USING THE LUNAR AUREOLE DERIVED FROM ALL-SKY CAMERAS FOR THE RETRIEVAL OF AEROSOL MICROPHYSICAL PROPERTIES

Size: px

Start display at page:

Download "USING THE LUNAR AUREOLE DERIVED FROM ALL-SKY CAMERAS FOR THE RETRIEVAL OF AEROSOL MICROPHYSICAL PROPERTIES"

Jerome Patterson
5 years ago
Views:

1 USING THE LUNAR AUREOLE DERIVED FROM ALL-SKY CAMERAS FOR THE RETRIEVAL OF AEROSOL MICROPHYSICAL PROPERTIES R. Román, B. Torres, D. Fuertes, V.E. Cachorro, O. Dubovik, C. Toledano, A. Cazorla A. Barreto, J.L. Bosch, T. Lapyonok, R. González, P. Goloub, M.R. Perrone, F.J.Olmo, A. de Frutos, L. Alados-Arboledas ACTRIS-. nd WP workshop Barcelona(Spain) November, 16

2 WHY AEROSOLS ARE IMPORTANT AT NIGHT? - Clouds absorb longwave radiation from Earth surface and reemit part of it to the Earth (warming the surface). It has more impact at night since clouds don t reflect solar radiation to the space (cooling). - Aerosol direct effect (extinction of incoming solar radiation) is not possible at night. But indirectly effect of aerosols (acting as cloud droplet nucleii) can vary the cloud properties and presence. Changes caused by aerosols on these cloud properties at night could contribute to global warming. - Winter seasons (especially polar areas) presents low sunshine duration, hence it is necessary the knowledge of aerosol properties also at night in order to improve the aerosol characterization in these seasons.

3 HOW AEROSOL CAN BE CHARACTERIZED AT NIGHT? - Lidar. Among others: Backscatter and Extinction profiles using elastic and Raman measurements. - In situ. Well characterization but at ground level. - Star/Lunar photometry. Aerosol Optical Depth at different wavelengths: *Discrimination between the extinction of fine and coarse mode (O Neill et al. 3). *Reliable characterization of the fine mode and less accurate information about coarse mode (Torres et al. 16). Spectral AOD plus sky radiance give adequate information to retrieve aerosol microphysical/optical properties (Abs+ Scat) (Dubovik & King, ) at day but measured sky radiance is not available at night yet.

4 LOOKING FOR SKY RADIANCE INFORMATION AT NIGHT We need information about scattered (by atmos.) radiance near an enough brillant celestial body at night. In this sense, the Moon was selected as target because: - Moon is too brillant, at least between quarters. - AOD from lunar photometers will be available. - Only ~15 days per month of measurements. What instrument can be used to detect Moon aureole? Sky cameras! They record the full hemispherical sky radiance. They measure different wavelength intervals. They are operative at night changing time exposure. High noise/signal ratio.

5 Objective OBJECTIVE AND METHOD To check if it is possible to obtain some information about sky radiance during the night observing moon aureole. Specifically, relative radiances from a sky camera. Steps - Methodology description: SONA sky camera is used; the effective wavelengths of its channels are calculated; sky geometry of each pixel is obtained; pixel sensitivity is retrieved; finally the camera is configured to take a set of multi-exposure night sky images to obtain a HDR image. - Validation of the obtained radiances: by comparison with those simulated by GRASP (Generalized Retrieval of Aerosol & Surface Properties) using the aerosol properties retrieved at day as input. - Retrieval of aerosol properties using camera information

), a ceilometer CHM-15k (Jenoptik) and a Triple (Sun/Sky/Moon) photometer (Cimel Electronique).

6 SITES This work has been developed with data measured at Valladolid and Granada (Spain). Both places are equipped with a SONA sky camera (Sieltec S.L.), a ceilometer CHM-15k (Jenoptik) and a Triple (Sun/Sky/Moon) photometer (Cimel Electronique). All instruments are installed at the rooftop of: Science Faculty of Valladolid and IISTA/CEAMA of Granada, respectively. 45. N 4.5 N Valladolid 4. N 37.5 N Granada 35. N 1. W 7.5 W 5. W.5 W..5 E 5. E

7 INSTRUMENTATION SONA sky camera (Automatic Cloud Observation System) is a device which provides full sky images at day and night. Is formed by a surveillance CCD camera with a fisheye lens inside a waterproof case and a dome with a shadow band blocking the Sun. It was mainly designed for cloud cover detection. CIMEL CE-318T photometer. AOD at 8 wavelengths (34-164nm) and sky radiance measurements at day. Well characterized aerosol properties at day from AERONET. AOD at 6 wavelengths at night (44-164nm) using ROLO model.

SONA ALL-SKY CAMERA (CHARACTERISTICS) Camera sky geometry (Azimuth, Zenith and FOV) are obtained using a dataset of

Granada Granada Pixel sensitivity is not linear with exposure.

Effective wavelengths are obtained weighting camera filters by Moon extraterrestrial spectrum. 4 6

8 SONA ALL-SKY CAMERA (CHARACTERISTICS) Camera sky geometry (Azimuth, Zenith and FOV) are obtained using a dataset of sky images under cloudfree conditions and wellknown coordinates of some visible stars/planets. Granada Granada Pixel sensitivity is not linear with exposure. Sensitivity function is calculated using images at different exposure times. Effective wavelengths are obtained weighting camera filters by Moon extraterrestrial spectrum Zenith (º) FOV(msr) log(relative Irradiance * Time) Camera Spectral Response Granada Red Channel Green Channel Blue Channel Pixel counts Valladolid 469nm 537nm 63nm Wavelength (nm)

HIGH DYNAMIC RANGE (HDR) IMAGES HDR can be

in a single image where the changes in pixel

For visualization Linear response Moon is

9 HIGH DYNAMIC RANGE (HDR) IMAGES HDR can be used to solve saturation non-linearity and sensitivity range. Multi-exposure technique: 16 images are taken doubling the time exposure. Sensitivity function converts the image set in a single image where the changes in pixel intensity is linear with incoming radiation (Debevec and Malik, 1997). Tone Mapping (Reinhard et al. ). For visualization Linear response Moon is too much intense log(relative Irradiance * Time) Red Channel Green Channel Blue Channel Pixel counts

EXTRACTING RELATIVE RADIANCES - Once the linear HDR image is obtained, the noise and background light from the city are removed subtracting an HDR image under clear conditions.

10 EXTRACTING RELATIVE RADIANCES - Once the linear HDR image is obtained, the noise and background light from the city are removed subtracting an HDR image under clear conditions. The weight of background image is scaled by the ratio between the dark signal on the original image and the dark signal on the background image. - Every pixel signal is divided by its FOV. This signal at averaged for each channel within 1º at several points of the lunar almucantar (both branch are averaged. - Finally, for each channel, the signals are normalized by the sum of all. Normalized Camera Radiance /5/16 :4 UTC 4/5/16 :4 UTC 4/5/16 :4 UTC cc(camera)= NaN oktas (NaN%); cc(ceilometer)= cc(camera)= NaN oktas (NaN%); cc(ceilometer)= Tone map of original HDR image Removed Noise and Background Selection of points of lunar almucantar 63 nm cc(camera)= NaN oktas (NaN%); cc(ceilometer)=

11 GRASP GRASP (Generalized Retrieval of Aerosol and Surface Properties) is an interesting algorithm for the retrieval of: - Optical and microphysical aerosol properties. - Optical surface properties. Its main characteristics are: - Versatile (gives the possibility of use different measurement kinds). - Flexible (Capacity to incorporate and exchange different methods, modules and libraries. Can be run as forward model). - The code is open and free. Lidar Satellite Sky camera Others GRASP Photometer

$.1 1 radius ( m) AERONET AOD 44 =.5 - Refractive Indices. - Sphere fraction. - Size Distribution. cc(camera)= NaN oktas (NaN%); cc(ceilometer)= NaN oktas.1 x 1 5 Granada.1 5 1 June 16 3:1 GMT.$ 8 4 Camera radiances during the night are compared with the forward simulated by GRASP (using as input AERONET info).

8 4 Camera radiances during the night are compared with the forward simulated by GRASP (using as input AERONET info).

12 Altitude (km) 6 4 dv(r)/dln(r) ( m 3 / m ) COMPARISON: CASE 1 GRANADA 1-13 JUNE 16 -Low aerosol load. -Low moon intensity -First quarter (phase=-83º). 18: : 6: Time (UTC)..1 1 radius ( m) AERONET AOD 44 =.5 - Refractive Indices. - Sphere fraction. - Size Distribution. cc(camera)= NaN oktas (NaN%); cc(ceilometer)= NaN oktas.1 x 1 5 Granada June 16 3:1 GMT.8 4 Camera radiances during the night are compared with the forward simulated by GRASP (using as input AERONET info). Differences are high due to the high noise caused by low moon intensity and low aerosol load. 3 1 Ratio Camera/GRASP Normalized Camera Radiance Normalized GRASP Radiance nm nm nm Ratio from Rayleigh simulation (no aerosol) to aerosol simulation.

$- Refractive Indices. 1 radius ( m) - Sphere fraction. - Size Distribution. cc(camera)= NaN oktas (NaN%); cc(ceilometer)= NaN oktas.$ 1 x 1 5 5 4 3 1 Valladolid 8 May 15 3:55 GMT Green and red ratio present ratio values near to 1.

1 x 1 5 5 4 3 1 Valladolid 8 May 15 3:55 GMT Green and red ratio present ratio values near to 1.

13 Altitude (km) COMPARISON: CASE VALLADOLID 8-9 MAY Fine aerosol (Moderate load). -1 st nd quarter (phase=-54º). -Medium moon intensity 18: : 6: Time (UTC) dv(r)/dln(r) ( m 3 / m ) AERONET AOD 44 =. - Refractive Indices. 1 radius ( m) - Sphere fraction. - Size Distribution. cc(camera)= NaN oktas (NaN%); cc(ceilometer)= NaN oktas.1 x Valladolid 8 May 15 3:55 GMT Green and red ratio present ratio values near to 1. Blue channel has more deviation but is similar to the deviation caused by the changes on aerosol along the nigth (between afternoon and next morning). Ratio Camera/GRASP Normalized Camera Radiance Normalized GRASP Radiance nm nm 63 nm Ratio from near-sunset aerosol simulation (afternoon) to near-sunrise (morning).

$36 = - Refractive Indices. 1 radius ( m) AERONET - Sphere fraction. - Size Distribution. cc(camera)= NaN oktas (NaN%); cc(ceilometer)= NaN.$

14 Altitude (km) 6 4 COMPARISON: CASE 3 GRANADA 3-4 MAY 16 -Coarse aerosol (high load). -After Full Moon (phase=5º). -Medium-high moon intensity 18: : 6: Time (UTC) dv(r)/dln(r) ( m 3 / m ).1.5 AOD AOD =.36 = - Refractive Indices. 1 radius ( m) AERONET - Sphere fraction. - Size Distribution. cc(camera)= NaN oktas (NaN%); cc(ceilometer)= NaN.1 oktas x Granada 4 May 16 :4 GMT All channels present ratio values near to 1. In fact, the differences in aerosol between afternoon and the next morning provides high differences in the relative radiances. Camera radiances looks good when moon intensity and aerosol load are enough. Ratio Camera/GRASP Normalized Camera Radiance Normalized GRASP Radiance nm nm 63 nm Ratio from near-sunset aerosol simulation (afternoon) to near-sunrise (morning).

15 GRASP Normalized Radiance.15 COMPARISON Camera Normalized Radiance Other four nights have been added.1 to the.1 dataset for comparison, given 7 nights (.5.5 clean cases, fine aerosol cases, and 3 coarse aerosol cases) nm Camera Normalized Radiance GRASP Normalized Radiance The camera radiance shows a high correlation with the simulated one for all nights, but clean cases present more deviation. The differences (ΔNR) between the camera and simulations have been calculated for each night, HDR image and channel as: NR % = 1% NR cam NR GRASP NR GRASP 4-5 Apr May 15-3 Aug May May Feb Jun 16 Line 1:1 Where NR cam is the normalized radiance of the camera and NR GRASP the simulated one. Camera Normalized Radiance Camera Normalized Radiance GRASP Normalized Radiance GRASP Normalized Radiance 63 nm GRASP Normalized Radiance

16 COMPARISON ΔNR has been calculated using all available data and its distribution is shown for different azimuth angle intervals. - Green channel presents the low bias, while camera over/underestimates the simulations near the moon at blue/red channel. - Standard deviation (uncertainty if gaussian) increases with azimuth. - Uncertainty is 9-13% for º intervals, for all channels. - The obtained uncertainty considering all data at each channel (5364 data) is 19%, 17% and 1% for blue, green and red channels. - If clean cases are excluded, total uncertainty goes down to 15%. NR (%) NR (%) NR (%) º º º 19.75º º º º º º 19.75º º º 63 nm º º º 19.75º º º

17 COMPARISON The obtained results were calculated using azimuth values from 18.5º to º each.5º. But if only use up to 19º: - Uncertainty is 13-16% below 183.5º, and 8-9% for º. - If clean cases are excluded, the uncertainty is between 6% and 8% for all channels and azimuth intervals below 186º. - The obtained uncertainty considering all data at each channel (384 data) is 14%, 13% and 16% for blue, green and red channels. - If clean cases are excluded, total uncertainty goes down to 11%, 9% and 1% for blue, green and red channels. NR (%) NR (%) NR (%) º º º º º º º º º º º º 63 nm º º º º º º

18 AEROSOL SIZE DISTRIBUTION RETRIEVAL The night of nd to 3 rd August 15, a Saharan dust plume came to Valladolid. GRASP was run to retrieve aerosol properties that 3/8/15 3:5 UTC night using AOD from lunar photomerer and sky radiances of Moon aureole from camera. amera)= NaN oktas (NaN%); cc(ceilometer)= NaN oktas 63 nm AOD /Aug 6: /Aug 1: /Aug 18: 3/Aug : 3/Aug 6: 3/Aug 1: 3/Aug 18: Angström Exponent The growing up of coarse mode along the night can be observed in the results, while fine mode is constant. 44 nm 5 nm 675 nm 87 nm 1 nm 164 nm /Aug 6: /Aug 1: /Aug 18: 3/Aug : 3/Aug 6: 3/Aug 1: 3/Aug 18:.1 dv(r)/dln(r) ( m 3 / m ) Angström44-87 day Angström44-87 night AERONET 18:13 UTC AERONET 6:4 UTC GRASP+cam+AOD 3:5 UTC

19 SUMMARY AND CONCLUSIONS - Sky cameras and HDR imagery have been proposed as a possible instrument capable to help in the nocturnal aerosol characterization. - The lunar almucantars from camera have been compared with simulated radiances, obtaining a better fit beween camera and model when the aerosol load is high and the moon enough bright. - Under non-clean conditions, the camera normalized radiance uncertainty for all channels is around 15% when the almucantar is up to º far the moon, and around 1% when the almucantar is up to 1º far the moon. - Aerosol size distribution has been retrieved with GRASP using lunar photometer and a sky camera during a dust episode over Valladolid, showing how the coarse (dust) mode was increasing along the night.

20 OUTLOOK 4/5/16 :4 UTC Normalized radiance at Moon aureole from sky camera will be included in the retrieval of aerosol profiles at night using lidar measurements. The results are expected to be shown at the Genereal Actris- assembly of 17 at Granada. cc(camera)= NaN oktas (NaN%); cc(ceilometer)= NaN oktas

21 THANK YOU! QUESTIONS?

Characterization of the atmospheric aerosols and the surface radiometric properties in the AGRISAR Campaign

Characterization of the atmospheric aerosols and the surface radiometric properties in the AGRISAR Campaign V. Estellés Solar Radiation Unit Universitat de València T. Ruhtz, P. Zieger, S. Stapelberg Institute