Clouds Affecting Total Ozone Column Measurements

Abstract

Total ozone column (TOC) measurements have been made remotely all over the globe using ground-based and satellite probing. Other atmospheric constituents such as clouds and aerosols can interfere significantly in such measurements. In this work, the influence of clouds on TOC ground-based measurements in a tropical site is investigated using 16 months of spectral direct Sun measurements from an ultraviolet narrow-band radiometer. The results show that, on average, TOC measurements through clouds blocking the Sun (TOC_C) can be 0.9% (ranging from −4.6 to 3.4%) larger than TOC measurements under cloud-free skies (TOC_CFS), and TOC measurements under all-sky conditions without clouds blocking the Sun can be 0.3% (ranging from −5.4 to 2.2%) larger than TOC_CFS. In addition, the variability rate (Dobson units/min) in TOC_C can be over five times that for TOC_CFS on average.

1 Introduction

Solar radiation interacts with the Earth's atmosphere through scattering and absorption. Ultraviolet (UV) radiation is of particular concern here. Ozone molecules in the stratosphere (the ozone layer) are formed by solar UV wavelengths < 280 nm. The ozone layer absorbs harmful longer UV wavelengths (< 315 nm), reducing their incidence on the ground (Lautenschlager et al., 2007; Whitten & Prasad, 1985).

Ozone is a trace gas in the terrestrial atmosphere, and its atmospheric columnar amount is given by the total ozone column (TOC), which has been measured remotely using ground-based and satellite probing (Whitten & Prasad, 1985). The most precise ground-based instruments produce TOC measurement with uncertainties < 2%, while similar uncertainties are found from satellite probing (McPeters et al., 2008; Silva & Tomaz, 2012; Silva, 2007, 2013a). Although satellites can fly over the whole Earth every day, their data are validated using ground-based equipment. Naturally, owing to its fundamental importance to both the biosphere and the atmosphere, ozone remains a matter of utmost interest (Bais et al., 2015).

Clouds and aerosols are other important constituents in the atmosphere. Clouds are the most important atmospheric modulator of ground incidence of solar radiation (Calbó et al., 2005; Silva, 2011). In addition, they interfere in TOC measurements. In general, clouds can be described as a cluster of droplets formed by water vapour condensed upon a tiny aerosol particle called the condensation nucleus. However, there is no broadly accepted consensus on what the exact number of such droplets gathered together should be to be considered a cloud, especially because an aerosol plume, although not usually composed of spherical particles like cloud droplets, can fit such a description perfectly (Chiu et al., 2009; Koren et al., 2007). Furthermore, cloud detection also depends on the radiation wavelength used.

The most well-known and perhaps frequent effect of clouds on solar radiation is attenuation (Calbó et al., 2005). However, some cloud genera are also closely related to the so-called radiation enhancement event (REE), in which there is a sudden increase in the flux of radiation towards the ground from radiation scattered from the thinner parts of clouds (Parisi & Dows, 2004; Silva et al., 2019; Silva, 2013b). Fundamentally, there exist three main mechanisms capable of yielding REEs:

• i. The Sun disk is not obscured by a cloud while its diffuse component is increased through forward scattering on the sides of vertically developed clouds, enhancing the global radiation on the ground.

• ii. The Sun disk is obscured by a thin-layered cloud or by the borders or thin parts of clouds, reducing the Sun’s direct beam, but the diffuse component is increased through forward scattering that more than compensates the reduction in the direct component, resulting in the enhancement of the global radiation on the ground.

• iii. Snow reflects the solar radiation to the atmosphere that (including clouds and aerosols) reflects the radiation back, enhancing the downwelling component of global radiation. Similarly, an extensive cloud layer reproduces this effect in relation to mountain peaks above such a layer.

Items (i) and (ii) are feasible for the tropical site considered here.

In this study, 16 months of ground-based TOC measurements from direct Sun (DS) spectral measurements using an ultraviolet narrow-band radiometer are presented for an urban, tropical site in Brazil. Sky images are gathered simultaneously by an imager, providing images of the sky during measurements. The effects from clouds on such measurements are determined.

2 Site, Instruments, and Methods

2.1 Site Description

Belo Horizonte (BH, 19.92° S, 43.94° W, 858 m a.s.l., 331 km², 7167 inhabitants/km² in 2010, Brazil) (IBGE, 2021) is the main city in the third largest Brazilian metropolitan area. As a municipality, it has over two million vehicles outfitted with catalytic converter devices and electronic injection engines. The predominant economic activities in the area are trade, industry, and mining. Settled in a hilly area of southeastern Brazil, the city features a tropical climate of altitude (milder weather due to the area’s altitude) between the savanna (Cerrado) and the remains of the Atlantic Forest (Mata Atlântica). A dry period develops from May to September, and a rainy one from November to March. Large amounts of smoke from biomass burning are blown into the area during dry periods, reinforcing the trend towards a dry scenario, where occasionally humidity can drop to around 10% during winter (June, July, and August) and early spring (September, October, and November).

2.2 Instruments

2.2.1 The Ultraviolet Rotating Shadowband Radiometer 4

The Ultraviolet Rotating Shadowband Radiometer 4 (UVMFR4, Yankee Environmental Systems, Inc., Turners Falls, MA, USA) (Harrison et al., 1994) is the ultraviolet narrow-band radiometer used in this work. It measures spectral irradiance of global (UV_G) and diffuse (UV_Dif) UV radiation at centre wavelengths of 300, 305.3, 311.5, and 317.5 nm with bandwidth of 2.3–2.4 nm full width at half maximum (FWHM) through filters. The direct component of the spectral solar UV radiation (UV_Dir, which can also be represented by DS) is calculated through

$${\text{UV}}_{{{\text{Dir}}}} = \frac{{{\text{UV}}_{{\text{G}}} - {\text{UV}}_{{{\text{Dif}}}} }}{{\cos \left( {{\text{SZA}}} \right)}},$$

(1)

where SZA stands for solar zenith angle, and the division by the cosine reminds us that the radiometer's measurement is the cosine projection of the direct component of the solar radiation through the SZA, as the radiometer is in the flat horizontal position.

The instrument uses a semicircular metallic strip to obstruct the direct incidence of solar radiation in its sensor. For this, firstly, the strip is set in the home position by a stepper motor, allowing the measurement of UV_G. Secondly, the strip is rotated to a central position, blocking the sensor from UV_Dir and allowing the measurement of UV_Dif. As a result, ≈ 3.3° of the sky is obstructed by the strip in addition to the solar disk. To compensate for the radiation lost from that part of the sky, another two measurements are made with the strip rotated 9º to the left and to the right off the central position. The difference between UV_G and the average between the two displaced-strip measurements represents the amount of diffuse radiation blocked from the sky by the strip. That amount is added to the original measurement of UV_Dif as a correction. The cycle of four measurements (UV_G, UV_Dif, and the two displaced-strip measurements) takes 20 s. Therefore, the minute-rated measurements of UV_G, UV_Dif, and their corresponding UV_Dir average on three cycles of measurements per minute. The UVMFR4 must be set up aligned with the meridian, where the accuracy in the positioning of the semicircular metallic strip is within 0.4º. The software controlling the instrument calculates both SZA and azimuth angle to set the strip correctly.

The instrument’s sensor, radiation filters, and the associated electronic circuit are housed and sealed against moisture and rain (the moisture seal of the stepper motor can fail within a few months). The instrument works with internal temperature controlled around 41–42 ºC for outdoor temperatures from −30 to +50 ºC. Cosine correction is automatic for UV_Dir with accuracy of 1% up to SZA = 80º, but no correction is attempted for UV_G or UV_Dif. Typical uncertainties in these measurements are around 6–8% (Hülsen & Gröbner, 2007). Annual calibration at appropriate facilities is recommended to maintain such standards.

2.2.2 The Total Sky Imager

The Total Sky Imager 440A (TSI, Yankee Environmental Systems, Inc., Turners Falls, MA, USA) measures cloud cover (cloudiness). It is composed of a digital camera facing down a rotating spherical mirror to photograph the reflected sky. A black strip is attached to the mirror in order to avoid direct reflection of sunbeams into the camera. Each snapshot results in a 24-bit color Joint Photographic Experts Group (JPEG) image with resolution of 352 × 288 pixels (101 kpixels). The software provided by the manufacturer and set up in a personal computer controls the TSI and processes captured images for cloudiness determination. Some of the software parameters must be adjusted by the operator in order to produce the best match between the sky observed by a trained operator and that determined by the equipment. Uncertainties in cloud cover measurements are greater than 10% in general (Silva & Souza-Echer, 2013).

2.2.3 The Ozone Monitoring Instrument (OMI)

The Ozone Monitoring Instrument (OMI) flies on board the Aura satellite (Levelt et al., 2006; Taskanen et al., 2006). Launched in 2004 to orbit the Earth at Sun-synchronous polar orbits of 705 km, 98° inclination, and 13:45 ± 15 min equator-crossing time, the OMI has a two-dimensional charged coupled device (CCD) to measure UV/visible radiation (270–500 nm at 0.5 nm spectral resolution) from both DS radiation and radiation reflected from the Earth in swathes of 2600 km (115°) perpendicular to the satellite's flight direction. A pixel of the two-dimensional CCD corresponds to a ground grid whose size depends on the OMI cross track position (CTP). This parameter increases with the swath-angle from CTP = 0 for the null swath-angle of nadir (exact overpass measurements) and ground-grid of 13 km × 24 km (along × cross track, 312 km²) to CTP = 59 for the swath-angle of 57° and ground-grid of 13 km × 150 km (1950 km²). Since local boundary layer parameters such as site altitude, clouds, aerosols, and surface albedo can affect OMI measurements significantly, another important OMI parameter is the distance (Dist) between the coordinates of the site and the CTP ground-grid.

Differences between TOC measurements from ground-based instruments and OMI are around 2% (McPeters et al., 2008; Silva & Tomaz, 2012).

2.3 Methods

Spectral DS measurements were conducted in BH at the Laboratório de Luz Ultravioleta (LLUV, 19.92° S, 43.99° W, 910 m a.s.l.) from June 2009 (JUN09) to September 2010 (SEP10) (16 months)—when the noon solar zenith angle (SZA_n) ranges from 0° in summer (December, January, and February) to 43.4° in winter—using the calibrated radiometer UVMFR4 #566. The TSI #157 was set up beside the radiometer, also providing one sky image per minute. Measurements were conducted for air mass m ≤ 2 (SZA < ≈ 60°).

Each of the four DS measurement wavelengths (λ₁ = 300 nm, λ₂ = 305.3 nm, λ₃ = 311.5 nm, λ₄ = 317.5 nm) can be applied to the fundamental atmospheric attenuation expression (or the Beer–Bouguer–Lambert law) (Fleagle & Businger, 1980; McCartney, 1976)

$$I_{{\lambda_{i} }} = I_{{o\lambda_{i} }} e^{{ - \left( {m_{R} \tau_{{R\lambda_{i} }} + m_{{O_{3} }} \tau_{{O_{3} \lambda_{i} }} + m_{p} \tau_{{p\lambda_{i} }} } \right)}} ,i \, = {1},{ 2},{ 3},{\text{ and 4}},$$

(2)

where $I_{{\lambda_{i} }}$ = spectral DS radiation in the atmosphere at λ_i; $I_{{o\lambda_{i} }}$ = spectral DS radiation at the top of the atmosphere at λ_i; $m_{{\text{R}}} = \sec \left( {{\text{SZA}}} \right)\left( {1 - 0.0012\left( {\sec^{2} \left( {{\text{SZA}}} \right) - 1} \right)} \right)$ is the relative optical air mass of the atmosphere for SZA < ≈ 75° (air mass < ≈ 4) (Young, 1994); $m_{{O_{3} }} = \frac{R + h}{{\sqrt {\left( {R + h} \right)^{2} - \left( {R + r} \right)^{2} \sin^{2} \left( {{\text{SZA}}} \right)} }}$ is the relative optical air mass for ozone (meaning the ratio between the actual path of the solar radiation and the vertical line through the ozone layer) (Gao et al., 2001) with R = 6371.229 km as the mean radius of the Earth, r = site altitude (in km), and h = 27.5 km as the ozone layer height for tropical sites (Kirchhoff & Guarnieri, 2002; NASA, 1976);

m_p = the relative air mass for particles (aerosols and cloud droplets) in the atmosphere that is considered equal to m_R for air mass ≤ 2; $\tau_{{R\lambda_{i} }} = 0.008569\lambda_{i}^{ - 4} \left( {1 + 0.0113\lambda_{i}^{ - 2} + 0.00013\lambda_{i}^{ - 4} } \right)\frac{P}{{P_{o} }}$ is the atmospheric optical depth for air molecules, where λ_i is in μm, P_o = 101,325 Pa is the sea-level atmospheric pressure, and P is the atmospheric pressure at the site (Teillet, 1990);

$\tau_{{O_{3} \lambda_{i} }} = \left( {C_{o} \left( {\lambda_{i} } \right) + C_{1} \left( {\lambda_{i} } \right)\left( {T_{{\text{e}}} - 273.15} \right) + C_{2} \left( {\lambda_{i} } \right)\left( {T_{{\text{e}}} - 273.15} \right)^{2} } \right){\text{TOC}}$ is the ozone optical depth, being TOC in cm, C_o(λ_i) (in cm⁻¹) the ozone absorption coefficient at normal temperature and pressure (NTP, 293.15 K and 1 atm), C₁(λ_i) (in cm⁻¹ K⁻¹) and C₂(λ_i) (in cm⁻¹ K⁻²) the ozone absorption coefficients for temperature correction as a function of the effective atmospheric temperature T_e (in K), which equals 229.6 K for the tropical atmosphere (Cervino et al., 1995); and $\tau_{{p\lambda_{i} }}$ = the particle optical depth corresponding to aerosols and cloud droplets in the atmosphere at λ_i.

Transforming Eq. (2) into

$$\ln \left( {I_{{\lambda_{1} }} } \right) = \ln \left( {I_{{o\lambda_{1} }} } \right) - m_{R} \tau_{{R\lambda_{1} }} - m_{{O_{3} }} \tau_{{O_{3} \lambda_{1} }} - m_{p} \tau_{{p\lambda_{1} }} ,$$

(3)

$$\ln \left( {I_{{\lambda_{2} }} } \right) = \ln \left( {I_{{o\lambda_{2} }} } \right) - m_{R} \tau_{{R\lambda_{2} }} - m_{{O_{3} }} \tau_{{O_{3} \lambda_{2} }} - m_{p} \tau_{{p\lambda_{2} }} ,$$

(4)

$$\ln \left( {I_{{\lambda_{3} }} } \right) = \ln \left( {I_{{o\lambda_{3} }} } \right) - m_{R} \tau_{{R\lambda_{3} }} - m_{{O_{3} }} \tau_{{O_{3} \lambda_{3} }} - m_{p} \tau_{{p\lambda_{3} }} ,$$

(5)

$$\ln \left( {I_{{\lambda_{4} }} } \right) = \ln \left( {I_{{o\lambda_{4} }} } \right) - m_{R} \tau_{{R\lambda_{4} }} - m_{{O_{3} }} \tau_{{O_{3} \lambda_{4} }} - m_{p} \tau_{{p\lambda_{4} }} ,$$

(6)

and making

$$\begin{gathered} \left( {\ln \left( {I_{{\lambda_{1} }} } \right) - \ln \left( {I_{{\lambda_{3} }} } \right) } \right) - \left( {\ln \left( {I_{{\lambda_{2} }} } \right) - \ln \left( {I_{{\lambda_{4} }} } \right) } \right), \hfill \\ \end{gathered}$$

(7)

to find TOC as (Dobson, 1931; Gao et al., 2001):

$${\text{TOC}} = \frac{{\left[ {\ln \left( {\frac{{I_{{o\lambda_{1} }} }}{{I_{{o\lambda_{3} }} }}} \right) - \ln \left( {\frac{{I_{{\lambda_{1} }} }}{{I_{{\lambda_{3} }} }}} \right)} \right] - \left[ {\ln \left( {\frac{{I_{{o\lambda_{2} }} }}{{I_{{o\lambda_{4} }} }}} \right) - \ln \left( {\frac{{I_{{\lambda_{2} }} }}{{I_{{\lambda_{4} }} }}} \right)} \right] - m_{R} \left[ {\left( {\beta_{{R\lambda_{1} }} - \beta_{{R\lambda_{3} }} } \right) - \left( {\beta_{{R\lambda_{2} }} - \beta_{{R\lambda_{4} }} } \right)} \right]\frac{{N_{A} {\text{HP}}}}{{V_{{{\text{STP}}}} P_{o} }}}}{{m_{{O_{3} }} \left[ {\left( {\alpha_{{O_{3} \lambda_{1} }} - \alpha_{{O_{3} \lambda_{3} }} } \right) - \left( {\alpha_{{O_{3} \lambda_{2} }} - \alpha_{{O_{3} \lambda_{4} }} } \right)} \right]\frac{{N_{{\text{A}}} }}{{V_{{{\text{STP}}}} }}}},$$

(8)

where N_A = 6.022 × 10²³ air molecules/mol is Avogadro's number, H = 8.4 × 10⁵ cm the terrestrial atmosphere scale height, V_STP = 22,400 cm³ the volume of 1 mol of ideal gas at standard temperature and pressure (STP, 273.15 K and 1 atm) (Walker, 1977), $\alpha_{{O_{3} \lambda_{i} }}$(see Eq. (10)) and $\beta_{{R\lambda_{i} }}$(see Eq. (11)) the effective cross sections (in cm²) for absorption by ozone and attenuation by the atmosphere, respectively, at λ_i.

The uncertainty in TOC ($\sigma_{{{\text{TOC}}}}$) was determined by the propagation analysis formula for uncertainties (ISO, 1995), considering $I_{{\lambda_{i} }}$(i = 1, 2, 3, and 4), $m_{R}$, $m_{{O_{3} }}$, and P as the independent variables x_j (j = 1, 2, 3, 4, 5, 6, and 7) and their corresponding uncertainties σ_j:

$$\sigma_{{{\text{TOC}}}} = \sqrt {\sum\limits_{j = 1}^{7} {\left( {\frac{{\partial {\text{TOC}}}}{{\partial x_{j} }}\sigma_{j} } \right)^{2} } } .$$

(9)

The four wavelengths of DS measurements centre filter bandwidths of 2.3–2.4 nm FWHM. Thus, all parameters depending on λ_i must be turned into an effective value by weighing the parameter with the transmission function ($T_{{\lambda_{i} }}$) of the radiometer filters (Komhyr et al., 1993). For this, we define both (i) the effective ozone absorption cross section

$$\alpha_{{O_{3} \lambda_{i} }} = \frac{{\sum\nolimits_{j = 1}^{{N_{\lambda i} }} {T_{{\lambda_{i} j}} {\text{CS}}_{{O_{3} \lambda_{i} j}} } }}{{\sum\nolimits_{j = 1}^{{N_{\lambda i} }} {T_{{\lambda_{i} j}} } }},$$

(10)

where ${\text{CS}}_{{O_{3} \lambda_{i} }} = \frac{{C_{o} \left( {\lambda_{i} } \right) + C_{1} \left( {\lambda_{i} } \right)\left( {T_{e} - 273.15} \right) + C_{2} \left( {\lambda_{i} } \right)\left( {T_{e} - 273.15} \right)^{2} }}{{N_{{\text{A}}} }}V_{{{\text{STP}}}}$ is the ozone absorption cross section (Cervino et al., 1995), and (ii) the effective atmospheric molecular scattering cross section

$$\beta_{{R\lambda_{i} }} = \frac{{\sum\nolimits_{j = 1}^{{N_{\lambda i} }} {T_{{\lambda_{i} j}} {\text{CS}}_{{{\text{AMS}}\lambda_{i} j}} } }}{{\sum\nolimits_{j = 1}^{{N_{\lambda i} }} {T_{{\lambda_{i} j}} } }},$$

(11)

where ${\text{CS}}_{{{\text{AMS}}\lambda_{i} }} = \frac{{0.008569\lambda_{i}^{ - 4} \left( {1 + 0.0113\lambda_{i}^{ - 2} + 0.00013\lambda_{i}^{ - 4} } \right)}}{{N_{{\text{A}}} H}}V_{{{\text{STP}}}}$ is the atmospheric molecular scattering cross section (Teillet, 1990). Figure 1 shows $T_{{\lambda_{i} }}$ for the radiometer UVFMR4 #566 filters, ${\text{CS}}_{{{\text{AMS}}\lambda_{i} }}$, and ${\text{CS}}_{{O_{3} \lambda_{i} }}$ for T_e = 229.6 K.

Fig. 1

a The ozone absorption cross section. Data points have been fitted by the curve $\theta_{{O_{3} }} = \left( {0.38 \pm 0.09} \right){\text{e}}^{{ - \left( {0.1384 \pm 0.0008} \right)\lambda }}$ with coefficient of determination r² = 0.999 and percent scattering SD% = 0.63% (${\text{SD}}\% = {{{\text{qmd}} \cdot 100\% } \mathord{\left/ {\vphantom {{{\text{sd}} \cdot 100\% } {\left( {{\text{CS}}_{{O_{3} \max }} - {\text{CS}}_{{O_{3} \min }} } \right)}}} \right. \kern-\nulldelimiterspace} {\left( {{\text{CS}}_{{O_{3} \max }} - {\text{CS}}_{{O_{3} \min }} } \right)}}$, where qmd is the quadratic mean difference between data points and fitting curve). b The atmospheric molecular scattering cross section. c The transmission function of the radiometer UVMFR4 #566 filters

From Eq. (7), the difference

$$- m_{p} \left[ {\left( {\tau_{{p\lambda_{1} }} - \tau_{{p\lambda_{3} }} } \right) - \left( {\tau_{{p\lambda_{2} }} - \tau_{{p\lambda_{4} }} } \right)} \right],$$

(12)

would come up in Eq. (8)'s numerator. Nonetheless, it has been neglected because the other differences in Eq. (8)'s numerator are much larger than it for the range of wavelengths (λ₄ − λ₁ = 17.5 nm) in this study (Dobson, 1931), as the dependence of the particle optical depth on radiation wavelength is usually low. Therefore, Eq. (8) indeed gives an approximation of the real value of TOC. However, neglecting that difference (Eq. (12)) is convenient, since we have no previous information regarding the particle optical depths.

The one-per-minute images yielded by TSI #157 gives a partial visual description of the sky and cloud cover during the measurement by the radiometer. In addition, the complete information for a given image is never obtained because the black strip attached to the TSI mirror blocks out the Sun and its immediate surrounding sky.

Regarding the sky cover during the radiometer measurements, three different scenarios are considered in this study. The first is the all-sky (AS) scenario with DS measurements occurring without any clouds blocking the Sun, but clouds are present in the sky. The investigation of the potential effect of such a scenario on TOC measurements is one of the goals here. Secondly, measurements occur with the Sun being blocked by some clouds (C) or, as a particular case of blocked Sun, a visible and identifiable aerosol plume (AP). And the third scenario is the cloud-free sky (CFS) that corresponds to a no-cloud sky period during which, naturally, the Sun is not blocked by any cloud, and therefore DS measurements are free from the potential influence of clouds.

Linear-fitting comparisons of TOC_CFS with TOC_OMI measurements and measurements from the ground-based instrument MICROTOPS II #8461 are also provided.

3 Results

Figure 2 depicts TOC measurements for AS, C, and CFS scenarios on 13 March 2010, while Table 1 presents the corresponding daily averages ± 1 standard deviation (sd) as $\overline{{{\text{TOC}}_{{{\text{AS}}}} }}$, $\overline{{{\text{TOC}}_{{\text{C}}} }}$, and $\overline{{{\text{TOC}}_{{{\text{CFS}}}} }}$ in addition to the corresponding time periods to obtain the measurements.

Fig. 2

Total ozone column (TOC) measurements for all-sky scenarios (TOC_AS), cloud blocking the Sun (TOC_C), and cloud-free skies (TOC_CFS) on 13 March 2010 in Belo Horizonte. Bars are 1σ uncertainty

Table 1 Daily averages ± 1 standard deviation (sd) as $\overline{{{\text{TOC}}_{{{\text{AS}}}} }}$, $\overline{{{\text{TOC}}_{{\text{C}}} }}$, and $\overline{{{\text{TOC}}_{{{\text{CFS}}}} }}$ on 13 March 2010 in Belo Horizonte and their corresponding measurement time periods (ΔT)

Based on sky images from the TSI, the sky cover predominance is depicted in Fig. 3 in terms of the identification of cloud cover and AP for the 87 days corresponding to TOC_C (or TOC_AP) values. Since the determination of the sky cover predominance was based solely on the analysis of TSI images, only four groups of cover (cloud or aerosol) have been selected (WMO, 1956): group 1 is low-altitude cumulus (Cu), cumulus fractus (Cu Fra), stratocumulus (Sc), and stratus (St) clouds; group 2 is middle-altitude altocumulus (Ac) and altostratus (As) clouds; group 3 is high-altitude cirrocumulus (Cc) and cirrostratus (Cs) clouds; and group 4 is for AP. In addition, there were also mixed scenarios where the sky was populated by similar shares of groups, resulting in subgroup representation by half-numbers, for example 1.5, which derives from a sky with comparable contributions from groups 1 and 2.

Fig. 3

Sky cover predominance from June 2009 (JUN09) to September 2010 (SEP10) in Belo Horizonte during total ozone column measurements with clouds blocking the Sun. Note that sky-cover predominance on some days was a mix between groups 1 and 2, resulting in subgroup 1.5. In such cases, cloud genera of group 2 are shown as Ac (altocumulus) and As (altostratus). Cumulus (Cu), cumulus fractus (Cu Fra), stratocumulus (Sc), stratus (St), cirrocumulus (Cc), cirrostratus (Cs), and aerosol plume (AP) belong to other groups

The daily averages of TOC_AS, TOC_C, and TOC_CFS from JUN09 to SEP10 are depicted in Fig. 4, with their variability represented by bars of 1 standard deviation. As shown by Silva (2013a), Usually, the diurnal variability in TOC measurements is small, being increased by the presence of clouds and aerosols. There were 138 days of measurements with AS and CFS scenarios for TOC_AS and TOC_CFS, respectively, and 87 days with C scenarios for TOC_C. The minimum-to-maximum ranges were 243–297 Dobson units (DU) for both AS and CFS scenarios and 247–302 DU for C scenarios. Maximum and minimum values occurred by September–October (early spring) and April, respectively, as expected for a low-latitude Southern Hemisphere site (Silva, 2007; Whitten & Prasad, 1985).

Fig. 4

Daily averages of total ozone column (TOC) measurements for all-sky conditions (TOC_AS), cloud or aerosol plume blocking the Sun (TOC_C or TOC_AP), and cloud-free skies (TOC_CFS) from June 2009 (JUN09) to September 2010 (SEP10) in Belo Horizonte. Bars are 1 standard deviation. Figure 3 shows when TOC_C or TOC_AP is to be considered

However, daily time periods of measurements varied greatly. Thus, to make reliable comparisons of the variability in TOC measurements among the three AS, C, and CFS scenarios, one must obtain a time-independent parameter regarding TOC from days where the three scenarios occurred. Thus, Table 2 depicts—for the 75 days where the three scenarios AS, C, and CFS occurred daily with at least three TOC measurements each—the minimum (T_min), the average ($\overline{T}$), and the maximum (T_max) elapsed times to obtain the TOC measurements for each of the AS, C, and CFS scenarios, the three corresponding average values of TOC for the three sets of 75 daily average values of TOC (TOC₇₅) ±  the corresponding average standard deviation (± sd₇₅), and the TOC variability rate ${{{\text{sd}}{}_{75}} \mathord{\left/ {\vphantom {{{\text{sd}}{}_{75}} {\overline{T} }}} \right. \kern-\nulldelimiterspace} {\overline{T} }}$ that represents the time-independent parameter.

Table 2 For the 75 days where the three scenarios AS, C, and CFS occurred with at least three TOC measurements each in Belo Horizonte, the minimum (T_min), the average ($\overline{T}$), and the maximum (T_max) elapsed times to obtain daily TOC measurements for the AS, C, and CFS scenarios, the three corresponding average values of TOC for the three sets of 75 daily average values of TOC (TOC₇₅) ± 1 corresponding verage standard deviation (sd₇₅), and the variability rate (${{{\text{sd}}_{75} } \mathord{\left/ {\vphantom {{{\text{sd}}_{75} } {\overline{T} }}} \right. \kern-\nulldelimiterspace} {\overline{T} }}$)

The effects of clouds on TOC measurements are better depicted through the ratios TOC_C/TOC_CFS and TOC_C/TOC_AS as shown in Fig. 5 for those 75 days. The average TOC_C/TOC_CFS equals 1.009 (0.9% increase in TOC measurement) with ratios from 0.954 (−4.6% reduction) to 1.034 (3.4% increase), while the average TOC_C/TOC_AS equals 1.003 (0.3% increase) with ratios from 0.946 (−5.4% reduction) to 1.022 (2.2% increase). From the 87 days of C, 10 had TOC_C/TOC_CFS < TOC_C/TOC_AS, and some were, in fact, days with an aerosol plume blocking the Sun instead of a cloud, as indicated in Fig. 3. For those days, TOC_C must be replaced with TOC_AP. Note also that 14 out of the 18 days with ratios < 1 were days within the dry period where AP scenarios are common, although not necessarily easy to identify.

Fig. 5

Ratios between daily averages of total ozone column measurements with cloud or aerosol plume blocking the Sun (TOC_C or TOC_AP) and cloud-free skies (TOC_CFS) or all-sky conditions (TOC_AS) and TOC_CFS from June 2009 (JUN09) to September 2010 (SEP10) in Belo Horizonte. Figure 3 shows when TOC_C or TOC_AP is to be considered

Since there are ground-based and satellite TOC measurements from other works referring to BH, it is interesting to compare them with TOC_CFS, as CFS scenarios do not present the interference from clouds, although aerosol effects on the measurements can be present. Therefore, Fig. 6 shows 138 days of daily average TOC_CFS, 77 days of TOC_OMI from OMI, and 84 days of average TOC_MII from the handheld MICROTOPS II #8461 operated from August 2009 (AUG09) to SEP10 by Silva and Tomaz (2012). The average TOC from those three sets of data for the whole period were 268 ± 10 DU, 264 ± 12 DU, and 259 ± 11 DU, respectively. To deepen the comparison, linear fittings were applied to TOC_OMI and TOC_MII as a function of TOC_CFS. Figure 7 depicts TOC_CFS vs. TOC_MII and Table 3 shows the corresponding linear fitting parameters.

Fig. 6

Total ozone column (TOC) measurements in Belo Horizonte from the UVMFR4 radiometer #566 for cloud-free skies (daily average TOC_CFS) and the Ozone Monitoring Instrument (TOC_OMI) from June 2009 (JUN09) to September 2010 (SEP10), and from the Microtops II #8461 (average TOC_MII) from August 2009 (AUG09) to SEP10

Fig. 7

Linear fitting between total ozone column measurements from the handheld Microtops II #8461 (average TOC_MII) and the UVMFR4 #566 under cloud-free skies (daily average TOC_CFS) from June 2009 (JUN09) to September 2010 (SEP10) in Belo Horizonte

Table 3 Parameters for the linear fitting (Y = a + bX) of TOC_CFS vs TOC_MII and TOC_CFS vs. TOC_OMI in Belo Horizonte. N and r refer to the number of data and the linear correlation, respectively, and SD% (${\text{SD}}\% = {{{\text{qmd}} \cdot 100\% } \mathord{\left/ {\vphantom {{{\text{sd}} \cdot 100\% } {\left( {Y_{\max } - Y_{\min } } \right)}}} \right. \kern-\nulldelimiterspace} {\left( {Y_{\max } - Y_{\min } } \right)}}$, where qmd is the quadratic mean difference between data points and fitting curve) is the percent scattering

4 Discussion

Both detection and identification of clouds remotely can produce a misleading result. In fact, it is not trivial to detect the presence of some clouds or reasonably identify them by remote observation, since clouds such as As, for instance, can be taken erroneously as an AP and vice versa. It must be taken into account that what is seen in the sky remotely from the ground can even depend on the angle of visualization in relation to the Sun's position in the sky: the remote perception of a cloud is merely the detection of a radiation scattering phenomenon. Hence, there have been many situations where there has been significant doubt from the analysis of TSI images regarding the precise identification of a stratus-genus cloud in relation to an AP. Such plumes identified from June 2010 (JUN10) to SEP10 (winter and early spring), as indicated in Fig. 3, were possibly mixed with As or other middle- and high-altitude stratus-genera in some days. The correct identification of C and AP scenarios is of utmost importance, as both can attenuate the solar radiation (Estupiñán et al., 1996). A study by Michelangeli et al. (1992) suggests that AP (or an aerosol layer) can also contribute to REEs.

In Table 1, the values of sd for $\overline{{{\text{TOC}}_{{{\text{AS}}}} }}$, $\overline{{{\text{TOC}}_{{\text{C}}} }}$, and $\overline{{{\text{TOC}}_{{{\text{CFS}}}} }}$ are much smaller than the uncertainties in each single TOC measurement in Fig. 2: the measurement uncertainty comes from Eq. (9) and represents ≈ 5% of TOC. Gao et al. (2001) found values of uncertainty of 2% in ${\text{TOC}}_{{{\text{AS}}}}$ considering a variability of 0.5% in DS measurements, while uncertainties of 8% (Hülsen & Gröbner, 2007) were assumed in DS measurements for this work. Nonetheless, from a statistical point of view, it can be said that $\overline{{{\text{TOC}}_{{{\text{AS}}}} }} = \overline{{{\text{TOC}}_{{{\text{CFS}}}} }}$ because of their 1–2 DU variability, while $\overline{{{\text{TOC}}_{{\text{C}}} }} \ne \overline{{{\text{TOC}}_{{{\text{CFS}}}} }}$ because of the variability of 1 DU in $\overline{{{\text{TOC}}_{{{\text{CFS}}}} }}$.

It seems that clouds cause such an effect. But how does it work? In the sky images produced by the TSI on 13 March 2010, there is a predominance of cloud genera Ac and As until 13:30 Universal Time (UT) and Sc and Cu Fra after 16:00 UT. Certainly, radiation attenuation played a role that day, but those cloud genera are also closely related to REE, yielding sudden (within a few tens of seconds) increases in the flux of radiation towards the ground from radiation scattering in clouds. Such sudden variations in the incidence of solar radiation could affect the values of TOC_C because of the process used to obtain UVMFR4 measurements: UV_G, UV_Dif, and DS average three cycles of measurements within 1 min. Thus, regarding DS measurements, radiation attenuation by clouds within a few tens of seconds during the three cycles of UV_G, UV_Dif, and DS measurements would reduce the DS measurement, leading to a misleading perception, a false effect, of increased TOC_C. On the contrary, the sudden augmentation of radiation yielded by an REE would increase the value of the DS measurement, leading to a misleading perception of decreased TOC_C. Both effects could be responsible for the varying TOC_C seen in Fig. 2 with 5 DU of standard deviation. Despite the outstanding characteristics of the TSI as an instrument, the black strip on its mirror represents a drawback that hinders this study, as the avoidance of direct reflection of sunbeams onto the TSI camera also unfortunately makes it impossible to exactly identify what (cloud or AP) and how (attenuation or REE) the Sun’s radiation is blocked during UVMFR4 measurements. As a result, improving the current investigation into how clouds affect the values of TOC_C is hampered for the time being. Nonetheless, Brogniez et al. (2005) and Houët and Brogniez (2004) in northern middle latitudes have observed results similar to those in Fig. 4, where in addition to larger values of TOC_C, larger variability in values was obtained.

The average ratio of 1.009 (0.9%) for TOC_C/TOC_CFS and 1.003 (0.3%) for TOC_C/TOC_AS from the daily ratios in Fig. 5 are an indication of cloud effects on DS measurements leading mainly to an increase in the value of TOC_C. Thus, radiation attenuation by clouds seems to be the preponderant factor here on average, as ratios tend to be > 1. On 13 March (Fig. 2), those ratios represented increases of 1–1.5% for mixed groups 1 and 2 of sky cover predominance. For the 18 days with ratio < 1, 44% were caused by group 1 sky cover predominance, 17% by group 2, and 39% by group 4. However, more investigation is needed, especially regarding the more precise identification of what really blocks the Sun during DS measurements. This is a fundamental aspect regarding the identification of situations where attenuation or REEs can take place or even when an AP blocks the Sun instead of a cloud. To address this problem, a new imager has been developed by the author to make images of the sky without obstructions like those that occur with TSI's black strip blocking out the Sun and its immediate surroundings.

In the comparisons of TOC_CFS vs TOC_MII and TOC_CFS vs TOC_OMI, whose linear fitting parameters are shown in Table 3, slope b and intercept a are all 95% confidence-level parameters, although the values of the linear correlations r are just fair. The reason for the poorer r is probably the difference among the data sets in terms of their particular meaning. While TOC_OMI considers a 13:45 ± 15 min equator-crossing time OMI satellite measurement and TOC_MII a ground-based DS measurement averaged over about 10 DS measurements as close as possible to the OMI overpass time (Silva & Tomaz, 2012), TOC_CFS is a daily average obtained within a measurement time period ΔT that does not necessarily correspond to the same times as the other two data sets. Restricting the satellite parameter Dist to impose only the use of satellite data corresponding to the site's territorial area leads to quite a superior value of r = 0.91 for TOC_MII vs TOC_OMI, as shown by Silva and Tomaz (2012) for the same site.

Silva (2013a) and Silva and Tomaz (2012) have shown that the variability in TOC depends significantly on cloudiness, as the sd can be < 1% of TOC for CFS scenarios and < 4% of TOC for AS ones. The average TOC_C/TOC_CFS larger than the average TOC_C/TOC_AS also implies that the very presence of clouds is enough to affect DS measurements, although no cloud had blocked the Sun for CFS or AS scenarios. Nonetheless, a true verification of the sky cover predominance effect on TOC measurements must naturally be time-independent. Thus, considering just the 75 days where each of the three scenarios AS, C, and CFS occurred daily with at least three TOC measurements, one calculates the ${{{\text{sd}}_{75} } \mathord{\left/ {\vphantom {{{\text{sd}}_{75} } {\overline{T} }}} \right. \kern-\nulldelimiterspace} {\overline{T} }}$ ratio shown in Table 2. As a TOC variability rate, this ratio is time-independent, and for C scenarios it is over five times larger than the ratio for the other two scenarios, implying a higher variability caused by clouds in the values of TOC measurements through attenuation and REE.

5 Conclusions

TOC measurements were drawn from DS measurements by a UVMFR4 radiometer for 16 months in a tropical site. Although uncertainties of about 5% were found in TOC measurements, their variability was generally smaller for all-sky conditions without clouds blocking the Sun (AS) and cloud-free skies (CFS) scenarios. The presence of clouds seems to affect not only the value of TOC, but also the variability of a set of measurements, especially if clouds block the Sun (C scenarios). Thus, the averages of TOC_C/TOC_CFS = 1.009 (0.9%), ranging from 0.954 (−4.6%) to 1.034 (3.4%), and TOC_C/TOC_AS = 1.003 (0.3%), ranging from 0.946 (−5.4%) to 1.022 (2.2%), indicate an average increase of 0.9% and 0.3% in the values of TOC measurements yielded by clouds in relation to CFS and AS scenarios, respectively, while the time-independent variability in TOC (the ${{{\text{sd}}_{75} } \mathord{\left/ {\vphantom {{{\text{sd}}_{75} } {\overline{T} }}} \right. \kern-\nulldelimiterspace} {\overline{T} }}$ ratio) was over five times larger for C scenarios than CFS scenarios on average. Thus, clouds affect TOC measurements with respect to both the measured values and, in particular, their variability.

Since the average TOC_C/TOC_CFS was larger than 1, it seems that radiation attenuation by clouds is the main factor responsible for the apparent increase in TOC. However, REEs can play an important role here, as 14 out of 18 TOC_C/TOC_CFS ratios < 1 were under AP scenarios, which sometimes come mixed with stratus-genera clouds, leading to the conclusion that the REE potential effects on TOC measurements deserve more investigation. Unfortunately, the determination of the real effect during DS measurements is not achievable for the time being due to the image partial blockage represented by TSI's black strip. However, the development of a new imager producing images without any blockage (like the black strip on the TSI mirror) of the sky, thereby showing the Sun disc and its immediate surroundings, will enable the improvement of such an investigation.