Ten wavelength channels of calibrated radiance image data from
the sunlit Earth are obtained every 65 min during Northern Hemisphere
summer from the EPIC (Earth Polychromatic Imaging Camera) instrument on the
DSCOVR (Deep Space Climate Observatory) satellite located near the Earth–Sun Lagrange 1
point (L
Measured backscattered radiances of the entire sunlit Earth were obtained
during the 21 August 2017 eclipse from EPIC (Earth Polychromatic Imaging
Camera) on the DSCOVR (Deep Space Climate Observatory) satellite. EPIC
obtains synoptic observations of the sunlit Earth from an orbit around the
L
Observations of total solar eclipses have been made with varying degrees of
sophistication for thousands of years as reviewed by Littman et al. (2008).
At a given location, observations of reduced irradiance reaching the Earth's
surface are limited to just a few minutes of totality and about 2 h of
partial obscuration (Meeus, 2003). The totality region (umbra) is an oval of
about 110–120 km in size near local noon at Casper, Wyoming, and Columbia,
Missouri, but will change size and shape as a function of local solar zenith
angle (
A detailed analysis of an eclipse that occurred in 2006 over southern Europe
includes both ground-based and space-based polar-orbiting MODIS (Moderate
Resolution Imaging Spectroradiometer) observations of cloud cover before
totality (Gerasopoulos et al., 2008) as well as theoretical modeling of the
eclipse, but unlike the present study it was largely limited to local
effects near the region of totality. A comparison from a meteorological
radiation model and measurements of total solar irradiance were made near
Athens, Greece (84 % of a total eclipse), which showed good agreement in the
presence of light clouds (Psiloglou and Kambezidis, 2007). A 3-D Monte Carlo
radiative transfer study (Emde and Mayer, 2007) was applied to the geometry
for the nearly overhead total eclipse of 29 March 2006 (13:20 LT in
Turkey) to estimate the downward global radiation at the surface, but
without the effect of clouds included in the calculation. An application of
the 3-D model to the 2006 eclipse over Kastelorizo, Greece, with fairly
cloud-free measurements (few cumulus, 1–2 octas, and scattered cirrus, 3–4 octas) at 380 nm showed good agreement for the ratio (ratio
Synoptic view of the sunlit Earth perturbed by the 21 August 2017
total eclipse centered over Casper, Wyoming, at 17:44:50 UTC. The black
region is the eclipse umbra centered over Casper, WY. The color image has
been adjusted from the images on
Synoptic view of the total eclipse centered over Columbia, Missouri,
at 18:14:50 UTC. The black region is the eclipse umbra centered over
Columbia, MO. The color image has been adjusted from the images on
The observations from the DSCOVR satellite are part of a larger project that
combines simultaneously obtained satellite and ground-based measurements
using a pyranometer (Ji and Tsay, 2000) and the Pandora spectrometer
instrument (Herman et al., 2009) at both sites. The combination will be used
to help validate 3-D radiative transfer models applicable to analysis of
eclipse effects on radiances reflected back to space and reaching the Earth's
surface. This study presents the only calibrated spectral synoptic satellite
data of the sunlit Earth ever obtained during an eclipse, which should place
tighter limits on validating radiative transfer studies under realistic
conditions. The data include EPIC-measured ozone amounts
DSCOVR/EPIC observations of the entire sunlit Earth from the eclipse day, 21 August 2017, are compared to those from two non-eclipse days to quantify the change of the global integral of reflected solar radiation caused by the eclipse. We present a potential validation test data set for the 21 August 2017 eclipse for 3-D radiative transfer models, namely the ratio of radiances without the eclipse on 20 and 23 August to the same regions that contained totality on 21 August 2017 (based on a suggestion in the paper by Emde and Mayer, 2007).
Section 2 describes the DSCOVR/EPIC instrument, available data, and
monochromatic images based on measured counts per second
(counts s
The EPIC instrument on board the DSCOVR spacecraft, in a 6-month orbit near
the L
Geolocated EPIC data (counts per second) from each set of 10 wavelengths are archived in an HDF5-formatted file available from the
permanent NASA Langley data repository center (
The EPIC HDF5 file names (e.g., epic_1b_20170821174450_02.h5) from the NASA data repository are interpreted as year 2017, month 08, day 21, 17:44:50 UTC, version 2, which is 11:44:50 local daylight savings time in Casper, Wyoming. The file name time refers to approximately the middle of the measurement sequence. Totality in Casper started at 11:42:39 MDT and ended at 11:45:05 MDT. Version 2 refers to the reprocessing of data with the latest CCD flatfielding and stray-light corrections (Herman et al., 2018; Marshak et al., 2018; Geogdzhayev and Marshak, 2018), and the geolocation algorithms.
Observing conditions for 21 August 2017 ranged from significant cloud cover
over the oceans to nearly clear skies over the United States (Figs. 1 and
2). The synoptic observations provided a unique opportunity to estimate the
fraction of reduced reflected radiation from the entire sunlit Earth caused
by a total solar eclipse. Two of the synoptic observations were timed so
that they centered on Casper, Wyoming (42.8666
The non-absorbed wavelength observations were combined to produce
eye-realistic color images (
Eclipse measurement timing and location details for five wavelengths. Eclipse maximum and EPIC image times. Total measurement duration: 2.7 min.
Table 1 summarizes eclipse timing and location details for Casper, Wyoming.
During the 2.7 min needed to obtain all 10 wavelength channel images, the
center of totality moves at about 46 km min
The timing and predicted shape of the Moon's shadow over Casper, Wyoming, and
Columbia, Missouri, can be seen at
Before quantitatively examining the EPIC data from the eclipse in units of counts per second or reflectance, the same data can be represented as monochrome grayscale images. The images (Fig. 3 with north down) range from 340 nm, with strong Rayleigh scattering effects and some ozone absorption, to 780 nm in the near infrared. North is selected as down to correspond to a 3-D projection image presented later where placing north down permits viewing inside of the umbral region. Because of the clarity of the atmosphere at 780 nm, the image serves as a geographic map of the Earth as viewed by EPIC where North and South America are clearly visible.
Atmospheric conditions during the eclipse at Casper, Wyoming, were almost cloud-free compared to Columbia, Missouri, which had optically thin low-altitude clouds (Fig. 2). Figure 4 shows the cloud cover on the day of the eclipse, 21 August 2017 (panel a), about 90 min before totality at Casper and about 2 h after totality, where the eclipse umbra is still visible over the Atlantic Ocean. The images (north is up) show that the skies remained relatively clear over the northern United States for the duration of the eclipse. A similar set of images (panel b) are shown for the day before (20 August) and 2 days after the eclipse (23 August). There were no useable data available on 22 August. Data obtained on 20 and 23 August at approximately the same UTC (backscatter phase angle for a given location on Earth) as occurred during the total eclipse are used as reference data to compare with the eclipse data on 21 August 2017. The basic global patterns of cloud cover are similar for all 3 days, but not identical. As shown later, the amount of light reflected back to space is approximately the same on the two non-eclipse days 20 and 23 August.
Grayscale images for six of the DSCOVR/EPIC channels for the eclipse over Casper, Wyoming, showing the blurring caused by Rayleigh scattering and the dark land and ocean surfaces at 340 nm to the almost clear atmosphere and bright continental surfaces at 780 nm. The images were obtained over a period of 2.7 min. North is facing down. The gray scale is linear, with black representing very low reflectivity and white very high reflectivity from high-altitude equatorial region clouds.
Figure 5a and b show longitudinal slices of 443 nm
reflected solar radiances in counts per second towards L
The minimum 443 nm values during totality are 16.6 counts s
The ratio
There is considerable variability in
For the eclipse study, the range of synoptically observed longitudes is
approximately from the international dateline (
Radiance ratio
The unique DSCOVR/EPIC measurements provide estimates of the fractional
reduction of sunlight from 388 to 780 nm reflected back to space for the
entire sunlit globe caused by the eclipse shadow on the Earth. To do this,
all of the light reaching EPIC in each of the five non-absorbed channels
(388, 443, 551, 680, and 780 nm) are integrated over the visible sunlit Earth
and compared (percent difference, PDF(
In the 3-D Fig. 7 for 443 nm, the nearly cloud-free eclipse region is the
blue area in the midst of greens, yellows, and reds. The high red values
correspond to fairly reflective clouds mostly seen near the Equator (Fig. 1).
The yellows and greens correspond to lower-altitude clouds that tend to have
smaller reflectivities. Integrating over all of the pixels for the eclipse on
21 August 2017, using the file named epic_1b_20170821174450_02.h5, we get
The counts per second observed by EPIC for the 443 nm channel
corresponding to the color image shown in Fig. 1. In the data file, the word
infinity has been replaced by the number zero. In this image there are
approximately Np
Measured counts-per-second images for six wavelength channels (340 to 780 nm)
on 20, 21, and 23 August (Fig. 8) were selected to be as close as possible to
the UTC time of the eclipse in Casper, Wyoming, keeping the scattering phase
angles nearly constant. Similar images for the strongly absorbed channels
317.5, 325, 688, and 764 nm are shown in the Appendix (Fig. A2). The
middle images in panels B and E of Fig. 8a, b, and c are for the eclipse over
Casper, Wyoming. These images are in the same format as Fig. 3, but rotated
with north up. Unlike Fig. 3, the scale in Fig. 8 was selected so that the
brightest clouds do not saturate the image. The increase in scale makes the
land surfaces less visible. While the figures are similar from wavelength to
wavelength, there are differences in the depth of the eclipse totality and
the reflectivities of the surrounding clouds. In general, equatorial clouds
with higher counts per second (reflectivities) tend to reach higher altitudes.
This is confirmed by examining the counts per second in the strongly absorbed
Average reflected light in counts per second for eclipse (21 August, red) and non-eclipse (20 and 23 August, black and gray) days from Table 3 and Eq. (1) for Casper and Columbia. The locations of the maxima are from curve fitting to the discrete wavelength measurements.
EPIC-measured
Table 3 and Fig. 9 show that the global reduction of backscattered light caused by the eclipse is similar for the two sites even though there is more cloud cover locally over Columbia than Casper. This is because the global reduction caused by the differing umbral regions is a small fraction of the total, and only 30 min has elapsed between the two measurements, which is not enough time for the global cloud cover to have significantly changed.
Figure 9 shows a plot of the data contained in Table 3 based on Eq. (1). The
two non-eclipse days are nearly identical, while the eclipse day (21 August) is
significantly lower at all wavelengths. The backscattered light (in counts per second)
peaks near 500 nm and then decreases toward longer wavelengths, since
For the 443 nm channel, the result is an approximate decrease of 9 % on
21 August at 11:44:50 MDT for Casper and 8 % at 13:14:50 CDT for
Columbia. As a reference, we compare two non-eclipse days (20 and 23 August).
The relative difference
Solar irradiance at 1 AU
PDF(
To estimate the fractional reflected radiance reduction for the wavelength
range from 388 to 780 nm, a polynomial interpolation
Figure 11b shows the product
TOA albedo measurements made by EPIC can be compared with reflectance
measurements made by the POLDER satellite instrument near the hot spot
backscatter direction (172
Measurements from the POLDER satellite over the Khingan Range, China (45.68
to 53.56
The EPIC instrument on board the DSCOVR spacecraft synoptically observes the
entire sunlit portion of the Earth from an orbit near the Earth–Sun
Lagrange-1 point. On 21 August 2017, EPIC was able to observe the totality
shadow from the lunar eclipse of the Sun with the Earth's surface for about
3 h (seven 10-channel measurements) as it crossed the United States from
west to east (about 1.5 h). When the region of totality was over Casper,
Wyoming, at 17:44:50 UTC, the reflected 443 nm TOA radiance was reduced to
16 counts s
All data authored by the
DSCOVR/EPIC project at the NASA Goddard Space Flight Center are archived at
the permanent NASA Atmospheric Science Data Center (ASDC) in Langley,
Virginia:
The course of the eclipse in the vicinity of Casper, Wyoming, and Columbia, Missouri, is shown in Fig. A1.
Grayscale images for the short UV wavelength channels (317.5, 325) with
strong ozone absorption and Rayleigh scattering, the longer-wavelength UV
channels (340, 388), and the strongly absorbed
The amount of ozone over the eclipse sites can be derived (Herman et al.,
2018) to produce ozone data that are stored in the NASA Langley archive.
During the eclipse, it is not possible to derive the amount of ozone from
either ground-based or satellite data. Ozone amounts do not change rapidly
from day to day except when major weather systems pass through a region,
which was not the case during the eclipse period, 20 to 23 August.
This is confirmed from OMI satellite data (Ozone Monitoring Instrument
on board the AURA satellite). Figure A3 shows the amount of ozone over the
eclipse trajectory obtained on 20 August. The values obtained (316 DU near
Casper, WY; 306 DU near Columbia, MO) compare well with ozone amounts derived
from OMI of 314 and 301 DU. The
The timing and shape of the Moon's shadow over Casper, Wyoming,
showing the relative location of Casper and Columbia (white circles) at
11:45 MDT (Mountain Daylight Time) and 13:15 CDT
(Central Daylight Time). The shadow is moving at about 46 km min
EPIC-measured ozone amounts from 20 August in the vicinity of Casper, WY, and Columbia, MO.
JH wrote most of the paper and performed most of the calculations. GW is the funded principal investigator of the project. AM provided the calibration coefficients for the visible and near-IR channels. KB provided the color images in Figs. 1 to 3. She was responsible for the geolocation of the 10 filter images on a common grid. LH provided the calibration coefficients for the UV channels. AC provided the flatfielding, stray-light correction, and dark-current analysis. NA helped with flatfielding and stray-light correction and was responsible for the ground-based portion of this research. MK provided the flatfielding, stray-light correction, and dark-current analysis.
The authors declare that they have no conflict of interest.
The author would like to thank the DSCOVR project for support in completing this study as well as financial support from an accepted NASA-ROSES proposal in response to NNH16ZDA001N-ISE. Edited by: Bernhard Mayer Reviewed by: two anonymous referees