Technical Note: Determination of aerosol optical properties by a calibrated sky imager
Metadata
Show full item recordAuthor
Cazorla Cabrera, Alberto; Shields, J. E.; Karr, M. E.; Olmo Reyes, Francisco José; Burden, A.; Alados Arboledas, LucasEditorial
Copernicus Publications
Materia
UV erythema irradiance Heat-wave Southeaster Spain Atmospheric aerosols Principal-plane Networks
Date
2009Referencia bibliográfica
Cazorla, A.; et al. Technical Note: Determination of aerosol optical properties by a calibrated sky imager. Atmospheric Chemistry and Physics, 9: 6417-6427 (2009). [http://hdl.handle.net/10481/31808]
Sponsorship
This work was supported by the Centro de Investigación Científica y Tecnológica (CICYT) of the Spanish Ministry of Science and Technology through projects CGL2007- 66477-C02-01 and CSD2007-00067 and the Andalusian Regional Government through project P06-RNM-01503 and P08-RNM-3568). First author has been funded by the Andalusian Regional Government and his research stay at University of California at San Diego has been also funded by the Andalusian Regional Government.Abstract
The calibrated ground-based sky imager developed
in the Marine Physical Laboratory, the Whole Sky Imager
(WSI), has been tested with data from the Atmospheric
Radiation Measurement Program (ARM) at the Southern
Great Plain site (SGP) to determine optical properties of the
atmospheric aerosol. Different neural network-based models
calculate the aerosol optical depth (AOD) for three wavelengths
using the radiance extracted from the principal plane
of sky images from the WSI as input parameters. The models
use data from a CIMEL CE318 photometer for training and
validation and the wavelengths used correspond to the closest
wavelengths in both instruments. The spectral dependency
of the AOD, characterized by the A° ngstro¨m exponent in
the interval 440–870 nm, is also derived using the standard
AERONET procedure and also with a neural network-based
model using the values obtained with a CIMEL CE318. The
deviations between the WSI derived AOD and the AOD retrieved
by AERONET are within the nominal uncertainty assigned
to the AERONET AOD calculation (±0.01), in 80%
of the cases. The explanation of data variance by the model
is over 92% in all cases. In the case of , the deviation is
within the uncertainty assigned to the AERONET (±0.1)
in 50% of the cases for the standard method and 84% for the
neural network-based model. The explanation of data variance
by the model is 63% for the standard method and 77%
for the neural network-based model.