Atmospheric Light Pollution by Proposed Reflect Orbital Space Mirrors

Abstract

ReflectOrbital is planning to launch a $18\times 18$ m mirror in space at a $\sim$600 km altitude to illuminate a $2.5$ km radius circular patch on Earth. The pilot satellite, called EARENDIL-1, is a pathfinder for a constellation of $\sim$50,000 larger mirrors, each with $54\times 54$ m in diameter, with the explicit goal to provide space-borne illumination on Earth. We calculate the sky brightness caused by the Rayleigh scattering and aerosol scattering of the incoming beam from one such satellite and by the light reflected back by the illuminated patch, for various atmospheric properties and ground albedos under cloud-free conditions. We demonstrate that the light pollution caused by one of these satellites is very significant, altering the nighttime environment up to distances of $\sim$30 kms. The results show that for an observer within the beam of a single 54 m satellite, the mirror will appear as a $-16.7$ mag point-like source, i.e. about 4 magnitudes brighter than the full moon. The diffuse sky background will be similar to the dusk sky shortly after sunset; too bright to see even the brightest stars. From a distance of 14 km, the glow from a single mirror will exceed the luminance of the full moon sky for majority of the sky. From a distance of 34 km the sky will still appear brighter than the moon-lit sky in the direction of the beam. If 400 such mirrors illuminate the same patch simultaneously, the glow will be obvious from 80 kms.

I Introduction

The idea of lighting up Earth from space is not new. The first documented idea came from the pioneering theorist of astronautics, Hermann Oberth, in the 1920s. Oberth described the concept of large orbital mirrors to reflect sunlight onto Earth for illumination (Oberth, 1923, 1929). In the 1990s, the Russian Znamya program was a set of experiments investigating deployable space mirrors for reflecting sunlight onto Earth (NASA, 1993; European Space Agency, 1990). In 1993, Znamya 2 successfully deployed a $\sim$20 m reflective membrane from the Mir space station, producing a moving illuminated spot on Earth with the illumination level roughly comparable to that of moonlight from the full moon. A subsequent attempt (Znamya 2.5) failed to deploy properly, underscoring the challenges associated with large, lightweight orbital reflectors.

Recently, the startup company Reflect Orbital filed a request to the United States Federal Communications Commission (FCC) under FCC File No. SAT-LOA-20230825-00198 (Reflect Orbital, 2023) to launch an $18\times 18$ m mirror in space as a pathfinder experiment toward establishing a giant constellation of 50,000 mirrors, each with $54\times 54$ m in size. The request was recently approved by the FCC without considering the environmental and safety impacts. Plans from Reflect Orbital lack the quantification of the resulting, potentially significant light pollution in Earth’s atmosphere caused by such mirrors. The light pollution comes from the following sources: i) the scattering of the incoming, $\sim 2.5$ km radius beam in the atmosphere (air molecules and aerosols), ii) the scattering of the light that is reflected back by Earth into the atmosphere, and iii) the Mie scattering of the light on clouds (with an estimated $\sim$67% of the Earth covered in clouds). The plans also lack consideration of the environmental damage caused by the light pollution due to the satellites.

In this paper, we describe the light pollution by performing a detailed calculation of the scattered light in Earth’s atmosphere using a 3-dimensional Radiative Transfer Equation model, and we characterize the night sky brightness distribution at varying distances from the center of illuminated area. We consider several combinations of the aerosol asymmetry parameter (g) and the surface albedo ($\alpha$). The calculations assume an aerosol optical depth of $0.30$ at the reference wavelength of 550 nm, a value that is typical for many regions of the world (Qin et al., 2021; Shikwambana and Sivakumar, 2018; Chin et al., 2004). For simplicity, we assume a cloud-free atmosphere with the understanding that under cloudy conditions, e.g. with a cloud deck at 10 km altitude illuminated by the beam, the overall illumination distribution will be starkly different. Since the prototype satellite is a pathfinder for the much larger mirrors, we performed calculations for the large mirrors. Scaling our results to the 18 m prototype mirror is trivial, as the light reflected by the 18 m mirror is 1/9th that of the proposed 54 m mirrors.

The paper is structured as follows. In Sec. II, we carry out a simple calculation of the light levels by the prototype (18 m) and large (54 m) mirrors without any scattering. For the rest of the paper we only consider the 54 m scenario. In Sec. III we show the sky brightness for observers outside the illuminated area, while in Sec. IV we carry out calculations for an observer standing in the illuminated patch. In Sec. V we compare the results to the sky background due to the full moon, and in Sec. VI we summarize our conclusions.

II Ballpark Figures

Figure 1: Schematic view of a proposed space-mirror illuminating the Earth. The observer is past astronomical twilight, with the Sun 18° below the horizon. “Mirror1” is seen at a 45° elevation from the observer, while “Mirror2” is directly overhead. Yellow arrows indicate the scattering of light in the atmosphere and reflected back from the ground.

First, we characterize the parameters of the 18 m prototype (EARENDIL-1) mirror. For simplicity, we assume that the observer is standing right on the terminator (right after sunset), with the mirror in zenith projecting light directly underneath. The mirror is tilted at $\epsilon=45$ degrees with respect to the direction of the Sun and also the direction of the observer. As we show later, different mutual positions of the observer and the mirror within reasonable limits do not fundamentally change the ballpark figures. We assume a mirror size of $18\times 18~\mathrm{m}$ and a reflectance of 0.9, an altitude of the mirror of $H=600~\mathrm{km}$, and an aerosol optical depth of $\tau_{a}=0.3$, a Rayleigh optical depth of $\tau_{R}=0.1$, leading to a combined optical depth of $\tau=0.4$ and an overall transmission of $T=e^{-\tau}\approx 0.67$. The solar irradiance is $F_{\odot}=1361~\mathrm{W\,m^{-2}}$, and the luminous efficacy is $110~\mathrm{lm/W}$ (Illuminating Engineering Society, 2020). The projected dimensions of the mirror are: $18~\mathrm{m}\times(18\cos 45\arcdeg)\,\mathrm{m}=18~\mathrm{m}\times 12.7~\mathrm{m},$ leading to an apparent angular size of $\approx 6.2^{\prime\prime}\times 4.4^{\prime\prime},$ and a projected collecting area of $A_{\rm proj}=18\times 18\times\cos 45^{\circ}\mathrm{m^{2}}\approx 229~\mathrm{m^{2}}$. To calculate the apparent brightness, we calculate the incident power using the solar constant and the projected area as:

$$P_{\rm in}=1361\mathrm{W/m^{2}}\times 229\mathrm{m^{2}}\approx 3.12\times 10^{5}~\mathrm{W}.$$

With the mirror reflectance of 0.9, the reflected power is:

$$P_{\rm refl}=0.9\,P_{\rm in}\approx 2.81\times 10^{5}~\mathrm{W},$$

and the transmitted power is:

$$P_{\rm ground}=P_{\rm refl}e^{-0.4}\approx 1.88\times 10^{5}~\mathrm{W}.$$

Using the claimed spot area of 2.5 km radius, and the area of the circle of $A_{\rm spot}=\pi(2500)^{2}\mathrm{m^{2}}\approx 1.96\times 10^{7}~\mathrm{m^{2}},$ the irradiance is

$$E=\frac{P_{\rm ground}}{A_{\rm spot}}=\frac{1.88\times 10^{5}\mathrm{W}}{1.96\times 10^{7}\mathrm{m^{2}}}\approx 0.0096~\mathrm{W/m^{2}}$$

and the illuminance is

$$E_{\rm lux}\approx 0.0096\mathrm{W/m^{2}}\times 110\mathrm{lm/W}\approx 1.05~\mathrm{lux}.$$

To place this into context, the approximate horizontal illuminance due to the full Moon near zenith is: $\sim 0.26~\mathrm{lux}$ (Kyba et al., 2017), leading to a horizontal ground illuminance ratio of the prototype satellite to the full Moon of: $1.05/0.26\approx 4.0$.

The approximate apparent magnitude of the mirror, using the full Moon for scaling, is:

$$m=-12.74-2.5\log_{10}\left(\frac{1.05}{0.25}\right)\approx-14.3\,.$$

Finally, with an orbital velocity of $v\approx 7.56~\mathrm{km/s},$ the angular speed of the mirror is

$$\omega=\frac{v}{H}\approx 0.72^{\circ}/\mathrm{s}\approx 2600~\mathrm{arcsec/s}.$$

In summary, the prototype mirror with $18\times 18$ m size will cause a $\sim 4$ times full-Moon illumination of $1.05~\mathrm{lux}$ over a $\sim 2.5~\mathrm{km}$ radius region, delivering $\sim 0.0096~\mathrm{W/m^{2}}$ of energy, and will appear as a $-14.3$ magnitude point-like source, moving at an approximate angular speed of $\omega\approx 0.72^{\circ}/\mathrm{s}$.

A similar calculation for the proposed, $54\times 54$ m square mirror leads to a $\sim 38$ times full-Moon illumination of $9.5~\mathrm{lux}$ (equivalent to street-light level lighting) over the same $\sim 2.5~\mathrm{km}$ radius region, delivering $\sim 0.086~\mathrm{W/m^{2}}$ of energy, and will appear as a $-16.7$ magnitude, still point-like ($18.6^{\prime\prime}\times 13.1^{\prime\prime}$) source.

If we now assume that the observer has just entered full darkness, i.e., past the astronomical twilight, with the sun elevation of $-18\arcdeg$, and we also assume that the mirror is at $45\arcdeg$ elevation (toward the West), simple geometry shows that the mirror is $\alpha_{M}\approx 13.25\arcdeg$ from the terminator as seen from the center of the Earth, lagging behind the observer (who is already $\beta_{O}=18\arcdeg$ past the terminator; see Fig. 1). The normal vector of the mirror is tilted by $\epsilon\approx 58.5\arcdeg$ with respect to the direction of the observer, yielding a projected surface area of $\cos(58.5\arcdeg)\approx 0.522$ that of the entire mirror area. Taking into account the slightly larger optical depth due to the slant angle (factor of $e^{-0.414\tau_{0}}=e^{-0.414\cdot 0.4}=0.847$), using the same atmospheric parameters, assuming that the size of the illuminated patch is still the same, we get that this tilted mirror will be $\sim 24$ times full-Moon illumination of $5.95~\mathrm{lux}$, delivering $\approx 0.054~\mathrm{W/m^{2}}$ of energy, and will appear as a $-16.19$ magnitude source.

In the same scenario, with the observer just entering full darkness ($\beta_{O}=18\arcdeg$), but with the mirror being in local zenith ($\alpha_{M}=18\arcdeg$), the normal vector of the mirror is tilted by $\epsilon=36\arcdeg$ with respect to the direction of the observer, yielding a projected surface area of $\cos(36\arcdeg)\approx 0.809$ that of the entire mirror area. The atmospheric optical depth is as in the first scenario, when the observer was straight on the terminator, because the mirror is still in zenith. In this situation the tilted mirror will be $\sim 44$ times full-Moon illumination of $10.9~\mathrm{lux}$, delivering $\approx 0.098~\mathrm{W/m^{2}}$ of energy, and will appear as a $-16.85$ magnitude source. In brief, the brightness of the satellite will barely change between these configurations, and it is reasonable to assume that the optimum satellite configurations would be used for illumination, i.e., for an observer at $\beta_{O}$ degrees past sunset, selecting the $\alpha_{M}$ longitude of the mirror such that illumination is performed with a relatively large projected mirror size and at a small optical depth. In summary, our baseline assumption of the mirror being in zenith and tilted at 45° is good to about $\sim 15-20$%, and slightly underestimates a situation whereby the observer is already past astronomical twilight and the mirror is overhead. The exact number will depend on whether the mirrors are deformable, and the patch size is kept constant.

Using 400 mirrors, all at 45° elevation, and assuming a mirror reflectance of 0.9, the combined brightness will be $\sim 2400$ lux, or about $\sim 10{,}000$ times brighter than the illuminance due to the full Moon, or about $2.4\%$ of the illuminance due to the Sun. These 400 mirrors can deliver roughly $22~\mathrm{W/m^{2}}$ power, which is close to the lower limit where solar panels can start operating.

III Sky Brightness Outside The Beam

We used the Modified Successive Orders of Scattering (MSOS) model (Kocifaj, 2018; Kocifaj et al., 2023), which is a 3D Radiative Transfer Equation model to compute the field of scattered radiation at any point in 3D space (on a chosen 3D grid), both for downward‑ and upward‑propagating radiation. The original code was developed for light emitted upward from ground-based sources and can have any angular emission pattern, such as a light cone. If the atmosphere was homogeneous throughout its entire volume, adapting the MSOS code to the present case of an incoming beam from space would be straightforward; we would simply compute the distribution of upward radiation at the top of the atmosphere and then rotate the coordinate system. This would effectively move the ground‑based source to the top of the atmosphere, and what we compute as upward radiation at the top of the atmosphere would become downward radiation at the Earth’s surface.

However, the atmosphere is a vertically stratified medium: air and aerosol densities decrease exponentially with altitude in a defined manner. In our calculations, we assumed a scale height for the air density of 8000 m, and a scale height for the aerosols of 2000 m (Qiu et al., 2005). We kept the MSOS model as it is, removed the ground‑based sources, and simulated the nadir‑pointing mirror by defining a brightness field at the top of the atmosphere (TOA) such that the radiance at zenith ($z=0\arcdeg$) is set to unity and is zero in all other directions. We applied the same procedure for all modeled altitudes above the Earth’s surface, with the brightness reduced according to atmospheric extinction. We then ran the computation for 9 scattering orders to ensure reliable convergence of the model (in principle, $\sim$3 orders are sufficient).

Because the MSOS model operates on a fixed spatial grid common to all scattering orders, applying it directly to a narrow collimated beam presents a resolution challenge. A highly directional source concentrates the optical signal within a small solid angle, meaning that at each altitude level the beam footprint would be contained within a single voxel of the standard grid. Such a grid may be insufficient to resolve the generating function of the first scattering order accurately (within the light cone).

We therefore adopted a hybrid approach: the contribution of higher scattering orders (as described in this section) were obtained from the MSOS model, and the single-scattered radiance within the light cone was computed analytically using the formalism described in Sec. IV, which permits arbitrarily fine angular sampling at modest computational cost. Because the directional preference of the photons is progressively lost with increasing scattering order, the higher-order radiance patterns become gradually smoother and form a slowly varying continuum. The fixed MSOS grid resolves this continuum adequately. Only the single-scattering (first-order) field varies rapidly across the narrow cone, so the fixed grid cannot capture its fine features. Refining the grid to do so would impose the same fine sampling on every order. This would be costly in memory and computing time, although only the first order requires it. For this reason, the first-order field within the cone is evaluated analytically, which permits arbitrarily fine sampling. The single-scattering and higher-order fields are then superimposed to give the total radiance. In practice this superposition is realized through a pre-computed ratio. From a full MSOS run, in which all orders share the same grid, we form the ratio of the total (all-order) radiance to the first-order radiance for each viewing direction. The total radiance is then recovered as the product of the accurate analytical first-order field and this smoothly varying ratio. The procedure retains the full angular detail of the direct-beam scattering signal while capturing the diffuse continuum produced by multiple scattering. Since the total is a single product of the accurate first-order field and a smooth ratio, the two contributions neither overlap nor introduce a discontinuity.

The conversion to absolute units was done taking into account the mirror area, its altitude above the Earth, the solar illuminance at the Sun–Earth distance, and the solid angle subtended by the solar disk. We assumed the parameters listed in Sec. II. In these calculations, the direct beam from the mirror contributes only at the single sky position corresponding to the mirror direction, and only for an observer located directly within the illuminated patch. At all other grid points, only the diffuse radiation field is computed, which is the primary focus of our calculations in this section.

The calculations include both scattered light and light reflected from the Earth’s surface (Kocifaj, 2019, see Eq. 1), where $\alpha$ is the surface albedo. We carried out calculations for albedos of $\alpha=0$ (zero reflection), 0.2 (typical Earth) and 0.8 (fresh snow) (Sailor and Fan, 2002; Liang, 2001). The zero albedo calculations help in establishing the fractional contribution of the light scattered by the incoming beam, as compared to the light reflected back from the Earth in a Lambertian manner. We also used two different values of $g=0.6$ and $g=0.8$, where $g$ is asymmetry parameter in scattering theory, i.e., the average cosine of the scattering angle over all directions. A value of $g=0.6$ is representative of fine-mode-dominated aerosol (e.g., urban/continental conditions), while $g=0.8$ corresponds to coarse-mode-dominated aerosol (Kinne, 2019) (e.g., maritime or dust-laden atmosphere); together they bracket the range of asymmetry factors reported for tropospheric aerosol (Kinne et al., 2013; Hatzianastassiou et al., 2007).

For each simulation we derive the map of the skyglow at various distances from the “epicenter” of the beam, with all distances being outside the beam. (Calculations for the sky brightness for an observer inside the beam are presented later in Sec. IV). The values of the skyglow in the figures is given in $\rm cd\,m^{-2}$, and can be easily converted (Bará et al., 2020) to magnitudes per square arcsecond via the approximate formula of $\mu\;[\rm mag/\Box^{\prime\prime}]=12.6-2.5\,\log_{10}\!\left(B\;[\rm cd\,m^{-2}]\right)$.

All calculations are performed at 550 nm, a wavelength near the maximum of the solar spectrum and near the maximum of the sensitivity of the human eye, and it therefore dominates the radiance pattern. Although blue light is scattered more strongly, its share in the solar spectrum is small. The conversion from radiance to luminance is made through a constant of proportionality (luminous flux per spectral radiant flux). This uses the fact that the relative angular distributions of sky radiance and luminance under daylight conditions are known to be nearly identical (Torres and Torres, 2008; Alshaibani et al., 2020), differing essentially only by such a constant. Igawa et al. (2004) confirmed that the relative radiance and luminance distributions of the sky can be described by the same equation across all sky conditions, so that the two differ only through the zenith-normalization term. Distribution models formulated for radiance usually reproduce measured sky luminance distributions with an accuracy comparable to that of models developed specifically for luminance (Alshaibani et al., 2020; Gracia et al., 2011). Consequently, a model derived for the radiance distribution can be taken to represent the corresponding luminance distribution up to a scaling factor. This is even more straightforward for the present case than for daylight. Unlike the daytime sky, where the radiance observed in a given direction also receives contributions from light scattered at very large distances, the relevant distances here are short.

The computed night-sky luminance maps for observer distances of 5.4 km and 14.1 km from the beam center are shown in Fig. 2 in Hammer–Aitoff equal-area projection, for three ground albedos ($\alpha=0.0$, $0.2$, and $0.8$) and aerosol asymmetry parameter $g=0.6$. These distances arise from the numerical configuration rather than from a deliberate selection of particular observing scenarios. The output nodes are defined on an exponential (logarithmic) rather than a linear scale over the modelling domain, which reflects the exponential character of the functions governing radiative transfer, for which such sampling is natural. For a perfectly absorbing idealized surface ($\alpha=0.0$), the sky luminance is exclusively due to light scattering in the atmosphere, and decreases steeply with increasing angular distance from the cone of light. With no light reflected from the surface, the contrast between the beam and its surroundings is particularly pronounced. The presence of ground-reflected light ($\alpha>0$) substantially modifies the luminance field. The Lambertian surface becomes a secondary source, redirecting diffuse radiation into the entire upper hemisphere, resulting in a significant brightening of the sky at the zenith and at intermediate elevation angles. Near the horizon the sky remains dark, consistent with the cosine law as the zenith angle approaches $90\arcdeg$. To facilitate comparison and to document the spatial decay of sky luminance with increasing observer distance ($R$), the results for horizontal distances of $R=5.4$ km (Fig. 2, left) and 14.1 km (Fig. 2, right) are displayed on the same color scale. At 14.1 km from the beam center, the luminance of the light cone near the horizon can reach approximately 0.15 $\rm cd\,m^{-2}$(14.7 $\rm mag/\Box^{\prime\prime}$) for a snow-covered surface, while for a perfectly absorbing surface the corresponding value is approximately half ($15.34\,\rm mag/\Box^{\prime\prime}$). These are an order of magnitude brighter than the sky would be due to the full moon, as we will show later in Sec. V.

Multiple scattering superimposes an additional, markedly more homogeneous continuum upon the single-scattering luminance field, yet does not alter the fundamental dominance of forward scattering in the overall angular distribution of the scattered light. When ground-reflected light is present, photons travel along trajectories spanning all directions, so that for any given viewing direction, there always exist photons undergoing predominantly forward scattering. This gives rise to a pronounced diffuse background extending well beyond the geometric boundary of the light cone, as clearly seen in Fig. 2 for $\alpha=0.8$.

Figure 2: Night-sky luminance maps in Hammer–Aitoff equal-area projection (in $\rm cd\,m^{-2}$, logarithmic color scale) for a perfectly absorbing surface ($\alpha=0.0$), a mean ground albedo ($\alpha=0.2$), and a highly reflective (snow-covered) surface ($\alpha=0.8$), computed for aerosol optical depth $\tau_{a}=0.3$ and aerosol asymmetry parameter $g=0.6$. Left: an observer located at a horizontal distance of $R=$ 5.4 km from the axis of the light cone produced on the ground by a 54 m mirror in Earth orbit, shining straight down onto the surface. The radius of the directly illuminated area is 2.5 km, so the observer is roughly 3 km from the edge of the lit portion of the ground. Right: The same situation, but for an observer 14.1 km from the center of the illuminated circle on the Earth’s surface.

The influence of ground-reflected light manifests not only in the emergence of a prominent luminance veil across the night sky, but also in an increase of luminance of the light cone. Although the contribution of reflected light decreases steeply toward the horizon, multiply scattered light modifies the light scene. The contribution from higher scattering orders increases with growing optical path (van de Hulst, 1980). While the optical path for rays traveling vertically is short, the integral of the atmospheric volume scattering coefficient for photons traversing the atmosphere horizontally can be orders of magnitude larger. The longer the optical path, the more scattering events occur along the photon trajectory from source to observer. This effect is well illustrated by the bottom right panel of Fig. 3, which shows the ratio of sky luminance for $\alpha=0.2$ to that for $\alpha=0.0$.

Figure 3: Top left: The sky luminance (in $\rm cd\,m^{-2}$) as a function of zenith distance for an observer located at different distances from the center of the beam produced by a 54 m mirror at zenith. We assumed $\alpha=0.0$ and $g=0.6$. Top right: The same as top left but for $\alpha=0.2$. Bottom left: The same as top left but for $\alpha=0.8$. Bottom right: The factor of the luminance increase when transitioning from $\alpha=0.0$ to $\alpha=0.2$.

Fig. 3 also shows the sky luminance for $\alpha=0.0$ (top left panel), $\alpha=0.2$ (top right panel), and $\alpha=0.8$ (bottom left panel) as a function of zenith angle, computed for discrete observer distances from the center of the illuminated circular patch of radius 2.5 km. Observers at $0$ km and $1.6$ km from the center are directly within the artificially lit area, which is discussed later in Sec. IV, while all remaining cases correspond to observers outside the directly illuminated area. Since the viewing direction is always toward the beam axis, the optical path within the light cone for an observer at $1.6$ km exceeds that for an observer at the center. The luminance is computed for the diffuse component of downward radiation only, with the direct beam contribution from the mirror excluded.

Therefore, the luminance profiles are primarily dictated by geometry, with a stronger luminance enhancement occurring wherever the line of sight intersects the axis of light cone. For an observer outside the beam and for $\alpha=0.0$, the zenith luminance is determined solely by multiply scattered light; first-order scattering contributes only when the viewing direction crosses the beam. The lower the altitude of this intersection, the stronger the luminance enhancement, since both aerosol and molecular concentrations decrease with height, reducing the scattering probability and increasing the mean free path between successive scattering events. The presence of a reflected component manifests as an additional optical signal over a relatively wide range of azimuths around the beam.

IV Calculations within the beam

For an observer located within the directly illuminated region on the Earth’s surface, the lighting conditions differ significantly from those outside the illuminated patch. In addition to the direct illumination estimated in Sec. II, the diffuse light field within the circular patch is dominated by single-scattered radiation. The higher scattering orders contribute at most 1–4% of the luminance and leave the angular radiance distribution unchanged. Light scattering in the vertical atmospheric column is governed by the volume scattering coefficient $k_{\mathrm{sca}}$ (unit m^-1), which depends on altitude $h$ but is independent of scattering angle (see, e.g., Ma et al. 2021, where it is denoted as $\beta$). In a cloud-free atmosphere, $k_{sca}$ is given by the sum of the aerosol component $k_{sca,a}$ and the Rayleigh component $k_{sca,R}$. Both components decrease exponentially with altitude, with the rate of decrease characterized by the molecular scale height ($H_{R}$) and the aerosol scale height ($H_{a}$), respectively. Since $H_{a}$ is typically several times smaller than $H_{R}$, the vertical gradient of aerosol concentration is much steeper than that of air density. As a result, the lower troposphere plays a key role in skyglow formation. The integral of the Rayleigh scattering coefficient over the vertical column from altitude $h$ to the top of the atmosphere defines the Rayleigh optical depth $\tau_{\mathrm{R}}(h)=\tau_{\mathrm{R,0}}\exp(-h/H_{\mathrm{R}})$, where $\tau_{\mathrm{R,0}}$ is the Rayleigh optical depth at ground level. Unlike molecules, which scatter conservatively (i.e. , the single scattering albedo is $\tilde{\omega}_{R}\approx 1$), aerosols both scatter and absorb radiation, so that $\tilde{\omega}_{a}<1$ in general. The aerosol volume scattering coefficient can therefore be written as $k_{\mathrm{sca,a}}=\tilde{\omega}_{\mathrm{a}}\,k_{\mathrm{ext,a}}$, where $\tilde{\omega}_{\mathrm{a}}$ represents the fraction of energy scattered relative to the total energy removed from the beam by aerosol particles. By analogy with the Rayleigh case, the aerosol optical depth is $\tau_{a}(h)=\tau_{a,0}\exp(-h/H_{a})$. The total optical depth at altitude $h$ is then $\tau(h)=\tau_{R}(h)+\tau_{a}(h)$.

The flux density of solar radiation at altitude $h$ is

$$F(h)=F_{\odot}\frac{A_{proj}}{\pi\psi_{0}^{2}(H-h)^{2}}\exp\{-\tau(h)\},$$

where $\psi_{0}$ is the beam divergence ($\psi_{0}\ll 1\arcdeg$). At the Earth’s surface ($h=0$ km), the denominator reduces to $\pi\psi_{0}^{2}H^{2}$, which equals $A_{spot}$. The amount of radiation scattered along a path element $d\ell$ is

$$dI=\frac{P(\theta)}{4\pi}\,k_{sca}(h)\,F(h)\,d\ell,$$

where $P(\theta)$ is the phase function describing the angular distribution of scattered radiation as a function of scattering angle $\theta$ (Horvath, 2014). In a vertically stratified atmosphere, it is convenient to substitute $d\ell=dh/\cos z$, where $z$ is the zenith angle of observation. The scattering source term can be expanded as

$$\frac{P(\theta)}{4\pi}k_{sca}=\frac{1}{4\pi}\left[k_{sca,R}(h)\,P_{R}(\theta)+k_{sca,a}(h)\,P_{a}(\theta)\right].$$

In the Henyey–Greenstein formalism,

$$P_{a}(\theta)=\frac{(1-g^{2})}{(1+g^{2}-2g\cos\theta)^{3/2}}$$

(Winkler, 2022). From the scattering geometry it follows that $\theta\approx z$, so the Rayleigh phase function reduces to $P_{R}(z)=\frac{3}{4}(1+\cos^{2}z)$. The night-sky brightness at the observer’s location is therefore

$$I=\int\exp\left\{-\frac{\tau_{a,0}}{\cos z}\left(1-e^{-\frac{h}{H_{a}}}\right)-\\ \frac{\tau_{R,0}}{\cos z}\left(1-e^{-\frac{h}{H_{R}}}\right)\right\}dI,$$

where $\tau_{a,0}$ and $\tau_{R,0}$ are the aerosol and Rayleigh optical depths at the ground, respectively. The integration extends from the Earth’s surface to the TOA. The computational results are demonstrated as luminance maps in Fig. 4 for an observer located directly beneath the mirror (left panel) and for an observer at approximately $1.6$ km from the beam axis (right panel).

As follows from the 3D geometry, the sky luminance distribution for an observer at the center of the illuminated patch ($R=0$ km) is azimuthally symmetric (Fig. 4, left). With increasing ground albedo ($\alpha$), the diffuse background at low elevation angles brightens noticeably. Near zenith, however, the sky luminance is much less sensitive to the ground albedo, altogether very bright, but with a small range: 2.6 $\rm cd\,m^{-2}$ ($11.56$ $\rm mag/\Box^{\prime\prime}$, for $\alpha=0.0$) to 2.9 $\rm cd\,m^{-2}$ ($11.44$ $\rm mag/\Box^{\prime\prime}$ for $\alpha=0.8$).

For an observer displaced from the beam axis ($R=1.6$ km), the sky brightness in zenith does not noticeably change. However, the luminance distribution becomes azimuthally asymmetric (Fig. 4, right). Assuming $\alpha=0.0$, the near-horizon luminance in the direction toward the beam axis (i.e. for azimuth angle $A=0\arcdeg$) reaches approximately 0.2 $\rm cd\,m^{-2}$, while in the opposite direction ($A=180\arcdeg$) it falls to below 0.05 $\rm cd\,m^{-2}$. This luminance disproportion broadly correlates with the ratio of the mean optical path lengths traversed by the beam in the respective viewing directions. The azimuthal asymmetry is somewhat reduced at higher albedo values such as for fresh snow cover ($\alpha=0.8$), though it remains clearly present. However, in the latter case the near-horizon luminance reaches approximately 1.4 $\rm cd\,m^{-2}$ toward the beam axis and $0.2$–$0.4$ $\rm cd\,m^{-2}$ in the opposite direction.

Figure 4: The same as in Fig. 2, but for an observer located inside the illuminated area. Left: an observer at the center of the illuminated region, directly beneath the mirror. Right: an observer 1.6 km from the axis of the light cone.

V Comparison to the illumination by the full moon

The lighting conditions of a night-time atmosphere illuminated by the Moon were modeled using the UniSky Simulator (Kocifaj, 2015), originally developed for daylight studies. This radiative transfer tool is based on the modified theory of successive orders of scattering (Kocifaj, 2012). The Moon is treated as a source of parallel rays, analogous to the Sun, given its distance from the Earth. This guarantees spatial illuminance homogeneity, so that the boundary condition for the incident radiation field at any point on the TOA can be written as $\delta(z-z_{0})\,\delta(A-A_{0})\,F_{\odot}$, where $z_{0}$ and $A_{0}$ are the zenith and azimuth angles of the Moon, and $F_{\odot}$ is the net flux density of moonlight. The radiative transfer problem is then solved subject to boundary conditions defined by this external source at the TOA and a Lambertian reflecting surface at the lower boundary.

To compare the diffuse light field produced by the mirror with that of the Moon, we performed the calculations with the following parameters: ground albedo $\alpha=0$, $0.2$, and $0.8$, aerosol asymmetry parameter $g=0.6$ and $g=0.8$, single-scattering albedo of aerosol particles $=0.9$, aerosol optical depth at the reference wavelength of $500$ nm $=0.3$, aerosol scale height $=2$ km, and scale height of the molecular atmosphere $=8$ km. The results for moonlit sky include first- and second-order scattered radiation, including the contribution from ground-reflected light.

We then derived the ratio of the sky brightness due to the mirror and the Moon (both at zenith) to perform a quantitative comparison. As we showed in Sec. 4, for an observer within the beam, the sky brightness ranges between 0.1 $\rm cd\,m^{-2}$ (15 $\rm mag/\Box^{\prime\prime}$) and 2.9 $\rm cd\,m^{-2}$ (11.4 $\rm mag/\Box^{\prime\prime}$), which exceeds the luminance of the moonlit sky (0.01 $\rm cd\,m^{-2}$ or 18 $\rm mag/\Box^{\prime\prime}$) by factors of 10 to 300. Thus, our analysis pertains to observers outside of the beam. To distinguish the results from the absolute luminance maps presented in Figs. 2–4, the ratio of diffuse sky luminances produced by the mirror and the Moon (both at zenith) is shown on a different color scale in Fig. 5, where blue represents cases in which the diffuse sky luminance due to the mirror is negligible compared to that of the moonlit sky, and red represents the opposite extreme. Cases where the two luminances are comparable are shown in white. Results are presented for the three albedo values considered in this work, i.e., for a perfectly absorbing surface (Fig. 5, left), a typical ground albedo (Fig. 5, middle), and a highly reflective snow-covered surface (Fig. 5, right). In general, the luminance of the artificial light from the 54 m mirror with a ground albedo of $\alpha=0.2$ far exceeds the luminance from the diffuse moonlit sky for most of the sky, except the horizon opposite to the beam, even at a distance of 14.1 km. This dominance is even more pronounced for higher albedos ($\alpha=0.8$). The luminance of the vertical beam of light against the diffuse moonlit sky background decreases with increasing distance from the center of the mirror-illuminated patch. At a distance of 5.4 km, the beam luminance exceeds that of the moonlit sky by a factor of $\sim 15$ in the beam direction; at 14.1 km this ratio drops to $\sim 4$; and at 34.1 km the beam luminance falls to roughly half the moonlit sky background. These ratios vary only slightly (less than 10%) with ground albedo at distances below $\sim 20-30$ km. However, at distances exceeding $\sim 30$ km, the beam-to-moonlight luminance ratio for a snow-covered surface can be up to 40% higher than for an ideally dark surface.

An elevated ground albedo causes the surrounding landscape to reflect a fraction of moonlight back into the atmosphere. On the other hand, the mirror in its orbit illuminates only a small patch of the Earth’s surface. As the observer distance from the beam increases, atmospheric extinction causes a more pronounced decay of beam intensity, while the higher diffuse background from ground-reflected moonlight lowers the luminance ratio and thus reduces the contrast of the light cone against the moonlit sky.

Lambertian reflection further disadvantages photons traversing the atmosphere along highly inclined trajectories, while the contribution of multiply scattered moonlight near the horizon produces a continuous luminance background. As a result, the ratio of the mirror-originated to moonlit sky luminance shows a minimum near the horizon, which gradually gives way to a broad luminance veil in the sky quadrant containing the beam. This luminance veil has an azimuthal width ranging from a few tens of degrees at a distance of $R\sim 30$ km to more than $100\arcdeg$ at the distance of $\sim 5$ km. The luminance within this veil receives comparable contributions from scattered moonlight and scattered beam light.

Figure 5: The same as in Fig. 2, but showing the ratio of the diffuse sky luminance produced by the mirror to that of the full-Moon-lit sky (blue: mirror $<$ Moon; red: mirror $>$ Moon; white: mirror $\approx$ Moon). Left: perfectly absorbing surface ($\alpha=0.0$); Middle: Typical ground albedo of ($\alpha=0.2$); Right: highly reflective snow-covered surface ($\alpha=0.8$). From top to bottom, the observer is located at horizontal distances of 5.4 km, 14.1 km, and 34.1 km from the axis of the light cone, all outside the directly illuminated area (radius 2.5 km).

VI Conclusions

In this paper, we calculated the sky brightness due to a single, $54\times 54$ m reflective space mirror, proposed by ReflectOrbital, as a long-term solution for illuminating $\sim$2.5 km radius spots on Earth. The satellite would appear as a -16.7 magnitude point-like source, i.e., about 4 magnitudes (40 times) brighter than the full moon, illuminating the ground at $\sim 9.5$ lux (street-light level). The brightness of the night sky will be increased due to Rayleigh and aerosol scattering of both the incoming beam and the light scattered back from the ground. Even without detailed calculations, by scaling from everyday experience, the resulting light pollution should be enormous due to 2.5 km spot evenly illuminated at street-light level, far exceeding the uplight of a town with a similar surface area (given that street lighting is highly localized and streets only occupy $\sim 10$% surface area). This is confirmed by our detailed calculations, which we carry out for observers both outside and inside the beam.

For observers outside the beam we used the Modified Successive Orders of Scattering (MSOS) model from (Kocifaj, 2018; Kocifaj et al., 2023) to compute the field of scattered radiation at any point in 3D space. We carried out these calculations for three different surface albedos ($\alpha=0,0.2,0.8$), and two scattering asymmetry parameters ($g=0.6,0.8$). We computed night-sky luminance maps for different observer distances, ranging from 5.4 km to 80 km from the center of the beam. We show that the sky brightness for average ground reflection ($\alpha=0.2$) will be much higher in almost all of the sky, except the horizon opposite to the beam, for observers located at 14 km from the center of the beam, meaning that the sky will be more light-polluted than due to the full moon in a circle with an approximate radius of 14 km (typical albedo) to 30 km (snow-covered terrain). The sky will be brighter than the full-moon lit sky at 34 kms from the center of the beam in the direction of the beam, appearing as a brightly lit pillar with large side-lobes ($\pm 45\arcdeg$ in azimuth and up to zenith).

For observers in the beam the diffuse light field is dominated by single-scattered radiation. We took into account scattering from both aerosols and atmospheric molecules. The sky brightness (up to 2.9 $\rm cd\,m^{-2}$ or 11.45 $\rm mag/\Box^{\prime\prime}$) will typically far exceed that of the full-moon lit sky down to very close to the horizon. The level of sky brightness will be similar to the dusk sky in zenith shortly after sunset, when the sun is at $\sim-5\arcdeg$ elevation (Patat et al., 2006). Not even the brightest stars remain visible.

R	Lum ($\alpha=0.2$)	Lum ($\alpha=0.2$)
(km)	($\rm cd\,m^{-2}$) ($\rm mag/\Box^{\prime\prime}$)	($\rm cd\,m^{-2}$) ($\rm mag/\Box^{\prime\prime}$)
0	2.71 (11.52)	2.89 (11.45)
1.6	2.62 (11.55)	2.77 (11.49)
5.4	0.25 (14.11)	0.59 (13.17)
14.1	0.07 (15.49)	0.13 (14.82)
34.1	0.011 (17.50)	0.015 (17.16)

Table 1: The maximum sky luminance in $\rm cd\,m^{-2}$ for selected scenarios of different observer distances from the center of the beam from the 54 m mirror. These calculations assume $g=0.6$. For comparison, the median sky brightness due to the full moon is 0.014 $\rm cd\,m^{-2}$ ($\alpha=0.2$) and 0.02 $\rm cd\,m^{-2}$ ($\alpha=0.8$).

Altogether, within the illuminated patch, the night sky is lost. Outside the beam, the night sky is seriously compromised to distances up to 30 kms, with the light pillar of the beam and the ground-reflection of light scatter visible from even larger distances. Our calculations present a best-case scenario by assuming a completely cloud-free atmosphere. With a cirrus cloud-deck at e.g. 10 km altitude illuminated by the beam, the light pollution could be visible from much larger distances.

The influence of clouds on night-sky brightness is complex. For an atmosphere illuminated from below by ground-based sources we have shown that this influence depends on the emission function of the source and on its distance from the observer. We have also shown that a transition zone exists, in which the amplifying effect of clouds turns into a weakening one (Kocifaj et al., 2025; Kocifaj and Falchi, 2025). However, for a light source above the atmosphere, such as from a satellite orbiting the Earth, the model would be substantially different and would require separate development.

Dense clouds, which for ground-based sources strongly reshape the night-sky brightness distribution, would here act as a barrier and would usually block the light from reaching the surface. Thin (cirrus) clouds would alter the radiation field computed for a clear sky in two ways, removing part of the radiation and scattering another part into other directions. The outcome would depend strongly on the actual source-cloud-observer geometry and on the shape and opacity of the cloud in its different parts. Such conditions can therefore be captured only through a statistical approach, describing the mean sky state expected for a given cloud type, sky coverage fraction, cloud height, vertical distribution, and opacity. These calculations are beyond the scope of the current paper.