Volume 119, Issue 5 p. 2429-2440
Research Article
Free Access

Observed influence of liquid cloud microphysical properties on ultraviolet surface radiation

D. Mateos

Corresponding Author

D. Mateos

Atmosphere and Energy Laboratory, University of Valladolid, Valladolid, Spain

Correspondence to: D. Mateos,

[email protected]

Search for more papers by this author
G. Pace

G. Pace


Search for more papers by this author
D. Meloni

D. Meloni


Search for more papers by this author
J. Bilbao

J. Bilbao

Atmosphere and Energy Laboratory, University of Valladolid, Valladolid, Spain

Search for more papers by this author
A. di Sarra

A. di Sarra


Search for more papers by this author
A. de Miguel

A. de Miguel

Atmosphere and Energy Laboratory, University of Valladolid, Valladolid, Spain

Search for more papers by this author
G. Casasanta

G. Casasanta

Institute of Atmospheric Sciences and Climate (CNR), Rome, Italy

Search for more papers by this author
Q. Min

Q. Min

Atmospheric Sciences Research Center, State University of New York, Albany, New York, USA

Search for more papers by this author
First published: 27 January 2014
Citations: 11


Measurements of different UV quantities (UV index, ozone photolysis rates, global and diffuse irradiances, and actinic flux spectra) and cloud properties were collected during a field campaign carried out in Southern Italy in May–June 2010. Independent measurements of cloud liquid water path and optical depth allowed retrieving the cloud effective radius. The cloud modification factor (CMF) is used to investigate the influence of liquid cloud properties on the UV radiation under overcast conditions. CMF was also simulated using a 1-D radiative transfer model. Experimental and simulated CMF values for UV index (under overcast conditions) show a normalized root-mean-square error around 11%. Clouds with small effective radius determine a higher UV radiation attenuation than clouds formed by large particles. The CMFs for the UV index and the global spectral irradiance show a very weak dependence on the solar zenith angle (SZA), while the CMFs for actinic flux (both integrated and spectral) and diffuse spectral irradiance show a variation with SZA. The irradiance is more effectively attenuated at low SZA, while the actinic flux at high SZA. These effects are due to the different weight given to the direct and the diffuse components.

Key Points

  • Experimental values of cloud microphysical properties and UV radiative flux
  • Small cloud particles cause stronger attenuation of UV radiative fluxes
  • Measurements and simulations (irradiance and actinic flux) are in good agreement

1 Introduction

UV radiation exerts a significant influence on the biosphere and atmospheric chemistry, and its propagation through the atmosphere is strongly modulated by clouds [e.g., Calbó et al., 2005; Thiel et al., 2008]. Few experimental studies on the cloud effects on UV solar radiation have been carried out so far, due to the lack of simultaneous measurements of UV radiation and cloud properties, which show a large temporal and spatial variability. Only in recent years new experimental and theoretical methods have been applied to this topic.

Most experimental and modeling studies on this topic used the cloud modification factor (CMF) [e.g., Seckmeyer et al., 1996; Kylling et al., 1997; Mayer et al., 1998; Crawford et al., 2003; Bernhard et al., 2004; Mateos et al., 2011; Antón et al., 2012], which is defined as UVcloudy/UVcloud-free, where UVcloudy and UVcloud-free are the UV radiation under cloudy and cloud-free conditions, respectively, for the same atmospheric conditions. The CMF can be evaluated for spectrally dependent or spectrally integrated quantities [Calbó et al., 2005] and for different radiative quantities (irradiance, weighted irradiance, actinic flux, and photolysis rate). Various studies have shown that the CMF for the UV irradiance displays a wavelength dependence, with a higher cloud transmission at 320 than at 400 nm [e.g., Seckmeyer et al., 1996; Kylling et al., 1997; Mayer et al., 1998]. This dependency is attributed to molecular scattering occurring above the cloud layer; the same effect is observed at high values of solar zenith angle (SZA), when the diffuse to direct ratio is large. On the other hand, attenuation at the shorter wavelengths is due to enhanced absorption by tropospheric ozone [e.g., Mayer et al., 1998; Mateos et al., 2011]. Kylling et al. [2005] showed that 1-D radiative transfer simulations for overcast situations can reproduce the overall behavior of airborne actinic flux measurements. They found an enhancement of the actinic flux above the cloud (~60–100% with respect to cloud-free conditions) and a reduction under the clouds (~−55–65% compared with cloudless conditions). Palancar et al. [2011] studied the vertical profile of both downwelling and upwelling actinic fluxes with airborne measurements for different cloudy situations and reported that the perturbations caused by clouds to these two components are not entirely independent. Although many authors investigated the role of cloud cover, very few studies take into account cloud types [e.g., Lubin and Frederick, 1991; Schwander et al., 2002] or other cloud properties [e.g., Mateos et al., 2011; Antón et al., 2012]. This study contributes to the characterization of the UV radiative flux under cloudy conditions by investigating the role of cloud optical and microphysical properties.

The analysis is based on experimental data obtained during a 2 month field campaign which took place in summer 2010 in southern Italy. Due to the high variability of cloud properties and the difficulty in reproducing complex three-dimensional cloud structures in broken conditions with radiative transfer models, the analysis done in this study is limited to overcast conditions. One-dimensional radiative transfer simulations are used to interpret the observations.

Spectra of global and diffuse irradiances, actinic flux, and spectrally integrated irradiances are used, together with radiative transfer simulations, to evaluate the role played by cloud droplet effective radius (reff) and liquid water path (LWP) on the radiative field under overcast conditions. To our knowledge, this is the first study showing the direct impact of these cloud properties on the UV integrated and spectral radiative flux (UV index, ozone photolysis rate, global and diffuse irradiance, and actinic flux) at the surface.

2 Instrumentation and Data

The observations used in this study were collected during a 2 month field campaign carried out in May and June 2010 at the Trisaia Research Centre (40.16°N, 16.64°E, 40 m above sea level (asl)) of ENEA (Italian National Agency for New Technologies, Energy and Sustainable Economic Development), in Southern Italy. The purpose of the campaign was to test and validate an improved version of the Italian National Integrated Model to support the international negotiation on atmospheric pollution (MINNI), which supports the air quality policy at national and regional levels (http://www.minni.org/). In conjunction with the MINNI validation campaign, the Marine Ozone and Radiation Experiment (MORE) airborne campaign took place to study the boundary layer vertical distribution of ozone and aerosol in a rural and coastal area. The Trisaia campaign involved several European Institutions: groups of Casaccia, Bologna, and Trisaia research centers of ENEA, the Atmosphere and Energy Lab of the University of Valladolid, the G-24 group of the University of Rome, the Institute of Meteorology and Climate Research-Atmospheric Environmental Research in Garmisch-Partenkirchen (Germany), the environmental consulting company Arianet in Milan (Italy), and the Atmospheric Modeling and Weather Forecasting Group of the University of Athens (Greece). All instruments used for this study were placed on the roof of a two-floor building and had the horizon free from obstructions.

A Yankee Environmental Systems ultraviolet multifilter rotating shadowband radiometer (UV-MFRSR) was used to measure solar irradiance at the nominal wavelengths of 299.0, 304.7, 310.7, 316.8, 323.7, 331.7, and 367.2 nm (each channel with ~2 nm full width at half maximum bandwidth). UV-MFRSR measures global and diffuse irradiances and derives the direct normal component. This instrument was calibrated by the manufacturer before the campaign. Krotkov et al. [2005] showed an uncertainty on the global signal of ~0.03 mV at high SZA and under high-turbidity conditions.

Downward spectral actinic fluxes were derived from observations of a METCON (Meteorology Consult Inc.) diode array spectrometer, which measures radiation between 280 and 700 nm at 0.83 nm steps. Further details about the instrument and the calibration processes were given by Casasanta et al. [2011]. The ozone photolysis rate, J(O1D), was calculated as
where σλ(T) is the ozone absorption cross section adopted from Daumont et al. [1992] and Malicet et al. [1995] and Φλ(T) is the ozone quantum yield taken from Matsumi et al. [2002]. Both variables present a spectral dependence, which also varies with the temperature. In order to highlight the radiative effects caused by atmospheric constituents (clouds, ozone, and aerosol), J(O1D) was calculated at a constant temperature. In this case, T = 295 K was used.

The UV index (UVI) was measured with a Yankee Environmental Systems UVB-1 radiometer, whose spectral sensitivity resembles the erythemal action spectrum, and with a cosine response better than ±5% for SZA < 60°. The instrument was calibrated at the Physikalisch-Meteorologisches Observatorium Davos World Radiation Center (Switzerland) before the field campaign. UVI was derived by a calibration matrix depending on total ozone column (TOC) and SZA. The expanded uncertainty of this type of radiometers can reach ~7–16% [Hülsen and Gröbner, 2007]. After the field campaign, the measurements of this instrument were compared with those of the double monochromator Brewer MKIII spectrometer installed at the ENEA Station for Climate Observations on the island of Lampedusa [see, e.g., di Sarra et al., 2008]. Differences of erythemal UV values estimated by UVB-1 and Brewer spectrometer are below 5%. More details about the quality control of UVB-1 data were given by, e.g., Bilbao et al. [2011] and Mateos et al. [2012].

A Solar Light Microtops-II manual Sun photometer was used throughout the campaign to measure total ozone column (TOC), precipitable water column, and aerosol optical thickness at 1020 nm. The Sun photometer Microtops-II is equipped with five optical collimators, with a field of view of 2.5°, to perform measurements of direct radiation at the following nominal wavelengths: 305, 312, 320, 936, and 1020 nm. The Microtops-II used in the Trisaia campaign is calibrated every 2 years at the Mauna Loa Observatory (Hawaii) and in the Solar Light Calibration Lab (Glenside, PA). More details were given by de Miguel et al. [2011]. After the field campaign, Microtops measurements were made also at Lampedusa during 5 days. More than 50 individual measurements at different SZAs were made and compared to total ozone data obtained by the Brewer spectrometer. The observed differences between the TOC estimations of the two instruments were below 2%. During the Trisaia campaign, whenever Microtops TOC measurements were not available because of cloudy conditions, data from the Ozone Monitoring Instrument (OMI, data available from http://macuv.gsfc.nasa.gov site) were used. TOC daily values derived by Microtops-II and OMI observations under cloud-free conditions during the Trisaia campaign were in good agreement showing a normalized root-mean-square error (RMSE) of 2.2% and a linear correlation coefficient of 0.95.

A MFRSR in the visible range (VIS-MFRSR: six channels at 415, 500, 615, 673, 870, and 940 nm nominal wavelengths, each with 10 nm full width at half maximum bandwidths, plus one photodiode covering the 300–1040 nm spectral range) provided measurements of the aerosol optical properties; the adopted retrieval methodology was described by Pace et al. [2006]. Hourly aerosol data obtained from this instrument showed a mean aerosol optical depth at 500 nm of 0.2 ± 0.11, with minimum and maximum hourly values of 0.06 and 0.6, respectively.

Continuous measurements of absolute humidity and temperature vertical profiles, integrated water vapor, and liquid water path were estimated by a Radiometer Physics GmbH (RPG) HATPRO (Humidity and Temperature Profile) microwave radiometer. This instrument measures the brightness temperature in the K and V bands sampling the water vapor absorption lines and oxygen complex, respectively. A statistical inversion methodology, similar to that described in Löhnert and Crewell [2003], was used to derive integrated water vapor and liquid water path as well as temperature and specific humidity profiles. Vertical profiles of specific humidity and temperature were calculated from quadratic and linear regressions, respectively. The coefficients used for the regressions were provided by RPG and derived by a group of radio-sounding data from the Mediterranean sites of Trapani and Cagliari. The estimated uncertainty on LWP was ~30 g m−2 [e.g., Löhnert and Crewell, 2003; Rose et al., 2005]. A detailed description of this instrument design was given by Rose et al. [2005]. Using LWP measurements and transmissivity at 415 nm obtained from a VIS-MFRSR, Min and Harrison [1996] developed a methodology to estimate the cloud optical thickness (COT) and effective radius by means of look-up tables (LUTs) obtained from radiative transfer simulations. Min et al. [2003] showed that the reff values retrieved with this method agree with in situ measurements to within 5.5%. The cloud base height was derived from measurements of a zenith-looking Heitronics KT 15.99 infrared pyrometer operating in the 9.6–11.5 µm band, installed on one side of the HATPRO radiometer.

3 Methodology

All radiometric data were acquired at 1 min time resolution. The reff retrieval from the HATPRO and VIS-MFRSR instruments requires a longer integration time. This is due to (a) the different field of view of the two instruments (the HATPRO looks at the zenith with a half-power beam width of 3.5°, while the VIS-MFRSR has a field of view of ~160°) and (b) the assumption of homogeneous and uniform clouds used in the retrieval. Thus, in some cases, although only overcast conditions were selected, the cloud portion viewed by the HATPRO instrument may have different properties with respect to whole cloud layer observed by the VIS-MFRSR. Min and Harrison [1996] derived retrievals of the effective radius using 5 and 30 min averaged data. In these periods, the cloud properties are assumed as stationary properties. In this study, the 5 min integration data were used to analyze the influence of cloud properties on cloud UV transmittance. Five minute averages of the cloud optical properties were already used by Min et al. [2003] and Nzeffe et al. [2008]. This time interval allows cloud variations to be monitored on a cloud lifetime scale.

To determine the cloud radiative fraction, two different automatic algorithms by Long et al. [2006] and Min et al. [2008] were used. The first one is based on measurements of global and diffuse shortwave radiation, while the latter uses the transmittance ratio at two different wavelengths (from VIS-MFRSR measurements). During the campaign, visual observations of cloud cover were also carried out every hour. These visual inspections were used to verify the accuracy of the two algorithms, both providing reliable result for the overcast conditions occurred during the campaign. For the purpose of this study only time intervals showing a radiative cloud fraction between 0.95 and 1.0 by both algorithms were selected as overcast conditions. Overcast conditions were frequently observed during the Trisaia campaign. During the whole campaign, 154.3 h were classified as cloud free (occurred in 37 days), while 119.4 h presented overcast conditions (occurred in 34 days). Most of the overcast conditions had a duration shorter than 2 h.

The number and frequency of occurrence of overcast cases during the Trisaia field campaign are presented in Table 1 for five cloud-based ranges, with the average and one standard deviation of COT, LWP, and reff values. The largest amount of cases under overcast conditions occurred for cloud base height between 2000 and 4000 m, while low clouds (cloud base up to 2000 m) exhibited the largest values of COT, LWP, and reff.

Table 1. Number and Frequency of Occurrence of Overcast Cases During the Trisaia Field Campaign Presented for Five Cloud-Based Rangesa
Cloud Base Interval Number of Data COT LWP reff
0–1000 m 22 (1.7%) 79.5 (37.7) 1.1 (1.0) 15.4 (9.2)
1000–2000 m 175 (13.6%) 55.4 (36.7) 0.5 (0.5) 11.6 (6.9)
2000–3000 m 427 (33.3%) 32.7 (18.1) 0.3 (0.3) 9.3 (5.4)
3000–4000 m 508 (39.6%) 23.8 (11.8) 0.09 (0.07) 6.5 (3.8)
4000–5000 m 150 (11.7%) 13.5 (5.9) 0.08 (0.06) 8.1 (5.8)
  • a Average and standard deviation values of COT, LWP, and reff are also shown.

The variable used to evaluate the cloud effects on the radiative field was the CMF. Hence, the estimation of the UV irradiance and actinic fluxes under cloud-free conditions was needed to calculate CMF. The cloud-free periods were identified by means of a modified version of the algorithm proposed by Long and Ackerman [2000], as described by Meloni et al. [2007]. Cloud-free UV radiation data were classified depending on SZA (between 15° and 80° at 1° steps) and TOC (between 280 and 400 Dobson units (DU), at 20 DU steps). UV data falling in each class of SZA and TOC were averaged. Five look-up tables of cloud-free UV radiation data, depending on SZA and TOC, are built, one for each variable: UV index, J(O1D), global and diffuse spectral irradiance, and spectral actinic flux. The empirical LUTs were compared with radiative transfer simulations (see the following paragraph) showing a normalized value of RMSE below 10%. The CMF is calculated using 5 min overcast UV data and the respective cloud-free LUT values for the corresponding closest values of SZA and TOC. The experimental CMF estimation minimizes the UV measurement uncertainties, affecting both the cloud-free and cloudy observations.

Radiative transfer simulations were carried out by using the library for radiative transfer (libRadtran) package [Mayer and Kylling, 2005]. The midlatitude standard atmosphere vertical profiles were used in the simulation. The aerosol optical properties were fixed to the average values occurring in the UV range during the field campaign. Taking into account rural, maritime, and desert dust aerosols, a single-scattering albedo of 0.85 and a asymmetry factor of 0.73, both wavelength independent, were used for an aerosol layer between 0 and 2 km. This latter assumption about the single-scattering albedo (SSA) can be considered as a good approximation when the global component of solar irradiance is analyzed [Román et al., 2013]. The surface wavelength-dependent albedo was estimated assuming a 50% grass and 50% sea water surface. The spectral dependencies of ocean and grass surfaces in the UV were taken from Doda and Green [1980] and Feister and Grewe [1995], respectively. For instance, at 280 and 400 nm, we estimated an albedo of 0.048 and 0.041, respectively. The model was run in 32 streams, with the discrete-ordinate method for radiative transfer DISORT2 [Stamnes et al., 2000] with a resolution of 0.5 nm for the wavelength range from 295 to 400 nm. To simulate the overcast conditions, a homogeneous cloud layer between 2 and 3 km asl was assumed [Mateos et al., 2011]. The parameterization of Hu and Stamnes [1993] was used to relate LWP and reff with optical properties. The simulations (under cloud-free and overcast conditions) were performed for several values of SZA, TOC, LWP, and reff. In this way, cloud-free and overcast simulations of the radiative fluxes were used to obtain the modeled CMF for the UV index, ozone photolysis rate, spectral irradiance (global and diffuse), and spectral actinic flux.

4 Broadband UV Radiative Fluxes

In this section, the effect of clouds on the spectrally integrated UV radiative fluxes (both irradiance and actinic flux) is considered.

The CMF for UVI (CMFUVI) was analyzed as a function of LWP and reff. Figure 1 shows the results obtained from measurements and simulations. Three intervals of particle radii were used in the calculations: 2.5–5 µm, 5–10 µm, and 10–15 µm. To exclude ice clouds from the analysis, only cases with cloud base (hcld) below 5000 m were selected. With this threshold, this analysis was only referred to liquid water clouds. As stratus or cumulus present similar values of the cloud asymmetry factor [Hess et al., 1998], effects of different cloud water types were minimized in this study.

Details are in the caption following the image
CMFUVI as a function of LWP for different cloud droplet radii. Symbols are the experimental data (different symbols correspond to different days), and curves are the radiative transfer simulations at SZA = 18° (solid lines) and SZA = 75° (dashed lines) with TOC = 300 and AOT550nm = 0.25.

The model calculations are shown for SZA = 18° and SZA = 75°, the smallest and the highest SZA occurring for overcast cases. As can be seen in the figure, the simulated values well reproduce the measurements, with a clear separation among the radius sizes. The three classes of radii include 16, 20, and 16 days of data, respectively. The model curves indicate that there is weak dependence on SZA. Both observations and simulations show that for a fixed LWP smaller size particles produce smaller CMFUVI values, i.e., the scattering phenomena occurring in the cloud are more effective for smaller radii. This effect is due to the relationship linking the cloud optical thickness to LWP and reff. For a vertically uniform cloud, COT can be expressed as (3 LWP)/(2 ρw reff) [Stephens, 1978], ρw being the water density. Smaller particles correspond to a higher COT for a fixed LWP. Using measurements and theoretical methods, Kim et al. [2003] obtained a linear relationship between COT and LWP, which is inversely proportional to reff.

Clouds with values of reff lower than 10 µm show LWP values lower than 0.4 kg m−2 that corresponds to a threshold value for nonprecipitating clouds [Löhnert and Crewell, 2003]. Larger values of LWP are observed for cloud droplet radii between 10 and 15 µm; in these cases neither drizzle nor rain reached the surface.

Crawford et al. [2003] observed a strong dependence of cloud transmittance on SZA for actinic flux. Hence, three intervals of SZA, each 12° wide were selected for CMFJ(O1D): (18°, 30°), (42°, 54°), and (63°, 75°). Figure 2 shows the obtained results. The simulations are plotted at 18°, 30°, 42°, 54°, 63°, and 75°. The overall agreement between measurements and simulations is also quite good. The number of days used for this analysis is shown in Table 2; the maximum number of days occurs for 5 < reff < 10 µm and moderate SZAs.

Details are in the caption following the image
CMFJ(O1D) as a function of LWP for different cloud droplet radii for three SZA intervals. Symbols are the experimental data (different symbols correspond to different days). In each plot, solid curves refer to results of simulations at the smallest SZA in the range (i.e., (a) 18°, (b) 42°, and (c) 63°), and dashed curves refer to simulations at the largest SZA in the range, with TOC = 300 and AOT550nm = 0.25.
Table 2. Number of Different Days (in Brackets the Total Number of Data) Used in Figure 2 With Respect to Intervals of SZA and Effective Radius
18° < SZA < 30° 42° < SZA < 54° 63° < SZA < 75°
2.5 < reff < 5 µm 5 (37) 8 (17) 3 (5)
5 < reff < 10 µm 9 (30) 11 (46) 7 (30)
10 < reff < 15 µm 7 (12) 7 (15) 4 (5)

As for CMFUVI, smaller particles cause stronger attenuation. CMFJ(O1D)values decreases with SZA for each interval of radii. The dependence on SZA is stronger for CMFJ(O1D)than for CMFUVI, which is in agreement with model simulations. CMFJ(O1D) decreases with SZA (Figures 2a and 2b), with the exception of high SZAs when there is a slight increase from 63° to 75° (Figure 2c). Thus, contrarily to CMFUVI, CMFJ(O1D) also depends on SZA.

In order to analyze the different behavior of CMFUVI and CMFJ(O1D), Figure 3 shows the simulations for radii of 2.5 and 10 µm at 18° and 75° SZA. As in Figure 1, CMFUVI exhibits weak dependence on SZA and no significant differences in the curve shape. For instance, the values of CMFUVI in Figure 3a at the lowest LWP value (0.01 kg m−2) decrease by 0.04 units between SZA = 18° and SZA = 75°. Conversely, CMFJ(O1D) shows notable dependence on SZA at fixed LWP values and changes in the curvature for varying LWP. For small particles (reff = 2.5 µm, Figure 3a), CMFUVI is smaller than CMFJ(O1D) at SZA = 18°, while the CMFUVI values are higher than the CMFJ(O1D) ones for all the LWP values at large SZA (SZA = 75°). The same behavior is observed for larger particles (reff = 10 µm, Figure 3b). The physical reason for this different behavior lies on the different role played by the direct and diffuse components at each SZA.

Details are in the caption following the image
Dependence of CMFUVI (purple curves) and CMFJ(O1D) (red curves) on LWP at 18° (solid curves), and 75° (dashed curves) of SZA for cloud droplet radii of (a) 2.5 µm and (b) 10 µm with TOC = 300 and AOT550nm = 0.25.

To clarify this point, simulations of the direct and diffuse components of UVI and J(O1D) for cloud-free conditions were analyzed. The simulations were carried out with the model setup described in section 2, with TOC = 300 DU and aerosol optical thickness at 550nm (AOT550nm) of 0.25. Figure 4 shows the ratio between the diffuse and global components versus SZA. The direct to global ratio can be evaluated as 1–diffuse/global. At low SZA, the direct component of the UVI is almost half of the global, while this fraction decreases to 35% for J(O1D). Therefore, the presence of clouds blocking the direct component affects the UVI much more effectively, and the CMFUVI values are smaller than the CMFJ(O1D) at these solar zenith angles. The J(O1D) and UVI diffuse/global ratio increases by ~43% and ~81%, respectively, for SZA varying from 18° and 75° SZA. The different definition of irradiance and actinic flux implies that at large SZA CMFUVI is higher than CMFJ(O1D). At low SZAs, conversely, UVI is attenuated more effectively by clouds than J(O1D). This behavior does not depend on cloud droplet radius.

Details are in the caption following the image
Dependence on SZA of the ratio between diffuse and global components for UVI (purple curve) and J(O1D) (red curve) calculated with the libRadtran model for cloud-free conditions.

Due to the small dependence of CMFUVI on SZA values, we decided to quantify the agreement between measurements and simulations only for this variable. In this way, we avoided the use of another classification as a function of the SZA. Therefore, we selected measured CMFUVI data presenting a reff between 5 and 10 µm at 0.5 µm steps to perform the validation. For this subset of data, the comparison between experimental and simulated CMFUVI was carried out for 169 LWP values showing a normalized RMSE of 11.5%. The linear correlation coefficient between the two series was 0.96, while the slope and the intercept of the linear fit were 0.88 and 0.04, respectively.

As the model reproduces the experimental data in a reliable way, we used the simulations to test the impact of aerosol and cloud properties. The analysis was performed only for CMFUVI data. The role of the AOT550nm, aerosol single-scattering albedo (SSA), and cloud asymmetry factor (gC) was evaluated by comparing the simulated CMFUVI values for two different values of these parameters. Realistic values of SSA for urban and maritime aerosols and of gC for water clouds were selected from Hess et al. [1998], and they are reported in Table 3. The CMF ratio defined as CMF(case1)/CMF(case2) was calculated for the AOT550nm, SSA, and gC. In the analysis of one particular variable, the other two were kept fixed to the reference value reported in Table 3. The SZA is 18.5°, while the other inputs are as described in section 3. The dependence of the CMF ratios on LWP is shown in Figure 5 as a function of reff. The length of the curves for each reff class ends at the maximum LWP value measured for that class. The mean LWP value for each reff class is shown as a symbol along the corresponding curve. Larger AOT550nmvalues produce lower CMFUVI values. The CMFAOT1/CMFAOT2 ratios show similar behaviors for each reff class (Figure 5a), with variations of 10% and 5% corresponding to maximum and average observed LWP values, respectively. The simulations show that absorbing aerosol (SSA1) determines lower values of CMFUVI; CMFSSA1/CMFSSA2 ratio variations of 5% occur for the maximum LWP value observed for each reff class (Figure 5a). It is worth mentioning here that SSA was found to be one of the key parameters in the retrievals of UV radiation at the surface from satellite observations [e.g., Li et al., 2000; Wang et al., 2000]. With respect to the cloud asymmetry factor for liquid clouds, stronger forward scattering (gC2) determines higher values of CMFUVI; the gC change between 0.85 and 0.867 causes variations of ~11% and 7%, respectively, for the maximum and average LWP values (Figure 5b). The aerosol asymmetry factor effect was also analyzed; it presented a very weak influence on CMFUVI evaluations, less than 0.5%, and the results are not shown in this study for brevity. These simulations demonstrate the weak CMFUVI sensitivity to the selected aerosol and liquid cloud parameters strengthening the reliability of the results presented in this study.

Table 3. Values of the Aerosol Optical Thickness at 550 nm (AOT550nm), Aerosol Single-Scattering Albedo (SSA), and Cloud Asymmetry Factor (gC) for the Reference and the Two Cases Considered in the Sensitivity Analysis
AOT550nm SSA gC
Case 1 0.1 0.81 0.85
Case 2 0.5 0.997 0.867
Ref. 0.25 0.90 0.86
Details are in the caption following the image
Impact of the AOT550nm, SSA, and gC values on the evaluation of CMFUVI. Cases 1 and 2 correspond to those shown in Table 3. The length of the curves is determined by the maximum of the LWP experimental data (see Figure 1), while the symbols point out the measured average value of LWP for each size interval.

5 Spectral UV Radiative Fluxes

5.1 Global and Diffuse Irradiances

The spectral irradiance measured by the UV-MFRSR was used to study the dependence of the global and diffuse irradiances on cloud microphysical properties. Two wavelengths were selected: one in the UV-B range (304.7 nm) and the other in the UV-A (367.2 nm). Following the classification mentioned in section 4, three groups of particle radii were considered. Due to the strong dependence of the diffuse CMF (CMFI-DIFFUSE) on SZA, only the interval 18°–30° was considered in this section. Similar results were found for SZA intervals of 42°–54° and 63°–75° that are not shown for brevity. Figure 6 shows the results obtained for the interval 18° < SZA < 30°. The model simulations are plotted for the smallest and highest SZAs, i.e., SZA = 18° and SZA = 30°. The agreement between measurements and simulations is quite good. For a given LWP, smaller particles produce stronger attenuation of both the global and the diffuse spectral irradiance, as also found for UVI. The CMF of the global component (CMFI-GLOBAL) does not display significant differences between the two wavelengths. However, when the diffuse component is investigated, the CMFI-DIFFUSE values at 367.2 nm are always higher than those at 304.7 nm. Moreover, a particular case occurs for very low LWP in the CMFI-DIFFUSE case: when the LWP is 0.01 kg m−2, the curves of reff = 10 µm are above those of reff = 15 µm. This effect is more evident at 367.2 nm. This behavior at low LWP suggested by the model results cannot be verified against observations because the HATPRO microwave radiometer is not sensitive to very low values of LWP. In addition, a case of homogeneous cloud with very low LWP and high reff (corresponding to COT around 1) is not easily observable.

Details are in the caption following the image
CMF for the global and diffuse spectral irradiances at two wavelengths as a function of LWP for different cloud droplet radii in the SZA interval 18°–30°. Dots are the experimental data, and curves are the radiative transfer simulations at SZA = 18° (solid curves) and SZA = 30° (dashed curves) with TOC = 300 and AOT550nm = 0.25.

CMFI-GLOBAL shows a small decrease with wavelength between 304.7 and 367.2 nm in the 18°–30° SZA interval. The observed stronger attenuation at longer wavelengths confirms previous results of CMFI-GLOBAL as a function of the cloud optical thickness [e.g., Seckmeyer et al., 1996; Bernhard et al., 2004; Mateos et al., 2011].

5.2 Actinic Flux

A similar analysis is carried out for the actinic flux. Figure 7 shows the dependence of actinic flux CMF (CMFF) on LWP for the same three groups of reff used in the previous figures. The wavelengths shown in the figure were chosen in order to match those used in the previous section. Due to the large dependence of CMFF on SZA, three intervals of SZA were shown: 18°–30°, 42°–54°, and 63°–75°. In general, CMFF decreases with SZA at 305.4 and 367.51 nm for all values of the radii. At low SZAs (SZA = 18°) there is a small increase in the CMFF from 305.4 to 367.51 nm; at intermediate SZA (SZA = 42°) no spectral dependence is evident, while at the highest SZA (SZA = 63°) a strong decrease occurs in the CMFF for all radii. Since in the actinic flux the radiation coming from different directions is uniformly weighted; CMFF is much more SZA dependent than CMFI-GLOBAL.

Details are in the caption following the image
CMF for the actinic flux at two wavelengths as a function of LWP for different cloud droplet radii in three SZA intervals. Dots are the experimental data, and curves are the radiative transfer simulations at the smallest (solid curves) and highest (dashed curves) SZAs in each panel with TOC = 300 and AOT550nm = 0.25.

6 Conclusions

Measurements of global and diffuse spectral irradiances, spectral actinic flux, UV index, ozone photolysis rates, and cloud properties, acquired during a 2 month field campaign in Southern Italy, were used to investigate the influence of liquid cloud on the UV radiation under overcast conditions.

The influence of the cloud liquid water path (LWP) and effective radius (reff) on broadband and spectral irradiances and actinic flux was studied for different solar zenith angles (SZA) in terms of cloud modification factor (CMF). Radiative transfer model calculations nicely agree with the measurements and were used to interpret the observations.

The main results can be summarized as follows:
  1. For fixed values of LWP, cloud droplets of small reff cause stronger attenuation of UV radiation than larger ones. This result is valid for all the UV quantities accounted.
  2. The CMFs for UVI and global spectral irradiance display very weak dependence on SZA, while the CMFs for actinic flux (both integrated and spectral) and diffuse spectral irradiance decrease with the increase of SZA.
  3. The CMFs for the global irradiance and for the actinic flux show different behaviors: irradiance is more efficiently attenuated at low SZAs than actinic flux, while the opposite occurs at high SZAs. These effects are due to the different ways in which the direct and diffuse components contribute to the irradiance and the actinic flux.
  4. At all SZAs the spectral global irradiance CMF does not appreciably change from 304.7 to 367.2 nm, while the diffuse irradiance CMF increases with wavelength. The actinic flux between 305.4 and 367.51 nm increases at low SZA remains constant at moderate SZA and decreases at high SZA.
  5. Measured and simulated UV index CMFs (CMFUVI) present a normalized root-mean-square error of ~11%. Sensitivity tests to study the CMFUVI variations for different values of aerosol optical depth and single-scattering albedo as well as of cloud asymmetry factor demonstrate the small impact on these parameters on the CMFUVI.


Measurements at Trisaia site were supported by the Italian Ministry for Environment through the MINNI project and by EUFAR through the MORE project. The authors gratefully acknowledge the financial support extended by the Spanish Government under the projects CGL2010-12140E and CGL2011-25363. David Mateos would like to thank the University of Valladolid for the PhD financial support (PIF-UVa grant).