The Temporal Brightening of Uranus' Northern Polar Hood From HST/WFC3 and HST/STIS Observations

Hubble Space Telescope Wide‐Field Camera 3 (HST/WFC3) observations spanning 2015 to 2021 confirm a brightening of Uranus' north polar hood feature with time. The vertical aerosol model of Irwin et al. (2023, https://doi.org/10.1038/s41550-023-02047-0) (IRW23), consisting of a deep haze layer based at ∼5 bar, a 1–2 bar haze layer, and an extended haze rising up from the 1–2 bar layer, was applied to retrievals on HST Space Telescope Imaging Spectrograph (STIS) (HST/STIS) observations (Sromovsky et al., 2014, 2019, https://doi.org/10.1016/j.icarus.2014.05.016, https://doi.org/10.1016/j.icarus.2018.06.026) revealing a reduction in cloud‐top CH4 volume mixing ratio (VMR) (i.e., above the deep ∼5 bar haze) by an average of 0.0019 ± 0.0003 between 40–80◦N (∼10% average reduction) from 2012 to 2015. A combination of latitudinal retrievals on the HST/WFC3 and HST/STIS data sets, again employing the IRW23 model, reveal a temporal thickening of the 1–2 bar haze layer to be the main cause of the polar hood brightening, finding an average increase in integrated opacity of 1.09 ± 0.08 (∼33% increase) at 0.8 µm north of ∼45°N, concurrent with a decrease in the imaginary refractive index spectrum of the 1–2 bar haze layer north of ∼40°N and longwards of ∼0.7 µm. Small contributions to the brightening were found from a thickening of the deep aerosol layer, with an average increase in integrated opacity of 0.6 ± 0.1 (58% increase) north of 45°N between 2012 and 2015, and from the aforementioned decrease in CH4 VMR. Our results are consistent with the slowing of a stratospheric meridional circulation, exhibiting subsidence at the poles.

This motivated Toledo et al. (2018), building on the previous work of Karkoschka and Tomasko (2009), to also probe CH 4 depletion as a possible source of the bright appearance of the polar hood.Their results concluded that the depletion in polar cloud-top CH 4 VMR was responsible for the hood's bright appearance, as opposed to latitudinal variations in the optical properties or abundance of aerosols.Later observations of Uranus obtained with HST/STIS in 2015 (Sromovsky et al., 2019) enabled a window into the temporal evolution of the hood by comparison with previous observations from 2012 (Sromovsky et al., 2014).An apparent brightening of the north polar hood with time was detailed by Sromovsky et al. (2019) who performed a retrieval study that concluded that the change was driven by increased aerosol scattering.Sromovsky et al. (2019) acknowledged that the persistent polar depletion in upper tropospheric CH 4 aids in the brighter appearance of the north polar hood, in agreement with Toledo et al. (2018).However, it was concluded that this depletion cannot account for the temporal change observed due to its spectral signature including wavelengths dominated by hydrogen absorption and due to the polar depletion in CH 4 VMR being found to be "generally stable" between the two observations.Toledo et al. (2019) subsequently carried out microphysical simulations to constrain the formation and structure of haze in Uranus' atmosphere, revealing timescales for haze particles to grow and settle out to be ≳30 years at pressure levels >0.1 bar.This is too long to explain the observed variations in the haze structure over Uranus' northern hemisphere post-equinox, hinting at a dynamical origin of the spatial and temporal distribution of the haze over the poles.To date, no study has managed to pinpoint the exact aerosol layer change(s) involved in the brightening of the polar hood within a chosen vertical aerosol model, which is what this study intends to address.Irwin, Teanby, Fletcher, et al. (2022), using an array of data sourced from HST/STIS, HST/WFC3, the Infrared Telescope Facility's SpeX (IRTF/SpeX) instrument (a medium resolution spectrograph) and Gemini-North's Near-Infrared Integral-Field Spectrograph (Gemini/NIFS), were able to produce an holistic vertical aerosol model simultaneously fitting these data sets for both ice giant atmospheres (i.e., Uranus and Neptune).The model consists of three layers (outlined in full in Section 2.3): a deep layer assumed to be composed of a mixture of H 2 S ice and photochemical haze, a layer of photochemical haze/ice at the methane condensation level, and an extended layer of photochemical haze extending from this level up through to the stratosphere.This holistic model was then improved upon by Irwin et al. (2023) by modifying the deep aerosol layer to be a compact layer based at ∼5 bar as opposed to the original diffuse layer.
We employ the HST/STIS observations from 2012 (Sromovsky et al., 2014) and 2015 (Sromovsky et al., 2019) in this investigation, in addition to yearly observations of Uranus spanning 2015-2021 obtained from the Outer Planetary Atmospheres Legacy (OPAL) program using HST/WFC3.We use the NEMESIS radiative transfer retrieval code (Irwin et al., 2008), combined with the modified version of the holistic ice giant aerosol model produced by Irwin et al. (2023), to probe the north polar hood brightening and identify its origin within this vertical aerosol scheme.The Minnaert limb-darkening approximation, which has previously been demonstrated to be an effective method to account for center-to-limb variations when fitting to Neptune VLT/MUSE observations (Irwin et al., 2021), is also utilized.We give a full description of the observations used in the investigation, followed by an outline of the aerosol model of Irwin et al. (2023) in Section 2. The results of a Minnaert limb-darkening analysis of the HST/WFC3 data set, and results from retrievals carried out on the HST/WFC3 and HST/STIS data 10.1029/2023JE007904 3 of 29 sets, are then presented in Section 3. A discussion of the results follows in Section 4 with plausible mechanisms explored.Finally a summary of our results, along with our conclusions, are given in Section 5.

HST/WFC3
Observations of Uranus were obtained using Hubble Space Telescope's (HST) Wide Field Camera 3 (WFC3) instrument in seven different spectral bands defined by WFC3's UVIS medium-band and narrow-band filters, displayed in Figure 1 and Table 1.
The observations span 2015-2021, each taken ∼1 year apart during 8 HST orbits and covering two full rotations of Uranus.Images were navigated using an ellipsoid limb-fitting technique with equatorial and polar radii of 25,559 and 24,973 km, respectively.The FQ727N data were fringe corrected post-pipeline, resulting in a ∼1.1% correction.Data were recalibrated in accordance with the latest WFC3 calibration pipeline (calwf3 3.6.2),accounting for time-dependent sensitivity changes, resulting in a ∼0.3%-0.8%correction (dependent on the filter and epoch) (https://github.com/spacetelescope/hstcal/releases).Images in each filter were selected in which discrete storms and clouds were absent.This was decided in order to avoid such bright discrete features from skewing our Minnaert latitudinal analyses.
The precise observations used in the analysis are available on the MAST archive (James, 2023).
Figure 2 displays the images from each WFC3 filter for the 2015 and 2021 data sets used.The progress of Uranus' orbit toward northern summer solstice (occurring in 2030) can be seen via the north polar hood swinging further into view as seen from the Earth.

HST/STIS
Observations of Uranus using the STIS instrument on HST were both made using four HST orbits, taken on 27-28 September 2012 (2012-09-27-21:38:11-2012-09-28-01:15:31 UT) (Sromovsky et al., 2014) and 10 October 2015 (14:09:48-17:44:54 UT) (Sromovsky et al., 2019).Two  of STIS's gratings, the G430L grating covering 290-570 nm with a 0.273 nm/pixel dispersion, and the G750L grating covering 524-1027 nm with a 0.492 nm/pixel dispersion, were utilized.Hubble Space Telescope's/STIS is a long-slit spectrometer, but the 51″ × 0.10″ long-slit was aligned with the planet's polar axis and stepped across the disc at intervals of 0.152″ from center to edge for the G430L grating (because the planet has no high spatial resolution center-to-limb features at these wavelengths, interpolation was used to fill in the missing columns of the mosaic), and stepped by 0.0562″ intervals for the G750L grating, forming a "cube" of half the planet, where at each location on the disc a complete spectrum covering 300.4-1,020.0nm was recorded at a resolution of 1 nm, sampled every 0.4 nm.The raw data underwent a thorough post-processing treatment described in Appendix A of Sromovsky et al. (2014), and were finally photometrically calibrated using simultaneously obtained HST/WFC3 supporting observations, producing the final data cubes.We smoothed the HST/STIS observations to IRTF/SpeX resolution of 2 nm and sampled on a regular grid of spacing 1 nm.This was due to the CH 4 absorption coefficients being used being at the same resolution as IRTF/SpeX (i.e., resolution of 2 nm) (Karkoschka & Tomasko, 2010), allowing us to achieve more efficient retrieval times after smoothing without losing any true resolution in the data.
It should be noted that for the 2015 cube, an error in the program resulted in half of the half-disc being covered with STIS's nominal 0.05″ slit instead.This produced a higher spectral resolution at the cost of a significant reduction in signal to noise ratio.The data were pipeline processed, extracted, calibrated, and combined with the uniform spatial and spectral sampling above, described in Sromovsky et al. (2019).
Uranus rotated more than 45° over the course of these observations, resulting in the assumption of a high degree of zonal symmetry in Uranus' atmosphere being embedded within our analysis from the outset.Finally, it should be noted that the data longward of ∼950 nm was significantly effected by noise for both observations, demonstrated by the plot and images in Figure 3.This region of the spectrum was therefore omitted from the analysis to avoid the low signal-to-noise ratio (SNR) skewing our retrieval results.

Aerosol Model
The vertical aerosol distribution model used in this investigation is that of Irwin et al. (2023) (IRW23), an improved iteration of Irwin, Teanby, Fletcher, et al. (2022)'s holistic aerosol model applying to both ice giant planets.The significant difference between the two models is that the IRW23 model replaces the diffuse deep aerosol layer with a compact one based at ∼5 bar, providing significant improvements to retrieved latitudinal profiles (Irwin et al., 2023).This consists of: (a) a deep compact aerosol layer centered at p ≃ 5 bar, assumed to be composed of a mixture of H 2 S ice and photochemical haze (aeorsol-1); (b) a layer of photochemical haze/ice, coincident with a layer of high static stability at the CH 4 condensation level at 1-2 bar (aerosol-2); and (c) an extended layer of photochemical haze, likely mostly of the same composition as the 1-2-bar layer, extending from this level up through to the stratosphere (aerosol-3) (Irwin et al., 2023).This is displayed in Figure 4.

Minnaert Analysis of HST/WFC3 Observations
The Minnaert limb-darkening approximation (Minnaert, 1941) provides an efficient way to model the center-to-limb variations of a given observation of a planet's full disc.The model approximates the reflectivity,    , for a given observation at a particular wavelength, taking the form: Where  (

𝐼𝐼 𝐹𝐹
) is the fitted nadir reflectance, k is the fitted limb-darkening parameter and μ and μ 0 are the cosines of the viewing and solar zenith angles, respectively.This model has been successfully vetted for work carried out on VLT/MUSE Jupiter observations (Pérez-Hoyos et al., 2020), and more relevant to our work, on HST/ STIS Neptune observations (Irwin, Teanby, Fletcher, et al., 2022) and VLT/MUSE Neptune observations (Irwin et al., 2021) in conjunction with the same NEMESIS radiative transfer retrieval code used in this investigation.
We first applied the approximation to our HST/WFC3 data set in order to carry out a latitudinal analysis of the fitted Minnaert parameters.The disc was divided into latitude bands of width 10°, stepped across the disc with step size 5°.These latitude bands were then fit to Equation 1 individually, fitting profiles for the Minnaert parameters    0 () and k(ϕ) as functions of latitude, ϕ, in each of the seven spectral filters.It should be noted that as we move to high northern latitudes, the lack of zenith angle coverage results in the limb of the planet no longer being captured by the latitude bands defined above, meaning the fits of the limb-darkening parameter k are less reliable (cutoff latitude bands centered at 60°N for 2015 and at 35°N for 2021).This increase in uncertainty is incorporated into the latter retrieval work in this paper by a scaling of the measurement errors with latitude, defined in Section 3.2.An alternative method to account for the added uncertainty by fixing k to the average value of all latitudes north of the cutoff latitude was attempted, but this did not produce physically plausible results near to the north pole.
The motivation behind this initial analysis is the Minnaert approximation's ability to provide a measure of brightness independent of viewing angle.This appears in the form of the fitted  (

𝐼𝐼 𝐹𝐹
) 0 parameter, enabling a direct comparison of the brightness between different regions on the disc.Figures 5 and 6 show the results for two of the spectral filters, 0.845 and 0.727 μm, chosen as representative filters in which the polar hood is distinctly visible and in which it is not visible, respectively (see Figure 2).
The latitudinal profiles of the fitted Minnaert parameters correlate very well with the north polar hood.Namely, for wavelengths in which the polar hood is distinctly visible (e.g., 0.845 μm), the observed increase in brightness at northern polar latitudes with time is significantly more pronounced than that of equatorial latitudes or the mid-latitudes (average increase in reflectivity of ∼0.0044/year at 65°N compared to ∼0.00084/year at 30°N, a ∼5.25-fold difference), and for wavelengths in which the polar hood cannot be clearly discerned (e.g., 0.727 μm), there is no such asymmetry in the change in brightness in the northern hemisphere, or one of a much lesser degree.
This result is significant insofar as it confirms a "true" brightening of the polar hood, ruling out any apparent brightening effects due to changes in the viewing geometry as significant contributors.This suggests that the main contributors to the polar hood brightening over time must originate from changes in the cloud-top CH 4 VMR (i.e., atop the ∼5 bar deep haze layer) and/or changes in the aerosol layers within Uranus' atmosphere (changes in scattering and/or optical properties).This is reflected by the fact that the 0.727 μm filter probes high up in the troposphere (≲ 1 bar) and shows no signs of the bright polar hood or its evolution with time, whereas the 0.845 μm, probing pressures of a few bars that coincides with the expected location of the aerosol layers and also the pressure sensitivity to CH 4 VMR, shows a strong signature of the hood and its brightening.) over the polar hood latitudes (45-90°N) between successive years for each WFC3 filter and plotted these along with linear fits to observe any potential trends (Figure 7).We indeed observe a tentative slowing in the rate of brightening for the filters in which the hood is distinctly visible over the timespan of the data set.This suggests that the change(s) responsible for the polar hood brightening, and the underlying mechanism(s) involved, could be slowing down with time.The filters in which the hood is not visible show tenuous trends by comparison.Note that the 0.547-μm filter (Figure 7) does in fact show an increased signature of brightening at polar hood latitudes (45-90°N) compared to equatorial and mid-latitudes despite the hood not being clearly visible.

Latitudinal CH 4 Profile From HST/STIS Observations
Due to the lack of spectral resolution in the HST/WFC3 observations, we cannot directly retrieve the latitudinally dependent cloud-top CH 4 VMR profile using the method of Karkoschka and Tomasko (2009).This utilizes the 818-832 nm region of Uranus' spectrum, which is dominated by hydrogen collision-induced absorption, to compare to regions dominated by CH 4 absorption, enabling a degeneracy that exists between cloud-top pressures and the cloud-top CH 4 mixing ratio within retrievals to be broken.Thus, HST/STIS observations, possessing a high spectral resolution, were employed in order to probe the temporal stability of the polar depletion in cloudtop CH 4 VMR in an attempt to test the conclusion of Sromovsky et al. (2019), and also to retrieve a latitudinal CH 4 profile to inform our latter retrievals on the HST/WFC3 data set.In order to carry out CH 4 retrievals we first need to assume a vertical profile for the VMR.A simple "step" model was adopted, similar to that of Irwin et al. (2021), with variable "deep" CH 4 VMR, limited to 100% relative humidity above the condensation level, but limited to a stratospheric VMR of 1 × 10 −4 .This model returns the mean CH 4 VMR in the region of sensitivity (∼2-4 bar), providing a parameter that is easy to interpret and compare with previous studies, as opposed to the more physically plausible descended profile (e.g., Sromovsky et al. (2019)'s recent model) (Irwin et al., 2021).
To efficiently model the observed center-to-limb variations of the HST/STIS observations, we again employed the Minnaert limb-darkening approximation described in Section 3.1.The observations were Minnaert-analyzed in latitude bands of width 10° across the disc with a step size of 5°, deriving spectra for  (

𝐼𝐼 𝐹𝐹
) 0 () and k(λ) as functions of wavelength, λ.Uranus is sufficiently distant from the Earth and the Sun such that the solar zenith angle ) 0 () and k(λ) were then used to generate reconstructed spectra at two viewing zenith angles (0°, 61.45°), corresponding to two of the five zenith angles of the Gauss-Lobatto multiple-scattering radiative transfer model used within the NEMESIS retrieval code.The higher angle is large enough to ensure that we fully capture the observed limb-darkening (or limb-brightening), and is coincident with one of the Gauss-Lobatto quadrature angles, saving the need for any interpolation (Irwin et al., 2021).These reconstructed spectra were then implemented as a set of "measured" observations to feed into NEMESIS retrievals.Using Equation 1, we can reproduce the observations at any other zenith angle, assuming that the Minnaert approximation holds at all other zenith angles in our zenith-angle quadrature scheme, which Irwin et al. (2021) found to be a good approximation for their analysis of VLT/MUSE Neptune observations.
The NEMESIS retrieval code was recently updated to incorporate the effects of Raman scattering and polarization (Irwin, Teanby, Fletcher, et al., 2022), important for wavelengths <0.55 μm in Uranus' atmosphere.To fully capture the effects of these phenomena on reflectivity at the limit of our wavelength range (0.3 μm), we extrapolated the reflectivity at 0.3 μm down to 0.2 μm with a 0.001 μm interval (setting each reflectivity value to that at 0.3 μm).These extrapolated reflectivity values were then multiplied by the solar spectrum at the corresponding wavelength and given very large errors (100%) to prevent NEMESIS from trying to fit to them and skewing our retrievals, whilst simultaneously incorporating effects from Raman scattering and polarization.
Similar to Irwin, Teanby, Fletcher, et al. (2022), we found that we were able to fit the HST/STIS reflectivity spectra to a precision of reduced-χ 2 ∼ 1 if we assumed the errors to be equivalent to 1/50 of the peak reflectivity in each wavelength bin (0.1 μm bin width with a step size of 0.05 μm).This estimated error covers unknown systematic uncertainties from sources such as the Lucy-Richardson spatial deconvolution applied to the HST/ STIS data, the inhomogeneous aerosol structure, and the methane absorption k-table parameters (Irwin, Teanby, Fletcher, et al., 2022).The spectral fit of the IRW23 model to the Minnaert-reconstructed disc-averaged spectrum is shown in Figure 8, having a reduced-χ 2 of 1.16.Retrievals were then carried out on the individual latitude bands over the full wavelength range, with the free parameters in our aerosol model being: aerosol-1 integrated opacity (τ 1 ), aerosol-2 integrated opacity (τ 2 ) and base pressure (p 2 ), aerosol-3 integrated opacity (τ 3 ) and cloud-top CH 4 VMR.All parameters were given 100% a priori errors.The imaginary refractive index spectrum of each layer, defining the reflection/absorption profile of its aerosols with wavelength, was also allowed to vary with 10% a priori errors (the real part of the refractive index was set to 1.4 at 0.8 μm for each layer).
When carrying out latitudinal analyses on these data, it is important to account for the fact that the viewing angle becomes limited as we move from the middle of the disc toward northern polar latitudes or southern mid-latitudes (due to the orbital phase moving toward northern summer solstice after northern spring equinox in 2007), resulting in reduced constraints on our retrievals and Minnaert fits.We accounted for this reduction in constraint within the errors extracted from our Minnaert analyses at each visible latitude band for the two data sets.We carried this out by implementing a scaling factor that was applied to the assumed errors described above (1/50 of the peak reflectivity in each wavelength bin).Our yardstick was chosen as the number of datapoints in each latitude band within our Minnaert analysis (i.e., the number of datapoints on the     vs. μμ 0 plot).This scaling factor was normalized such that it is equal to unity for the visible latitude band with the largest number of datapoints, and therefore increases as we move away from the center of the disc.This scaling factor is defined in Equation 2: Where    is the scaling factor (    ≥ 1) for a given latitude band ϕ band , n is the number of datapoints in the latitude band being considered, and n max denotes the largest number of datapoints found within a single latitude band across all visible latitude bands.
The resulting spectral fits of these retrievals at representative latitude bands, in the equatorial and polar regions, are shown in Figures 9 and 10. and in which it is not (bottom) (see Figure 2).Linear fits to the datapoints from each filter are included.These fits display trends of a decrease in the rate of brightening with time within the hood for the filters in which it is distinctly visible (0.619, 0.657, 0.763, and 0.845 μm), and trends of a virtually constant rate of brightening in comparison for those in which the hood is not (0.467 and 0.727 μm).The 0.547-μm filter does in fact show an increased signature of brightening at polar hood latitudes (45-90°N) compared to equatorial and mid-latitudes despite the hood not being clearly visible (see Figure 2), and shows a trend of a decrease in the rate of brightening over time here. 10.1029/2023JE007904 10 of 29 The resulting retrieved latitudinal profiles of cloud-top CH 4 VMR are displayed in Figure 11.It can be seen that the retrieved profiles for the two data sets are largely similar.However, a small but distinct reduction in cloud-top CH 4 mixing ratio is observed between ∼40°N and 80°N, coinciding with the north polar hood.This suggests that a contributing factor to the temporal brightening between the two observations does indeed originate from a further depletion in polar cloud-top CH 4 mixing ratio with time.However, due to the fact that it is not entirely clear how the latitudinally dependent cloud-top CH 4 VMR varies in the northern hemisphere over short (≤1 year) or long timescales (≥ a few years), further observations and retrieval work would be required to confirm whether this is a continued trend over time.It is entirely plausible that there may be interseasonal variability in the cloud-top CH 4 VMR over short or long timescales in this dynamically active region.The conclusion of this depletion being "generally stable" over time by Sromovsky et al. (2019) appears to be a safe and conservative one in light of this.We thus assert that this conclusion holds for the timespan of our HST/WFC3 data set (2015-2021), enabling us to fix the latitudinal cloud-top CH 4 VMR profile for our HST/WFC3 retrievals to that retrieved on the 2015 HST/STIS observation (Figure 11), the closest available high resolution observation in time to these data and therefore our best estimate.

Brightening of the North Polar Hood From HST/WFC3 Observations
We applied the IRW23 model to our HST/WFC3 data set with the imaginary refractive index spectra and cloud-top CH 4 VMR profile fixed to that retrieved from the 2015 HST/STIS observation at each latitude.We Minnaert-analyzed the observations and, due to WFC3's lack of spectral resolution, set the assumed errors to  1 50 of the peak reflectivity within 0.1 μm wavelength bins, stepped by 0.05 μm, across the wavelength range of each WFC3 filter (reflectivity values were taken from the disc-averaged 2015 HST/STIS observation to determine these errors).We used the Minnaert fits to generate reconstructed spectra at the two chosen zenith angles (0°, 61.45°).The scaling factor from Equation 2 was also implemented in an identical manner to capture the increase in uncertainty toward the extremities of the disc.10.1029/2023JE007904 13 of 29 Retrievals were then carried out on the individual latitude bands using all seven WFC3 filters, with the free parameters in our aerosol model being the: aerosol-1 integrated opacity (τ 1 ), aerosol-2 integrated opacity (τ 2 ) and base pressure (p 2 ), and finally aerosol-3's integrated opacity (τ 3 ).All parameters were given 100% a priori errors.Spectral fits of the resulting retrievals at representative latitudes for the 2015 observations are shown in Figure 12.
Comparable fits were achieved for retrievals on the observations in each of the proceeding years.
The retrieved latitudinal profiles of the free parameters were then plotted for the northern hemisphere to pinpoint the exact aerosol layer changes responsible for the polar hood brightening.The results are displayed for all years, and for the 2015 and 2021 data sets only, in Figures 13 and 14 respectively.Our retrieval results clearly point to the integrated opacity of the 1-2 bar haze layer within the IRW23 model as the main culprit of the brightening, retrieving an average increase in the integrated opacity between 45-90 • N from 2.9 ± 0.2 to 4.1 ± 0.2, a 1.2 ± 0.3 increase (i.e., a ∼41% increase) at 0.8 μm, between 2015 and 2021.All other retrieved profiles for the free parameters are consistent with no change across all northern latitudes over the entire timespan of the data set.Note the large values of τ 1 and its sharp increase at ∼30°N were a consistent pattern seen in all retrievals carried out on this data set.This is a result of τ 1 having a strong correlation with the imaginary refractive index spectrum of aerosol-2, as well as the CH 4 VMR (i.e., changing the a priori values for these parameters significantly affects the retrieved τ 1 profile), which are both fixed for these retrievals, resulting in large retrieved values of τ 1 to compensate.The values retrieved at the equator (∼15-20) compared to the peak at 30°N (∼30-60) are observationally indistinguishable.
The limited spectral resolution of the WFC3 instrument means that we are unable to reliably retrieve any change in the imaginary refractive index spectra of our aerosol layers with any confidence.As these spectra describe the optical properties of the aerosols, detailing the degree to which they are absorbing or reflecting for a given layer, this remains an important missing piece of the puzzle.

Further Latitudinal Analysis From HST/STIS
We sought to clarify whether or not an increase in τ 2 alone was sufficient to account for the spectral fingerprint of the polar hood brightening observed.This fingerprint is displayed in Figure 15 in which an increase in radiance between the two polar spectra longwards of ∼0.6 μm is observed, with the brightening occurring predominantly  at continuum wavelengths, suggesting that its main source originates from aerosol layer changes as opposed to changes in cloud-top CH 4 VMR (equatorial spectra extracted between 20°S-20°N show no difference between the two observations within error (Figure 16)).We employed a forward model approach with the NEMESIS retrieval code to carry out this analysis.In the scenario that an increase in τ 2 alone is not sufficient to match the observations, this approach affords the opportunity to narrow down additional parameters within our model that may be responsible for the observed brightening's spectral signature.
Upon running a forward model with an increase in τ 2 alone (see Figure 17), it became evident that this scenario was very unlikely to be the full picture.Although this solution was able to reproduce a similar spectral brightening longwards of ∼0.6 μm, it also produced a darkening at wavelengths short of ∼0.55 μm.In light of the conclusions reached in Section 3.3, the simplest and most logical next step was to test changing the imaginary refractive index spectrum of aerosol-2 (n i2 ), in addition to an increase in τ 2 .To counteract the short-wave darkening caused by the increase in τ 2 , the aerosol particles need to be more reflective (i.e., a decrease in n i2 ).We therefore ran a forward model with a decrease in n i2 at all wavelengths and an increase in τ 2 , displayed in Figure 18.
This forward model matches the observed spectral fingerprint of the polar hood brightening fairly well.For completeness' sake, we tested forward models for all other combinations of parameters (including changes in the imaginary refractive index spectra of aerosol-1 and aerosol-3) involving an increase in τ 2 , as well as models without increasing τ 2 .The combination of increasing τ 2 and decreasing n i2 was the simplest model that matched the observations fairly well.Informed by this forward model study, latitudinal retrievals were then carried out in an identical manner to those in Section 3.2, but in this case, with n i2 the only imaginary refractive index spectrum free to vary.
The results of these retrievals are projected onto synthetic discs of Uranus in Figures 19 and 20, where the halfdisc has been mirrored to create a full synthetic image of Uranus' disc.The trends observed in τ 1 , τ 2 and cloud-top CH 4 VMR are consistent with strong subsidence in the polar domain in both the middle and upper tropospheres (Fletcher et al., 2020), resulting in the observed polar depletions (relative to the equatorial region).The τ 3 profile displays a concentration of haze in the equatorial region, suggesting a transport of haze from here to higher latitudes (Fletcher et al., 2020).The results are also plotted in Figure 21 displaying the comparison between the retrieved latitudinal parameter profiles in the northern hemisphere, and Figure 22 displays the retrieved imaginary refractive index spectrum profiles at wavelength intervals of 0.1 μm.We see the same distinct increase in integrated opacity north of 45°N that we found from our WFC3 retrievals (see Figure 14), here by an average of 1.09 ± 0.08 at 0.8 μm (a ∼33% increase).We also see the same decrease in cloud-top CH 4 VMR as Figure 11, here by an average of 0.0019 ± 0.0003 between 40°N-80°N.
However, in addition to these changes noted from the aforementioned retrieval analyses of the HST/STIS and HST/ WFC3 observations, we also see an increase in integrated opacity by an average of 0.6 ± 0.1 north of 45°N at 0.8 μm, incident with the polar hood.It should be noted that there is also a signal of increasing τ 1 south of the polar hood between ∼27-40°N.This could suggest an origin of the thickening stemming from south of the polar hood, or that the thickening is completely unrelated to the hood.We will comment on this further in the discussion in Section 4.
Finally, the retrieved latitudinal profiles of aerosol-2's imaginary refractive index spectrum show a marked decrease between the 2012 and 2015 observations (Figure 22) in the 0.7, 0.8, 0.9 and 1.0 μm channels north of ∼40 • N, and between 60-80 • N at 0.5 μm, with the largest change mostly occurring at 60 • N.This result denotes an increase in the scattering of the aerosols at these wavelengths, localized to the north polar hood (i.e., more reflective aerosols at polar hood latitudes within the 1-2 bar haze layer).No clear latitudinal trends in n i2 are observed at the four other, largely short-wave, wavelength channels (i.e., for 0.2, 0.3, 0.4, and 0.6 μm).
When comparing the latitudinal profiles of Figure 21 with those of Figures 13 and 14 obtained from the HST/ WFC3 data set, we see a number of similarities as well as differences.A direct comparison of the latitudinal profiles obtained from the 2015 observations of each data set, show very similar profiles for the opacity (τ 2 ) and pressure (p 2 ) of the 1-2 bar haze, with the p 2 profile displaying slightly more variation in the case of HST/STIS.However, there are distinct differences in the deep haze opacity (τ 1 ) and extended haze opacity (τ 3 ) profiles.The τ 1 profiles display similar latitudinal trends, however, the peak in opacity is shifted equatorwards in the HST/ STIS, peaking at ∼15°N as opposed to ∼25°N for HST/WFC3.There is also a distinct difference in the magnitude of τ 1 , with the HST/WFC3 profile being ∼3-4× larger than that of HST/STIS.Our HST/STIS retrievals revealed a strong correlation between τ 1 , the imaginary refractive index of the 1-2 bar haze layer, and the CH 4 VMR.With these parameters able to vary in the HST/STIS retrievals when fitting to the much higher resolution spectra, the τ 1 profile is ultimately better constrained, leading to the smaller opacity values observed.Note the large equatorial opacity values observed in both profiles are observationally indistinguishable and therefore not contradictory.
A comparison of the τ 3 profiles the most significant difference.The HST/STIS profile shows a concentration of haze at the equator, reducing toward the pole, whereas the HST/WFC3 profile shows a marginal concentration of haze at the equator which flattens out in the midlatitudes and begins to rise north of 40°N, reaching a maximum at the pole.In this case, the HST/STIS results present a much more reliable profile.The greater spectral resolution of the data set, combined with the additional free parameters (CH 4 VMR and aerosol-2 imaginary refractive index spectrum, n i2 ), allowed us to run retrievals fixing each of the additional parameters that appear to contribute to the polar hood brightening (i.e., τ 1 , CH 4 VMR and n i2 ) individually, as well as running a retrieval with all three of these parameters fixed.In each case, the latitudinal trend in τ 3 was almost identical, consistent with the profile in the final retrieval shown in Figure 21.Looking at the 0.727 μm WFC3 filter image in Figure 2, which probes high in the atmosphere, we observe a bright equatorial band displaying a clear concentration of haze in this region for both data sets.
A comparison of the retrieved imaginary refractive index spectra for aerosol-2 at tropical and polar latitude bands is shown in Figure 23, demonstrating the latitudinal variation across time.

Stability of the Polar CH 4 Depletion
As stated in Section 3.2, despite our retrievals showing a small decrease in the polar cloud-top CH 4 VMR at latitudes coincident with the polar hood, we settled on the conclusion of a "general stability" in the polar depletion VMR in the north polar region, this would support the trend returned by our retrievals, suggesting that some sort of dynamical change is occurring (likely a seasonal phenomenon linked to the orbital phase) that is driving a further depletion with time, calling into question its purported stability.One plausible candidate for this is a meridional circulation, previously presented as a potential mechanism to explain the existence of the polar depletion (e.g., Fletcher et al., 2020;Karkoschka & Tomasko, 2009;Sromovsky et al., 2014), that is changing over time.The proposed circulation leads to upwelling at equatorial latitudes and downwelling near the poles.This results in CH 4 -rich gas from the lower troposphere rising up near the equator with CH 4 condensing out as it passes its condensation level (∼1-2 bar).This CH 4 -poor gas is then transported polewards where it descends and likely decreases convection from below the polar hood (Fletcher et al., 2020), leading to the observed upper tropospheric depletion in CH 4 (see Figure 8 of Fletcher et al. (2020)).
A change in the meridional circulation presented above would likely be driven by the increasing insolation experienced by the northern hemisphere as the north pole shifts with time and points more directly toward the Sun as Uranus' orbit progresses toward northern summer solstice.This could result in a slowing of the meridional circulation due to a heating of the north pole, resulting in a decrease in the rate of downwelling there.However, intuition tells us that this would lead to the opposite trend in cloud-top CH 4 VMR than we retrieved, namely an increase in polar cloud-top CH 4 VMR with time.This could hint at a more complex interplay between CH 4 , aerosols and any potential meridional circulation.We will comment further on this in Section 4.2.
If future observations were to instead show an increase in cloud-top CH 4 VMR compared to the retrieved profile produced by our retrievals on the 2015 HST/STIS observation, this would provide further support to the conclusion of a general stability within the polar depletion, hinting at interseasonal variability in the polar cloudtop CH 4 VMR over time.VMR, in line with the degree found from our HST/STIS retrievals (average decrease of ∼10% between 45-90°N between the 2012 and 2015 profiles), would account for a portion of the brightening attributed to the increase in the 1-2 bar haze layer's integrated opacity (τ 2 ).We would expect this to reduce the retrieved increase in τ 2 between the 2015 and 2021 HST/WFC3 observations.To put this prediction to the test, we re-ran our retrievals for the 2021 HST/WFC3 observations with a 20% reduction in cloud-top CH 4 VMR (in line with the degree of reduction retrieved between HST/STIS observations, i.e., 10% over 3 years) between 45-90°N.A comparison of the latitudinal profiles of τ 2 is shown in Figure 24.All other retrieved profiles were omitted as they showed very little to no change within error, being practically identical to those seen in Figure 14.
We see a relatively small reduction in τ 2 north of 45°N as predicted.We therefore assert with confidence that the question of stability in the polar cloud-top CH 4 VMR depletion, in line with the degree of change observed from our HST/STIS retrievals with time, does not have a major bearing on the conclusions of our results.

Brightening of the North Polar Hood
Observations of Uranus' polar hood over the past ∼37 years, first in the southern polar region and, post-equinox, in the northern polar region, suggest that the hood is a seasonally dependent phenomenon (i.e., linked to the orbital phase of the planet).The analysis and results presented in this paper are in alignment with this picture, leading us to a solution with which to explain the observed post-equinox brightening of this long-lived large-scale latitudinal feature.Our initial HST/WFC3 Minnaert analysis confirmed an absolute brightening of the hood, ruling out any viewing geometry-related brightening effects.The observed depletion in cloud-top CH 4 VMR at polar latitudes Karkoschka and Tomasko (2009) and Sromovsky et al. (2014Sromovsky et al. ( , 2019)), resulting in less absorption in the polar region, is one of the main contributors to the brighter appearance of the north polar hood when compared to equatorial latitudes Toledo et al. ( 2018).However, it was not entirely clear whether or not this depletion played a significant role in the observed brightening of the polar hood across time.Sromovsky et al. (2019), analyzing HST/STIS and near-infrared images from Keck's Near Infrared Camera-2 (NIRC-2) instrument, refuted this as a significant source of the temporal brightening, finding wavelengths dominated by hydrogen absorption to also be increasing in brightness.In addition to our Minnaert analysis, we also found the spectral signature of the brightening to include wavelengths dominated by hydrogen absorption (818-832 nm) (see Figure 19) in agreement with Sromovsky et al. (2019), leading us to the same conclusion that changes in the aerosol layers were the predominant cause of the brightening.
In light of this, within the framework of our vertical aerosol model (IRW23 model), we find that the 1-2 bar haze layer (aerosol-2 layer) is the main contributor to the observed brightening of the north polar hood over time via: 1.A thickening north of ∼45°N.
2. An increase in reflectivity of the aerosols north of 40°N and longwards of ∼0.7 μm, and also between 60-80°N at 0.5 μm.
In addition to these changes, we see small contributions to the brightening stemming from an average reduction in cloud-top CH 4 VMR of ∼10% north of ∼40°N, and a thickening of the deep aerosol-1 layer north of ∼45°N.As noted in Section 3.4, the thickening of the aerosol-1 layer is not strictly localized to latitudes coinciding with the polar hood.There is a marked difference between the two regions of thickening established from our retrieval results.South of the polar hood between ∼27-40°N, the average increase in τ 1 retrieved for each latitude band is ∼20%, whereas for the thickening coincident with the hood (north of 45°N), the average increase is ∼56%, almost 3 times as great.It is not clear whether or not these two regions of thickening are linked, but we do see a trend of increasing integrated opacity as we move toward the pole, rising from a ∼23% increase in the retrieved value at 30°N, to a ∼79% increase at 90°N.It is plausible that this could be the result of a dynamical transport of haze polewards at the pressure level of aerosol-1 (p ∼ 4-7 bar).It is noteworthy that these two regions of thickening are separated by a region of no change in the integrated opacity of aerosol-1.This region lies within the two latitude bands that simultaneously capture the boundary of the polar hood and the darker mid-latitudes (bands centered at 40°N and 45°N).
Figure 25 displays a comparison between a forward model with a 56% increase in τ 1 compared to the baseline IRW23 model.We see that a thickening of the aerosol-1 layer, to the degree seen in our HST/STIS retrievals,  does not result in much change at all in the spectrum, apart from two spikes in radiance (brightness) at the edge of the CH 4 absorption bands located at the long end of our wavelength range.When compared to the residuals in Figure 15, we see that although this contribution is small, it is still significant (∼2 sigma at 61.45°).
The combination of the HST/WFC3 and HST/STIS data sets resulted in an ideal complementary set of observations with which to carry out our analysis, providing both a wide temporal window and sufficient spectral resolution and coverage between the two data sets.It is important to note that with only two HST/STIS observations (i.e., only two snapshots in time), it is difficult to assert with a high degree of confidence that the parameters uniquely probed by these retrievals (i.e., cloud-top CH 4 VMR and n i2 ) reflect their trends over longer timescales.
Further observations with comparable spectral resolution and spectral coverage are necessary to provide further context and support for our results.
Toledo et al. ( 2019)'s study constraining the structure, formation and transport of haze in Uranus' atmosphere, via microphysical simulations, found that timescales for haze production are far too slow to account for the post-equinox where the aerosols were seeded in an environment with concentrations of 2% and 10% CH 4 gas respectively.It can be seen that higher CH 4 gas concentrations lead to a decrease in n i for the tholins, that is, more reflective particles, in the laboratory, whereas our retrieved HST/STIS n i spectra decrease by a comparable degree for a relatively small decrease in CH 4 volume mixing ratio.This highlights the added complexity of in situ atmospheric processes and is discussed in Section 4.2.rate of change in haze observed in the northern hemisphere.With simulations returning timescales for haze particles to grow and settle out to be ≳30 years at pressure levels >0.1 bar, Toledo et al. ( 2019) concluded that dynamics are likely the main factor controlling the spatial and temporal distribution of the haze over the poles.In addition to this, the observed changes through equinox are much faster than would be expected from radiative timescales (Conrath et al., 1990) and photochemical timescales (Moses et al., 2018).Toledo et al. ( 2019) put forward a meridional stratospheric transport of haze particles as a plausible explanation for the observed variations in haze structure.As mentioned earlier in Section 4.1, a meridional circulation has also been put forward in the literature to explain the observed polar depletion in cloud-top CH 4 VMR (e.g., Fletcher et al., 2020;Karkoschka & Tomasko, 2009;Sromovsky et al., 2014).One such circulation model is given by Fletcher et al. (2020) (see their Figure 8), which takes into account findings from studies of Uranus' zonal winds, latitudinal variations and stratospheric chemical tracers, amongst other considerations.This model consists of a number of vertically stacked circulation cells, including a stratospheric meridional circulation with upwelling at mid-latitudes and downwelling at the equator and poles.It is plausible that the thickening of the 1-2 bar haze layer that we find could be driven by such a circulation, with the transport of haze particles from equatorial regions to the mid-latitudes and onto the poles, where upon descent, they settle and accumulate at the 1-2 bar region of static stability (see Irwin, Teanby, Fletcher, et al. (2022) Figure 23).It is important to note that although Fletcher et al. ( 2020)'s model is consistent with the findings from the observational studies considered in its construction, and despite it also being consistent with the trends observed in our Figures 19 and 20, it is yet to be reproduced by global circulation models of Uranus' atmosphere.
Returning to the discussion from Section 4.1, we speculated that as the season progresses and Uranus' northern hemisphere receives an increasing amount of insolation (i.e., increasing amounts of energy), this would result in a slowing of any meridional circulation that involves downwelling at the north pole via heating.This would be compounded by the fact that Uranus has a low internal source of heat (Pearl et al., 1990), providing relatively little energy to drive atmospheric dynamics from below, meaning any changes in solar flux are likely to have greater influence.However, a scenario in which a stratospheric meridional circulation cell undergoes this slowing effect via heating, but the tropospheric cell below is not effected would eliminate the contradiction within our results.
In this scenario, a slowing in the rate of transport of stratospheric haze polewards would be expected, resulting in less haze particles downwelling at the pole with time to seed the 1-2 bar haze layer, and a reduction in the rate of thickening of the aerosol-2 haze layer within the polar hood.Analyzing the average retrieved rate of change in τ 2 between consecutive years at polar hood latitudes (45-90°N) resulted in no definitive trend being observed in the rate of brightening over time, but did not rule this out.In Section 3.1 we found a tentative decrease in the rate of polar hood brightening over the timespan of the HST/WFC3 data set (see Figure 7).Although these results are consistent with the slowing of a stratospheric meridional circulation, we highlight that this model is speculative.Future work is certainly needed to shed more light on the circulation of the atmosphere of Uranus.
The 1-2 bar haze layer is coincident with pressures in which CH 4 is predicted to condense.Therefore, we expect the aerosols at 1-2 bar to act as cloud condensation nuclei (CCN), accumulating CH 4 ice as it condenses.This condensation results in an increase in mass of the aerosols as they accumulate more and more CH 4 , as well as a change in their optical properties (e.g., making them more absorbing or reflective at particular wavelengths).Unlike Neptune, Uranus' atmosphere does not seem to produce any stable, long-lasting CH 4 clouds (Irwin, Teanby, Fletcher, et al., 2022;Irwin et al., 2017).One plausible explanation for the absence of CH 4 clouds predicts that as the aerosols accumulate CH 4 , they quickly reach a "critical mass" and begin to sink, precipitating out of the 1-2 bar haze layer (Irwin, Teanby, Fletcher, et al., 2022).These precipitating CH 4 ice aerosols then sublimate as they fall deeper into the atmosphere where temperatures are higher, releasing the photochemical haze particles that served as CCN, making them readily available to seed H 2 S condensation in the 4-7 bar region (Irwin, Teanby, Fletcher, et al., 2022).It is important to note that the polar depletion in upper tropospheric CH 4 mixing ratio would alter its saturation vapor pressure curve, resulting in its condensation level being pushed higher in the atmosphere given the same temperature (i.e., to a lower pressure level).This would result in less condensation, assuming significant condensation is still occurring, for the same pressure level when compared to equatorial latitudes, resulting in this process of accumulation and precipitation being slower at high latitudes.In the extreme case, if the atmospheric circulation exhibits strong subsidence at the latitude in question, CH 4 condensation may be suppressed entirely.
We thus posit the question, what effect would a thickening of the aerosol-2 layer have on this proposed process?CH 4 condensation is predicted to happen rapidly, with estimates of CH 4 ice particles at 1-2 bar in Ice Giant atmospheres growing to a size of ∼5 mm in as little as ∼100s (Carlson et al., 1988).The thickening of the aerosol-2 layer with time (i.e., the accumulation of haze particles between 1-2 bar) within the polar hood would provide further CCN for CH 4 condensation.This could result in an increase in the number of particles precipitating out per unit of time, given a constant rate of accumulation of CH 4 , resulting in more haze particles being readily available below this level as CCN for H 2 S condensation.This could explain the thickening of the aerosol-1 layer found from our HST/STIS retrievals.However, relaxing our assumption of a general stability in the polar cloud-top CH 4 VMR depletion, the retrieved decrease in polar cloud-top CH 4 VMR would likely act to slow down the rate of accumulation of CH 4 onto the haze particles.This is not only due to less CH 4 being available for condensation but also the lower VMR leading to the condensation level itself being slightly higher in the atmosphere.This would result in the aerosols remaining within the 1-2 bar layer for a longer period before precipitation occurs, if at all.These longer-lasting aerosols would likely be smaller in size compared to at other latitudes, and could result in a change in the aerosol optical properties that matches the retrieved signature of a decrease in the imaginary refractive index spectrum (i.e., an increase in reflectivity of the aerosols).This is speculative as it is not currently known what the spectral identity of an increase in reflectivity of these aerosols would look like, and therefore how it would match up to the retrieved signature of an increase in reflectivity at ∼0.5 μm and longwards of 0.7 μm.Work carried out by Mahjoub et al. (2012) on the influence of CH 4 gas concentration on the optical properties of Titan aerosol analogs does, however, provide plausibility to this speculative claim.They found that an increase in CH 4 concentration caused a decrease in the imaginary refractive index spectrum across the 0.37-1.0μm wavelength range for their aerosol analogs.A scenario in which the thickening of the aerosol-2 layer, combined with a decrease in polar cloud-top CH 4 VMR, slows the rate of accumulation of CH 4 onto the haze particles, but to a degree whereby the number of aerosols precipitating out per unit of time is still increasing with time, could simultaneously explain the thickening of the aerosol-1 layer and the increase in reflectivity of the aerosol-2 particles.

Conclusions
We used observations of Uranus from the HST/WFC3 and HST/STIS instruments to analyze its north polar hood, a large-scale latitudinal atmospheric feature.The main objective was to narrow down the predominant cause of the hood's recent evolution (i.e., an apparent brightening).By combining complementary data sets, and adopting the IRW23 holistic vertical aerosol model, the work presented in this paper provides a solution using a retrieval and forward-model approach with the NEMESIS retrieval code.Our conclusions can be summarized as follows: 1.A Minnaert limb-darkening analysis of our HST/WFC3 data set confirmed an absolute brightening of the polar hood, ruling out viewing geometry-related brightening effects as the main cause.We also observed a tentative reduction in the rate of polar hood brightening with time.2. Retrievals on our HST/STIS supporting data set revealed a further depletion in cloud-top CH 4 VMR, by an average of 0.0019 ± 0.0003 between 40°N and 80°N (∼10% average reduction), between 2012 and 2015.However, more work is needed to reveal whether this is a long-term or short-term trend.3. A combination of HST/WFC3 and HST/STIS latitudinal retrieval analyses found the main contributions responsible for the observed brightening of the polar hood to be, within the framework of the IRW23 vertical aerosol model: • A thickening of the aerosol-2 layer-finding an average increase in integrated opacity of 1.09 ± 0.08 north of ∼45°N at 0.8 μm (a ∼33% average increase).• An increase in reflectivity of the aerosol-2 particles north of ∼40 • N and longwards of ∼0.7 μm, and also between 60°N and 80°N at ∼0.5 μm. 4. From our HST/STIS retrievals, we also find small contributions to the polar brightening stemming from a thickening of the aerosol-1 layer, finding an average increase in integrated opacity of 0.6 ± 0.1 north of 45°N at 0.8 μm (i.e., a ∼56% average increase) between 2012 and 2015, and from the aforementioned average decrease in cloud-top CH 4 mixing ratio of 0.0019 ± 0.0003 between 40°N and 80°N (a ∼10% decrease) 5. Our results are consistent with the slowing of a stratospheric meridional circulation, similar to that posited by Fletcher et al. (2020), exhibiting downwelling at the poles.6. Future observations with HST/STIS, or an instrument with comparable spectral resolution and coverage (e.g., the VLT Multi Unit Spectroscopic Explorer), would enable more light to be shed on the question of the trend in the latitudinal distribution of cloud-top CH 4 VMR over short and long timescales.These observations would either provide support for a further depletion in polar cloud-top CH 4 VMR with time, or an interseasonal variability in the VMR with time, supporting the picture of a general stability.

Figure 1 .
Figure 1.The seven Hubble Space Telescope's (HST)/Wide-Field Camera 3 filters used for 2015-2021 observations of Uranus on a plot of throughput against wavelength in microns.The dashed gray line displays the disc-averaged reflectivity (I/F) spectrum of the 2015 HST/Space Telescope Imaging Spectrograph cube as a reference for the spectral features captured by each filter.

Figure 2 .
Figure 2. Hubble Space Telescope's/Wide-Field Camera 3 images in each of the seven filters for the 2015 (top) and 2021 (bottom) data sets.The 0.467-μm images include superimposed latitude-longitude grids with spacing of 15° between lines of latitude and 30° between lines of longitude.The north pole (at upper right of disc) is seen to swing further into view over the 6-year timespan between the two sets of observations as Uranus approaches northern summer solstice in 2030.

Figure 3 .
Figure 3.A plot of the reflectivity spectrum extracted from a point at the center of the disc of our 2015 Hubble Space Telescope's/Space Telescope Imaging Spectrograph observation, demonstrating the significant increase in noise (decrease in signal-to-noise ratio) longwards of ∼0.95 μm.Images at 0.552, 0.62, 0.83, and 0.958 μm are shown above to visually demonstrate the increase in noise, covering representative wavelengths from both continuum and methane-absorbing regions.

Figure 4 .
Figure 4. Plot of particle density against pressure retrieved from the disc-averaged Hubble Space Telescope's/Space Telescope Imaging Spectrograph 2015 observations using the IRW23 vertical aerosol model.

Figure 5
Figure 5 also appears to show signs of a decrease in the amount of brightening year on year.To investigate this, we took the average change in brightness (i.e., change in  (   ) 0

Figure 5 .
Figure 5. Plot of the fitted Minnaert parameters  (   ) 0 and k versus latitude for the 0.845 μm filter observations of our Hubble Space Telescope's/Wide-Field Camera 3 data set of Uranus.A clear trend is observed of a relatively large increase in brightness (i.e.,  (   ) 0 ) north of ∼45°N over time.A very similar trend is also observed in the limb-darkening parameter, k.The legend applies to both plots.The black dashed line on the lower plot represents the transition point between limb-darkening (k > 0.5) and limb-brightening (k < 0.5).Errors are shown to 3σ and have been scaled according to Equation 2 to account for the lack of constraint at high latitudes.

Figure 6 .
Figure 6.Plot of the fitted Minnaert parameters  (   ) 0 and k versus latitude for the 0.727 μm filter observations of our Hubble Space Telescope's/Wide-Field Camera 3 data set of Uranus.No clear, simple trend in the Minnaert parameters is observed with time, but an overall small but fairly consistent brightening across latitudes is seen across the entire timespan of the data.The legend applies to both plots.Errors are shown to 3σ and have been scaled according to Equation 2 to account for the lack of constraint at high latitudes.

Figure 7 .
Figure 7. Plots of the average rate of change in brightness (i.e., the rate of change in  (   ) 0 ) against the 1-year time window considered.Datapoints are calculated from the average change across polar hood latitudes (45-90 • N) between consecutive years in the Hubble Space Telescope's/Wide-Field Camera 3 data set.Plots are separated by the filters in which the hood is distinctly visible (top) and in which it is not (bottom) (see Figure2).Linear fits to the datapoints from each filter are included.These fits display trends of a decrease in the rate of brightening with time within the hood for the filters in which it is distinctly visible (0.619, 0.657, 0.763, and 0.845 μm), and trends of a virtually constant rate of brightening in comparison for those in which the hood is not (0.467 and 0.727 μm).The 0.547-μm filter does in fact show an increased signature of brightening at polar hood latitudes (45-90°N) compared to equatorial and mid-latitudes despite the hood not being clearly visible (see Figure2), and shows a trend of a decrease in the rate of brightening over time here.

Figure 8 .
Figure 8. Plot of disc-averaged radiance against wavelength for the 2015 Hubble Space Telescope's/Space Telescope Imaging Spectrograph cube at the two chosen Minnaert zenith angles (0°, 61.45°), displaying the spectral fit to the disc-averaged reconstructed Minnaert spectrum of the IRW23 aerosol model.The fit has a reduced-χ 2 value of 1.16.Note, residuals are divided by the error on the measured spectrum.

Figure 9 .
Figure 9. Plots of radiance against wavelength showing the spectral fits at the two chosen Minnaert zenith angles (0°, 61.45°) for retrievals carried out on the 2012 Space Telescope Imaging Spectrograph observation at representative latitude bands of 10-20°N (top) and 50-60°N (bottom).Note, residuals are divided by the error on the measured spectrum.

Figure 10 .
Figure 10.Plots of radiance against wavelength showing the spectral fits at the two chosen Minnaert zenith angles (0°, 61.45°) for retrievals carried out on the 2015 Space Telescope Imaging Spectrograph observation at representative latitude bands of 10-20°N (top) and 50-60°N (bottom).Note, residuals are divided by the error on the measured spectrum.

Figure 11 .
Figure 11.Plot of volume mixing ratio against latitude displaying the latitudinal profiles of the cloud-top CH 4 mixing ratio retrieved from the 2012 and 2015 Space Telescope Imaging Spectrograph observations.The two profiles appear largely similar, however, the 2015 profile shows a small reduction in cloud-top CH 4 mixing ratio between ∼40-80°N.The established negative gradient in cloud-top CH 4 mixing ratio when moving from equator to pole is clearly seen.The black dashed line denotes the a priori value of 2.97%.

Figure 12 .
Figure 12.Spectral fits at the two chosen Minnaert zenith angles (0°, 61.45°) for retrievals carried out on the 2015 Wide-Field Camera 3 observations at representative latitude bands of 10-20°N (top) and 50-60°N (bottom), with reduced-χ 2 values of ∼1.53 and ∼0.66 respectively.Note, residuals are divided by the error on the measured spectrum.

Figure 14 .
Figure 14.Latitudinal profiles in the northern hemisphere of the free parameters in the IRW23 vertical aerosol model retrieved from the 2015 and 2021 Hubble Space Telescope's/Wide-Field Camera 3 observations only: aerosol-1 integrated opacity (τ 1 ), aerosol-2 base pressure (p 2 ) and integrated opacity (τ 2 ), and aerosol-3 integrated opacity (τ 3 ).This plot demonstrates the overall change over the 6-year timespan of the data set, showing an increase in τ 2 northwards of ∼45°N as the only definitive change.

Figure 15 .
Figure 15.Plot of the 2012 and 2015 Hubble Space Telescope's/Space Telescope Imaging Spectrograph polar spectra extracted between 50°N and 90°N for our two chosen Minnaert zenith angles (0°, 61.45°).A clear increase in the radiance (i.e., a brightening) can be seen longwards of ∼0.6 μm between the two spectra.The residuals display the spectral fingerprint of this brightening.The brightening occurs predominantly at continuum wavelengths, suggesting that its main source originates from aerosol layer changes as opposed to changes in cloud-top CH 4 volume mixing ratio.Note, residuals are divided by the error on the measured 2012 spectrum.

Figure 16 .
Figure 16.As Figure 15 but for the equatorial region (20°S to 20°N).Very little difference between the two measured spectra is observed.Note, residuals are divided by the error on the measured 2012 spectrum.

Figure 17 .
Figure 17.Plot comparing the retrieved spectrum of the IRW23 best-fit model at polar latitudes (50-90°N), carried out on the 2012 Hubble Space Telescope's/ Space Telescope Imaging Spectrograph observation, to a forward model spectrum of a modified version of this model, testing an increase in the integrated opacity of aerosol-2 (τ 2 ).The residuals show a clear difference in profile from those of Figure 15, suggesting that an increase in τ 2 alone is not sufficient to explain the nature of the observed polar hood brightening.Note, residuals are divided by the error on the best-fit model's measured spectrum.

Figure 18 .
Figure 18.Plot comparing the retrieved spectrum of the IRW23 best-fit model at polar latitudes (50-90°N), carried out on the 2012 Hubble Space Telescope's/Space Telescope Imaging Spectrograph observation, to a forward model spectrum of a modified version of this model, testing an increase in the integrated opacity of aerosol-2 (τ 2 ) and a decrease at all wavelengths in the imaginary refractive index spectrum of aerosol-2.The residuals display a close resemblance to those from Figure 15.Note, residuals are divided by the error on the best-fit model's measured spectrum.

Figure 19 .
Figure19.Synthetic images displaying the retrieved profiles of the free parameters (except the aerosol-2 imaginary refractive index spectrum shown in Figure22) within our 2012 Hubble Space Telescope's/Space Telescope Imaging Spectrograph retrievals: aerosol-1 integrated opacity (τ 1 ), aerosol-2 integrated opacity (τ 2 ) and pressure (p 2 ), aerosol-3 integrated opacity (τ 3 ), and cloud-top CH 4 volume mixing ratio (labeled above each image).The north pole is located near to the top of the disc, with the red dashed line signifying the equator and the blue dashed lines signifying ±45°.A comparison of the τ 2 image with the corresponding one in Figure20reveals an increase in τ 2 in the north polar region.Note the sharp decrease in τ 2 at the pole is likely an artifact of the lack of constraint here.

Figure 20 .
Figure 20.Similar to Figure 22 but displaying retrieved profiles from the 2015 Hubble Space Telescope's/Space Telescope Imaging Spectrograph retrievals.Again, the aerosol-2 imaginary refractive index spectrum for these retrievals is shown in Figure 22.Note the sharp increase in τ 2 at the pole is likely an artifact of the lack of constraint here.

Figure 22 .
Figure 22.Retrieved latitudinal profiles of the aerosol-2 imaginary refractive index spectrum (n i2 ) at each wavelength interval in the IRW23 vertical aerosol model from our 2012 and 2015 Hubble Space Telescope's/Space Telescope Imaging Spectrograph retrievals.A clear decrease in n i2 is observed north of ∼40 • N for wavelengths longwards of ∼0.7 μm.A decrease in n i2 is also seen at 0.5 μm between ∼60-80°N.For the most part, the largest reduction occurs for the latitude band centered at 60°N.The black dashed lines represent the a priori values.

Figure 23 .
Figure 23.A comparison of the retrieved imaginary refractive index spectra (n i2 ) for the aerosol-2 layer at representative tropical (25-35°N) and polar (55-65°N) latitudes from our 2012 and 2015 Hubble Space Telescope's (HST)/Space Telescope Imaging Spectrograph (STIS) retrievals.A decrease in n i2 is observed with time here for wavelengths longwards of ∼0.6 μm.The imaginary refractive index spectra of Mahjoub et al. (2012)'s Titan aerosol analogs (tholins) are included for comparison, where the aerosols were seeded in an environment with concentrations of 2% and 10% CH 4 gas respectively.It can be seen that higher CH 4 gas concentrations lead to a decrease in n i for the tholins, that is, more reflective particles, in the laboratory, whereas our retrieved HST/STIS n i spectra decrease by a comparable degree for a relatively small decrease in CH 4 volume mixing ratio.This highlights the added complexity of in situ atmospheric processes and is discussed in Section 4.2.

Figure 24 .
Figure24.Plot of opacity against latitude displaying the retrieved latitudinal profiles of the integrated opacity of aerosol-2 (τ 2 ) produced from retrievals on the 2015 and 2021 Hubble Space Telescope's (HST)/Wide-Field Camera 3 observations.The green line represents retrievals carried out on the 2021 observations in an identical manner to those of Section 3.3 but with the addition of a 20% reduction in cloud-top CH 4 volume mixing ratio (consistent with the average decrease with time found from our HST/Space Telescope Imaging Spectrograph retrievals) between 45-90°N.A relatively small reduction in τ 2 north of 45°N can be seen as predicted.

Figure 25 .
Figure 25.Plot comparing the retrieved spectrum of the IRW23 best-fit model at polar latitudes (50-90°N) carried out on the 2012 Hubble Space Telescope's/Space Telescope Imaging Spectrograph observation to a forward model spectrum of a modified version of this model, testing an increase in the integrated opacity of aerosol-1(τ 1 ) only by 56%.We see from the residuals that there is only significant change to the spectrum at two small well-defined regions between 0.8-0.95μm.Note, residuals are divided by the error on the best-fit model's measured spectrum.