dark}+\mathrm{Bkgd}}}^2\sim 5.53/20=0.27$, the uncertainty in D′(j) can be estimated as

$${\sigma }_{D\text{'}\left(j\right)}^2=\left[D\left(j\right)·G+6.0\right]{ \mathrm{DN}}^2.$$(14)

The analytic estimate in equation (12) can be validated using an independent random uncertainty estimate calculated from the variance in the differences of consecutive spectra. On average, the difference of a pair of spectra, D(j, n) − D(j, n − 1), is a single sample from the signal distribution of N spectra multiplied by $\sqrt{2}$ and σ(j, N) = standard deviation $\left(D\right(j,n)-D(j,n-1\left)\right)/\sqrt{2}$, n = 1, N − 1. The difference removes the contribution from solar variability on time scales greater than 3 s. Figure 8 compares the two uncertainty estimates for GOES-16 EUVS-C observations made on February 19, 2017, a day with weak particle background activity. Estimates computed from difference spectra increased by ~0.3 DN on September 10, 2017, after an X8.2 solar flare. This increase is consistent with the inability of the simple operational background removal algorithm that ignores particle hits that produce data spikes that are smaller than ParticleThreshold = 17 DN (Sect. 3.2.4).

Figure 8 The black curve is an analytic uncertainty estimate for a single spectrum observed on February 19, 2017 and calculated from photon counting statistics, and detector read noise and digitization noise (Eq. (12)). The red curve is an estimate based on the differences of consecutive spectra obtained during the entire day. Uncertainties estimated from differencing spectra are 3% larger than those of the analytic calculation.

4.1.2 MgII_EXIS index

The fractional random uncertainty in MgII_EXIS, which is calculated from uncertainties in the D″_x terms (McClintock et al., 2025), is

$$\frac{{\sigma }_{{\mathrm{MgII}}_{\mathrm{EXIS}}}^2}{{{\mathrm{MgII}}_{\mathrm{EXIS}}}^2}=\frac{{\sigma }_{{D^{″}}_{\mathrm{h}}}^2+{\sigma }_{{D^{″}}_{\mathrm{k}}}^2}{{({D^{″}}_{\mathrm{h}}+{D^{″}}_{\mathrm{k}})}^2}+\frac{{\sigma }_{{D^{″}}_{\mathrm{BlueWing}}}^2+{\sigma }_{{D^{″}}_{\mathrm{RedWing}}}^2}{{({D^{″}}_{\mathrm{BlueWing}}+{D^{″}}_{\mathrm{RedWing}})}^2}.$$(15)

$${\sigma }_{{D^{″}}_x}^2=\frac{\sum _j{\sigma }_{D_j}^2·{W\left(j\right)}_x^2}{{\left[\sum _j{W\left(j\right)}_x\right]}^2}+{\sigma }_{D_{\mathrm{Dark}+\mathrm{Scatter}}}^2.$$(16)

In equation (16), ${\sigma }_{{D^{″}}_x}^2$is calculated from ${\sigma }_{D_j}^2$ because the same value of D_Dark + D_Scatter is subtracted from each D(j) to produce D′(j); therefore, the D′(j) values are not strictly uncorrelated.

Evaluation of equations (6) and (8) using a single, arbitrarily chosen, GOES-16 EUVS-C 3-second observation taken at 00:05:02 UT on February 19, 2017 gives D″_BlueWing + D″_RedWing = 55584.16 DN, D″_h + D″_k = 16234.50 DN and MgII_EXIS = 0.29207. Precision follows from equations (12), (15), and (16) resulting in σ_MgIIEXIS/MgII_EXIS = 1.01 × 10⁻⁴, which is a factor of 10 better than the EUVS-C precision requirement (McClintock et al., 2025). MgII_EXIS is relatively insensitive to errors in background + offset subtraction. Adding ±1 DN (approximately 3.5 times the magnitude of the value of ${\sigma }_{D_{\mathrm{Dark}+\mathrm{Bkgd}}}^2$ from Eq. (13)) changes the value of the index by ~±0.45 × 10⁻⁴.

4.1.3 Validating the analytic estimates for a single spectrum

The plots in Figure 7 indicate that the solar irradiances in the blue and red wings of EUVS-C spectra are correlated with each other and that the cores of the h and k lines are also correlated, both on the time scales of seconds to minutes. This provides a means for validating the preceding analytic uncertainty analysis by using the observed ratios of red-wing to blue-wing signals and of k-line to h-line signals after removing diurnal variations as a direct measure of the random errors in EUVS-C spectra. In Figure 9, top panels, plots of the GOES-16 and GOES-18 EUVS-C normalized red-wing/blue-wing ratio for spectra observed on August 9, 2022, are shown as black curves (left abscissae). The red curves are values of the ratio averaging over 30 min (600 measurements) that track variations with diurnal time scales (see discussion of Fig. 7 and Sect. 4.2.1). Differences between the raw ratios and diurnal trends are plotted as a blue curve offset by 1.002 for clarity (right abscissa). Analogous to equation (15), the uncertainty in the blue/red ratio for a single 3-second observation is

$$\frac{{\sigma }_{\mathrm{BlueWing}/\mathrm{RedWing}}^2}{{\mathrm{BlueWing}/\mathrm{RedWing}}^2}=\frac{{\sigma }_{{D^{\mathrm{″}}}_{\mathrm{BlueWing}}}^2}{{D^{″}}_{\mathrm{BlueWing}}^2}+\frac{{\sigma }_{{D^{″} }_{\mathrm{RedWing}}}^2}{{D^{″}}_{\mathrm{RedWing}}^2}.$$(17)

Figure 9 In the top panels (left abscissae), the black curves are plots of the normalized blue wing signal/red wing signal ratio observed by GOES-16 and GOES-18 EUVS-C On August 9, 2022. Red curves are 30-minute-long running averages that track the diurnal variation of the ratios. Blue curves (right abscissae) are black-red. Their standard deviations (σb/r = 3.1 × 10−5 and 2.4 × 10−5) for GOES-16 and GOES-18, respectively, are measures of the random error in the blue/red ratios after removal of diurnal trends. The bottom panels are plots of the h line/k line ratios (black) and 3-minute running means (red), along with the differences (blue). Their standard deviations are σh/k = 2.5 × 10−4 and 1.9 × 10−4. Values for both σb/r and σh/k are consistent with the errors calculated using equations (9), (11), and (17). Larger values for GOES-16 reflect its relatively larger loss of radiometric sensitivity with solar exposure (Sect. 4.2.4).

Evaluating the two terms on the right using equation (11) results in $\frac{{\sigma }_{\mathrm{BlueWing}/\mathrm{RedWing}}}{\mathrm{BlueWing}/\mathrm{RedWing}}={2.8 \times {10}^{-5} \mathrm{and} 2.4 \times 10}^{-5}$ for GOES-16 EUVS-C and GOES-18 EUVS-C, respectively. These can be compared to the standard deviations of the blue-wing/red-wing ratios, σ_b/r = 3.1 × 10⁻⁵ and 2.4 × 10⁻⁵). The identical analytic analysis for the h/k ratio results in $\frac{{\sigma }_{\mathrm{h}/\mathrm{k}}}{\mathrm{h}/\mathrm{k}}={2.3 \times 10}^{-4} \mathrm{and} {1.7\times 10}^{-4}$ for GOES-16 EUVS-C and GOES-18 EUVS-C compared to the standard deviation of ratio plots (bottom panel, blue curve) σ_h/k = 2.5 × 10⁻⁴ and 1.9 × 10⁻⁴.

In general, errors derived from the ratio plots are ~10% larger than those derived from equations (9), (11), and (17), indicating that the analytic calculation appears to underestimate the random errors. Also GOES-16 EUVS-C sensitivity had declined by ~30% relative to its immediate post-launch value by mid-2022 (Sect. 4.2.4) leading to ~20% larger errors relative to those for GOES-18 EUVS-C based on the analytic estimates.

4.2 Systematic errors in the MgII index

Sources of systematic errors in the MgII_EXIS index include shifts in the wavelength scale relative to the set of fixed masks used in the operational algorithm to calculate the components of the index and wavelength dependent degradation in instrument responsivity. Diurnal and seasonal wavelength shifts are caused by Doppler motion, thermal distortion of the optical path, and pointing errors. Wavelength-dependent degradation impacts index values on yearly time scales.

4.2.1 Doppler shifts caused by diurnal instrument orbital motion

The GOES spacecraft’s orbital speeds are ~3.07 km/s. This introduces an associated diurnal sinusoidal Doppler shift in the solar spectrum on the EUVS-C detector with a maximum displacement of ±2.86 × 10⁻³ nm = ±0.136 pixels. This is sufficient to cause ~±0.1% variations in the D″ wing values calculated using equation (9) with fixed pixel masks (e.g., Fig. 7, top panel). Similar variations are also likely to occur for the h and k lines but these are not obvious due to the relatively large intrinsic solar variability in these emissions.

Assuming that spacecraft-Sun distance is fixed, the impact of Doppler shifts on the index and its components can be estimated from a simulation using a baseline GOES EUVS-C spectrum observed at noon when the Doppler shift is zero. A series of Doppler shifted spectra are then calculated on a 3-second cadence using linear interpolation to convert from the noontime wavelength scale to a shifted wavelength scale. At each time step, the pixels in the shifted scale are displaced by an amount

$$\Delta p\left(n,t\right)=v\left(t\right)·\lambda (n,0)/c·\mathrm{Disp}\left(n\right),$$

where v(t) is the spacecraft radial velocity relative to the Sun, λ(n, 0) is the wavelength value for pixel n in the unshifted spectrum, c is the velocity of light, and Disp(n) is the EUVS-C spectrograph dispersion for pixel n. If λ is measured in nm, then Disp is measured in nm/pixel. The operational algorithm is applied to the shifted series of spectra to produce values for the index and its components. Results are plotted in Figure 10 for a GOES-16 EUVS-C simulation. Values for the individual wing and line components are plotted as a function of time in the top panel. Terms in the numerator and denominator of equation (8) are plotted in the lower panel along with MgII_EXIS. Doppler shifts introduce an approximate ±0.6% diurnal error in the individual wing components. Because the two wings are anti-correlated the error in their sum has an amplitude <0.01% and the impact on the denominator of the index (Eq. (8)) is negligible. A GOES-18 EUVS-C simulation has an identical shape to that of GOES-16 but is shifted in time by 4.13 h (62° in longitude).

Figure 10 Simulation of Doppler shift on effects on the MgIIEXIS index and its components. Top panel: Blue- and Red-wings relative values for spectra that have been shifted in wavelength to simulate effects of spacecraft orbital velocity relative to the Sun. Bottom panel: orbital velocity simulation for h line, k line, and the index. Doppler shifts introduce a diurnal systematic error of ~ ±0.03% in the simulated index.

Doppler errors in the flight data can be corrected by shifting spectra to the pixel scale of the spectrum observed at noontime before calculating the index. Shifts are determined by fitting Gaussian profiles to the h and k emission lines and calculating the differences in centers of the fitted profiles to the centers for fits to the noontime spectrum. Since diurnal solar variability is negligible, this approach should produce spectra with equal blue-wing and red-wing signals throughout the day. In addition, although the corrected values of h and k and the corrected index exhibit significant solar-induced variability, the ratio of index values observed with one instrument to those observed by another EXIS instrument should be constant. (Because the index is calculated from corrected counts (D″) rather than irradiance (Eq. (13)) and each instrument has a slightly different radiometric sensitivity (McClintock et al., 2025), the absolute value of the index differs for each instrument). Plots in the top panel of Figure 11 are blue wing, red wing, and blue wing + red wing signals for GOES-16 and GOES-18 calculated from observations made on November 3, 2022, after removing the spacecraft-Sun distance variation and shifting them to their respective noontime pixel scales using linear interpolation. Although shifting significantly reduced the diurnal amplitudes in the red wing and blue wing (see, e.g., Figs. 7 and 10), residuals remain for both instruments. GOES-17 plots are nearly identical in shape to those of 18, except the residual amplitudes after shifting are 50% smaller, but the phases are nearly the same because these instruments are located at nearly equal longitudes. The ratios of GOES-16 MgII_EXIS to the GOES-18 MgII_EXIS for 3-second observations, which have spacecraft-Sun distance variation removed and are normalized to a median value of one but are not shifted to common pixel scales, are plotted in black in the lower panel. These indicate that there is an approximate ±0.1% diurnal variation in the index resulting from the use of fixed masks. Ratios computed after pixel shifting to correct for fixed masks are plotted in blue. Comparison of the blue and black curves indicates that shifting to a common pixel scale largely removes the ±0.1% velocity-dependent bias from the MgII_EXIS index. The source of the remaining variation has not been identified.

Figure 11 Plots in the top panel are normalized blue wing, red wing, and blue + red wing values that have been computed after removing spacecraft-Sun distance variation and shifting the observed spectra to a common wavelength scale to account for Doppler effects. The bottom panel compares values of the MgIIEXIS index before and after spectral shifting (black and blue, respectively). Shifting to a common wavelength scale largely, but not completely, removes a ±0.1% velocity dependent bias from the index. Solid curves are 30-minute averages of the raw spectra that have a 3-second integration periods.

4.2.2 Spectral displacement during eclipse

For approximately 100 days each year, the absolute declination of the Sun is <9°, and the Earth occults the Sun as viewed from geostationary orbit for up to 72 min per day. Eclipse entry leads to a rapid 0.25 °C drop in case temperature accompanied by a displacement of the spectrum on the detector. The effect can be quantified by comparing MgII_EXIS values for GOES-16, which enters eclipse near 5 h UTC, with MgII_EXIS values for GOES-17 or GOES-18, which enters eclipse near 9 h UTC. In the top panel of Figure 12, the normalized ratios of the GOES-16 MgII_EXIS relative to the GOES-18 MgII_EXIS are plotted for an eclipse day (black) and a non-eclipse day (red), both calculated using the fixed-mask operational algorithm. Both observations match until GOES-16 enters the eclipse. The 0.05% red-black displacement observed at the exit most likely results from temperature-induced instrument housing distortion during the eclipse, which leads to spectral shifts. A second deflection, which is larger than the first, takes place as GOES-18 exits the eclipse. Recovery requires approximately 5 h. The comparison of ratios in the bottom panel demonstrates that the eclipse effect can be removed by shifting the spectra from each instrument independently to its noontime pixel scale using Gaussian fits to the positions of the h and k lines before computing the index as described in Section 4.2.1.

Figure 12 Top panel: Comparison of the GOES-16 MgIIEXIS/GOES-18 MgIIEXIS ratio for an eclipse day (black) and non-eclipse day (red). Indices were calculated using the fixed-mask operational algorithm. Displacements or the black curve relative to the red result from abrupt changes in instrument temperature that occur during eclipse entry. Lower Panel: Ratio comparisons for eclipse and non-eclipse days with indices calculated using the operational algorithm after shifting the spectra from each instrument to its noontime pixel scale as described in Section 4.2.1.

4.2.3 Pointing errors

The alignment experiments, performed during instrument commissioning and repeated on a 3-month cadence, are used to track index variations associated with pointing angle displacements relative to the Sun center. These arise because displacements lead to wavelength scale shift-stretch (Sect. 3.2.3), which is not corrected by the MgII_EXIS operational algorithm, and because exposure to solar radiation causes the instrument response function at each wavelength to vary differently with viewing angle (Sect. 4.2.4). Maintaining accuracies of −0.02° < α < 0.02° and −0.04° < β < 0.08° provides better than ±0.1% relative accuracy in the MgII_EXIS index for all three GOES EUVS-C instruments. Reducing those ranges by a factor of two improves the relative accuracy to ±0.05%. During routine observations, α and β remain within 0.0033° of the center, indicating that systematic errors introduced into MgII_EXIS by pointing errors are negligible.

4.2.4 Wavelength-dependent radiometric responsivity degradation

Heath & Schlesinger (1986) argued that defining an index that contained terms from both the blue and red wings minimizes instrument responsivity degradation for scanning instruments like SBUV and SOLSTICE, which use a single detector element to observe the entire spectral range. This assumption may not apply equally to the GOES EXIS EUVS-C instruments, which use an array of fixed detector elements (pixels) that are uniquely assigned to specific wavelengths. For this reason, the field of view experiment described in Section 3.2.3 is repeated on a 3-month cadence to track potential spectrally dependent changes in radiometric responsivity as a function of solar exposure. The top panel of Figure 13 summarizes the results for GOES-16 EUVS-C. Total data numbers for each spectrum, corrected for distance to 1 AU, are plotted as a function of the β (cross dispersion) axis scan for each field of view experiment. These indicate that from launch until December 2023, the instrument has experienced an approximate 35% decline in sensitivity at its FOV center (nominal observing position) relative to angles near β =±0.9°. The magnitude of this decline is not concerning because the lifetime of a GOES satellite is 10 years, and the total decline projected for an additional 4 years is less than 50%. A factor of 2 loss in sensitivity decreases precision by $\sqrt{\left(2\right)}$. This reduces the precision at the beginning-of-life performance, ~0.01% (Sect. 4.1.2) to ~ 0.014%, which is a factor of 7 better than the 0.1% mission requirement.

Figure 13 GOES-16 average radiometric responsivity degradation with solar exposure. Upper panel: FOV scans in the β axis (cross dispersion axis) indicate that the average sensitivity has declined ~35% from February 9, 2017, until November 1, 2023. Lower panel: normalized ratios of spectra observed at β = 0° relative to those observed at β = 0.9°.

Spectra taken with β = ±0.9° exhibit a much smaller sensitivity loss since launch that is not evident in the plots because the scaling is set to capture the loss at β = 0°. Comparing β = +0.9° calibrated irradiances (Sect. 3.2.5) with TSIS Spectral Irradiance Monitor (Richard et al., 2020) and SORCE Spectral Irradiance Monitor (Harder et al., 2022) irradiances after convolution with the TSIS and SORCE instruments line spread function (~1.1 nm FWHM), reveals an average sensitivity loss of ~0.25% from February 2017 until December 2023. The shape as a function of wavelength is nonlinear, with the midpoint of the blue wing declining by ~0.5% and the midpoint of the red wing and h and k cores declining by ~0.1% and ~0.2%, respectively.

Plots in the lower panel of Figure 13 are β = 0°/β = 0.9° ratios for EUVS-C spectra that have been convolved with the TSIS and SORCE line spread function. Each has been divided by the β = 0°/β = +0.9° ratio observed on February 9, 2017, when the degradation is assumed to be 0 (red curve in the upper panel) and then normalized to 1 at 283.6 nm in order to track the evolution of wavelength-dependent shape of the degradation at each epoch. (The wavelength-independent component of degradation does not affect the value of the index, although it does lower the precision of the measurement (Sect. 4.1).) This result differs from the Heath & Schlesinger (1986) linearity assumption and is observed, with essentially identical wavelength dependence, in all four EUVS-C instruments.

Referring to the lower panel, the midpoint of the blue wing declines ~3% faster than the midpoint of the red wing and ~2.5% faster than the h and k emission cores. Results for β = 0°/β = −0.9° are essentially identical. α (dispersion) axis scans are not included in this analysis because regions of the FOV for α = ±0.9° are continuously exposed by the spectrally dispersed light within the spectrograph. The mechanism for this nonlinear degradation is not understood.

Plots in the upper panel of Figure 14 are the ratios of the four GOES-16 EUVS-C components used to calculate the MgII_EXIS index from equation (8) for spectra observed at β = 0° to those observed at β = +0.9°. These are consistent with the spectral ratios in the lower panel of Figure 13, which indicate that the red wing and the h and k lines degrade at similar rates that are significantly less than the rate for the blue wing. If spectra taken from β = ±0.9° were not degraded, the ratios of the index values calculated for these two FOV locations, which are plotted as the black line in the lower panel of Figure 14, would represent the relative impact that spectrally dependent degradation has on the MgII_EXIS index indicating that the reported numbers overestimate the true index value by ~1% after ~7 years exposure. The red curve in the lower panel of Figure 14 is a plot of the relative impact that spectrally dependent degradation for β = 0° has on the index ratio after correcting for the β = 0.9° degradation. Lower radiometric sensitivity also reduces precision. A factor of 2 loss in sensitivity decreases precision by $\sqrt{\left(2\right)}$.

Figure 14 Top panel: Change in the components used to calculate the GOES-16 MgIIEXIS index (Eq. (8)) as a function of time, which is a proxy for solar exposure. The blue wing signal declines more rapidly than the red wing and h and k line signals. Bottom panel: Comparing index values calculated from observations taken quarterly from an offset position from β = 0.9° (black) track the impact of signal decline on the absolute index value. Because spectra taken at β = 0.9° also show very weak sensitivity loss, the β = 0°/β = 0.9° ratio differences must be increased by ~10% (red) in order to calculate the total impact of instrument degradation.

GOES-17 EUVS-C and GOES-18 EUVS-C exhibit nearly identical spectrally dependent declines and correction factors with solar exposure.

5 Summary

Snow & McClintock (2005) used SOLSTICE II data to show that MgII_{SORCE/SOLSTICE} fluctuates ~1% 1 − σ in a typical 24-hour period. They also found index variability of ~0.2% on time scales of 6–10 min, increasing to ~ 0.3% and ~ 0.55% for 30 and 80 min, respectively. These studies informed the measurement objectives for the EXIS EUVS-C, which are to provide daily measurements of the solar MgII index with a spectral resolution Δλ ~ 0.1 nm and a precision better than 0.1% with a cadence of 30 s. Evaluation of flight data from GOES-16, -17, and -18 EUVS-C instruments demonstrates that they all meet or exceed these objectives. Analytic calculations and intercomparison of simultaneous observations demonstrate that random errors, which determine the precision, are σ_MgIIEXIS/MgII_EXIS ~ 1.01 × 10⁻⁴ with a 3-second measurement cadence.

The operational algorithm for calculating MgII_EXIS employs fixed pixel masks that enable rapid data processing. This simplification introduces systematic errors in index values as the spectrum is displaced from its nominal position on the detector. Spectral shifts arise from spacecraft orbital motion, which introduces an approximate ±0.1% diurnal variation, and from instrument mechanical distortion during the eclipse, which also introduces an approximate ±0.05% variation that recovers in approximately 5 h. Additionally, wavelength-dependent radiometric responsivity degradation leads to a systematic increase in the reported index on a timescale of years at an average rate of 0.2% per year. A more sophisticated index algorithm, which incorporates wavelength shifting, can significantly reduce these diurnal variations. Including the effects of radiometric sensitivity degradation, which is determined from quarterly alignment experiments, can also provide correction factors for long-term errors in calculated index values.

Acknowledgments

Many researchers contributed indirectly to this work. They include the GOES-R EXIS team at the Laboratory of Atmospheric and Space Physics at the University of Colorado Boulder, the Synchrotron Ultraviolet Radiation Facility team at the National Institute of Standards and Technology, where the EXIS-C flight instruments were calibrated, and staff at the Cooperative Institute for Research in Environmental Sciences at the University of Colorado/NOAA NCEI who have contributed to understanding to their flight measurement performance. This research is supported by NASA Contract NNG07HW00C. The editor thanks two anonymous reviewers for their assistance in evaluating this paper.

References

All Tables

All Figures

	Figure 1 GOES-16 raw counts (D(j) in equation (4) observed on February 14, 2017, and corrected for particle backgrounds, electrical offset, linearity, and flat field. Dark counts and scattered light have not been subtracted from this spectrum. Pixels 0–60 are masked and used for real-time dark and background correction.
In the text

	Figure 2 GOES-18 EUVS-C observation made on June 17, 2022. Solid blue and red lines mark the GOES trapezoidal weighting functions (right ordinate). Rectangles, which are 9 pixels wide and 8 pixels wide, respectively, represent the weighting functions for the k and h emission lines used in the current operational algorithm implementation.
In the text

	Figure 3 GOES-16 EUVS-C signal rate in the masked pixels (0–59) for the average of 1000 standard 2.934 s integrations on February 14, 2017. Counts in pixels 11–59 arise from sunlight multiply reflected between the front surface of the detector and the surfaces of its window in a “walk” pattern. The dashed line is the mean of the first 10 pixels (1.30 DN/s).
In the text

	Figure 4 EUVS-C field of view relative to the SPP Guide Telescope Assembly as measured after launch. GOES-16 and -17 EUVS-C FOVs are centered within approximately 0.01°. The GOES-18 and 19 EUVS-C FOVs are displaced relative to their GTAs by Δα = +0.065°, Δβ = +0.07° and Δα = +0.03°, Δβ = +0.09°, respectively. Their plots in this figure have been offset by those amounts relative to 16 and 17.
In the text

	Figure 5 Comparison of SOLSTICE daily-averaged irradiance on February 14, 2017, with the EUVS-C GOES-16 daily average, shown in blue; the SOLSTICE spectrum has been convolved with a triangular profile with a FWHM = 0.025 nm to match the EXIS-C spectral resolution. SOLSTICE comparisons with GOES-17, -18, and -19 exhibit nearly identical agreement as that with GOES-16.
In the text

Figure 6 GOES-16/SOLSTICE (black), GOES-17/SOLSTICE (red), GOES-18/SOLSTICE (blue), and GOES-19/SOLSTICE (orange) sampled at 0.1 nm per spectral bin. Also shown are linear fits to the ratios and a dashed line for Ratio = 1. The ~10% and 25% larger MgII h and k emissions observed for GOES-18 and GOES-19 are consistent with the uncertainty in extrapolating MgII emissions in the SOLSTICE spectra from July 4, 2019 and September 18, 2019 to 2022 and 2024 (Woods et al., 2018).

In the text

Figure 7 Diurnal relative variability in the GOES-16 and GOES-18 EUVS-C MgIIEXIS index components (Eq. (8)) for August 9, 2022. Diurnal variability in the blue and red wing irradiances arises from Doppler shifts caused by the spacecraft orbital motion that is not corrected in the operational data reduction algorithm (Sect. 4.2.1). Differing satellite longitudes (GOES-16 at 75.2° W and GOES-18 at 137.0° W) cause the profiles of the two instruments to peak at different UTC positions.

In the text

Figure 8 The black curve is an analytic uncertainty estimate for a single spectrum observed on February 19, 2017 and calculated from photon counting statistics, and detector read noise and digitization noise (Eq. (12)). The red curve is an estimate based on the differences of consecutive spectra obtained during the entire day. Uncertainties estimated from differencing spectra are 3% larger than those of the analytic calculation.

In the text

Figure 9 In the top panels (left abscissae), the black curves are plots of the normalized blue wing signal/red wing signal ratio observed by GOES-16 and GOES-18 EUVS-C On August 9, 2022. Red curves are 30-minute-long running averages that track the diurnal variation of the ratios. Blue curves (right abscissae) are black-red. Their standard deviations (σb/r = 3.1 × 10−5 and 2.4 × 10−5) for GOES-16 and GOES-18, respectively, are measures of the random error in the blue/red ratios after removal of diurnal trends. The bottom panels are plots of the h line/k line ratios (black) and 3-minute running means (red), along with the differences (blue). Their standard deviations are σh/k = 2.5 × 10−4 and 1.9 × 10−4. Values for both σb/r and σh/k are consistent with the errors calculated using equations (9), (11), and (17). Larger values for GOES-16 reflect its relatively larger loss of radiometric sensitivity with solar exposure (Sect. 4.2.4).

In the text

Figure 10 Simulation of Doppler shift on effects on the MgIIEXIS index and its components. Top panel: Blue- and Red-wings relative values for spectra that have been shifted in wavelength to simulate effects of spacecraft orbital velocity relative to the Sun. Bottom panel: orbital velocity simulation for h line, k line, and the index. Doppler shifts introduce a diurnal systematic error of ~ ±0.03% in the simulated index.

In the text

Figure 11 Plots in the top panel are normalized blue wing, red wing, and blue + red wing values that have been computed after removing spacecraft-Sun distance variation and shifting the observed spectra to a common wavelength scale to account for Doppler effects. The bottom panel compares values of the MgIIEXIS index before and after spectral shifting (black and blue, respectively). Shifting to a common wavelength scale largely, but not completely, removes a ±0.1% velocity dependent bias from the index. Solid curves are 30-minute averages of the raw spectra that have a 3-second integration periods.

In the text

Figure 12 Top panel: Comparison of the GOES-16 MgIIEXIS/GOES-18 MgIIEXIS ratio for an eclipse day (black) and non-eclipse day (red). Indices were calculated using the fixed-mask operational algorithm. Displacements or the black curve relative to the red result from abrupt changes in instrument temperature that occur during eclipse entry. Lower Panel: Ratio comparisons for eclipse and non-eclipse days with indices calculated using the operational algorithm after shifting the spectra from each instrument to its noontime pixel scale as described in Section 4.2.1.

In the text

	Figure 13 GOES-16 average radiometric responsivity degradation with solar exposure. Upper panel: FOV scans in the β axis (cross dispersion axis) indicate that the average sensitivity has declined ~35% from February 9, 2017, until November 1, 2023. Lower panel: normalized ratios of spectra observed at β = 0° relative to those observed at β = 0.9°.
In the text

Figure 14 Top panel: Change in the components used to calculate the GOES-16 MgIIEXIS index (Eq. (8)) as a function of time, which is a proxy for solar exposure. The blue wing signal declines more rapidly than the red wing and h and k line signals. Bottom panel: Comparing index values calculated from observations taken quarterly from an offset position from β = 0.9° (black) track the impact of signal decline on the absolute index value. Because spectra taken at β = 0.9° also show very weak sensitivity loss, the β = 0°/β = 0.9° ratio differences must be increased by ~10% (red) in order to calculate the total impact of instrument degradation.

In the text