Nofel Lagrosas 1,2, Gerry Bagtasa3, Naohiro Manago1, Hiroaki Kuze1

Center for Environmental Remote Sensing (CEReS), Chiba University, Chiba 263-8522, Japan
Manila Observatory, Ateneo de Manila University Campus, Quezon City 1108, Philippines
Institute of Environmental Science and Meteorology, University of the Philippines Diliman, Quezon City 1101, Philippines


Received: July 19, 2018
Revised: April 11, 2019
Accepted: April 21, 2019

Download Citation: ||https://doi.org/10.4209/aaqr.2018.07.0267  

  • Download: PDF


Cite this article:

Lagrosas, N., Bagtasa, G., Manago, N. and Kuze, H. (2019). Influence of Ambient Relative Humidity on Seasonal Trends of the Scattering Enhancement Factor for Aerosols in Chiba, Japan. Aerosol Air Qual. Res. 19: 1856-1871. https://doi.org/10.4209/aaqr.2018.07.0267


HIGHLIGHTS

  • Monthly trends of ambient f(RH) are determined from ground instruments.
  • Monthly ambient f(RH) values with RH are modeled.
  • Ambient f(RH) measurements are higher than f(RH) values from sampling measurements.
  • Scattering enhancement ratios from dusts have low response to RH.
 

ABSTRACT


In this study, we used ground instruments, namely, a visibility meter, an integrating nephelometer, an aethalometer, a lidar, and a weather monitor, to measure the scattering enhancement factor, f(RH), which quantifies the effect of ambient relative humidity (RH) on aerosol light-scattering, and to generate a model of its annual variation in the city of Chiba, Japan. First, the f(RH) values were calculated with chemical analysis data. Second, visibility-meter and aethalometer data were used to model the monthly trends of f(RH) at 550 nm. The f(RH) values were higher during summer than during the other three seasons, which can be attributed to the general pattern of the regional climatology as well as the loading of different particle types into the lower troposphere. Third, the f(RH) values at 532 nm were obtained from lidar and aethalometer measurements. Low and constant f(RH) values were observed during November, whereas higher and increasing f(RH) values were observed during May. Also, dust events during March 2015 showed decreasing f(RH) with increasing RH .


Keywords: Scattering enhancement factor; Relative humidity; Scattering coefficient; Dust aerosol; Ångström coefficient.


INTRODUCTION


Relative humidity (RH) strongly affects the scattering properties of aerosols in the troposphere. For hygroscopic aerosols, the increase of ambient RH in the atmosphere also increases particle sizes and their scattering coefficients (Tang, 1996). One of the optical parameters that is used to characterize changes in scattering coefficients with RH is the scattering enhancement factor, f(RH). This is defined as the ratio between the average scattering coefficient at a particular RH and the scattering coefficient at a dry RH. Aerosol sampling and optical measurements reported in previous works have revealed that as RH increases, the aerosol mass extinction efficiency and scattering enhancement factor increase (Bagtasa et al., 2007; Zieger et al., 2013). The seasonal trends of the effects of RH on aerosols have been reported from the measurements of direct and scattered solar radiation using a spectroradiometer (Manago et al., 2011). Different aerosol compositions have different optical scattering responses to RH change. For instance, because of their hygroscopic nature, sea-salt aerosols can have higher increase in scattering coefficient with increasing RH compared to dust particles (Zieger et al., 2013). Hygroscopic growth of aerosols has also been estimated from chemically resolved samples assuming size distribution measurement is known (Lin et al., 2014). Other measurements of aerosol hygroscopic growth involve active and passive remote sensing and collocated radio sounding data (Granados-Muñoz et al., 2015).

From the viewpoint of remote sensing, the parameter f(RH) is important in assessing the influence of aerosols on atmospheric radiation budget calculation, especially when the dry-state information from sampling instruments is exploited for simulating optical scattering properties under humid ambient conditions. Such simulation is indispensable, for instance, when calibrating multi-wavelength lidar data or carrying out atmospheric correction on satellite imagery (Kuze, 2012). The understanding of f(RH) is also required for describing the onset of cloud particle formation (Hobbs, 1993). From the viewpoint of optical modeling, the estimation of this parameter on a monthly basis, for instance, can lead to the understanding of the seasonal variation of aerosol scattering coefficient in the regional atmosphere. Measurements of this parameter are usually performed by air sampling using two nephelometers operated under wet and dry conditions (Fierz-Schmidhauser et al., 2010). This approach, however, is not always effective for studying the actual values of f(RH) under ambient conditions because of the inherent difference in aerosol size distribution under ambient and sampling conditions, in addition to the difficulty in precisely controlling the RH values in the scattering volumes in the instruments.

In Chiba, a city located along the east side of Tokyo Bay, large monthly variations of RH occur. This RH variation and its effect on different aerosol types (natural and anthropogenic) provide the complexity involved in measuring and characterizing aerosol optical properties in Chiba (Fukagawa et al., 2006; Manago et al., 2011). Previous spectroradiometer measurement in Chiba under clear sky conditions has shown that Ångström coefficient exhibits a seasonal trend that ranges from 0.5 to 1.6 for aerosol optical thickness (AOT; at 550 nm) greater than 0.06 (Manago et al., 2011). The constraint on AOT (> 0.06) represents the condition that aerosol scattering significantly affects the intensity of directly transmitted solar radiation as compared with molecular (Rayleigh) scattering. The change in Ångström coefficient, on the other hand, suggests the occasional dominance of relatively coarse particles, such as sea salt (with a value of ~0.5), or that of fine particles due to urban activities, such as combustion (~1.6). In addition, sun photometer measurements have shown similar seasonal trends but with a slightly wider range of the Ångström coefficient (0.5–2) (Fukagawa et al., 2006). Winter months are characterized by high Ångström coefficient (i.e., dominance of fine particles) but low AOT values (i.e., good visibility). On the contrary, summer months tend to show relatively low Ångström coefficient and high AOT values (Manago et al., 2011). The factors contributing to low values of Ångström coefficient during summer are the predominance of south to southwesterly winds blown from the Tokyo Bay, coupled with the effect of aerosol growth due to high RH during these months.

The objectives of the present study are threefold. First, we simulate the functional form of f(RH) from September 1998 to February 2004 expected for Chiba area by reanalyzing the chemical sampling data reported in our previous paper (Yabuki, 2003; Fukagawa et al., 2006). Second, we develop a measurement technique to evaluate and model f(RH) values of aerosols using extinction coefficient from visibility-meter data and absorption coefficient from aethalometer data. This method produces f(RH) at wavelength 550 nm. Third, we quantify f(RH) values using the data from lidar and ground-based instruments to derive f(RH) at wavelength 532 nm, for which our slant-path lidar provides the aerosol profile data in the lower troposphere. The lidar data reported here are taken from our campaign measurements made in Chiba during two particular months of November 2014 and May 2015. These two months represent a relatively dry month (November) and a transitional month from spring to summer (May). 


SIMULATION OF F
(RH) FROM CHEMICAL ANALYSIS DATA


Different aerosols exhibit different responses when they interact with water vapor in the atmosphere. This leads to different f(RH) curves when this parameter is plotted as a function of RH. The change in optical properties of aerosols is quantified using the scattering enhancement factor (Zieger et al., 2013) and is defined as: 

where αscat is the scattering coefficient at a particular RH and wavelength λ; αscat(RHdry, λ) is the dry scattering coefficient. This parameter relates the ambient value of RH to the scattering coefficient of aerosols measured with a sampling instrument under dry conditions. Although Eq. (1) has commonly been used as the definition of f(RH), different forms of empirical equations have been derived for representing functional forms of f(RH) (Day et al., 2000; Chen et al., 2014; Titos et al., 2014). These empirical equations are inevitable and indicate diverse responses of aerosols to RH and can be attributed to different physical properties of aerosols, such as water activity.

In the present theoretical evaluation of f(RH) of aerosols in Chiba, size distributions from six aerosol types (elemental carbon, organics, (NH4)2SO4, NH4NO3, NaCl, and soil) are used to compute the aerosol scattering coefficients. These size distributions are obtained from chemical measurements conducted in Chiba from September 1998 to February 2002 (Yabuki, 2003). Aerosol growth factor of hygroscopic aerosols (i.e., (NH4)2SO4, NH4NO3, and NaCl) due to increasing RH is simulated using Tang’s method (Tang, 1996). As RH is simulated to increase, the growth factor is used to adjust the mean radii of aerosols in the size distribution, and Mie theory is used to calculate the scattering coefficient at 550 nm. Thus, scattering coefficients for each RH are obtained. Adding the results for all components provides the total scattering coefficients based on the aerosols observed in Chiba, and this provides the estimate of αscat(RH)/αscat(RHdry).


VISIBILITY-METER
 AND lIDAR MEASUREMENTS OF F(RH)


In our analysis, values of αscat(RH, λ) are obtained from either a visibility-meter (λ = 550 nm) or a lidar (λ = 532 nm). Here, we define the dry scattering coefficient as the values observed when RH is below approximately 30%. This criterion of 30% has been chosen in view of the growth/evaporation hysteresis behavior of hygroscopic particles (Tang, 1996). For studying the monthly change of f(RH), the scattering coefficients are estimated by calculating the difference between the extinction coefficient (from the visibility meter) and the absorption coefficient (from the aethalometer). In this approach, the value of dry scattering coefficient is obtained from the condition of RH < 30%. To obtain the value of f(RH) from lidar measurements, αscat(RHdry, λ) is obtained from the nephelometer data with RHinst < 30%. Here, RHinst stands for the RH value inside the nephelometer. The value of RHinst is normally controlled 

in the range between 15% and 30%, though occasionally values exceeding 30% are observed for higher values of ambient RH (RHamb). Hourly values of visibility, V, in Chiba are obtained from the website of Japan Meteorological Agency (JMA) (http://www.data.jma.go.jp/). The continuous observation of visibility is made with a Present Weather Detector (PWD22, Vaisala) operated on the premises of Chiba’s district office building, located 2.5 km south of Chiba University. The visibility meter uses a near infrared light source of 875 nm to measure the visibility but reports the visibility value at 550 nm (according to personal communication with Vaisala’s helpdesk on 12 July 2017). Accuracy of this instrument is within ±10% for 1–10 km visibility and ±15% for 10–20 km visibility. Hereafter, RH indicates the value of ambient RH, RHamb, throughout this paper unless otherwise noted. The values of RH are taken from the JMA website where the visibility data are also reported. Knowledge of visibility change on an hourly basis is desirable for studying the effect of RH on aerosol scattering coefficient. However, in Chiba, visibility values are not always less than 20 km, which is the measurable range of PWD22. There are days in a month with hourly visibility greater than 20 km when examining variation of visibility with RH on an hourly or diurnal scale proves to be impossible. Thus, only the monthly variation of visibility with RH is employed to investigate the trend of RH on aerosol scattering coefficient. This approach can help elucidate the monthly variation of f(RH) with RH in a year and can improve current understanding of energy balance in the local atmosphere. To model the RH dependence on the monthly average of visibility, a 7-point running average is applied to the hourly average visibility. The resulting running average is fitted using the power-law equation of:

 

The parameter V0 indicates the limiting value of visibility under dry conditions (x < 0.3). This implies that a similar power-law relation holds also between the scattering coefficient and ambient RH. Subsequently, the visibility values are converted to extinction coefficients at 550 nm using the following equation:

where V (in units of km) and αext (km–1) are the visibility and the ambient extinction coefficient, respectively. The factor 2.996 represents the logarithmic attenuation of ln(1/0.05) (5% of the original intensity). Since the maximum visibility measurable with the instrument (PWD22) is 20 km, data for an actual visibility of 20 km or more are not included in the present calculation of f(RH). In effect, we lose data for low aerosol loading.

A three-wavelength integrating nephelometer (Model 3563, TSI Inc.) and other sampling instruments are routinely operated at the Center for Environmental Remote Sensing (CEReS), Chiba University (35.63°N, 140.10°E). In order to protect the instrument from harmful dew formation, the nephelometer keeps the inside temperature at around 302 K to provide relatively dry conditions for particles under optical measurement. The nephelometer in CEReS is regularly calibrated. The accuracy of scattering coefficient measurements from this instrument is dependent on the nephelometer calibration, and it is in general within 10% (Anderson et al, 1996). The three-wavelength nephelometer and a seven-wavelength aethalometer (Model AE31, Magee Scientific) are installed on the rooftop of the CEReS building. Table 1 shows the specifications of the nephelometer and aethalometer, as well as those of the slant-path lidar system operated at the elevation angle of 30° toward the north. The ambient air is sampled through a vertical, 3-m long stainless pipe, and the nephelometer and aethalometer output scattering coefficients and black carbon (BC) concentrations every 1 and 5 min, respectively. BC concentrations are converted to absorption coefficients, αabs, as:

 

where C is the measured BC concentration; Cref (=2.14) is the multiple scattering correction factor (Müller et al., 2011); and λ is the wavelength between 370 and 950 nm. Average absorption coefficients (~1 × 10–5 m–1) at 550 nm during months with high RH are lower compared to the average absorption coefficients (~2 × 10–5 m–1) during months with low RH.


Table 1. Specifications of nephelometer, aethalometer, and lidar instruments.

Here, we derive the seasonal change of f(RH) for the year 2014 in the following manner. The absorption coefficients at 550 nm are interpolated using the power law relating the absorption coefficient and the wavelength by employing the Ångström coefficients from the data (Moosmüller et al., 2011). The hourly values of ambient scattering coefficient are obtained by subtracting the aethalometer-based absorption coefficient from the JMA extinction coefficient calculated using Eq. (3). The single-scattering albedo, a parameter defined as the ratio between the scattering coefficient and the extinction coefficient (i.e., the sum of scattering and absorption coefficients), takes a value of 0.85 to 1.0 in Chiba (Manago et al., 2011). This indicates that the correction due to absorption coefficient is relatively small (~10%). Moreover, BC particles are normally non-hygroscopic, and hence, their contribution to f(RH) is not significant. To obtain the dry scattering coefficient at 550 nm on a monthly basis, absorption coefficients are averaged for RH < 30% and subtracted from the extinction coefficient (Eq. (3)) for RH < 30%. This approach can facilitate the understanding of monthly trends of f(RH) with RH. However, in short-term measurements, variability of the atmosphere is often incorporated in the data, as in the case of lidar, and often provide variable f(RH). In this situation, in situ measurements of scattering coefficient at low RH provide better approximation of f(RH).

For optical modeling purposes, the monthly f(RH) values were fitted with the ambient RH using the parameter x (= RH/100). Ordinary power law fitting shows poor correlation between f(RH) and x due to the presence of outlier data. However, employing a robust power-law fitting on the data solves this problem (Verardi and Croux, 2009). The most obvious equation that can be used to fit the relationship between f(RH) and RH (for moderate RH) is of the form (Liu et al., 2013):

 

However, the application of Eq. (5) to our data shows that f(RH) values observed at high RH are sometimes larger than what can be derived from Eq. (5), though it fits well when RH is low or moderate. In order to improve the fitting for high RH (> 80%), we adopt the following modification by changing the second term of Eq. (5) to an exponential function of x:

 

Here, the variables p, q, and r are the coefficients that will be determined from the robust fitting. The first and second terms in Eq. (6) can be interpreted as the response of f(RH) at low and high RH, respectively. We choose to use the form of Eq. (6) to model monthly trends and patterns of f(RH) in Chiba. Previous works have described several equations that best fit their average data (Kotchenruther et al., 1999; Carrico et al., 2003). These previous equations are the results from measuring the effects of RH on non-deliquescent aerosols, hygroscopic aerosols, and smoke from biomass burning. In the present case, the analysis described below will reveal that Eq. (6) provides the best fit, efficiently describing hygroscopic growth of aerosols.

A measurement campaign was conducted at CEReS using the slant-path lidar and ground instruments from 13 to 23 May 2015 to measure the effects of RH on the local aerosols and other observed phenomenon, e.g., dust occurrence during the end of spring. The transition of season from spring to summer occurs during this time interval, and occasionally strong winds cause the outbreak of local dust events. Our previous study has shown that from April to June, as water vapor concentration gradually increases during the transition from spring to summer, AOT values increase (i.e., more aerosol loading) and Ångström coefficient decrease (i.e., coarser particles) (Manago et al., 2011). The simultaneous ground measurements provide an opportunity to study the effects of such a gradual change in season between April and May on the optical properties of aerosols. In the present work, the occurrence of local dust is verified using an optical particle counter (OPC; KC-01D, Rion Co., Ltd.) and depolarization data from National Institute of Environmental Studies (NIES) lidar (http://www-lidar.nies.go.jp) located in Chiba University.

A similar campaign was conducted from 14 to 20 November 2014. During this time, both the slant-path lidar and the ground instruments were simultaneously operated. The month of November is considered a relatively dry month in Chiba. Values of AOT are at their lowest and Ångström coefficient values are high (Manago et al., 2011). Thus, data collected from this campaign period show the response of aerosol optical properties to changing RH for relatively small (fine-mode) particles.

Using the slant path lidar and ground instruments, we propose another method to measure f(RH) at λ = 532 nm to look at the dependence of f(RH) on RH in a short-term measurement. Ambient scattering coefficients are obtained from the lidar extinction coefficient by subtracting relatively small absorption from aethalometer data. Dry scattering and absorption coefficients at 532 nm were derived from nephelometer and aethalometer data, respectively, using the power law relationship between wavelengths. During the November 2014 and May 2015 campaigns, hourly visibility values are almost 20 km. Thus, extinction coefficients from visibility meter cannot be used. The truncation error of the nephelometer (Anderson and Ogren, 1998) due to the limited capability in measuring the scattered light, especially from relatively coarse particles, is taken into consideration (see “f(RH) measurements from lidar: November 2014 and May 2015 campaigns”). Dry scattering coefficients are chosen when the ambient RH is 31% and 32% for the November 2014 and May 2015 campaigns, respectively. This condition usually occurs in the afternoon when the sky is clear. This may not be the case during summertime in Chiba when, most of the time, the RH is greater than 50% even in the afternoon. Extinction coefficients from the slant-path lidar at the full-overlap range of ~600 m (altitude of ~300 m) were measured and compared with ground measurements from both nephelometer and aethalometer. The ratio between the ambient and reference dry scattering coefficients can provide the lidar derived f(RH). Inversion of lidar data was implemented using Fernald’s algorithm (Fernald, 1984) with extinction-to-backscatter ratio, S1, set to 50 sr. Extinction coefficient at the reference altitude was obtained by taking the slope of the range-corrected signal at an altitude of 8–15 km. This method is effective if the atmosphere is homogeneous, and provides an accuracy of around 10% if the extinction coefficients are less than 1 × 10–3 m–1 and signal to noise ratios are better than 1000 (Kunz and de Leeuw, 1993). If clouds exist in the atmosphere, lidar data exhibit strong backscattering signals from clouds with extinction coefficients greater than 1 × 10–3 m–1, and the resulting extinction coefficients are not included in the present analysis of aerosol growth.


RESULTS AND DISCUSSION



Theoretical Evaluation of f(RH) from Chemical Analysis

Simulation of f(RH) is implemented by employing the results of sampling and chemical analysis of aerosols in Chiba. In simulating f(RH), the growth factor (D/Do) is computed using the equation:

 

where D and ρ are the aerosol diameter and density at a solubility of w weight %, respectively (Tang, 1996). The density of each type of hygroscopic particle is modeled from Tang et al. (1996). Then, the growth factor calculated using Eq. (7) is used to adjust the mean radii of the number size distribution of aerosols obtained from the chemical sampling data (Yabuki, 2003). Aerosol scattering coefficients are evaluated through the Mie scattering calculation. The scattering enhancement factor is computed as the ratio of the scattering coefficients to the scattering coefficient at low RH. Fig. 1 shows the annual average f(RH) simulated for aerosols in Chiba. The drastic increase of f(RH) can be seen for RH > 80%. This behavior is similar to the observed trend of average monthly f(RH) deduced from the visibility meter and ground instruments as presented and discussed below.


Fig. 1. Simulated f(RH) for aerosols in Chiba assuming (NH4)2SO4, NH4NO3 and NaCl are the dominant species that contribute to hygroscopic growth of aerosols. The original sampling measurement was made during a time period of September 1998–February 2002 (Fukagawa et al., 2006). The solid line is the robust fit of the modeled f(RH), Eq. (6), with p = 6.429, q = 36.35, and r = 2.242 with R2 = 0.99.Fig. 1. Simulated f(RH) for aerosols in Chiba assuming (NH4)2SO4, NH4NO3 and NaCl are the dominant species that contribute to hygroscopic growth of aerosols. The original sampling measurement was made during a time period of September 1998–February 2002 (Fukagawa et al., 2006). The solid line is the robust fit of the modeled f(RH), Eq. (6), with p = 6.429, q = 36.35, and r = 2.242 with R2 = 0.99.


f(RH) Values Based on Visibility Data

Increase in RH generally leads to highly varying but decreasing values of visibility. Fig. 2 shows the variation of average visibility with RH in the month of January 2014. In this figure, the average visibility is plotted for each value of RH (43, 44, …, 93%) that appeared in the RH record during this month. To study the trend of the variation of visibility with RH, a 7-point running average has been applied to the average visibility. Since the visibility rapidly decreases with high RH, the power-law relationship given by Eq. (2) is used to fit the visibility with RH. The resulting fit yields the monthly trend of visibility given as a function of ambient RH.


Fig. 2. Average visibility plotted against the ambient RH (%) for the month of January 2014. The trend of decreasing visibility with RH is shown in the modeled equation, Eq. (2), of V = a·xb + V0 with a = –18.28, b = 4.692, and V0 = 17.95 (with R2 = 0.9423).Fig. 2. Average visibility plotted against the ambient RH (%) for the month of January 2014. The trend of decreasing visibility with RH is shown in the modeled equation, Eq. (2), of V = a·xb + V0 with a = –18.28, b = 4.692, and V0 = 17.95 (with R2 = 0.9423).

Fig. 3 shows the daytime and nighttime average wind speeds and directions with the box plot of the ambient RH in 2014. The original dataset is the hourly data provided by the JMA Chiba station. The monthly average of hourly RH values increases from 47% in January to 80% in June and decreases to 52% in December. High and low RH values are more frequent during summer and winter months, respectively. A closer inspection of the RH values indicates that during the month of February, hourly RH values range from 20% in the late afternoon to 95% in the early morning. The months of February, May, August, November, and December show RH values peaking at 95% (at 07:00), 94% (at 08:00), 93% (at 03:00 and 10:00), 92% (at 19:00), and 93% (at 06:00), respectively. The average and standard deviation (±1σ) of RH in February, May, August, November, and December are 54.8 ± 20.3%, 65.0 ± 17.5%, 74.0 ± 10.4%, 63.3 ± 18.7%, and 52.2 ± 18.0%, respectively. Wind direction during the daytime (10:00–14:00) during winter, spring, and summer months (January–September) usually exhibit winds predominantly coming from the north-northwest, west-southwest and southwest, while fall months (October–December) are dominated by northeast and north-northeast wind (Fukagawa et al., 2006). In 2014, however, the predominant daytime and nighttime winds from April to August are from the southwest, south, and southwest, as indicated in the upper part of Fig. 3. The predominant winds from October to December are from the northeast, northwest, and west. February and August have higher daytime and nighttime wind speeds compared to other months.


Fig. 3. Box plot of hourly RH values in Chiba and vector plot of daytime and nighttime wind speed and wind direction for the months of 2014.Fig. 3. Box plot of hourly RH values in Chiba and vector plot of daytime and nighttime wind speed and wind direction for the months of 2014.

Fig. 4(a) shows the contour map of the average measured visibility with RH. The white sections in the graph represent no visibility data that correspond to the RH in the vertical axis. Higher visibility is seen in the months of spring and autumn. This is due to the fact that the atmosphere is relatively dry in these months when compared to summer months, as illustrated in Fig. 3.


Fig. 4. Average monthly visibility with RH (%) in the year 2014: (a) visibility-meter (PWV22) measurements and (b) modeled visibility. Both panels show lower visibility during summer months and higher visibility during spring and autumn months.Fig. 4. Average monthly visibility with RH (%) in the year 2014: (a) visibility-meter (PWV22) measurements and (b) modeled visibility. Both panels show lower visibility during summer months and higher visibility during spring and autumn months.

Fig. 4(b) shows the result after modeling the visibility with RH. The modeling was carried out with a procedure shown in Fig. 2 (after taking a running average) by using Eq. (2). The modeled visibility strongly indicates lower values during summer months. This degradation of visibility is caused by higher RH during summer months, leading to the increase in the size of aerosols in the atmosphere. Table 2 shows the parameters resulting from the present model fitting. This indicates that for a particular month, when RH becomes less than ~30%, the parameter V0 can be interpreted as the characteristic dry visibility for that month. The dry extinction coefficient obtained from this characteristic visibility can be ascribed to the prevailing aerosol type and temporal variation of aerosol loading for that month. By using Eq. (3), the average of the visibility values for RH < 30% (or extrapolated to RH < 30%, if necessary) provides the representative mean of dry extinction coefficient of that month. Then, dry scattering coefficients are obtained by subtracting the monthly averages of dry absorption coefficients from the values of dry extinction coefficient. The resulting dry scattering and absorption coefficients range from 1.5 × 10–4 to 2 × 10–4 m–1 and from 0.4 × 10–5 to 1 × 10–5 m–1, respectively. Monthly changes in the dry scattering coefficients are minimal, and these indicate that temporal changes in αscat(RHdry) do not significantly influence the calculation of f(RH). In the case of July and August, the dry extinction coefficient is regarded as the dry scattering coefficient, since the aethalometer measurements do not have records for RH < 30%. This leads to an error of approximately less than 10% when compared to using actual values of dry scattering and absorption coefficients.


Table 2. Fitting coefficients resulting from the fit of average monthly visibility with RH using Eq. (2) (V = a·xb + V0, x = RH/100).

We present here the f(RH) values from measurements with the visibility meter and aethalometer in the year 2014. Fig. 5 shows the measured monthly variation of f(RH). Higher f(RH) values at higher RH are observed during winter and summer months due to the high RH values that occur, especially during nighttime and early morning. In summer, the dominance of southwesterly winds results in the abundance of sea salt particles, since Chiba is on the east coast of Tokyo Bay (Fukagawa et al., 2006).


Fig. 5. Variation of average f(RH) measured in 2014.Fig. 5. Variation of average f(RH) measured in 2014.

The modeled and average monthly variation of f(RH) at 550 nm for 2014 are presented in Fig. 6. The data are the result from fitting Eq. (6) on the average values of measured f(RH). Both values show rapid increase at high RH. The error bars are the one standard deviation of the average scattering coefficients in a month, representing the expected range of f(RH) at a particular RH, which is an integer between 50 and 95 in Fig. 6. At high RH, magnitudes of these error bars change, which can be attributed to the temporal variation of aerosol loading with different types. The result of applying the robust power-law fit on the average f(RH) captures the trend of increase of f(RH) with RH. The resulting equation now describes the monthly behavior of f(RH) with RH.


Fig. 6. Measured and modeled f(RH) at 550 nm derived from visibility-meter and aethalometer measurements in each month of 2014. The model curves are based on Eq. (6) in the form of f(RH) = pxq + exp(xr) with x = RH/100.Fig. 6. Measured and modeled f(RH) at 550 nm derived from visibility-meter and aethalometer measurements in each month of 2014. The model curves are based on Eq. (6) in the form of f(RH) = pxq + exp(xr) with x = RH/100.

The model also shows that f(RH) curve of February sharply increases before RH ≈ 90%. Average daytime and nighttime wind speeds during this month are generally higher (Fig. 3) than those of January and March, implying more aerosol influx. The winds are generally from the northwest (Tokyo area), which indicates that transported aerosols have contributions from the urban area. At around RH > 85% in early morning, a noticeable increase of f(RH) is observed as compared to other months. Fine particles dominate during February, with ammonium sulfate and ammonium nitrate contributing to at least 50% of the volume ratio of fine particles (Fukagawa et al., 2006). Thus, the high RH values in the early morning of this month can effectively contribute to aerosol growth and scattering properties. The f(RH) curves of January, March, April, May, June, July, August, and September exhibit similar degree of steepness at RH < 90%. However, the f(RH) curves of October, November, and December show a more gradual increase in f(RH) for the range of RH > 95%. Such “delayed” response of f(RH) at high RH during these fall months can plausibly be attributed to increasing elemental carbon, decreasing sea salt concentrations due to the change in wind direction, and relatively higher concentrations of organics in both the fine and coarse modes (Fukagawa et al., 2006). The effects of RH on the optical properties of elemental carbon and organics are reported to be minimal even at high RH (Zieger et al., 2013).

Table 3 summarizes the coefficients that result from the fitting process, including the root mean square error (RMSE). The values of both coefficients p and q show increasing trend from June to September. These are also the months when hydrophilic aerosols (ammonium sulfate, ammonium nitrate, and sea salt) are consistently high (Fukagawa et al., 2006). The sudden decrease of these coefficients after September is consistent with the change in wind direction, which brings in less hydrophilic aerosol to the Chiba area. It is worth mentioning that the small number of data points (~26) observed in September cast doubts on the significance of the high value of the coefficient p. This is the result of poor correlation when a simple power law fitting process (Eq. (5)) is applied. Moreover, September is the transition month from summer to fall. From the computational point of view, these overall high values of fitting coefficients (R2 ≥ 0.9) are essential characteristics for the modified power law (Eq. (6)) to fully model high f(RH) values at high RH.. This can presumably be attributed to the high volume ratio between sea salt and organics, especially for coarse particles (Fukagawa et al., 2006). Under conditions of high concentration of sea salt and low concentrations of organics, f(RH) increases at high RH (Randles et al., 2004; Garland et al., 2007). In the case of NaCl, f(RH) can even reach 20 at RH ≈ 90% (Fierz-Schmidhauser et al., 2010). A simple correlation computation of the parameters p, q, and r shows that an increase in p increases with q but not necessarily r (Table 4). The relationship of p and q is expected, since p facilitates the increase of f(RH) through x (< 1). The low correlation between parameter r and parameters p and q, on the other hand, indicates that exp(xr) in Eq. (6) is the adjustment needed to properly characterize the dependence of scattering coefficients on RH for ambient aerosols.


Table 3. Fitting coefficients from the fit of f(RH) = pxq + exp(xr) with x = RH/100 in the year 2014.


Table 4. Correlation coefficient values among the fitting coefficients p, q, and r for the case of all months in 2014.

As seen from Figs. 1 and 6, the f(RH) values derived from the simulation based on chemical sampling and the visibility-meter measurement show similar increasing trend with RH. Although the simulation result in Fig. 1 represents annually averaged aerosol property in Chiba, we have conducted a separate simulation using the monthly data taken during the period of seven years (1998–2004) (Fukagawa et al., 2006). It is found that the resulting values of f(RH) generally exhibit an onset of increase even at lower RH values, as low as RH ≈ 30% (figures not shown). The lower ambient f(RH) from visibility meter indicates the possible influence of aerosol washout in the real environment. Such higher values of f(RH) from chemical sampling are consistent with a previous work comparing f(RH) values from chemical and optical measurements (Brock et al., 2016). In addition to the inherent difference between the chemical and optical approaches (Brock et al., 2016), it is noted that the monthly data (Fukagawa et al., 2006) were obtained from sampling measurements conducted during 96 h at the end of each month, irrespective of weather conditions. Therefore, it is likely that the present methodology using a whole month’s data from the visibility meter is more suitable for characterizing the f(RH) of ambient aerosol particles.


f(RH) Measurements from Lidar: November 2014 and May 2015 Campaigns

To investigate lidar-derived f(RH) at 532 nm, two measurement campaigns were carried out in November 2014 and May 2015. These two months correspond to a relatively dry season and a transition of seasons from spring to summer, respectively. For the analysis of campaign data, the dry scattering coefficients at 532 nm were obtained from the nephelometer data for RH closest to 30%. This is to evaluate the dry scattering coefficient parameter just at the start of aerosol growth due to the subsequent increase in ambient RH. If the V0 values in Table 2 for the months of May and November were used instead, the resulting f(RH) values would not reproduce the actual behavior in the atmosphere because of the deviations of visibility values with RH, as indicated in Fig. 2. It is noted that this method of using the dry scattering coefficient from the nephelometer data can be applied to approximate f(RH) for the data obtained only in a limited time span (e.g., a few hours) when the air-mass property is stable (slow wind speed or constant wind direction, as in the cases discussed below). Under such conditions, it can be assumed that the scattering property of the pertinent air mass is governed mainly by the change in ambient RH.

Scattering coefficients from nephelometer measurements are known to suffer from truncation of light inside the instrument. This can be corrected by first calculating the scattering Ångström coefficient, δ, from the uncorrected scattering coefficients at 450, 550, and 700 nm. Second, the relationship between Ångström coefficient and the correction factor, Ccor, measured for the nephelometer (Model 3563, TSI Inc.) is given as:

where c and d are the fitting coefficients (Anderson and Ogren, 1998). In this work, the values of c and d at 532 nm are linearly interpolated from the fitting coefficients at 450, 550, and 700 nm provided by Anderson and Ogren (1998). The corrected scattering coefficient is the product of Ccor and nephelometer-measured scattering coefficient. The trends and range of f(RH) values under different atmospheric conditions are then investigated to look at the behavior of aerosol optical properties and the typical values of scattering enhancement factor for each condition.


Conditions of High and Low RH

During the campaign periods, cases with high and low RH were observed. Low ambient RH (< 50%) existed from 15:25 of 14 November to 08:00 of 15 November 2014 (Japan Standard Time; JST). The RH values ranged from 31% to 48% from 15:00 to 01:00 and stayed relatively constant (44–47%) until 08:00. The temperature ranged from 14°C to 20°C, and the wind speed was below 3 m s–1. Since RH is relatively higher at nighttime, corrected dry scattering coefficient at 532 nm of 3.9 × 10–5 m–1 was obtained at around 15:25 of 14 November when ambient RH was 31%. The estimated correction factor due to truncation error is around 1.05. Fig. 7(a) shows the time-height indication of aerosol extinction coefficient during this period. In the region below 1 km, a conspicuous aerosol layer was observed from 15:00 to 01:00. After 01:00, the aerosol layer dissipated, leaving the region below 1 km almost aerosol free. The lidar-derived extinction coefficients at an altitude of ~300 m are used to calculate f(RH) at ground level by assuming aerosol vertical homogeneity up to this height. The low RH in the lower troposphere results in relatively low scattering coefficients (< 2.5 × 10–4 m–1) in ambient aerosols. Fig. 7(b), then, shows the temporal change of RH and average f(RH) at 532 nm, which was more or less stable at ~1.2, from 00:50 to 08:00 of 15 November, when the lower atmosphere was almost aerosol free. However, from 15:25 to 00:50, high f(RH) values ranging from 1.3 to 3.8 were observed even at relatively low RH. This unusual event of augmented aerosol scattering can possibly be traced back to the high loading of air pollutants after ~15:00 (Fig. 8). Fig. 8(a) shows the locations of the 11 stations, at which the hourly values of various pollutants are routinely measured and reported as Atmospheric Environmental Regional Observation System (AEROS) data (http://soramame.taiki.go.jp/). Figs. 8(b) and 8(c) indicate that unusually high concentrations of NOx are observed in the city area around Chiba University, while suspended particulate matter (SPM) concentration is moderate. The plotted data were obtained from the CEReS database (http://quicklooks.cr.chiba-u.ac.jp/~soramame/image/all_JP/2014/11/). In some areas, NOx and SPM levels reached 0.14 ppm and 0.05 mg m–3, respectively. The nocturnal chemistry of NOx often produces highly scattering nitrate aerosols or secondary organic aerosols, as reported in some previous literatures (Petkewich, 2004; Benton et al., 2010; Joyce et al., 2014). When these aerosols are formed, surface deposition is the major mechanism that explains aerosol loss in the atmosphere. In the campaign period described here, this loss mechanism can explain the decrease in SPM concentration after 01:00, as manifested in Fig. 8(c), especially at Stations 9, 10, and 11. These sites are located north of CEReS and are near the path of the slant-path lidar.


Fig. 7. (a) Time-height indication of aerosol extinction coefficient from slant-path lidar measurements from 15:30 of 14 November 2014 to 08:00 of 15 November 2014 and (b) corresponding RH and f(RH) values.Fig. 7. (a) Time-height indication of aerosol extinction coefficient from slant-path lidar measurements from 15:30 of 14 November 2014 to 08:00 of 15 November 2014 and (b) corresponding RH and f(RH) values.


Fig. 8. (a) Eleven AEROS stations near CEReS (diamond) and the corresponding temporal change of (b) NOx and (c) SPM concentrations from 15:00 of 14 November to 05:00 of 15 November 2014. Stations 9, 10, and 11 are located north of CEReS and near the light path of the lidar measurement.
Fig. 8. (a) Eleven AEROS stations near CEReS (diamond) and the corresponding temporal change of (b) NOx and (c) SPM concentrations from 15:00 of 14 November to 05:00 of 15 November 2014. Stations 9, 10, and 11 are located north of CEReS and near the light path of the lidar measurement.Fig. 8. (a) Eleven AEROS stations near CEReS (diamond) and the corresponding temporal change of (b) NOx and (c) SPM concentrations from 15:00 of 14 November to 05:00 of 15 November 2014. Stations 9, 10, and 11 are located north of CEReS and near the light path of the lidar measurement.

High RH conditions were evident during the evenings of May 2015 measurement campaign (Fig. 9). At times, high RH occurs during daytime with cloud presence. The low solar radiation due to the presence of clouds decreases ambient temperature, resulting in increasing RH. The occurrence of low-lying clouds sometimes produces high values of extinction coefficient. Therefore, data under such conditions are not included in the present calculation of f(RH). Fig. 9(a) shows the time-height indication of extinction coefficient from our slant-path lidar on 18 May 2015. During the observation period shown in Fig. 9(a) (00:00–18:00 on 18 May 2015), RH and temperature values ranged from 43% to 92% and from 17°C to 27°C, respectively. Wind speed was below 1 m s–1 from 00:00 to 08:00, and it increased up to 6 m s–1 from 08:00 to 18:00, though the wind direction (southeast) was stable. As shown in Fig. 9(a), layers of clouds were observed in the altitude range of 0–6 km. Extinction coefficients from low-lying clouds occurring at 03:00, 11:00, and 14:00–16:00 are not included in the calculation of f(RH). Dry scattering coefficient of 5.7 × 10–5 m–1 (truncation error corrected with a factor of 1.067 and interpolated at 532 nm) was used from previous day (17 May 2015 at 12:35) when the ambient RH was 32%. Fig. 9(b) shows the f(RH) values representing the high RH conditions. Measured f(RH) shows a maximum value of 19 at a high RH of ~92%.


Fig. 9. Time-height indication of aerosol extinction coefficient from the slant-path lidar measurements from 00:00 to 18:00 on 18 May 2015 and (b) corresponding RH and f(RH) values.Fig. 9. Time-height indication of aerosol extinction coefficient from the slant-path lidar measurements from 00:00 to 18:00 on 18 May 2015 and (b) corresponding RH and f(RH) values.


Observation of Dust and Effects on f(RH)

During the measurement campaign in May 2015, the slant-path lidar was operated from 10:30 to 20:30 (JST) on 13 May 2015. Fig. 10(a) shows the time-height indication of aerosol extinction coefficient. Maintenance of the laser head prevented data acquisition from around 13:30 to 15:30. From 13 to 14 May 2015, the predominant wind directions were southwest and south-southwest, i.e., from the Tokyo Bay area. This implies possible transport of sea salt particles. At around 12:00 of 13 May 2015, updrafts near the ground occurred during relatively high maximum wind speeds of 8–12 m s–1. These updrafts could possibly aid in the occurrence of a local dust event. The aerosol size distribution is obtained from the OPC (KC-01D, Rion Co., Ltd.) that counts particles larger than 0.3, 0.5, 1, 2, and 5 µm. Here, the OPC bins for 2–5 µm and for 0.3–0.5 µm are regarded as indicating the concentration of coarse and fine particles, respectively. Fig. 10(b) shows the temporal change of the number of particles in the 2–5-µm size range (dN/dlnD), together with the change of f(RH) and RH. The dry scattering coefficient of 3.1 × 10–5 m–1 at 532 nm (truncation error corrected with a factor of 1.07) used in the calculation was obtained at 12:00 of 13 May.


Fig. 10. (a) Time-height indication of aerosol extinction coefficient from slant-path lidar measurements (10:30–20:30 JST on 13 May 2015) showing updrafts (encircled) near the ground and (b) corresponding temporal variation of f(RH), ambient RH, and dN/dlnD values (for particles with diameters between 2 and 5 µm). Increasing dN/dlnD values during 12:30–14:00 indicate presence of local dust (see text). Temporal variation of (c) single scattering albedo (SSA) values from sampling instruments (nephelometer and aethalometer) and slant-path lidar and (d) dN/dlnD for fine particles with diameters between 0.3 and 0.5 µm.
Fig. 10. (a) Time-height indication of aerosol extinction coefficient from slant-path lidar measurements (10:30–20:30 JST on 13 May 2015) showing updrafts (encircled) near the ground and (b) corresponding temporal variation of f(RH), ambient RH, and dN/dlnD values (for particles with diameters between 2 and 5 µm). Increasing dN/dlnD values during 12:30–14:00 indicate presence of local dust (see text). Temporal variation of (c) single scattering albedo (SSA) values from sampling instruments (nephelometer and aethalometer) and slant-path lidar and (d) dN/dlnD for fine particles with diameters between 0.3 and 0.5 µm.Fig. 10. (a) Time-height indication of aerosol extinction coefficient from slant-path lidar measurements (10:30–20:30 JST on 13 May 2015) showing updrafts (encircled) near the ground and (b) corresponding temporal variation of f(RH), ambient RH, and dN/dlnD values (for particles with diameters between 2 and 5 µm). Increasing dN/dlnD values during 12:30–14:00 indicate presence of local dust (see text). Temporal variation of (c) single scattering albedo (SSA) values from sampling instruments (nephelometer and aethalometer) and slant-path lidar and (d) dN/dlnD for fine particles with diameters between 0.3 and 0.5 µm.

When the updraft started, the counts of coarse particles gradually increased and remained almost steady after 15:00, which continued until the midday of 15 May 2015. Fig. 10(d) shows that the counts of fine particles continuously increased. As shown in Fig. 10(b), this increase at the start of the updraft is accompanied by increasing RH and f(RH), indicating that the coarse and fine particles are hygroscopic and highly scattering. Fig. 10(c) shows the temporal variation of single scattering albedo (SSA). The SSA values from nephelometer and aethalometer (dotted line) are lower than those from lidar and aethalometer (solid line). During the updraft period of 12:30–14:00, increase in the SSA from lidar is seen in Fig.10(c), which is consistent with the increase in f(RH).

After 15:30, on the other hand, a downdraft starting from a height of ~2 km is observed. This downdraft mixed with the ground dust after ~17:00, as seen in Fig. 10(a), even though the counts of fine particles increased up to 2500. The data from polarization lidar from National Institute of Environmental Studies (NIES) (http://wwwlidar.nies.go.jp/Chiba/archives/140513-140517-abc-bl.png), located also in Chiba University, indicates that the observed downdraft has depolarization ratios between 0.1 and 0.2, implying the presence of non-spherical particles. From 15:30 onwards, decrease of f(RH) is seen in spite of the high dN/dlnD of fine particles and increasing RH. This implies that the gradual mixing of downdraft particles with the dust and sea salt near the ground effectively decreased the scattering effect of the overall dust. Fig. 10(c) shows that the ambient SSA (from the slant-path lidar data) decreased. This implies that the aerosols composing the downdraft may have interacted with urban aerosols, such as black carbon, during the descent, effectively increasing the absorptive properties and decreasing values of f(RH). The black carbon concentrations from aethalometer measurements during the downdraft are between 1500 and 2000 ng m–3. Although f(RH) from dust usually exhibits relatively low response with RH (Howell et al., 2006; Pan et al., 2009; Zieger et al., 2013), some studies utilizing satellite remote sensing data have shown that when desert dusts are transported to locations having anthropogenic activities, they can become more absorptive due to mixing with other aerosols (e.g., black carbon) (Li-Jones et al., 1998; Deepshikha et al., 2005; Yi et al., 2014). 


CONCLUSION


This work describes a methodology for obtaining f(RH) using three different approaches. The first approach estimates the humidification of aerosol particles using previously recorded chemical sampling data collected at the target location. Applying Mie theory to the chemical sampling data shows that f(RH) increases with RH. The computed f(RH) values, however, are lower than the measured ambient values.

The second approach calculates f(RH) using visibility-meter and aethalometer data. The f(RH) at λ = 550 nm is calculated because the visibility is measured only at this wavelength. The absorption coefficient values at this wavelength are interpolated from the aethalometer data using the power law relationship between the absorption coefficient and the wavelength. Using data collected by these instruments in 2014, the monthly f(RH) values are obtained by fitting visibility data with RH and approximating visibility values for RH < 30% to obtain monthly dry scattering coefficients as references. Fitting f(RH) with RH using a modified exponential function, y = pxq + exp(xr), produces monthly variations of f(RH) that can be used for modeling and calculating the optical response of ambient aerosols based on ground measurements. Compared to the rest of the year, the f(RH) curves for October, November, and December exhibit a more gradual increase in f(RH) after RH exceeds 95%, suggesting the dominance of hydrophobic type aerosols (e.g., elemental carbon) in the atmospheric boundary layer during these months. This approach to measuring f(RH) with RH can clarify the monthly variations of f(RH) and be useful in understanding and modeling the energy balance in the atmosphere.

The third approach determines the f(RH) values at λ = 532 nm from measurements conducted with a slant-path lidar and other ground instruments, viz., a nephelometer, an aethalometer, a particle counter, and a weather monitor, in November 2014 and May 2015. These measurements represent near real-time values for the scattering enhancement factor, in contrast to the averaged values obtained via the first and second approaches, and display higher values (e.g., ~15 at 80% RH; Fig. 9(b)) than those reported in the literature (~2–3 at 80% RH). This approach can be used to identify systems causing observed phenomena (e.g., dust events), assess the aerosol optical response during low and high RH, and examine changes in f(RH) and local aerosol loading over short durations, i.e., on a scale of hours or days.

Accurate f(RH) values are important for understanding the initial step in cloud formation. The methods presented in this work prove the feasibility of deriving regional f(RH) values to illustrate the monthly variation in f(RH). Future research will use these results to compare and calibrate optical parameters measured by other instruments, e.g., sun photometers or sky radiometers, as well as satellite sensors. 


ACKNOWLEDGMENTS


This work was financially supported by the CEReS Overseas Joint Research Program during 2014–2016.



Don't forget to share this article 

 

Subscribe to our Newsletter 

Aerosol and Air Quality Research has published over 2,000 peer-reviewed articles. Enter your email address to receive latest updates and research articles to your inbox every second week.

Latest coronavirus research from Aerosol and Air Quality Research

2018 Impact Factor: 2.735

5-Year Impact Factor: 2.827


SCImago Journal & Country Rank

Aerosol and Air Quality Research (AAQR) is an independently-run non-profit journal, promotes submissions of high-quality research, and strives to be one of the leading aerosol and air quality open-access journals in the world.