Unveiling Illumination Variations During a Lunar Eclipse: Multi-Wavelength Spaceborne Observations of the January 21, 2019 Event

Abstract

Space-based observations of the total lunar eclipse on 21 January 2019 were conducted using the geostationary Earth-orbiting satellite Gaofen-4 (GF-4). This study represents a pioneering effort to address the observational gap in full-disk lunar eclipse photometry from space. With its high resolution and ability to capture the entire lunar disk, GF-4 enabled both quantitative and qualitative analyses of the variations in lunar brightness, as well as spectra and color changes, across two spatial dimensions, from the whole lunar disk to resolved regions. Our results indicate that before the totality phase of the lunar eclipse, the irradiance of the Moon diminishes to below approximately 0.19% of that of the uneclipsed Moon. Additionally, we observed an increase in lunar brightness at the initial entry into the penumbra. This phenomenon is attributed to the opposition effect, providing scientific evidence for this unexpected behavior. To investigate detailed spectral variations, specific calibration sites, including the Chang’E-3 landing site, MS-2 in Mare Serenitatis, and the Apollo 16 highlands, were analyzed. Notably, the red-to-blue ratio dropped below 1 near the umbra, contradicting the common perception that the Moon appears red during lunar eclipses. The red/blue ratio images reveal that as the Moon enters Earth’s umbra, it does not simply turn red; instead, a blue-banded ring appears at the boundary due to ozone absorption and the lunar surface composition. These findings significantly enhance our understanding of atmospheric effects on lunar eclipses and provide crucial reference information for the future modeling of lunar eclipse radiation, promoting the integration of remote sensing science with astronomy.

1. Introduction

The phenomenon of a lunar eclipse has captivated human curiosity and scientific inquiry for centuries, offering an exceptional opportunity to study the interplay of celestial bodies and the intricate dynamics of light and shadow within our solar system. Lunar eclipses occur when the Earth, positioned between the Sun and the Moon, casts its shadow upon the lunar surface, causing a temporary dimming or reddening of the Moon’s luminosity. In the past, due to the limited observation techniques and insufficient understanding of space, lunar eclipses were often considered an omen or supernatural event, and were explained through myths, legends, and religious interpretations [1]. Over time, with the invention of the telescope and the accumulation of scientific knowledge, the study of lunar eclipses began to take shape [2]. In the three centuries prior to the advent of rockets and artificial satellites, lunar eclipses were utilized to determine the relative size of the Sun–Earth–Moon system, measure geographical longitudes [3], study lunar motion [2], and research the thermophysical properties of lunar surface soil [4,5,6]. While reports of lunar eclipses have been documented for millennia, the comprehensive investigation of their brightness characteristics and their applications in advancing research across various fields, such as the study of the composition and thickness of the Earth’s atmosphere, has significantly progressed over the past century. Numerous scholarly works have contributed to our understanding of lunar eclipse phenomena, including studies on eclipse brightness [7,8,9,10,11,12,13], infrared analyses of lunar eclipses [4,14,15,16,17], the impact of volcanic aerosols on lunar eclipses [18,19,20,21], measurements of composition of the Earth’s atmosphere using lunar spectra during an eclipse [22,23], and the development of lunar eclipse models [24,25,26,27].

Traditionally, lunar eclipse studies have heavily relied on ground-based measurements, which are susceptible to atmospheric interference and urban light pollution. Additionally, ground-based observations of a lunar eclipse are only possible in regions within the eclipse path and during the appropriate time, under favorable weather conditions. With the development of observation techniques, lunar orbiting satellites have also conducted some lunar eclipse observations, but their results may be affected by striping and color variations introduced by local observations [5,28]. Compared to ground-based telescopes and lunar orbiting satellites, Earth-orbiting platforms, being away from the atmosphere, have emerged as a solution to avoid these issues, enabling us to obtain accurate measurements of lunar brightness or temperature variations. To leverage these advantages, we conducted a photoelectric photometry of the Moon’s surface during the total lunar eclipses on 21 January 2019, using the Gaofen-4 (GF-4) satellite [29]. GF-4, China’s first high-resolution Earth observation satellite in geosynchronous orbit, views the Moon through its normally nadir-viewing optics by executing spacecraft attitude maneuvers. Based on staring mode imaging with a focal plane array, GF-4 captured the entire lunar disk with a single exposure.

A significant lunar eclipse commonly known as “the Super Blood Wolf Moon” took place on 21 January 2019, capturing the attention of astronomers and enthusiasts worldwide with its celestial grandeur. According to the astronomical ephemeris (accessible at: https://www.timeanddate.com/eclipse/list-lunar.html, accessed on 21 January 2020), this eclipse began with the first penumbral contact at 02:36:30 Coordinated Universal Time (UTC) (P1), followed by the onset of the partial eclipse nearly an hour later, marked by the first umbral contact at 03:33:57 UTC (U1). Between 04:41:20 UTC (U2) and 05:43:16 UTC (U3), the Moon remained completely inside the umbra. The second partial phase ended at 06:50:40 UTC (U4), and finally the Moon exited the penumbral shadow at 07:48:05 UTC (P4). This event provided an exceptional opportunity for comprehensive scientific investigation.

Benefiting from the high-resolution and full-disk imaging capabilities of GF-4, we utilized GF-4 for the first time to observe the lunar eclipse on 21 January 2019. This study includes a detailed analysis of the time-dependent brightness distribution, spectral variations, and color changes of the lunar surface, demonstrating the potential of space-based observations in advancing lunar eclipse research.

2. Data Description

The data utilized in this study were acquired from the geostationary satellite GF-4 during the total lunar eclipse event that occurred on 21 January 2019. GF-4’s advanced capabilities enabled the acquisition of high-resolution full-disk images of the Moon in five visible and near-infrared (VNIR) bands, in addition to one mid-infrared band. The VNIR detector onboard GF-4 features a 0.8° × 0.8° field of view (FOV) and employs a CMOS image sensor with a 10,240 × 10,240 pixel array, resulting in a spatial resolution for lunar imagery of ∼500 m/pixel [30]. Detector counts (DN) are transmitted to the ground station, and represented by 16-bit binary integers (0 ≤ DN ≤ 65,535). The five VNIR bands, designated as B1 to B5, comprise a panchromatic band (B1) and bands for blue, green, red, and near-infrared wavelengths (B2 to B5, respectively). The lunar eclipse data captured by GF-4’s spectral bands B2 to B5 were utilized. These bands correspond to lunar effective wavelengths of 491.17 nm, 560.58 nm, 653.53 nm, and 809.43 nm, respectively. The characteristics of GF-4 PMS are shown in

Table 1. Specifications of GF-4.

Satellite Characteristic	VNIR	MWIR
Lunar Effective Wavelength (nm)	B1: 611.93 nm B2: 491.17 nm B3: 560.58 nm B4: 653.53 nm B5: 809.43 nm	3.77 μm
FOV	0.8°	0.66°
IFOV ($\mathsf{\mu }$ rad/pixel)	1.363	11.249
Ground sampling distance (m/pixel)	50	400
Spatial resolution for the Moon (m/pixel)	∼500	∼4000
Swath width (km)	500	400

.

Table 2. Lunar observation geometry for GF-4 during the lunar eclipse on 21 January 2019.

Time (UTC)	Sun-Moon Distance (AU)	Viewer-Moon Distance (ME)	Selenographic Subsolar Longitude (°)	Selenographic Subsolar Latitude (°)	Selenographic Sub-Observer Longitude (°)	Selenographic Sub-Observer Latitude (°)	Phase Angle (°)
2:30	0.9856	1.0149	359.9029	−0.1051	354.9822	2.2756	−5.47
2:40	0.9849	1.0172	359.8187	−0.1054	355.2189	2.2620	−5.17
3:00	0.9871	1.0214	359.6503	−0.1059	355.7053	2.2214	−4.58
3:10	0.9874	1.0233	359.5661	−0.1062	355.9527	2.1945	−4.28
3:30	0.9864	1.0267	359.3977	−0.1067	356.4575	2.1276	−3.69
3:40	0.9853	1.0281	359.3135	−0.1070	356.7137	2.0877	−3.40
3:50	0.9856	1.0298	359.2293	−0.1072	356.9269	2.0432	−3.10
4:00	0.9883	1.0306	359.1451	−0.1075	357.2304	1.9954	−2.84
4:10	0.9897	1.0316	359.0609	−0.1078	357.4902	1.9432	−2.58
4:20	0.9845	1.0324	358.9767	−0.1080	357.7516	1.8868	−2.34
4:30	0.9873	1.0331	358.8925	−0.1083	358.0142	1.8264	−2.12
6:00	0.9831	1.0319	358.1347	−0.1107	0.3471	1.1189	2.53
6:20	0.9882	1.0299	357.9663	−0.1112	0.8456	0.9255	3.06
6:40	0.9875	1.0272	357.7979	−0.1118	1.3316	0.7207	3.63
6:50	0.9879	1.0256	357.7137	−0.1120	1.5690	0.6143	3.92
7:00	0.9858	1.0239	357.6295	−0.1123	1.8019	0.5057	4.22
7:20	0.9844	1.0201	357.4611	−0.1128	2.2558	0.2809	4.81

provides the geometric information pertinent to the lunar eclipse observation. These measurements allow for both quantitative and qualitative analyses of variations in lunar brightness across the entire lunar disk and within resolved regions.

3. Methods

3.1. Data Processing

In scientific analysis, lunar images are typically assessed in terms of radiance or reflectance units. Achieving precise calibration is paramount to maximizing scientific return and allowing for confidence in the derived photometric measurements. Essential image calibration procedures include dark current subtraction, and the removal of bad pixels and bad columns, as well as flat field correction and absolute radiometric calibration.

3.1.1. Dark Current Subtraction

The CMOS sensor continuously responds to thermal electrons, even in the absence of incident radiation, generating a dark current that should be subtracted from the observed images to accurately determine reflectance levels. However, because the GF-4 satellite is not equipped with a physical shutter, a direct assessment of the dark current during each observation is not feasible. Before the lunar eclipse observation mission began, GF-4 conducted deep-space observations. The signal levels recorded in deep space, unaffected by the Moon, provide a reliable estimate of the dark current from thermal electrons, along with a bias in the output to prevent negative values, and other periodic noise. Given that the bias has already been subtracted from the GF-4 output images, the deep-space images are treated as a direct reflection of dark current. By averaging multiple frames of these dark images, a representative dark current image was generated, which was then subtracted from the lunar eclipse observation data.

3.1.2. The Removal of Bad Pixels and Bad Columns

Bad pixels are those that exhibit abnormal responses during the imaging process, often displaying extremely low responses or no response, and are therefore unable to provide valid light signals. These may occur as individual pixels or as clusters of pixels (e.g., lines, rows, or areas). Once identified, the locations of bad pixels are marked, and correction is applied. In the GF-4 images, only bad pixels and bad columns are observed. For bad individual pixels and clusters, correction is performed by replacing the DN of the bad pixels with the average of the valid DN values from the eight neighboring pixels. For bad columns, the correction is made using a weighted averaging method, as two neighboring columns are typically problematic. The weights assigned to the columns immediately adjacent to the bad column (i.e., the nearest valid columns on either side) are set at 0.75, while the weights of the columns one step further away are set at 0.25. This processing approach can be represented by the following equation:

$$\begin{array}{c}DN(i,j)=0.75\times DN(i,j-1)+0.25\times DN(i,j+2),\hfill \\ DN(i,j+1)=0.25\times DN(i,j-1)+0.75\times DN(i,j+2),\hfill \end{array}$$

(1)

where i is row number and j is the column number.

Figure 1a,b illustrate the correction applied to a cluster of bad pixels, with Figure 1a showing the image before correction and Figure 1b displaying the post-correction result. Similarly, the correction of bad columns is presented in Figure 1c,d, depicting the images before and after correction, respectively.

3.1.3. Flat-Field Correction

To enhance the readout speed, the VNIR sensor of GF-4 comprises five readout circuits. Due to minor gain discrepancies among the circuits, this configuration results in five striped regions in the captured images. The non-uniformity characterized by the stripes was first observed in GF-4 images after lunar observations. To the best of our knowledge, the previous literature has not specifically addressed this issue in relation to GF-4. The large field of view of the CMOS sensor on GF-4 presents a challenge in finding a sufficiently uniform field for traditional flat-field correction. To resolve this, we employed a statistical approach to perform flat field calibration.

Given the FOV of the GF-4 at 0.8°and the angular diameter of the Moon as observed from Earth at approximately 0.5°, the Moon’s imaging diameter on GF-4 is estimated to be around 6400 pixels, spanning over three stripe regions. Based on a statistical analysis of the existing data, parts of the Moon consistently fall within the third stripe region. Therefore, the methodology involves using the third stripe as a reference to adjust the DN of the other four stripes. However, in situations where the third stripe is unable to capture an image due to large phase angle conditions, an alternative stripe can be chosen as the reference instead. Based on the assumption that the average pixel values of two adjacent columns are equal at the boundary of each stripe, the adjustment coefficient for gain is determined by the ratio of the DN values of these two columns. Due to the varying contrasts observed across different rows within the same stripe, a sliding window of $1001\times 2$ was employed for coefficient determination in practical correction. Taking the correction of the second stripe as an example, its boundary with the third stripe lies between the 4096th and 4097th columns. The gain coefficient, $C_{gain}\left(i\right)$, for the i-th row of the second stripe, and the corrected DN, are calculated as follows:

$$\begin{array}{c}{C}_{\mathrm{gain}}\left(i\right)={\displaystyle \frac{\frac{1}{1001}{\displaystyle \sum _{i=i-500}^{i+500}}DN(i,4097)}{\frac{1}{1001}{\displaystyle \sum _{i=j-500}^{j+500}}DN(i,4096)}},\hfill \\ D{N}_{\mathrm{corrected}}(i,j)=C_{\mathrm{gain}}\left(i\right)\times DN(i,j).\hfill \end{array}$$

(2)

This same approach was applied to correct the first, fourth, and fifth striped regions. Notably, during the correction of the first stripe, the corrected DN values from the second stripe were used as a reference to calculate the coefficients. Similarly, when correcting the fifth stripe, the corrected DN values from the fourth stripe served as the reference.

By applying the destriping method, we can ensure more consistent image quality across the entire field of view. Figure 2 presents the results after flat field correction. A visual inspection of the destriped image clearly demonstrates a significant improvement in image quality. The boundary lines between stripes are effectively eliminated, and the transitions between stripes in the color composite image are now smooth, with consistent tones. The overall visual quality is markedly enhanced.

To further evaluate the effectiveness of the flat-field correction, a quantitative assessment was performed using the full lunar disk region from the 02:30 UTC image, acquired before the lunar eclipse commenced. The results are shown in

Table 3. Quality assessment of flat field correction.

	B2		B3		B4		B5
	Before	After	Before	After	Before	After	Before	After
Mean	7339.69	7311.82	13,952	13,816.6	10,961.6	10,765.3	12,191.3	11,882.4
Standard deviation	3064.15	2953.4	5205.08	4999.63	4142.58	3935.5	4431.31	4203.1
Image entropy	8.46	12.25	9.23	13.05	8.94	12.79	9.02	12.95

. The image entropy H is defined as follows:

$$H=-\sum _{DN=min}^{max}{p}_{DN}{log}_2\left(p_{DN}\right),$$

(3)

where $min$ and $max$ represent the minimum and maximum DNs of the lunar disk, respectively, and $p_{DN}$ denotes the probability of each DN value, i.e., the frequency of occurrence of that gray-level pixel relative to the total number of pixels. The standard deviation reflects the variation in individual pixel values from the mean, while image entropy represents the complexity of pixel value distribution.

As shown in the standard deviation and image entropy results in Table 3, the standard deviation was higher before the flat-field correction, which is attributed to the increase in the range of pixel values caused by the stripe non-uniformity, thus elevating the standard deviation. After correction, the standard deviation decreased significantly, indicating that the DN distribution across the lunar disk became more uniform, and the striping noise was effectively suppressed, resulting in a smoother and more consistent image. The post-correction image entropy increased, indicating that the removal of the five stripes enhanced detail representation in the image. Both subjective and objective evaluations demonstrate that the applied flat-field correction algorithm effectively improves image quality and is well-suited for processing GF-4 lunar eclipse data.

3.1.4. Absolute Radiometric Calibration

The last step is absolute radiometric calibration, the conversion of DN levels to radiance, aradiance factor (often called $I/F$), or other physical quantities. The absolute radiometric calibration procedure for GF-4 using the Moon is extensively detailed in the work of Cai et al. (2024) [31]. Visible spectral reflectance data from the Interference Imaging Spectrometer (IIM) aboard Chang’E-1 and Moon Mineralogy Mapper (${\mathrm{M}}^3$) aboard Chandrayaan-1 were compared with GF-4 images to provide absolute calibration coefficients. Cai et al. (2024) [31] established the relationship between DNs and physical units of radiance, denoted in $\mathrm{mW}·{\mathrm{m}}^{-2}·\mathsf{\mu }{\mathrm{m}}^{-1}·\mathrm{s}{\mathrm{r}}^{-1}$, laying the foundation for quantitative analyses of lunar radiance during lunar eclipses. The absolute calibration coefficients of B2 to B5 are shown in

Table 4. Absolute calibration coefficients of B1 to B5 for GF-4.

Band	The Expression for Radiometric Calibration
B1	$\left\{\begin{array}{c}L=2.2774DN+6062,DN\ge 2000;\hfill \\ L=-0.0011D{N}^2+7.5158DN,DN<2000.\hfill \end{array}\right.$
B2	$\left\{\begin{array}{c}L=4.05DN+7500,DN\ge 800;\hfill \\ L=-0.0055D{N}^2+18DN,DN<800.\hfill \end{array}\right.$
B3	$\left\{\begin{array}{c}L=2.48DN+5800,DN\ge 1500;\hfill \\ L=-0.00165D{N}^2+8.9DN,DN<1500.\hfill \end{array}\right.$
B4	$\left\{\begin{array}{c}L=3.09DN+6196,DN\ge 2000;\hfill \\ L=-0.002D{N}^2+10.13DN,DN<2000.\hfill \end{array}\right.$
B5	$\left\{\begin{array}{c}L=2.203DN+5405,DN\ge 2500;\hfill \\ L=-0.001D{N}^2+6.8DN,DN<2500.\hfill \end{array}\right.$

.

Figure 3 displays images of the Moon acquired by GF-4. These images underwent the aforementioned preprocessing steps and were converted into disk-resolved radiance.

3.2. Disk-Integrated Irradiance and Photometric Model

In the field of lunar eclipse research, the assessment of lunar surface brightness within the Earth’s umbral region during eclipses often relies on the Danjon scale or magnitude system. To ensure precise the measurement and representation of lunar luminosity, we emploedy lunar disk-integrated irradiance, expressed in physical radiation units: $\mathsf{\mu }\mathrm{W}·{\mathrm{m}}^{-2}·\mathrm{n}{\mathrm{m}}^{-1}$. The lunar irradiance was obtained by integrating GF-4 lunar images calibrated to radiance, as defined by Equation (5) in Kieffer and Stone (2005) [32] and shown in Equation (4).

$$I_k^{\prime }={\Omega }_p\sum _{i=1}^{N_p}{L}_{i,k},$$

(4)

where $L_{i,k}$ represents the radiance measurement of a single pixel on the Moon in band k, ${\Omega }_p$ denotes the solid angle of one pixel, and $N_p$ is the total number of pixels illuminated in the lunar image.

Due to the natural variations in the Earth’s and the Moon’s orbits, coupled with the orbital constraints of GF-4 and practical considerations regarding the observational procedures, the geometric configuration of the Sun, Moon, and satellite constantly changed with each observation of the Moon. In order to accurately depict the fluctuations in irradiance during lunar eclipse, the measured irradiance $I^{\prime }$ was standardized to the corresponding irradiance I at the standard distances: a Sun–Moon distance $D_{S-M}$ of $1\mathrm{AU}$ (Astronomical Unit) and a viewer–Moon distance $D_{V-M}$ of 384,400 km. The distance correction is represented by the following equation:

$$I_k=I_k^{\prime }·{\left({\displaystyle \frac{D_{S-M}}{1\mathrm{AU}}}\right)}^2·{\left({\displaystyle \frac{D_{V-M}}{384,400\mathrm{km}}}\right)}^2,$$

(5)

where the units of $D_{S-M}$ and $D_{V-M}$ are in km, and $1\mathrm{AU}=1.495978707\times {10}^8\mathrm{km}$.

To analyze the the reasons for the phase-dependent brightness changes across the penumbra, we employed the lunar photometric model proposed by Xu et al. (2023) [33]. This model uses the Hapke disk-integrated photometric approach [34] with GF-4 observations to characterize the lunar reflectance phase function. By modeling the reflectance of the lunar disk integral (or named effective disk reflectance) at the GF-4 wavelengths for varying phase angles, the model enables us to estimate the lunar irradiance at any specific phase angle. Using Equation (6), we calculated the expected disk-integrated irradiance for each phase angle, considering it as the irradiance under uneclipsed conditions. Thus, by normalizing the lunar irradiance I during the eclipse to the predicted uneclipsed irradiance $I^{″}$ at corresponding phase angles, we successfully minimized the brightness variations in lunar eclipse observations attributable to phase effects. The conversion between disk-integrated irradiance and effective disk reflectance is formulated as follows:

$$I_k^{″}=r{\left(g\right)}_k{\Omega }_M{E}_k/\pi ,$$

(6)

where $r{\left(g\right)}_k$ is the effective disk reflectance computed using the disk-integrated photometric model at phase angle g, ${\Omega }_M$ is the solid angle of the Moon (=$6.4177\times {10}^{-5}sr$), and $E_k$ is the solar spectral irradiance derived from the model of Gueymard (2004) [35] at 1 AU.

4. Results

Given the limited radiance of refracted sunlight passing through Earth’s atmosphere, the penumbra is predominantly illuminated by the unobscured portion of the Sun. As presented in Figure 3, perceptible changes during the penumbral phase are minimal until the penumbra covers approximately 70% of the lunar disk. Through a statistical analysis of GF-4’s radiance images, it was determined that at 03:10 UTC, the image shows that the luminosity of the Moon’s left-hand limb diminished to 40% of its initial intensity. Similarly, an image captured at 07:20 UTC, near the eclipse’s conclusion, indicates that the luminosity of the Moon’s right-hand limb rose to half of its original intensity.

4.1. Disk-Integrated Irradiance Curve During the Lunar Eclipse

Figure 4 illustrates the irradiance curve at standard distances plotted against UTC. GF-4 observations began at 2:30 UTC, 6 min before the onset of the lunar eclipse, and concluded at 7:20 UTC, 28 min before the eclipse’s conclusion. At 2:40 UTC, the Moon had recently entered the penumbral shadow, showing a subtle increase in lunar irradiance. A parallel occurrence was also observed in studies by Schmude et al. (1999) [9], Dvorak (2005) [10] and Birriel et al. (2020) [12]. Notably, a significant decline in lunar irradiance became evident at 3:10 UTC, with each spectral band experiencing an approximately 6% decrease. At 3:30 UTC, as the left edge of the Moon entered the umbra, there was a decrease in irradiance by nearly 47%. Subsequently, the lunar irradiance experienced a continuous decline until the Moon completely entered the umbra after 4:30 UTC, with the Moon’s brightness reaching approximately 0.19% of that observed during the uneclipsed phase. At this stage, the blue band experienced the most significant decrease, dropping to as low as 0.14%. However, inadequate consideration of the faintness of lunar brightness when fully immersed in the umbra during the total lunar eclipse resulted in a lack of long-exposure observations, leading to gaps in the observational data. Therefore, in the present study, the observations of total eclipse phase were ruled out of the discussion. The lunar irradiance began to increase as the Moon exited the umbra at 6:00 UTC, transitioning into the brighter outer region of the penumbra.

It is evident that the gradient of the light-change curve remains relatively stable at the boundary between the penumbra and the umbra. This can be attributed to the indistinct nature of the transition between the penumbra and the umbra, resulting in a gradual impact on the Moon’s brightness. This observation aligns with findings from the photometric observation of the October 2004 lunar eclipse [10].

4.2. Variations in Disk-Resolved Radiance Spectra During the Lunar Eclipse

To investigate the spectral variations in the lunar surface during a lunar eclipse, we selected specific points for spectral comparison. As the visible area of the lunar surface decreases during the eclipse, several commonly used calibration sites were chosen to facilitate the comparison. These calibration sites are the Chang’E-3 (CE-3) landing site (44.1°N, 19.5°W) [36], MS-2 in Mare Serenitatis (18.7°N, 21.4°E) and the Apollo 16 highlands (9.0°S, 15.5°E) [37]. The calibration sites, characterized by their homogeneity, uniformity, relative flatness, and representation of significant lunar terrains and geologic processes, provided an ideal means to track spectral changes. To mitigate the influence of random noise, a 3 × 3 pixel window was used to calculate the average radiance, which was then used as the radiance value for the calibration sites. Figure 5 depicts the chosen calibration sites as observed by GF-4.

From the numerous images captured during lunar eclipses, we selected data to analyze the spectral variations in the resolved regions. We specifically chose images that demonstrated significant brightness variations as the Moon progressed through the stages of outer penumbra, inner penumbra, and the side of the inner penumbra closest to the umbra. This selection method highlights the most pertinent changes and offers a clearer representation of the observed phenomena. Figure 6 illustrates the variations in radiance spectra at different stages of the lunar eclipse for three calibration sites: CE-3, MS-2, and Apollo 16. The left column displays the spectra as the Moon progresses from the outer penumbra towards the inner penumbra approaching the umbra. As shown, there was a noticeable decrease in radiance, accompanied by increasingly flat spectra and decreasing slopes, indicating a diminishing ratio of red to blue light. Conversely, the right column shows the spectra as the Moon moves from the inner penumbra near the umbra towards the outer penumbra. Here, the radiance increases with steeper slopes. The red-to-blue ratio varies between 1.14 and 0.86 among these calibration sites. Interestingly, near the umbra, the red-to-blue ratio drops below 1, contrary to the common perception that the Moon appears red during lunar eclipses. This will be discussed further in the subsequent sections.

5. Discussion

5.1. Increased Lunar Brightness upon Initial Entry into Penumbra

Schmude et al. (1999) [9], Dvorak (2005) [10], and Birriel et al. (2020) [12] have documented unexplained increases in lunar brightness upon initial entry into the penumbra, just as we observed in our data, even though our observational methods and platforms differ. While Dvorak (2005) [10] suggested that this increase may result from the gradual increase in the Moon’s altitude angle, as observed from ground-based telescopes, empirical evidence supporting this hypothesis has yet to be provided.

In our study, we conducted a comparison of lunar irradiance during the eclipse with uneclipsed irradiance at corresponding phase angles, utilizing the lunar photometric model provided by Xu et al. (2023) [33]. This methodology effectively isolates the impact of phase angle on lunar brightness, as depicted in Figure 7. Following the elimination of phase angle effects, our findings diverge from the observations presented in Figure 4, which show an increase in brightness during the initial stages of the lunar eclipse. Specifically, upon initial entry into the penumbra, lunar brightness no longer exhibits an increase but instead demonstrates a decreasing trend.

Through an analysis of the ratio of irradiance, we can reasonably infer that the increase in lunar brightness due to the opposition effect (OE), where lunar brightness increases sharply as the phase angle decreases at small phase angles, is more pronounced than the attenuation caused by Earth’s obstruction. Despite the ongoing reduction in phase angle, as the lunar surface area entering the penumbra expands to around 3:00 UT, the influence of Earth-obstructing sunlight appears significant enough to counterbalance the OE.

5.2. Blue-Banded Moon at Penumbra–Umbra Border

In the community of astronomy enthusiasts, collections of lunar eclipse observations are plentiful and are often appreciated for their aesthetic appeal. These collections, however, are rarely subjected to rigorous scientific analysis. To elucidate the detailed variations in blue and red light across different regions of the lunar surface during the lunar eclipse, ratio images were generated by dividing the radiance of the red band (i.e., B4) by that of the blue band (i.e., B2). Figure 8 illustrates the ratio images at 03:30 UTC, 03:40 UTC, 03:50 UTC, and 04:10 UTC. These ratio images highlight the boundary between the umbra and penumbra with a distinctive ribbon, approximately 120 to 190 km in width, where the proportion of blue light is higher than that of red light. This leads to the appearance of a blue ribbon on the lunar surface with a true color representation, consistent with the “Blue-Banded Blood Moon” observed by astronomy enthusiasts (accessible at: https://apod.nasa.gov/apod/ap211201.html, accessed on 21 January 2020). Their HDR images, which digitally process and equalize the Moon’s brightness while exaggerating the colors, similarly reveal this phenomenon.

Using the ratio image at 3:30 UTC as an example, the red–blue ratio on the right side represents the Moon’s natural features, directly illuminated by the sunlight. It is evident that the Mare Tranquillitatis appears blue due to its high titanium contents. The left part of the Moon is within the Earth’s umbra and is not directly illuminated by the Sun. Although the faint brightness in this region is insufficient to allow for GF-4 to clearly capture the full details of the Moon in deep shadow, there is a subtle reddening trend because Earth’s dense lower atmosphere scatters more blue light than red.

The pronounced reduction in the red–blue ratio in the center region at the boundary between the penumbra and the umbra is distinctly observable and can be ascribed to the presence of ozone within the atmosphere at altitudes exceeding 15 km. This phenomenon results from the absorption band of ozone in the Chappuis region (430–750 nm), which absorbs red light, particularly around the central wavelength of 600 nm. Consequently, a portion of the red light from sunlight is absorbed by the ozone, leaving the remaining blue-shifted light to be refracted onto the Moon, thereby forming a distinct blue region at the edge of the Earth’s umbra [24,25]. Although this bluing effect is not perceptible to the naked eye, it is evident in the spectral data as a shift towards the blue wavelengths. A quantitative analysis revealed that the red-to-blue ratio at the boundary between the penumbra and the umbra typically ranges from approximately 0.8 to 1.0, correlating with the ozone content and geological composition. For instance, at 03:50 UTC, the blue ribbon exhibits a higher red–blue ratio, appearing less blue compared to other lunar mare regions at the boundary. This is because, generally, the titanium content in mare regions is higher than that in the highlands.

It has long been assumed that the shape of the “blue ribbon” directly corresponds to the shape of the Earth’s shadow during a lunar eclipse. However, this assumption is not shown to be strictly true in Figure 8. The formation of the blue ribbon is influenced not only by the Earth’s shadow but also by the composition of the lunar surface. As illustrated in Figure 8, the blue-banded outer ring shape is not strictly parallel to the Earth’s shadow. Instead, it demonstrates a significant correlation with the local titanium content of the lunar surface. In general, the titanium content within the lunar maria exceeds that of highlands, resulting in a lower spectral slope for the lunar maria compared to highlands. Therefore, the combined effects of titanium contents and ozone absorption contribute to the bluer appearance of the lunar maria relative to highlands at the penumbral boundary during a lunar eclipse.

6. Conclusions

Advancements in observational techniques are continuously enhancing our understanding of lunar eclipses. Unlike earlier approaches, whih relied primarily on ground-based observations or described lunar eclipses using empirical coefficients such as the Danjon scale or magnitude estimations, this study adopts a novel perspective by analyzing lunar irradiance during an eclipse through remote-sensing data. This research marks the first exploration of lunar irradiance during a lunar eclipse using data from an Earth-orbiting satellite. Utilizing observations from the GF-4 satellite during the lunar eclipse on 21 January 2019, we investigated the characteristics of lunar brightness, spectral variations, and color changes from multiple perspectives. We conclude the following:

• Based on the observations, the luminosity of the Moon decreases to below about 0.19% before it fully enters the umbra.
• The observed increase in Moon brightness at the onset of the penumbral phase before 3:10 UTC is attributed to the small phase angles during the observation period. During this phase, the phenomenon of the opposition effect exerts a greater influence on brightness compared to the obscuration of Earth’s shadow.
• The presence of ozone absorption in Earth’s upper atmosphere and the local titanium content of the lunar surface contribute to the formation of the “blue ribbon” at the umbra boundary, which remains imperceptible to naked-eye observers. The lunar mare regions generally appear bluer compared to the highlands at the boundary.

Consequently, utilizing space-based observations of lunar eclipses presents a significant and innovative approach for advancing research in the field of lunar eclipses. This research promotes the integration of remote-sensing science with astronomy, paving the way for future interdisciplinary studies. It is crucial to meticulously consider factors such as exposure time, temporal resolution, and spectral resolution in future lunar eclipse observations from space. These considerations are essential for advancing research on lunar geological features, Earth’s atmosphere, optical phenomena, and related areas.