Resumo:
The ultraviolet radiation (UVR) influences life on Earth and its monitoring is relevant and
necessary for studies on different areas of knowledge. UVR modelling by radiative transfer
models (RTM) is always an alternative for lack of surface and/or remote measurements.
However, RTM estimates may present significant errors related to the UVR attenuation
processes in the atmosphere. Cloud cover (CC) effects represent one of the most important
source of RTM errors. In general, sky images are taken by all-sky cameras as a tool for
automating cloud quantification, but these instruments are still an expensive option for poorer
countries. Thus, we develop a low-cost and affordable technique for determining CC using
photography. 230 sky images were collected using a smartphone (Samsung Galaxy A5) and
photographic camera (Sony Cyber-Shot – DSH-H7) simultaneously. This high-quality set of
sky photos covers different types and amounts. A set of algorithms processed the images,
classifying and counting the pixels of the clouds. 13 thresholds were tested, among the ones
found in the literature (0,12, 0,65, and 0,90.), to classify the pixels of the clouds. For
validation, the photographs were visually assessed by an observer, and statistically, by the
methods of analysis of linear regression and coefficient of determination (R²). UVR estimates
were performed using the Tropospheric Ultraviolet and Visible (TUV) RTM model. The
following conditions were observed for atmospheric vertical profile: a) tropical atmosphere
profile; b) total ozone content (TOC); and c) aerosol optical properties at Itajubá’s site
(22.4°S, 45.5°W, 850m.). Previously, RTM parametrization and sensitivity tests were
performed involving different optical depths (OD), vertical thicknesses, cloud base height and
cloud top height for winter and summer solstices. Then, UVR data provided by a Kipp &
Zonen SUV-E radiometer were used to compare UV Index (UVI) estimates from RTM by
statistical methods: a) the root-mean-square error (RMSE); b) correlation coefficient (r); c)
BIAS and; index of agreement (d). To evaluate the effect of the clouds, the Cloud
Modification Factor (CMF) Index was used. Results showed satisfactory performance of the
technique (p < 0.0001) with the utilization of the thresholds of 0.18 (smartphone) and 0.26
(photographic camera), being higher the ones obtained with the utilization of the threshold of
0.12. About the other thresholds found in the literature, the results reveal equivocally, values
of CC corresponding to 100% for every case. Sensitivity tests showed that the vertical
position and the thickness of the clouds represent only a source of minor errors lower than
3.2% for UVI calculations. Thus, the cloud OD is the most important input value to be
considered for the TUV RTM. In general, the performance of the TUV MTR in cloudiness
events was satisfactory (p < 0.0001), with strong positive correlations (r > 0.91), high
agreements (d > 0.92) and tendency of overestimation of the model. The maximum UVI
observed was 13.7 in the presence of Altocumulus (Ac) clouds. In case of strong UVI
attenuation caused by overcast skies with Nimbostratus (Ns) clouds, the mean CMF value was
0,2 (± 0,1). On the other hand, UVI increase occurred in Cumulus (Cu) clouds conditions,
with mean CMF of 1,1 (± 0.1). It is worth pointing out that the occurrence of clouds does not
indicate safe sun exposure.