A WISPR of the Venus Surface: Analysis of the Venus Nightside Thermal Emission at Optical Wavelengths

WISPR consists of two telescopes, WISPR-Inner and WISPR-Outer (Vourlidas et al. 2016), hereafter referred to as WISPR-I and WISPR-O, respectively. Both telescopes use Active Pixel Sensor detectors with 2048 × 1920 pixels and have broad optical bandpass covering roughly 0.48–0.80 μm (Vourlidas et al. 2016; Hess et al. 2021).

In this paper, we focus exclusively on the 2021 February fourth PSP flyby of Venus. Figure 1 shows a top-down view of the PSP trajectory during VGA4 (in the J2000 SPICE reference frame). Based on the relative field of views (FOVs) of each WISPR telescope, WISPR-O began observing Venus prior to WISPR-I, and WISPR-I continued observing Venus after WISPR-O. For this work, we study only the WISPR-O images of Venus during VGA4. Although we were able to generally reproduce the major findings in this paper using WISPR-I data, errors in the projection of reference data onto the image plane (see Section 2.3) limit the accuracy of our latitudinal and longitudinal knowledge on Venus in the WISPR-I images such that it is difficult to precisely identify correlations between WISPR brightness and known surface properties (e.g., elevation). An analysis of the WISPR-I data will therefore be the subject of future work. Figure 2 shows the sequence of WISPR-O images observed during the 2021 Venus flyby (VGA4). A figure set showing analogous flyby images for WISPR-I is available in the online journal.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (2 panels)

Tarball of figure set images

As with Wood et al. (2022), we use the WISPR level 2 images of the flyby. These images are calibrated in units of mean solar brightness (MSB) according to the procedure detailed in Hess et al. (2021). To convert between MSB and DN s⁻¹, a calibration factor of C_f = 9.24 × 10⁻¹⁴ is used for WISPR-O based on the in-flight star calibration (Hess et al. 2021), but modified to account for the gain setting used for the Venus observations.

Beginning with the level 2 calibrated data products, we remove residual striping in the images along the readout direction, likely due to light smearing during the readout, since WISPR cameras lack a shutter. We first mask the pixels that contain Venus' disk, as well as 20 pixels on each side around the edges of the frame that are impacted by reflection off the protective barrier at the edge of the detector (edge effects were treated similarly for WISPR-I images in Stenborg et al. 2021). The median of the remaining pixels in each row is then calculated and subtracted from that row, significantly improving the apparent striping effect, which appears constant across an entire row. This step also effectively removed the sky background from each image, which appeared due to excess scattered sunlight that was particularly apparent in the first two WISPR-O frames, but decreased as PSP entered the Venus penumbra.

We use a series of reference radar data sets from the Magellan mission (frequency = 2.4 GHz, λ = 12 cm, Ford & Pettengill 1992; Ford et al. 1993) to explore correlations with known surface conditions. These include Magellan Global Topographic Data Records, Magellan Global Emissivity Data Records; each of these mosaics are resampled to a spatial resolution of 4.6 km pixel⁻¹, which is the standard for Magellan data products as described in the data reference ⁵ (Ford & Pettengill 1992; Ford et al. 1993). We also use the Magellan Synthetic Aperture Radar (SAR) FMAP Left Look Global Mosaic with a spatial resolution of 75 m pixel⁻¹. Surface geological units were identified in the Magellan data on the basis of morphological characteristics in the SAR data (e.g., Brossier & Gilmore 2020). These were then mapped to exclude pixels at the edges of the units that might contain multiple types of surface materials given the spatial scale of the WISPR-O observations and motion blur. No presupposition was made of any correlation between geology and WISPR-O; instead, we aim to test for evidence of correlations between WISPR-O signatures and radar-derived quantities for surface units previously identified to be geologically distinct. Throughout this work, elevation values are quoted relative to 6051.8 km, the mean planetary radius.

The Venus reference data products are projected onto the WISPR FOV to facilitate scientific investigations. The projection follows exactly what was done by Wood et al. (2022), but now expanded to include more data products described using physical units. Since PSP is a solar orbiter, the fits files use the WCS Helioprojective-Cartesian (HPC; Thompson & Wei 2010). The date and time for each exposure is used with SPICE (Acton 1996; Acton et al. 2018) kernels to extract the HPC coordinates of each pixel. We use the Python port SpiceyPy (Annex et al. 2020). The SPICE kernels allow us to track spacecraft location and location, Venus' location, shape, and size, as well as conduct coordinate and reference frame transformation. From HPC, we convert the coordinates to Venus mean equator (VME) of date coordinates (which requires a transformation from HPC to heliocentric Aries ecliptic, and then to VME). The result is that each pixel has a resulting latitude and longitude on Venus from the computed line of sight—surface intersection point. Sky values are treated as NaNs. The maps, e.g., Venus Magellan Global Topography 4641m v2, are projected in simple cylindrical coordinates, and thus, a projected image is built up by filling in the requisite data at each latitude and longitude pixel coordinate. We use the IAU rotation period for Venus of 243.0185 days as is given in the SPICE kernels. Mueller et al. (2012) suggest that longitude offsets are on the order of −0.3°–0.08°, which is a smaller effect than the motion blur in the WISPR-O images.

The flyby begins with an excess of scattered sunlight in both WISPR-O and WISPR-I as PSP enters the Venus shadow. The spacecraft orientation is such that WISPR-I is essentially looking nadir at Venus' barycenter. The flyby occurs over a timescale of roughly 500 seconds, and the trajectory is shown in Figure 1. Exposures vary between roughly 3–7 s for the WISPR-I camera and roughly 4 s for the WISPR-O camera during the flyby. The spacecraft orientation remains fixed during the flyby and does not remain targeted at Venus' barycenter causing Venus to progressively slip out of the FOV. Thus, throughout each image, a certain amount of blurring is visible, which we had originally attributed solely to atmospheric scattering, but is also caused by the motion of the spacecraft during the duration of the exposure. Thus, we project each map at 9 different times within a single exposure time duration and average them together to reproduce this motion blur. For the maps of individual surface units, we combine all the maps to one single map where each value represents a different geologic unit. Instead of averaging the 9 subtime projections, we instead only count pixels to be in the geologic unit if they fall consistently within the unit in all subtime frames. This helps to ensure that when we evaluate by geologic unit we are only considering pixels that are not contaminated with background sky or other geologic units.

Figure 3 shows one of the WISPR-O flyby images along with the Magellan elevation and radar emissivity data projected into the frame. By visual inspection, it is clear that the WISPR-O intensity correlates (inversely) with the elevation data. The subsequent analysis of correlations between the WISPR and Magellan data sets is presented in Section 3.2.

Zoom In Zoom Out Reset image size

We use the Spectral Mapping Atmospheric Radiative Transfer code (SMART; Meadows & Crisp 1996; Crisp 1997) to simulate the top of atmosphere Venus radiances for comparison with the WISPR measurements given contemporary knowledge of the Venus atmosphere and surface. SMART solves the radiative transfer equation for one-dimensional plane-parallel atmospheres using line-by-line, multi-stream, multi-scattering calculations. All radiative transfer calculations in this work were made at 1 cm⁻¹ wavenumber resolution and used 8-streams (four upward and four downward) at Gaussian quadrature computational points, except where specified for finer sampling of observer emission angles. We used spatially averaged Venus International Reference Atmosphere (VIRA) thermal and molecular profiles from Moroz & Zasova (1997) using the lower atmosphere updates from Arney et al. (2014). Line-by-line rovibrational molecular absorption coefficients for CO₂, H₂O, CO, H₂S, HF, and HCl were calculated using the LBLABC code (Meadows & Crisp 1996; Crisp 1997) using the HITRAN2016 line list (Gordon et al. 2017). Sulfuric acid clouds were simulated using the nominal model from Crisp (1986) containing Mode 1, Mode 2, Mode 2', and Mode 3 particles, which are all assumed to be 75% H₂SO₄ by weight (Arney et al. 2014). Following the approach of Crisp (1986), we include the unknown UV absorber as a modified version of the submicron Mode 1 particles that match dayside observations of the Venus spherical albedo in the optical. All of the vertically resolved atmospheric profiles (cloud optical depths, gas volume mixing ratios, and the thermal profile) were held static at the nominal global profiles throughout this work. We explore the effect of changes in cloud opacity in Appendix, but find that large changes in cloud optical depths lead to relatively small changes in WISPR brightness. This is because the cloud model is roughly 35% more transparent in the optical than in the 1 μm window. We further discuss the implications and caveats of these atmospheric assumptions in Section 4.2.

Given the nominal radiative transfer modeling setup described above, we run a series of models to simulate the expected sensitivity of the WISPR observations to known and/or predictable variations in the surface conditions. We produce thermal radiances across a three-dimensional grid in (1) the elevation (temperature) of the surface, (2) the emissivity of the surface at relevant WISPR wavelengths, and (3) observer zenith angles. For the surface elevation and temperature, we use the VIRA thermal profile to define the relationship between altitude and temperature, and then, we truncated the atmosphere accordingly across a grid in elevation from −2 to 20 km at 1 km intervals. This procedure ensures that the surface temperature scales physically with systematic changes in surface elevation following the atmospheric thermal structure. While the majority of the Venus surface has elevations ≥0 km, some of the lowest elevation regions lie below this level and therefore have negative elevation values relative to the zero-point. For the surface emissivity, we assume wavelength independent values that range from 0.5 to 1.0 at intervals of 0.05. For the observer zenith angles, we ran simulations at the finite angles of 860, 707, 593, 215, and 00.

We followed the same procedure as Wood et al. (2022) to convert spectrally resolved top of atmosphere radiances to photometric counts in units of digital number per second (DN s⁻¹) within the WISPR bands. This conversion is given by the following convolution integral over wavelength:

$$\begin{eqnarray}&&\mathrm{DN}\,{{\rm{s}}}^{-1}=\displaystyle \frac{{\rm{\Omega }}}{g}{\int }_{{\lambda }_{0}}^{{\lambda }_{1}}\displaystyle \frac{{A}_{\mathrm{eff}}R}{{E}_{\mathrm{phot}}}d\lambda \end{eqnarray} \tag{ 1 }$$

where A_eff is the WISPR effective area sensitivity curve for WISPR-O (or WISPR-I), R is the Venus thermal emission radiance spectrum, E_phot photon energy of each spectral interval, Ω is the angular extent of each pixel, and g is the detector gain.

The stark sensitivity to the Venus surface in the WISPR images initially raised concerns that it could have been explained by emission from well-known opacity windows, for example, at 1 μm being picked up by excess sensitivity beyond the nominal WISPR bandpass. Wood et al. (2022) reported on lab measurements from the spare WISPR optics that showed no signs of a red light leak. The consistency of the observed counts with thermal emission models provides a second line of evidence supporting the in-band nature of the WISPR Venus flux (Wood et al. 2022). We now present a third line of evidence to evaluate whether any light leaks might be present by looking at the background stars in each WISPR image.

Using SPICE, we query for which stars of sufficient brightness from the Hipparcos catalog (Perryman et al. 1997) should have fallen within the FOV and determined their pixel coordinates. Looking up the stars in Simbad (Wenger et al. 2000) allows us to select stars with known effective temperatures and gravities. We then compare the flux detected by WISPR with that predicted by the appropriate PHOENIX model (Husser et al. 2013) convolved through the WISPR bandpass. We show an example of a WISPR-O image and stellar spectra in Figure 4. If the WISPR bandpass has a previously uncharacterized light leak, additional flux would be measured for all stars. If the light leak were bluer than the known bandpass, only hotter bluer stars would show additional flux, because the flux from redder stars is proportionally less at these wavelengths. If the light leak were redder than the known bandpass, redder stars would be proportionally more affected than bluer stars. We conduct aperture photometry on the stars in the WISPR image and show the results of the measure versus PHOENIX model expected values in Figure 4. The lack of clear trend here is suggestive that there is no light leak. Adding a faux light leak to the WISPR transmission function used to predict the Phoenix model fluxes shows a clear trend with stellar temperature that cannot be confused with random noise. Thus, we can be confident that the observations are only sensitive to in-band photons, and a window through Venus' atmosphere to its surface is in fact present at these wavelengths.

Zoom In Zoom Out Reset image size

Figure 5 shows the relationships between WISPR-O counts, elevation, and radar emissivity for a variety of surface geological units. Only the pixels with emission zenith angles <60° are used in this and the subsequent surface analyses unless otherwise stated to reduce the confounding effects of limb brightening, as discussed later. Five units are mapped, including the tessera terrains Ovda Regio, Thetis Regio, and Haasttse-baad tessera, and two plains units: "Lava," which has a high WISPR-O brightness signature and "Plains," which has WISPR-O values typical of regional plains. "Lava" corresponds roughly with the undivided smooth flow and shield terrains Sogolon Planitia of the Niobe Planetia quadrangle of Venus (Hansen 2009), whereas the "Plains" to the northeast are Niobe Planitia proper. The image and analysis shown is for the first frame in the 2021 flyby. Multiple interesting correlations are evident. Together, the combination of mapped geological units span a large range in altitude from about 0 to 5 km and exhibit a strong negative correlation with WISPR brightness. This is a well-known feature of thermal emission from the hot Venus surface (e.g., Mueller et al. 2008) and is a result of the Venus temperature profile, and its characteristic decrease with altitude (Seiff 1987; Lorenz et al. 2018). Ovda and Thetis are the highest elevation regions in view and correspondingly have the lowest WISPR brightness, whereas the region labeled "Lava" is the lowest elevation region in view and has the highest WISPR brightness. Although also tessera terrain, Haasttse-baad has a much lower maximum elevation than either Ovda or Thetis, and correspondingly higher WISPR-O values. The WISPR-O elevation trends of Haasttse-baad and Thetis appear colinear, and divergent from the Ovda trend.

Zoom In Zoom Out Reset image size

Radar emissivity appears positively correlated with WISPR brightness; however, the relationship between elevation and radar emissivity (e.g., Figure 5 lower left) exhibits a characteristic negative correlation known for some regions of the Venus surface (e.g., Klose et al. 1992) that can propagate the surface elevation (temperature) trend into radar emissivity space, complicating the assessment of trends with radar emissivity. The negative trend of radar emissivity with elevation is ascribed to the volume and type of high dielectric minerals in surface rocks (Klose et al. 1992), while the VNIR emissivity is a function of FeO content (Dyar et al. 2020); it has not been yet shown that there is a systematic relationship between these two characteristics.

Similar to Figure 5, Figure 6 shows the relationships between WISPR-O counts, elevation, and radar emissivity for the same geological units, but now shown using 11 frames from the 2021 Venus flyby. Many of the same features of Figure 5 are seen in Figure 6 although now supported by a significantly larger quantity of measurements. In Figure 6, the relationships between the tessera units are even more distinct, including the Haasttse-baad–Thetis colinearity trend in WISPR-O versus elevation, and the divergence of Ovda from that trend. The distinction between the Lava unit and the Plains is also emphasized in this expanded data set. The relationship between WISPR-O counts and the emission zenith angle is also shown in the upper right of Figure 6. In general, WISPR brightness for a given terrain remains relatively flat at emission angles up to 60°–70° and then rises sharply toward the limb at 90°. This limb brightening trend motivates our choice to restrict the surface analyses to emission zenith angles of 20°–40°.

Zoom In Zoom Out Reset image size

Figure 7 is a restricted subset of Figure 6, highlighting very narrow regions of the parameter space to facilitate comparison between geological units. We compare units at the same elevation to remove the temperature effect on emissivity and limit WISPR-O emission angles to 20°–40° to consider surface emission viewed with a similar path geometry through the atmosphere. When Ovda and Thetis regio are compared this way, Ovda has a brighter WISPR-O signature than that of Thetis by ∼20% at 1.6σ (or 3.6 DN s⁻¹ on average) where the two tesserae have both similar elevation and radar emissivity. The Lava unit is brighter than the Plains unit by ∼35% at 4σ (or 9.6 DN s⁻¹ on average) at similar elevations and with only small differences in radar emissivity (∼0.05).

Zoom In Zoom Out Reset image size

Figure 8 shows the WISPR-O brightness as a function of surface elevation compared to our baseline nadir thermal emission models (orange lines) over the same range of elevations. WISPR-O data are used from all VGA4 flyby images, but only those with emission zenith angles <40°. The WISPR-O counts are visibly negatively correlated with elevation and yield a correlation coefficient of −0.81 with a p-value of zero. Although the models exhibit the same trend with elevation, they are clearly offset relative to the measurements by an apparent constant offset, indicating either an additional source of light or an overestimation of the atmospheric opacity in the models. While thinner clouds than those from the nominal model used in our radiative transfer calculations could cause the thermal flux from the surface to be enhanced, as shown in Appendix, the effect is minimal for modest changes in cloud opacity. We note that this could also be a product of imperfect calibration factors used to convert between physical flux units and observed counts. However, since the WISPR images in Figure 2 clearly show a bright limb and limb brightening is apparent at high emission angles in Figure 6, likely due to O₂ and O i nightglow (Wood et al. 2022), we entertain the hypothesis that there is additional nightglow across the entire Venus disk.

Zoom In Zoom Out Reset image size

The pixels with an expected low brightness from the thermal emission models provide a means to measure the average nightglow brightness at relatively low emission angles. Figure 9 shows the WISPR-O counts for all high elevation (>5 km), low emission angle (<40°) points from all VGA4 flyby images as a function of elevation, radar emissivity, and emission angle, and compressed into a one-dimensional histogram. This subset of points with expected low intensity surface thermal emission have a WISPR-O count rate of 16.1 ± 1.4 DN s⁻¹. Since our thermal models still predict a measurable surface emission for regions at and above 5 km, we subtract off the 5 km predicted flux (6.6 DN s⁻¹) from the mean brightness excess derived in Figure 9 to obtain an offset of 9.5 ± 1.4 DN s⁻¹. We take this value to be our empirical estimation for the excess brightness near the Venus disk center, which may ultimately be due to a combination of nightglow emission, thinner clouds than our nominal model, scattered sunlight from the dayside, or a flux calibration offset.

Zoom In Zoom Out Reset image size

Using the empirical offset to correct missing radiative processes in our thermal emission spectral models can help to explain the brightness levels observed with WISPR. Figure 8 shows nadir thermal emission models convolved with the WISPR-O bandpass with the empirical brightness correction added (blue and teal lines). The overlapping lines indicate a range caused by thermal emissivity variations between the models, whereas the faint blue models show the ±3σ bounds around the median empirically corrected models. The fit to the elevation trend is significantly improved using the empirical offset. Therefore, the empirical brightness correction derived using only high elevation pixels helps to explain both the offset and breadth in the scatter of the measured trend with elevation relative to our thermal calculations. Furthermore, the consistency between the thermal models and the WISPR observations robustly validates the nature of the observations as thermal emission from the surface.

Extending beyond the insights gleaned from our analysis of empirical trends in the WISPR-O data, we use our precomputed grid of radiative transfer simulations of the Venus thermal emission to construct a full model of the WISPR-O images. We linearly interpolate the thermal radiance models onto the WISPR image projections, taking two-dimensions into account. The first is the WISPR-projected Magellan altitude data to capture the temperature-dependence of the radiance, and the second is the map of subspacecraft emission zenith angles to capture the angle dependence of atmospheric path lengths for rays emitted from the surface. We hold the thermal emissivity of the surface fixed at 0.9 consistent with a surface albedo of 0.1 measured by Venera 9 and Venera 10 (Ekonomov et al. 1980).

Figure 10 shows the third WISPR-O image from the 2021 flyby on the leftmost panel compared to our thermal emission only WISPR-O image model in the second panel. The thermal model captures much of the spatial variability in brightness contrast, which is consistent with the previously determined strong correlation between the altitude (temperature) of the surface and the WISPR brightness. However, the thermal model exhibits limb darkening rather than the stark limb brightening seen in the true image. These characteristics are consistent with the lack of nightglow emission in the spectral models.

Zoom In Zoom Out Reset image size

A small linear correction was applied to our simulated surface thermal emission image to optimally account for the unknowns associated with cloud opacity and a baseline flux offset. We determine a scale factor (slope) and baseline offset (intercept) that when applied to our nominal thermal emission models minimizes the residuals with the WISPR-O image for pixels with emission angles <35° of the disk center. This yielded a thermal flux scale factor 0.95 and an offset 11.18 DN s⁻¹. This offset is similar to the empirically derived flux offset in Section 3.3. The thermal model panel in Figure 10 incorporates the multiplicative factor, while the the additive offset is visualized in the third panel along with the nightglow component discussed next.

To account for the missing nightglow component at the limb, we fit a simple model to the residuals between the WISPR-O image and the thermal model. We take the residuals as a function of emission angle and smooth them using a rolling median with a window of size 150 points. The smoothed residuals exhibit a slight linear increase with emission angle from 0° to 65°, before sharply increasing >65°. We use a piecewise analytic model composed of a linear portion and a polynomial portion to fit for the average nightglow contribution. The functional form of the nightglow model is

$$\begin{eqnarray}&&{y}_{\mathrm{glow}}=\left\{\begin{array}{l}m(\theta -{\theta }_{0})+b,\mathrm{for}\,\theta \leqslant {\theta }_{0}\\ \displaystyle \sum _{k=0}^{N}{a}_{k}{\mu }^{k},\mathrm{for}\,\theta \gt {\theta }_{0}\\ \end{array}\right.\end{eqnarray} \tag{ 2 }$$

where θ is the emission angle, μ is the cosine of the emission angle, θ₀ is the angle of the break from linear to polynomial, and m (the linear slope) and the set of polynomial coefficients a_k are all fitting parameters. The y-intercept b is determined by the value of the polynomial function at θ₀. We determined that θ₀ = 67°, and N = 7 best captures the average residual trend without overfitting. The relatively high polynomial order is required to fit the sharp rise in brightness. Table 1 lists the best fitting model parameters determined using a nonlinear least squares fit to the smoothed residuals using the Python code scipy.optimize.curve_fit.

The third panel in Figure 10 shows the residual fit to the median excess emission projected back onto the image plane. The fitted nightglow emission model is able to capture the baseline flux contributions near the center of the Venus disk at low emission angles (as discussed in Section 3.3) and, critically, the stark brightening seen at the Venus limb.

The fourth panel in Figure 10 displays our final WISPR-O image model with the thermal only model from the second panel added to the fitted model to the mean excess brightness from nightglow. This model demonstrates that the WISPR-O image is well explained by emission from the hot Venus surface escaping through the new atmospheric window in the optical, with an overlying emission component from the atmosphere that is dominant at high emission angles, but present across the entire disk.

The fifth and final panel of Figure 10 shows the residuals calculated as the WISPR-O image from the first panel minus the final image model in the fourth panel. A diverging colormap is used to highlight regions of Venus where the WISPR data are brighter than the model (red) and darker than the model (blue). Overall, however, the residuals are Gaussian distributed around zero with a standard deviation of around 3.5 DN s⁻¹ at 1σ (−0.3 ± 3.5 DN s⁻¹). This indicates that the fit is quite good at capturing the average behavior of the images.

The regions of Venus with >2σ deviations from the model are interesting to note and may point to a variety of different factors. First, the brightest surface region in the WISPR-O image—a low elevation lava plain—is also the brightest region of the surface thermal model, but the model is unable to reach the brightness levels seen in the data. This could be due to surface compositional information, cloud optical depth variations, or nightglow spatial structure, which we discuss in Section 4.2. Second, there are ridges in the lower third of Venus that are not well fit, and this could result from atmospheric or motion blurring, or small errors in the model used to project the Venus reference data sets onto the WISPR-O images. For particularly small surface features, small shifts in the relative position of the WISPR images and reference data can cause excess residuals. Third, the limb of Venus exhibits some of the largest residuals. This could be explained by a combination of the aforementioned image-model offsets, but the limb residuals also show flux excesses (red) and flux deficits (blue) indicating that brightness variations along the limb that cannot be captured by our average limb brightening model also contribute to the residuals. Such brightness variations along the limb may indicate a spatial structure in the nightglow emission, cloud opacity, scattered sunlight, or a complex combination of these factors. Finally, the rightmost part of the image, particularly the upper right, shows a brightness deficit compared to the models that could be caused by vignetting imperfections not fully removed by the data calibration procedure.

WISPR-O measurements of geologic units at similar elevations show distinct differences in brightness that can be attributed to factors that are independent of temperature and emission angle (Figure 7). The Lava unit has the highest brightness values of all mapped units, including the Plains unit (30% at 4σ) and Haasttse-baad tessera (25% at 2.9σ). Although less statistically significant, Ovda tessera is notably brighter than Thetis tessera by 20% at 1.6σ where both units cover the same elevations and emission angles. These observations demonstrate that the WISPR-O data may be sensitive to compositional or grain size variations on the Venus surface.

Laboratory work shows a positive correlation between 1 μm emissivity and FeO (ferrous iron) content in rocks at Venus temperatures (440°C; Dyar et al. 2020; Helbert et al. 2021). These measurements are made at longer wavelengths (0.86–1.18 μm) than the WISPR-O broadband spectral range (0.48–0.80 μm; Wood et al. 2022). However, Helbert et al. (2021) demonstrate that the high temperature laboratory emissivity measurements of basalts correlate with emissivity derived from photometer data of the Venus surface collected by the Venera 9 and 10 landers at 5 channels over the range 0.5–1.1 μm (Ekonomov et al. 1980). This implies that higher values of WISPR-O brightness should correspond to greater FeO content in observed rocks. The weathering processes in the deep atmosphere of Venus are modeled to convert ferrous iron to ferric iron on geological timescales of days to tens of thousands of years (e.g., Smrekar et al. 2010; Berger et al. 2019; Filiberto et al. 2020; Radoman-Shaw et al. 2022; Santos et al. 2023). If so, the high WISPR-O brightness of the Lava unit shows that it has the greatest FeO values of the studied units. The Lava unit having greater FeO values than the Plains unit, which are also interpreted to be basaltic lava flows (Hansen 2009), indicates that the Lava unit has experienced less weathering and is thus younger than the Plains flows, or that the Lava flows has an intrinsically higher FeO content. Geomorphologically, the Sogolon Planitia region that corresponds to the Lava unit has smooth, homogeneous volcanic flow materials with a lower density of small volcanic shield features than the adjacent shield terrain to the north. Niobe Planitia to the northeast corresponds to the Plains unit, and is consistent with a different mode and/or timing of emplacement. The high WISPR-O brightness of the smooth flow region may suggest that the Lava unit has a higher FeO content than surrounding shield plains, as well as the Haasttse-baad tessera (see Figure 7). Northern Sogolon Planitia lies partially beneath the parabolic ejecta of the impact crater Merit Ptah. Parabolic ejecta deposits are ephemeral, with a mean crater retention age of ca. 10s Ma (Phillips et al. 1991; Izenberg et al. 1994; Campbell et al. 2007). It is possible that Merit Ptah impact ejecta is geologically young and thus has suffered less weathering than surrounding plains units resulting in a slightly higher FeO content and corresponding WISPR-O brightness.

Thetis and Ovda tesserae have distinct WISPR-O brightness values indicating that these regions, of unknown rock type, have different compositions. The lower brightness values for Thetis indicate a lower FeO content than those of Ovda Regio. This may be due to intrinsic differences in the rocks themselves, where Thetis has a more silica-rich composition than does Ovda, or it may suggest that ferrous iron in Thetis has been preferentially consumed due to differences in the style, rate, or duration of surface-atmosphere weathering of these rocks as compared to Ovda (e.g., via the production of ferric hematite as has been suggested to explain spectral measurements collected by the Venera 9 and 10 landers; Pieters 1983), or enhanced distribution of basaltic crater ejecta on the surface of Ovda as compared to Thetis (Whitten & Campbell 2016). The colinearity of the Thetis–Haasttse-baad WISPR-O versus elevation trends may imply that the two tessera regions are more similar in composition than either are to Ovda. Variations in the radiophysical properties of tesserae also show that tessera composition is not uniform across the planet (Whitten & Campbell 2016; Brossier & Gilmore 2020); the WISPR-O data provide an independent method to assess this variability and its causes.

Particle size has a demonstrated effect on NIR reflectance, where finer grain sizes at the 10s–100s μm scale will increase reflectance and therefore lower emissivity (e.g., Pieters 1983). If grain size is the dominant cause of WISPR-O brightness, this would suggest that the materials of the surface of the Lava unit have a larger average grain size in the uppermost 10s of μm than both the Plains and Haasttse-baad units and that Ovda Regio has a larger average grain size than Thetis Regio. Grain sizes are reduced by chemical and physical weathering and/or the addition of sediment, such as from impact ejecta. If due to weathering, this would imply that the Plains and Haasttse-baad and/or Thetis have undergone more extensive weathering due to friability, age, and/or topography than the Lava and/or Ovda units. However, we note that limited measurement of the 1 μm emissivity of powders and slabs at Venus temperatures to date shows no systematic dependence on particle size, concluding that FeO content is the dominant contributor to emissivity (Helbert et al. 2021).

A few factors limit the alignment precision between our reference Venus maps and the observed WISPR images, including blurring from nonnegligible spacecraft motion and atmospheric scattering, and uncertainty in the pointing and/or image projections. Although we introduced a blur in the direction of spacecraft motion based on the duration of the exposures and we were conservative in selecting the interiors of known geological units, residual errors may remain and are difficult to quantify. While we did account for motion blur, we did not correct for the atmospheric scattering footprint, and as a result, our image model appears more sharp than the real images. Additionally, trends with surface quantities are likely broadened by this additional and unavoidable scattering uncertainty. Since shorter wavelengths experience greater levels of Rayleigh scattering (Knicely & Herrick 2020), we would expect these optical measurements to have a scattering footprint around or above the nominal 50–100 km footprint in the NIR measured from orbit (Moroz 2002). However, our finding that the cloud opacity in the WISPR band is distinctly lower than the cloud opacity in the 1 μm window could lead to slight improvements in the spatial resolution, but Rayleigh scattering will still be a limiting factor.

We also identified a characteristic of the image projections wherein the alignment between the Venus reference data projections onto the WISPR images were seen to become progressively more misaligned as the flyby advanced. This could be caused by minor pointing errors, SPICE kernel errors, or errors in the FOV projection. No corrections were applied to the observed images or in our analyses to account for alignment artifacts because this effect was insignificant in the WISPR-O images that had Venus centered in the frame, and primarily affected the frames late in the flyby with only the Venus limb visible. This issue also affected the WISPR-I images, and to a larger degree than WISPR-O. Although we were able to reproduce many of our results with the WISPR-I images, the projection alignment issue was severe enough in WISPR-I to warrant further investigation that is beyond the scope of this paper.

There remain degeneracies between whether nightglow or clouds are responsible for the excess flux required across the Venus disk relative to our thermal emission models. For example, we attributed the excess flux seen at low emission angles and high elevations in Figure 9 to nightglow at disk center that rises sharply toward the limb (e.g., in Figure 10). However, the bright limb may be a red herring. The excess flux near disk center could also be, in part, due to suboptimal cloud opacity in the thermal emission models. Specifically, if the optical depth of the H₂SO₄ clouds is substantially decreased relative to our nominal cloud model, then the net result would be stronger emission from the surface seen in the WISPR band (see Appendix). However, our radiative transfer model is sufficiently insensitive to small changes in cloud opacity. If the flux offset were entirely attributed to optically thinner clouds, the entire cloud model would need to have roughly half the nominal optical depth to provide a sufficient brightness enhancement to match the WISPR measurements. Moreover, these relative opacity changes linearly impact the thermal emission from the surface, which provides a poor fit to the observed WISPR-O versus surface elevation trend, whereas a constant flux offset that is more indicative of an emission source provides an excellent fit to the measurements (Figure 8). Therefore, the additional flux component near the Venus disk center (∼10 DN s⁻¹) is unlikely to be entirely attributable to thinner-than-expected clouds.

Nightglow is clearly needed to capture the limb brightening, and is a plausible explanation for the excess disk brightness. Wood et al. (2022) predicted the WISPR brightness of O₂ nightglow at 0.76 μm should be around 1.6 DN s⁻¹ based on the typical limb-to-disk brightness ratio of the 1.27 μm feature (Gérard et al. 2008). Our derived brightness excess of ∼10 DN s⁻¹ well exceeds this estimate. This could indicate a few different effects such as (1) higher intensity disk emission relative to the limb for the 0.76 μm line, or (2) a combination of multiple in-band nightglow lines, for example, from both O₂ and the atomic O i 5577 Å green line, or (3) unidentified systematic effects. Observations of oxygen nightglow show substantial spatial and temporal variability (Allen et al. 1992; Crisp et al. 1996; Hueso et al. 2008). This variability could help to explain spatial residuals seen in Figure 10, but the residuals remain relatively small—about 35% intensity variations on a baseline effect of order 10 DN s⁻¹ (at 1σ)—compared to previously observed nightglow heterogeneity with up to 10× contrasts across the nightside (Gérard et al. 2008). A perfect explanation remains elusive. Ultimately, the reliance on only a single photometric band in this work provides limited evidence with which to break the degeneracies with nightglow. Additional observations, particularly concurrent spectroscopy or narrowband imaging, as well as further analyses would be helpful in the future.

PSP will perform its final flyby of the Venus nightside on 2024 November 6 (VGA7). During this encounter, PSP will pass at ∼340 km of the Venus surface at the closest approach, and the event will last approximately 8.5 minutes. VGA7 is expected to probe in the general vicinity of Phoebe Regio, a completely different area of the Venusian surface compared to the VGA4 images studied in this paper. While there may be challenges in directly comparing VGA4 (this work) with VGA7 due to the differing planetary and spacecraft environments, relative differences may still be quite informative. For example, the different surface coverage under VGA7, and the flyby's lower closest approach may allow a comparison of new territory—some quite uniform—over a range of emission angles, which, when compared to Magellan radar data, may aid in deconvolving variable atmospheric effects from the WISPR images. These upcoming WISPR observations will provide one final opportunity to access the Venus surface through the unique WISPR window and to further test and validate the findings presented here in advance of dedicated forthcoming Venus missions.

Additionally, upcoming Venus missions will have an incredible opportunity to advance the state of knowledge of the Venus surface. The regions imaged by WISPR-O were not included in the NIR survey of the southern hemisphere provided by Venus Express; thus, the variations in the NIR properties of geologic units imaged by WISPR-O indicate that the global multiband mapping to be provided by the DAVINCI, VERITAS, and EnVision missions will critically advance our understanding of the diversity, origin and relative age of geologic units on Venus.