During this 28-day study, the sun was nearly at the zenith at solar noon. While the UVR measured at and near sea level was significantly modulated by clouds and vog during many days, sky conditions allowed mannequin head deployments during some or all of 14 days. Some measurement days at MLO were made through an exceptionally clear sky in which UVR was modulated only by the ozone layer and Rayleigh scattering. During other days at MLO, clouds, vog, and a rare smoke event significantly altered the UVR. At or near sea level only a few days were free of clouds for most of a day. The results that follow reflect these very different sky conditions at the best sites where we made measurements.
4.1. Mannequin Head Results
When all seven head sensors were in use, the head’s two data loggers collected 420 samples each minute and 252,000 samples per 10-h day.
Figure 4 shows the ozone-corrected UVI measured by the left cheek sensor during three midmorning rotations of the head. The peaks in the data occurred when the cheek was pointed toward the sun. The minima occurred when the cheek was pointed opposite the sun. While the minima have a much lower UVI than the peaks, these data provide important information about diffuse UVR scattered from the sky and the surface.
Figure 5 shows the variation in the ozone-corrected UVI received each second and plotted in blue for the left cheek, left ear and left eye during ~800 hundred 36-s rotations of the mannequin head at MLO on 6 August 2018. The forehead sensor was not often used, for this location is often covered by hair or shaded by a hat. The global UVR measured by the PMA1102 (red) is superimposed over the mannequin head data in
Figure 5, which makes the very significant reduction in the UVI at solar noon especially obvious. At least as significant is that (1) the horizontal UVI is much higher than that measured by each of the three sensors during most of the time data was collected, and (2) the UVI measured by all three sensors is much higher at midmorning than at solar noon. As in
Figure 4, the upper extreme of the ~25,000 samples plotted for each sensor in
Figure 5 represents when the sensor was pointed in the direction of the sun, while the lower extreme represents when the sensor was pointed opposite the sun’s position.
Table 3 tabulates what is visually obvious in
Figure 5. Note that the noon minimum UVI represents the UVR received from the sky and the lava surface, with each filling half the sensor’s FOV when the sensor was pointed opposite the direction of the sun. We address the potential significance of this scattered UVR and the cosine error of the sky in
Section 4.2.2.
The UVR sensitive surfaces of the head sensors are perpendicular to the surface and pointed at the horizon, and this explains the sharp drop in the UVI at noon when the sun is almost directly overhead. While the cheek and ear sensor data form similar patterns, the eye sensor was partially shaded by the brow, and this resulted in the prolonged reduction before and after noon shown in
Figure 5. Conversely, the eye sensor measured the highest UVI, and we attribute this to scattering of UVR toward the sensor by the brow. Note that the diffuse UVR scattered by the lava and sky opposite the direction of the sun was too low to be detected during the first hour or so during this measurement session.
Figure 6 shows the global UVI measured by the zenith looking PMA1102 and the mannequin head’s left cheek UVI sensor for the most cloud-free days at four of the measurement sites. The mannequin head’s left cheek was selected for comparison with the UVR measured by the PMA1102, since the cheek sensors were subject to less shading from the mannequin head’s facial features than the ear and eye sensors, which were slightly shaded by, respectively, the brows and the ear lobes.
The UVI data plotted in
Figure 6 clearly show significant differences among the three sites.
Table 4 summarizes the most pertinent data for the voggy Kona site in (a) and the clear MLO site in (c). The noon peak and minimum UVI is higher for Kona than MLO, which is to be expected in view of scattering of UVR from the hazy sky over Kona caused by a thick layer of vog as opposed to the very clear sky at MLO.
Note that the data plotted in
Figure 6 have a much tighter waist at solar noon on 22 July and 6 August than on 12 and 13 August. This suggests that the cheek sensor was perpendicular to the surrounding surface at solar noon, while the wide waist data suggests that the sensor was looking slightly upward. This was confirmed for the 12 August session by photographs of the mannequin head, which revealed that the head was tilted upward ~10° toward the west during the morning. While the turntable was always made level before the mannequin head was placed atop it, its angle likely shifted when the mannequin head was pushed down over its holder more tightly than usual due to windy conditions.
The data in
Figure 4,
Figure 5 and
Figure 6 are plotted against local time to provide a more intuitive depiction of the UVI magnitude over the course of a day than the SZA. The data can also be plotted against the SZA as shown for the 22 July morning measurements in
Figure 7. At solar noon, the PMA1102-measured UVI reached 12.8 (extremely high) while the UVI at the left cheek was only 3.3 (moderate). When the mannequin head’s left cheek was facing toward the sun, the UVR at the cheek exceeded its noon UVI (3.3) from a SZA of from 56° to 3°, reaching a peak UVI of 5.5 at a SZA of 32°, which clearly demonstrates why facial protection from UV is more important at midmorning and midafternoon than at noon for this and other tropical sites and probably temperate sites during summer.
4.3. UVR and Volcanic Sulfur Dioxide
From 1986 to 2018, continuous emissions from the Kilauea volcano caused hazy vog and respiratory distress across Hawai’i Island. Vog is formed from the aerosolization of SO
2 that has combined with water vapor in the emission plume and the atmosphere [
36]. Residual SO
2 in vog absorbs UVR, while the aerosols in vog both absorb and scatter UVR. The latter effect significantly increases the sky’s diffuse UVR. During the first 19 days of our field study, the historic eruption of the Kilauea volcano that ended 5 August blanketed much of Hawai’i Island with considerably more vog than usual. On 22 July at the Old Kona Airport Park beach, the cloud-free sky was veiled with volcanic SO
2 at solar noon. As explained above, byBy shading a zenith-facing, global UVI sensor with a black 2-cm diameter disk mounted on a 1-mm rod and held approximately 20 cm over the sensor, we found that 62% of the surface UVR was scattered from the hazy sky and 38% arrived from the direct sun. The sky was much cleaner at MLO on 12 August, and the same method found that only 31% of the UVR was scattered from the sky (half the sea level fraction) and 69% arrived from the direct sun. While haze can alter a UVR sensor’s cosine response, we were unable to incorporate appropriate corrections, for we did not regularly measure the diffuse/direct UVR.
During the main phase of this eruption, Kilauea’s fissure 8 (
Figure 9) was emitting up to 50,000 tons of SO
2 per day [
37], and this provided a serendipitous opportunity to expand the measurement objectives for the Hawai’i UV study, particularly since the UVR absorption spectrum of SO
2 emitted by coal-burning power plants and industrial activity matches the UVR spectral absorption by volcanic SO
2. Natural and anthropogenic aerosols mixed with the SO
2 may cause minor differences.
The Fissure 8 eruption plume created a continuous cumulus cloud infused with significant SO
2 that we measured by pointing a PiSpec spectrometer at the zenith sky during one of two vehicular traverses along Pāhoa-Kalapana Road (Highway 130) under and adjacent to the eruption plume (
Figure 10). The PiSpec is a novel UV spectrometer that has a resolution of 1 nm and a bandpass of 60 nm centered at 310 nm [
38]. Specifically designed to measure volcanic SO
2, the PiSpec is built from a 3D-printed housing, off-the-shelf optical components and a low-cost sensor primarily designed for the smartphone market. The latter is from a modified Raspberry Pi camera module, making sensor control and data acquisition relatively straightforward using the Python programming language.
Because SO
2 efficiently absorbs UVR wavelengths also absorbed by ozone, SO
2 from industrial sources can inflate measurements of total column ozone made by Dobson, Brewer and other spectroscopic instruments. Zerefos et al. [
39] have observed that: “… SO
2 has a number of strong UV absorption bands in which the cross-sections are three to four times as large as those for ozone.” This interference has long been known to the ozone monitoring community, and Komhyr and Evans [
40] have reported that Dobson spectrophotometer measurement of total ozone in regions with high levels of anthropogenic SO
2 can cause errors as high as +25%.
We observed examples of this while using a handheld Microtops II
® (Solar Light Microtops II
®) (Solar Light, 100 East Glenside Ave., Glenside, PA 19038, USA) [
41]) to make ozone measurements where the SO
2 was especially concentrated near the Fissure 8 eruption plume. Microtops II measures direct UVR at 300, 305 and 312 nm (5 nm FWHM) in W/m
2 when pointed directly at the sun. Microtops II infers the ozone column from ratios of direct sunlight intensity at these wavelengths, where some of the key absorption bands of SO
2 and ozone overlap. Thus, SO
2 can cause erroneously high ozone measurements by Microtops II. The instrument also measures total water vapor at 940 nm/1000 nm and the optical depth at 1000 nm.
Microtops II measurements indicated excessively high levels of total ozone over Pāhoa on 31 July and over both Kalapana and the Seaview Estates Park on 4 August. On 4 August, Microtops II measured up to 410 DU of ozone through openings to the sun in the volcanic eruption cloud over Seaview Estates Park 8 km SSW of Fissure 8. This measurement was clearly erroneous, for the average of 99 total ozone measurements from several sites well away from the eruption cloud on 24–26 July was 270 DU. Thus, it’s reasonable to conclude that the ozone error caused by SO
2 on 4 August was 410–270 or 140 DU, an error of +52%. On 5 August at the Seaview Park site, the fraction of sky covered by the volcanic cloud was significantly reduced. This permitted 128 Microtops II total ozone measurements, the average of which was only 2% over the background ozone measured on 24–26 July. All Microtops II ozone measurements during this time are plotted in
Figure 11.
The Fissure 8 eruption ceased on 5 August, and the Hawai’i Volcano Observatory described the dramatic drop in SO2 emissions that followed:
“When lava output from fissure 8 suddenly declined in early August, SO
2 emission rates dropped precipitously as well. Emissions on Aug. 3 indicated tens of thousands of tons of SO
2 coming from the fissure 8 vent, but just two days later, on Aug. 5, the emission rate was only about 200 tons per day.” [
42]
4.4. UVR during Major Smoke Event at MLO
Brush fires caused by lightning, human activity, and military exercises are relatively common on Hawai’i Island during drought years. A rare biomass smoke event occurred over MLO from 11:45 to 12:25 on 7 August 2018 when dense smoke from the Keauhou Fire (5–15 August 2018), a major brush and forest fire on the southeast slope of Mauna Loa [
43], covered most of the sky and provided a second serendipitous measurement opportunity. While this was a highly unusual event, the staff at the adjacent Mauna Loa Solar Observatory reports it has occasionally observed similar smoke events over MLO [
44]. Smoke from the fire was also present over MLO on 10 August.
UVR is strongly absorbed by smoke [
45,
46], and this was dramatically illustrated by the UVI measured during this event by the PMA1102 sensor and confirmed by data from a CSU broadband UVR radiometer and the pair of Canadian Brewer spectroradiometers (Nos. 009 and 119) at MLO.
Figure 12 shows the UVI measured by these four sensors during the smoke event, which fortuitously arrived just before solar noon (12:30) after a very clear morning. The PMA1102 UVI data are plotted as a continuous line at one-second intervals. The CSU broadband UVI data are plotted as the average of 3-min intervals, and the preliminary UVI measured by each of the two Brewers is plotted at greater intervals. The minimum UV occurred at 12:07, when the full sky UVR was only 8.3% of the UVR before the smoke arrived.
Figure 12 includes a 190° photograph of the smoky sky 1 min past the peak blockage of UVR (12:08).
The smoke was also measured by all 7 channels of a CSU UVR radiometer (Yankee UV-1; HI22). and a CSU 7-channel Ultraviolet Multifilter Rotating Shadowband Radiometer (Yankee UVMFR-7). The latter provides measurements at 300, 305, 311, 317, 325, 332, and 368 nm (5 nm full width, half maximum (FWHM)) of the total, direct sun, and diffuse UVR at 3-min intervals.
Figure 13 shows significant attenuation of UVR at all 7 wavelengths during the smoke event. UVB wavelengths can suppress or kill exposed viruses and bacteria that might cause infectious diseases in people and animals. Thus, the persistent suppression of natural UVB by biomass smoke or clouds might play a role in the incidence of infectious respiratory diseases [
46,
47]. Therefore, the MLO smoke event provides relevant data for future studies of this possible connection. After the smoke drifted away, the sky became sufficiently pristine to permit Langley calibrations of 5 Microtops II sun photometers.
4.5. Photography
Photography provides an important and inexpensive method for recording sky conditions, yet few papers on UVR and the UVI include photographs of the sky, much less the surrounding landscape or body of water. We employed several conventional and specialized digital cameras to record the sky, landscape, rotating mannequin head, PMA1102 sensor, ocean and landscape during our measurement sessions. The principle cameras we used are listed in
Table 6.
These cameras provided superb results. While their total cost was approximately $2000 US, investigators with a limited budget can collect reasonable sky images with an inexpensive mobile phone camera equipped with a clip-on, wide-angle lens available for as little as $15 US.
During deployments of the rotating mannequin head, photographs were made of the solar aureole, which provide a visually obvious view of atmospheric haze that forms a glow around the sun (
Figure 14). The aureole was captured by a digital camera (Canon G9) mounted on a fixture equipped with a black ball centered in the camera’s field of view (FOV) that blocks the direct sun (0.5°) with a 4° black circle [
48]. The camera settings (f4, 1/600 s, ISO-500) were identical to those used for solar aureole photographs on all days when atmospheric measurements were made at Geronimo Creek Observatory (GCO), Texas (29.6° N, 97.9° W), since 3 October 2008. These fixed settings permit quantitative comparisons of sky conditions recorded in thousands of solar aureole images from Hawai’i Island and GCO.
The sky over each measurement site was also photographed at intervals with a digital camera (Sony α6300) equipped with a lens having a 190° (2.17π sr) FOV (Meike 6.5 mm f/2.0 Circular Fisheye Lens). These images, four of which are shown in
Figure 15, show the entire sky and provide a visually obvious depiction of haze. Their outer edges show surrounding trees, soil, water, and structures.
Virtual reality (VR) cameras produce omnidirectional images with a spherical, 360° (4π sr) FOV of view of the entire sky and surrounding terrain, which can be depicted in various formats (
Figure 16). We employed a miniature VR camera (Xiaomi Mijia Mi Dual-Lens Sphere Action Cam) with a pair of 190° (2.17π sr) lenses, one on each side of this shirt-pocket camera. There is sufficient overlap in the FOV of the two lenses to block a rod used to support the camera, thereby providing images that suggest the camera is floating in the air.
Our cameras captured images of the sky intended for a webpage for the public that describes how shade and a hazy sky affect UVI. While the role of shade in reducing the UVI is intuitive, the image of the sky through coconut palms in
Figure 17 shows that significant UVR can leak through overhead branches. The public may not know that diffuse UVR scattered from a sky polluted with aerosols can exceed direct sun UVR. This is also illustrated in
Figure 17, a photo of the open sky veiled by thick haze caused by volcanic vog. UVR from the diffuse sky in this image was 62% of the total UVR, while only 38% arrived from the direct sun.
A novel, low-cost UV camera [
49] that could be fitted with various UVR filters was used to make several hundred UV photos of the mannequin head at the Old Kona Airport Park beach. The ozone column blocks virtually all sunlight below 295 nm. Thus, the UV camera’s spectral response with a 300-nm filter (40 nm FWHM) approximated that of the UVB portion of the erythemal action spectrum. The color-coded versions of these images (TW) in
Figure 18 clearly indicate significant variations in the intensity of the UVR illuminating the mannequin’s head’s face at midafternoon when the head was pointed north, east, south, and west. While we have yet to analyze these images, we include them here to illustrate their potential merit.
Finally, the historic eruption of Kilauea provided an opportunity to employ the UV camera technique commonly applied in volcanology [
50,
51,
52] to make UV photographs of the major emission plume from what was then the world’s single largest emitter of SO
2, which strongly absorbs UVR. The plume from Kilauea’s Fissure 8 is clearly visible in the UVR photographs in
Figure 19a,b, both of which were made at 10:39 on 31 July 2018. Photograph (a) was made through a 310-nm filter (10 nm FWHM), a wavelength strongly absorbed by SO
2. Photograph (b) was made through a 330-nm filter (10 nm FWHM), a wavelength weakly absorbed by SO
2. This explains why the plume appears darker and larger in the 310-nm image. Some of the dark cloud to the left (east) of the plume is caused by SO
2 rising from the lava flowing to the ocean from Fissure 8.
Figure 19c is a visible light photograph made at the same time as the two UVR images. SO
2 does not significantly absorb visible light. Thus, the plume appears much narrower and less dense than in the UVR images.
Horizontal scans with ImageJ image processing software across the plume between the tops of the palm tree at left and the utility pole at right in
Figure 19a,b reveal the optical density of the 310- and 330-nm images. The resulting plots in
Figure 20 show the enhanced absorption by SO
2 at 310 nm that quantifies what is visually evident in
Figure 19.