Formaldehyde and Acetaldehyde at Different Elevations in Mountainous Areas in Hong Kong

Intensive field measurements of formaldehyde (HCHO) and acetaldehyde (CH3CHO) were concurrently conducted at a mountain site (TMS) and an urban site (TW) at the foot of the same mountain in Hong Kong from September to November 2010. The spatiotemporal variations of HCHO and CH3CHO, the correlation between HCHO and CH3CHO and the ratios of HCHO/CH3CHO indicated different impacts of primary emissions and secondary formation at the two sites. The source apportionments of HCHO and CH3CHO at both sites were investigated using the Positive Matrix Factorization (PMF) model, while the in-situ formation of HCHO and CH3CHO was estimated using a Photochemical Box Model coupled with Master Chemical Mechanism (PBM-MCM). At TMS, the in-situ formation was the most significant contributor to ambient HCHO and CH3CHO, accounting for 51 ± 5 and 32 ± 2%, respectively, followed by the transport of photochemicallyformed aldehydes, vehicular emissions, biogenic emissions, biomass burning and solvent usage. On the other hand, at TW, the in-situ formation and vehicular emissions explained 55 ± 7% and 18 ± 1% of ambient HCHO, respectively, while vehicular emissions and in-situ formation made comparable contributions to CH3CHO (~35%). The findings are helpful for the formulation and implementation of appropriate control strategies for aldehydes and their precursors in Hong Kong.


INTRODUCTION
Formaldehyde (HCHO) and acetaldehyde (CH 3 CHO), key components of airborne carbonyl compounds, play important roles in atmospheric photochemistry and air quality because of their abundance and photochemical reactivity (Atkinson et al., 2006).They account for a significant fraction of the reactivity of the total volatile organic compounds (VOCs) in different environments.In addition, HCHO and CH 3 CHO are hazardous air pollutants for their dermatological, optometric and nasopharynegeal irritating effects on human beings with carcinogenic posing risks (WHO, 2000;US EPA, 2011).
HCHO and CH 3 CHO are emitted directly from various primary sources, including biomass burning, fossil fuel combustion, industries and vegetation sources (Liu et al., 2006;Seco et al., 2007;Parrish et al., 2012).They are also produced as intermediate from photooxidation of methane and non-methane hydrocarbons (Atkinson, 2006;Seinfeld and Pandis, 2006).The knowledge of the mechanisms for the photochemical formation and removal of HCHO and CH 3 CHO has been significantly advanced over the past years.Firstly, the hydrogen abstraction from hydrocarbons and the following addition of oxygen to the alkyl radical produce an alkyl peroxy radical (RO 2 ), which can react with NO to form an alkoxy radical (RO) and NO 2 , leading to O 3 formation after photolysis (Atkinson et al., 2006;Tseng et al., 2009).Through further reaction of the alkoxy radical with oxygen, HCHO and CH 3 CHO, together with HO 2 radical can be obtained.HCHO and CH 3 CHO can also be formed from alkoxy radicals after the scission of a C-C bond through photolysis and alkoxy radical formation when the radicalradical reactions dominate the radical removal while the nitrogen oxides (NO x ) are insufficient (Luecken et al., 2012).For the removal of HCHO and CH 3 CHO, photolysis and reaction with OH radical are the major pathways.The photolysis of HCHO and CH 3 CHO can result in the formation of HO 2 radicals, which have significant influence on the photochemical reactivity, while the reactions with OH radical contribute substantially to ambient O 3 and secondary organic aerosol (SOA) in polluted urban environments (Carlton et al., 2009;Cheng et al., 2010).It should be noted that the relative importance of primary emissions and secondary formation of HCHO and CH 3 CHO varies with time and locations.Therefore, investigating the contributions of primary emissions and secondary formation is prerequisite for the formulation and implementation of HCHO and CH 3 CHO control measures to alleviate photochemical pollution (Louie et al., 2013;Zhong et al., 2013).
In recent years, efforts have been made to investigate the contributions of primary emissions and secondary formation of HCHO and CH 3 CHO in different environments by using various methods, including the ratios of HCHO/CH 3 CHO, the correlation of HCHO and CH 3 CHO with other tracers, emission inventories, emission-based measurements, and receptor models such as Positive Matrix Factorization (PMF) and PCA/APCS models (Guo et al., 2004;Wei et al., 2008;Li et al., 2010;Wang et al., 2010a;Ho et al., 2012;Parrish et al., 2012;Louie et al., 2013;Dong et al., 2014).For example, by comparing the ratios of HCHO/CH 3 CHO, it was found that the extremely high levels of HCHO (17.3 ppbv) and CH 3 CHO (8.0 ppbv) at a sampling site in the Pearl River Delta region (PRD) on the evening of 15 November 2008 were related to the emissions of biomass burning (Yuan et al., 2012).Through emission-based measurements, Ho et al. (2012) investigated the emissions of carbonyls from commercial cooking sources and vehicular exhaust in Hong Kong, reporting that HCHO was the most abundant species emitted from vehicular exhaust, with the range of 55-61%.By using the PMF model, Guven and Olaguer (2011) conducted source apportionment of HCHO in Houston during summer of 2006.Secondary formation was the most important source of HCHO, with the average contribution of 36%, followed by biogenic sources, mobile sources, and industrial emissions.
Hong Kong and the inland PRD region have had dramatic economic and industrial growth, which was accompanied by severe photochemical pollution due to the air pollutants emitted from various anthropogenic sources (Guo et al., 2009;Wang et al., 2009).Previous studies have emphasized that HCHO and CH 3 CHO considerably contributed to O 3 and SOA in the PRD region, but their sources and characteristics, especially the contributions of primary emissions and secondary formation to HCHO and CH 3 CHO, which could provide scientific support for effectively alleviating photochemical pollution, were not well understood in this region (Louie et al., 2013).Hence, in this study, we conducted an intensive field measurement at two different locations in Hong Kong to characterize the relative contributions of primary emissions and secondary formation to HCHO and CH 3 CHO.To our best knowledge, this is the first attempt on the quantification of primary emissions and secondary formation of HCHO and CH 3 CHO using a combined field measurements and model simulation approaches.

Sampling Site
In this study, the field measurements were conducted simultaneously at different altitudes of the highest mountain in Hong Kong, i.e., Mt.Tai Mo Shan (TMS), from 28 September to 21 November, 2010 (Fig. 1).The detailed description of the two sites was provided in Guo et al. (2013a).The Hong Kong Environmental Protection Department (HKEPD) air quality monitoring station at Tsuen Wan (TW), a mixed residential, commercial and light industry area, was selected as the measurement site at the foot of the mountain.The highelevation site was set at the rooftop of a building at the waist of the mountain (640 m a.s.l.), the highest possible observation location, beyond which to the mountain summit was only natural territory with shrubs and grasses (AFCD, 2008).The TW site was located on the rooftop of a building approximately 15-20 m above the ground level.The linear distance between TMS and TW sites was about 7 km and the difference in elevation between the two sites was 630 m.

Sampling and Analysis of HCHO and CH 3 CHO
Carbonyl samples were collected simultaneously at both sites on selected O 3 episode (e.g., October 23-24 and 29-31, November 1-3 9, and 19) and non-O 3 episode days  (e.g.,September 28,October 2,8,14,(18)(19).The potentially-high O 3 episode days, which were further confirmed by the observed concentrations (i.e., maximum hourly concentration ≥ 100 ppbv, the China's Ambient Air Quality Standard (Grade II)), were selected based on weather prediction and meteorological data analysis.The O 3 episode days were usually related to stronger solar radiation, weaker wind speeds, prevailing northerly winds and less vertical dilution of air pollution compared to non-O 3 episode days.Air samples were collected into the acidified 2,4-dinitrophenylhydrazine (DNPH) silica cartridges with the flow rate of 1 L min -1 for 2 hours.During non-O 3 episode days, carbonyls were consecutively collected every 2 hours from 7 a.m. to 9 p.m. (a total of 8 samples per day) at both sites.On O 3 episode days, carbonyl samples were simultaneously collected every 2 hours from 7 a.m. to 9 p.m., with additional samples collected at midnight and 3 a.m.(a total of 10 samples per day).Totally 180 and 178 carbonyl samples were collected at TMS and TW, respectively.The unequal sample size was because some samples at the two sites were contaminated by the malfunction of O 3 scrubber.The procedures for the pretreatment, the configuration of the analysis system, and the methods of the quality control and quality assurance for carbonyl samples were provided in details elsewhere (Guo et al., 2009;Ho et al., 2012).Briefly, identification and quantification of carbonyl compounds were based on retention times and peak areas of the corresponding calibration standards analyzed by high performance liquid chromatography (HPLC) system through an auto-sampler.The instrument was calibrated using five standard concentrations covering the concentrations of interest for ambient air.There were good linear relationships (R 2 > 0.998) between the concentrations and responses for all carbonyls identified.Cartridge collection efficiency was determined with two cartridges in series; over 98% of carbonyl compounds were found in the first cartridge.Relative percent differences for duplicate analysis were within 10%.Typically, C 1 -C 6 carbonyl compounds were analyzed effectively by this technique with a detection limit of ∼0.2 ppbv (Guo et al., 2009;Ho et al., 2012).

Sampling and Analysis of Methane, Non-methane Hydrocarbons and Halocarbons
In this study, volatile organic compounds (VOCs), including methane (CH 4 ), non-methane hydrocarbons (NMHCs) and halocarbons were collected simultaneously at both sites on the same sampling days as those for the carbonyl samples by using pre-evacuated 2-L electro-polished stainless steel canisters.The sampling period for each canister sample was 60 minutes by using a flow-controlling device connected to the canister inlet.At both sites, onehour integrated sample was collected at 2 h intervals from 0700 to 1900 LT (a total of 7 samples per day) during non-O 3 episode days, while one-hour integrated sample was collected from 0900 to 1600 LT at 1 h interval with additional integrated samples collected at 1800, 2100, 0000, 0300 and 0700 LT on O 3 episode days (a total of 13 samples per day).Totally, 202 and 182 canister samples were collected at TMS and TW, respectively.After sampling, the canister samples were sent to the laboratory at University of California, Irvine for chemical analysis.Methane and 54 NMHCs, oxygenated VOCs (OVOCs) and halocarbons were identified and quantified.The detection limit of aromatics was 3 pptv and for other VOCs was 5 pptv.The accuracy of the measurements was 5% and the measurement precision was 0.5-16%.Detailed descriptions of the analytical system and the quality control, detection limits and analysis precision of the VOC samples are provided in Colman et al. (2001) and Simpson et al. (2010).

Continuous Measurements of O 3 , CO, NO x and SO 2
At TW, the hourly data of O 3 , NO-NO 2 -NO x , CO and SO 2 collected continuously from 06 September to 22 November were obtained from Hong Kong Environmental Protection Department (HKEPD, http://epic.epd.gov.hk/ca/uid/airdata), as well as the meteorological data, including temperature, relative humidity, solar radiation, wind speeds and directions.Detailed information about the measurements, quality assurance and control protocols can be found in the HKEPD report (HKEPD, 2012).At TMS, O 3 was measured using a commercial UV photometric instrument (Advanced Pollution Instrumentation (API), model 400E) with a detection limit of 0.6 ppbv, while CO was measured with a gas filter correlation, nondispersive infrared analyzer (API, Model 300E) with a heated catalytic scrubber to convert CO to carbon dioxide (CO 2 ) for baseline determination.
The detection limit was 30 ppbv for a 2-min average, while the 2 s precision for the CO analyzer was about 1% for a level of 500 ppbv (2-min average) and the overall uncertainty was estimated to be 10%.NO, NO 2 and NO x were detected with a chemiluminescence NO-NO 2 -NO x analyzer (API, Model 200E) that had a detection limit of 0.5 ppbv.The detailed information about the analysis and quality control of trace gases was overviewed in Guo et al. (2013a).Furthermore, the meteorological parameters at TMS were monitored by a weather station (Vantage Pro TM & Vantage Pro 2 plus TM Weather Stations, Davis Instruments).

Positive Matrix Factorization (PMF) Model
PMF was used to resolve the source contributions of HCHO and CH 3 CHO at TMS and TW.Detailed description for the model configuration has been provided in our previous studies (Ling and Guo, 2014;Ling et al., 2011).In this study, the concentrations of 13 NMHCs and halocarbons, HCHO and CH 3 CHO as well as CO and total oxidant O x (O x = O 3 + NO 2 ) in 164 and 150 samples at TMS and TW, respectively, were used for model simulation.Species that are important tracers of sources were selected, including CO, C 2 Cl 4 , CH 3 Cl, ethane, ethene, ethyne, propane, propene, isoprene, benzene, toluene, ethylbenzene and xylenes.For example, CO, ethane, ethene and ethyne are typical tracers for combustion, while benzene, toluene, ethylbenzene and xylenes can be emitted from vehicular emissions and solvent usage.C 2 Cl 4 is a tracer for urban emissions, including solvent usage, while isoprene is from biogenic emissions.As the secondary photochemical product, O 3 could be titrated by NO in the atmosphere.Therefore, total oxidant O x (O x = O 3 + NO 2 ) was used as the tracer of photochemical formation and/or aged air masses in this study.The uncertainty for each species was determined as the sum of 10% of its concentration and two times the method detection limit (MDL) of the species (Paatero, 2000).In this analysis, different numbers of factors were tested, and an optimum solution was determined based on both a good fit to the observed data and the most meaningful results.Many different starting seeds were tested and no multiple solutions were found.All the scaled residuals were between -3 and 3 and the Q values in the robust mode were approximately equal to the degrees of freedom.The profiles of factors extracted from the PMF model were chosen and identified by comparing the results with previous studies based on model simulations (PMF and PAC/APCs), emission-based measurements and tunnel measurements (Tsai et al., 2007;Ho et al., 2009;Guo et al., 2011a, b).

Photochemical Box Model Coupled with Master Chemical Mechanism (PBM-MCM)
A Photochemical Box Model coupled with Master Chemical Mechanism (PBM-MCM) was used to simulate the in-situ formation of HCHO and CH 3 CHO at TMS and TW on all the sampling days.The PBM-MCM was developed by assuming that it was a well-mixed box without consideration of physical processes, such as wind speeds/directions and/or vertical/horizontal transport, and air pollutants in the model were homogeneous.The MCM (version 3.2) used in this study is a state-of-the-art chemical mechanism, which describes the degradation of 143 primary VOCs including methane and contains around 16,500 reactions involving 5,900 chemical species (Jenkin et al., 2003;Saunders et al., 2003).The measurement data, including O 3 , CO, NO x , SO 2 , 54 VOCs and methane, together with the meteorological conditions in the region, i.e., temperature, relative humidity and the heights of boundary layer, were used to constrain the model.The photolysis rates of different species in the model were parameterized as suggested by the previous study (Pinho et al., 2009) using the photon flux determined from the Tropospheric Ultraviolet and Visible Radiation (v5) model based on the actual conditions, such as meteorological conditions, location and time period of the field campaign in Hong Kong (Lam et al., 2013).The dry deposition rates and aerosol uptake of HCHO and CH 3 CHO were configured in the PBM-MCM model according to the parameterization of previous studies (Margaret and Jeanne 1993;Zhang et al., 2003;Li et al., 2014).The detailed information for the model frameworks, the model development and the evaluation for the model performance has been reported in our previous studies (Lam et al., 2013;Ling et al., 2014).

General Characteristics HCHO and CH 3 CHO at Different Altitudes
Table 1 presents the average concentrations of HCHO and CH 3 CHO at TMS and TW, together with the average HCHO and CH 3 CHO levels measured at different locations in previous studies.In general, the average levels of HCHO were comparable at TMS and TW (p > 0.05), with the average values of 2.7 ± 0.2 ppbv and 2.9 ± 0.2 ppbv (mean ± 95% confidence interval), respectively, whereas the average CH 3 CHO concentration was higher at TW (p < 0.05).The average concentrations of HCHO at TMS and TW were similar to that in PRD region (2.5 ± 0.3 ppbv, mean ± 95% C.I.) based on the data collected at a recently-established regional air quality monitoring network (Louie et al., 2013), while the average concentrations of CH 3 CHO were relatively higher in this study.Compared to other mountainous/high elevation sites, the HCHO concentration at TMS (640 m a.s.l) was similar to those at Mt. Dignhu (320 m a.s.l, 2.7 ± 2.5  Chi et al., 2008; 3 Villanueva et al., 2014; 4 Yu et al., 2008; 5 Müller et al., 2006; 6 Ho et al., 2002; 7 Guo et  al., 2009; 8 Guo et al., 2014; 9 Wang et al., 2010a; 10 Wang et al., 2010b; 11  Evtyugina et al., 2006; 12 Sakai et al., 2004.ppbv, mean ± s.d.) and a forest park in southern China (572 m a.s.l, 2.8 ppbv), while the CH 3 CHO concentration (1.4 ± 0.1 ppbv) at TMS was lower than those at Mt. Dinghu (2.2 ± 1.9 ppbv) and the forest park (1.7 ± 1.8 ppbv).Furthermore, HCHO and CH 3 CHO levels were higher at high altitudes in China than the value observed in Cabañeros National Park (617 m a.s.l) in Spain, but comparable to that reported in a coniferous forest in Germany (Bavaria, 776 m a.s.l),where the carbonyls were formed via photooxidation of biogenic hydrocarbons (Müller et al., 2006).
For the coastal urban TW site, the mean HCHO and CH 3 CHO concentrations measured at TW in this study were higher than those in winter 1999/2000 (2.1 ± 1.3 ppbv and 0.89 ± 0.3 ppbv for HCHO and CH 3 CHO (mean ± s.d.), respectively) at the same site, which might be related to the annual and/or seasonal variations of primary emissions, secondary formation and regional impact (Guo et al., 2013b;Ling et al., 2014), while their levels were lower than those at Tung Chung (TC) (7.5 ± 1.9 ppbv and 4.9 ± 1.1 ppbv for HCHO and CH 3 CHO (mean ± 95% C.I.), respectively), a polluted site influenced by the air masses from inland PRD region and Hong Kong urban areas (Guo et al., 2009;Cheng et al., 2010).Furthermore, the HCHO and CH 3 CHO concentrations at TW were comparable to those at the coastal urban site in southwestern China, i.e., Qingzhou (2.5 ± 1.9 ppbv and 1.6 ± 2.2 ppbv for HCHO and CH 3 CHO (mean ± s.d.), respectively), but much lower than those in Kaohsiung, a coastal city in southern Taiwan.In addition, the C 1 -C 2 carbonyl concentrations measured at the coastal urban sites in Japan, Portugal and Sweden were in the ranges of 0.4-4.3ppbv for HCHO and 0.2-0.4ppbv for CH 3 CHO, much lower than the levels in China.Nevertheless, the concentrations of HCHO and CH 3 CHO at a given location were related to sampling and analytical conditions, meteorological parameters, geographic features, primary emission sources, secondary formation and levels of precursors (Ling et al., 2014).

Temporal Variations
Diurnal patterns provide further information on the variations of primary emissions and secondary formation of HCHO and CH 3 CHO.Fig. 2 shows the average diurnal patterns of HCHO and CH 3 CHO, together with wind speeds and wind directions during O 3 and non-O 3 episode days at the two sites.Please note each value at each hour was obtained by averaging the mixing ratios of the samples collected at the same time at both sites.For example, the average mixing ratios of HCHO and CH 3 CHO at 0000 LT on O 3 episode days at TMS were calculated based on all the 0000 LT samples of O 3 episode days at that site.On O 3 episode days, a clear shift from southerly winds during 0700-1600 LT to northerly winds at 1700-0600 LT was found at TW. Since the TW site was located in an urban area near the foot of the mountain and close to the South China Sea, this site was influenced by the combination of synoptic winds and mesoscale circulations, such as mountain-valley and sealand breezes.The wind shift was likely related to the dominant influence of mesoscale circulations as the synoptic winds were weak on O 3 episode days (Guo et al., 2013a).This also explained why there was no delayed peak of HCHO and CH 3 CHO observed at TMS on O 3 episode days.
HCHO presented similar diurnal patterns at the two sites, with the lowest mixing ratio at midnight and the highest value in the afternoon through a daytime buildup, followed by a slow decrease until midnight on both O 3 and non-O 3 episode days.These were consistent with the patterns of O 3 at the two sites, suggesting that photochemical formation had significant influence on the distributions of HCHO at both sites.Furthermore, similar value of HCHO observed at both sites (p > 0.05) regardless of different elevations of the sites indicated that the TMS site was also affected by regional transport/mesoscale circulations while the TW site was influenced by local emissions, apart from the influence of in-situ formation at both sites.
On the other hand, broad humps of CH 3 CHO were observed in the afternoon at TMS, and the difference of CH 3 CHO between daytime and nighttime hours was less significant than that of HCHO.For example, the average difference of CH 3 CHO on O 3 episode days was 0.31 ± 0.20 ppbv, lower than that of HCHO (0.81 ± 0.36 ppbv) on O 3 episode days.This feature was mainly related to the difference in photochemical reactivity of and/or the contributions of primary emissions to HCHO and CH 3 CHO (Guo et al., 2014).Moreover, as TMS was a mountain site and there were no anthropogenic emissions nearby, the buildup of HCHO and CH 3 CHO in the afternoon was more likely related to the insitu formation and the transport of air pollutants from urban areas at the foot of the mountain due to valley breeze, and/or from inland PRD under synoptic winds (Guo et al., 2013a;Lam et al., 2013).
The diurnal patterns of CH 3 CHO at TW showed no differences between daytime and nighttime concentrations during O 3 and non-O 3 episode days (p = 0.13 and 0.93, respectively), consistent with the patterns of acetone at that site (Guo et al., 2013b), indicating the complexity of CH 3 CHO sources at the urban site.In addition, different from HCHO, no obvious peaks were found for CH 3 CHO on O 3 and non-O 3 episode days, implying that there might be other factors that have concealed the CH 3 CHO peak at noon/in the afternoon caused by photochemical production.On the other hand, higher average mixing ratios of CH 3 CHO, together with large 95% interval confidence values, were observed at midnight and 0700 LT on non-O 3 episode days and at 1900 LT on O 3 episode days, suggesting the large variations of primary emissions of CH 3 CHO.The potential influence of primary emissions was examined using a linear regression analysis.The highest association was found between CH 3 CHO and i-butene (R 2 = 0.78, p < 0.05), followed by n-pentane (R 2 = 0.68, p < 0.05) and 1butene (R 2 = 0.41, p < 0.05) at the urban site.The moderate to good correlations of CH 3 CHO with 1/i-butenes and npentane suggested a significant contribution of vehicular emissions to the observed CH 3 CHO levels at the urban site (Guo et al., 2011a, b).

The Correlation and Ratios between HCHO and CH 3 CHO
To further investigate the potential sources, the correlation of HCHO with CH 3 CHO and the ratios of HCHO/CH 3 CHO during O 3 and non-O 3 episode days were examined at both sites (Fig. 3).Good correlations of HCHO with CH 3 CHO were found on O 3 and non-O 3 episode days at TMS, with the R 2 values of 0.66 and 0.61 (p < 0.05), respectively, indicating their common sources, i.e., photo-oxidation of hydrocarbons (Pang and Mu, 2006) and/or primary emissions from similar sources (Guo et al., 2013a).However, though good correlation between HCHO and CH 3 CHO was found at TW on O 3 episode days (R 2 = 0.63, p < 0.05), the correlation on non-O 3 episode days was poor.The poor correlation was mainly caused by the samples with relatively high CH 3 CHO values (> 2.3 ppbv) (totally 17 samples).The average mixing ratio of CH 3 CHO in the 17 samples was 3.7 ± 0.6 ppbv, 1.8 times that in all the samples on non-O 3 episode days.To investigate the potential sources of the high CH 3 CHO mixing ratios, a linear regression analysis of CH 3 CHO with other species in the 17 samples was performed.It was found that CH 3 CHO experienced some correlation with ethane, propane and n/ibutanes (R 2 = 0.30-0.36,p < 0.05), suggesting that emissions associated with ethane, propane and n/i-butanes were responsible for the high CH 3 CHO levels in the 17 samples.Overall, the above features further confirmed the different impacts of primary emissions and secondary formation on ambient HCHO and CH 3 CHO at the two sites.
The ratio of HCHO/CH 3 CHO has been used as an indicator of the contribution of photochemical conversion of biogenic hydrocarbons to carbonyls in previous studies, which reported the ratio of HCHO/CH 3 CHO varied from 1.0 to 2.0 in urban environments to ~10 in forested rural areas (Liu et al., 2006;Santarsiero and Fuselli, 2008).The higher ratio of HCHO/CH 3 CHO in the forested areas is because photooxidation of biogenic hydrocarbons, i.e., isoprene, yields more HCHO than CH 3 CHO.In this study, the average ratios of HCHO/CH 3 CHO at both sites were < 2, with the values of 1.32 ± 0.05 and 0.97 ± 0.05 (µg m -3 )/(µg m -3 ) at TMS and TW, respectively, consistent with the ratios in urban areas in previous studies (Feng et al., 2005;Moussa et al., 2006;Cerón et al., 2007).The results suggested that anthropogenic emissions, including primary emissions and the photo-oxidation of anthropogenic hydrocarbons likely played an important role in the secondary formation of carbonyls at both sites.In addition, the higher ratio of HCHO/CH 3 CHO at TMS than at TW (p < 0.05) might be related to the higher contribution of photochemical oxidation to carbonyls, including photochemical in-situ production and transport of photochemically-formed HCHO and CH 3 CHO in urban areas and/or during transit from urban areas to TMS (Section 3.2).To further investigate the origin of HCHO and CH 3 CHO, correlations of HCHO and CH 3 CHO with propionaldehyde (CH 3 CH 2 CHO) were explored since CH 3 CH 2 CHO was only related to anthropogenic emissions (Wang et al., 2010a, b).The correlation coefficients (R 2 ) of HCHO and CH 3 CHO were only 0.18 and 0.36 at TMS, respectively, while they were 0.36 and 0.40 at TW, respectively, further confirming the importance of secondary formation of HCHO and CH 3 CHO at both sites.On the other hand, the relatively lower correlations at TMS were due to the more aged air masses arriving at that site (Guo et al., 2013a).

Source Apportionments
Fig. 4 shows the explained variations of species in the identified sources extracted by the PMF model.
HCHO and CH 3 CHO fairly correlated with ethene, propane, propene, and to a lesser extent, ethane, ethyne and CO in the first factor.In addition, certain levels of benzene, toluene and ethylbenzene were presented in this factor.It is suggested that this factor was related to vehicular emissions as ethene, ethyne, propane and propene are typical tracers of vehicular exhaust in Hong Kong (Tsai et al., 2007;Ho et al., 2009;Guo et al., 2011a, b).Furthermore, the emissionbased measurements demonstrated that vehicular emission is also an important source of carbonyls (Ho et al., 2012;Dong et al., 2014).At TMS, 10% HCHO and 32% CH 3 CHO were associated with vehicular emissions, while they were 18% HCHO and 35% CH 3 CHO at TW.
The second factor showed high loadings of toluene, ethylbenzene, xylenes and C 2 Cl 4 .C 2 Cl 4 is often used as a dry cleaning solvent and degreasing agent, while the aromatic species are released from the solvent of paints, inks, sealant, varnish and thinner for architecture and decoration (Borbon et al., 2002;Liu et al., 2008), apart from vehicular emissions.As poor correlations were found among the aromatic species and other combustion tracers, this factor was identified as solvent usage.Previous studies have identified that carbonyls could be used as solvents in industrial processes and household products (Yuan et al., 2012;Chen et al., 2014).Nevertheless, minor contributions of source usage to ambient HCHO (2% at TMS and 4% at TW) and CH 3 CHO (3% at TMS and 4% at TW) were found in this factor.High percentage of O x was assigned to factor 3, indicating that this source was more related to photochemical processes.Indeed, certain amount of the relatively-unreactive species, i.e., C 2 Cl 4 , CH 3 Cl, ethyne, benzene and CO, were presented in this factor.After emitted from the sources and/or during the transport of air masses, VOCs with higher photochemical reactivity, i.e., alkenes and aromatics, would decay more rapidly, resulting in the survival of relatively-unreactive species and the enrichment of photochemical products, i.e., O 3 , HCHO and CH 3 CHO, in the aged air masses.Therefore, this factor was assigned as secondary formation at both sites (Li et al., 2010).This factor was responsible for 76% HCHO and 53% CH 3 CHO at TMS, and 60% and 45%, respectively, at TW.It should be noted that secondarilyformed HCHO and CH 3 CHO at a sampling site includes in-situ photochemical production and the transport of photochemically-formed HCHO and CH 3 CHO from the upwind areas (Li et al., 2010).
The fourth factor was assigned to biogenic emissions as the tracer isoprene dominated (87 and 85% for TMS and TW, respectively) this source at both sites.This factor accounted for 7% HCHO and 6% CH 3 CHO at TMS, and 13% and 12% at TW, respectively.
Factor 5 was characterized by high percentages of CH 3 Cl (52% and 46% for TMS and TW, respectively), with certain amount of ethane, ethene, ethyne, benzene and CO, consistent with the profile associated with biomass burning in Hong Kong (Guo et al., 2011b).Therefore, this factor was assigned to biomass burning.Indeed, our previous studies have confirmed that biomass burning was the source of VOCs at both sites (Guo et al., 2013b;Ling et al., 2015).This factor made insignificant contributions to HCHO (5% at TMS and TW, respectively) and CH 3 CHO (4% at TMS and 6% at TW).

In-situ Photochemical Formation
To estimate the contributions of in-situ formation, the secondary HCHO and CH 3 CHO formed from the in-situ photo-oxidation of their precursors were explored by the simulation of PBM-MCM model on all the sampling days at TMS and TW.The average mixing ratios of the in-situ phtotochemically-formed HCHO and CH 3 CHO at TMS were 1.37 ± 0.13 and 0.45 ± 0.03 ppbv, respectively, accounting for about 51 ± 5% and 32 ± 2% of the observed HCHO and CH 3 CHO, respectively, while the levels of in-situ secondarily formed HCHO and CH 3 CHO at TW were 1.61 ± 0.21 and 0.77 ± 0.06 ppbv, with the percentages of 55 ± 7% and 35 ± 3% of the observed HCHO and CH 3 CHO levels, respectively, consistent with the results at an urban site in Hong Kong in previous study (Cheng et al., 2010), in which the average proportions of secondary HCHO and CH 3 CHO production were about 50% and 28%, respectively, estimated based on the relationship between each individual carbonyl and CO.The higher levels of in-situ photochemically-formed HCHO and CH 3 CHO at TW were due to the higher values of hydrocarbons, which provided more precursors for the photochemical formation of HCHO and CH 3 CHO (Guo et al., 2013b;Ling et al., 2014).On the other hand, the lower in-situ secondarily-formed HCHO and CH 3 CHO levels at TMS were related to the lower levels of hydrocarbons (Ling et al., 2014).
By subtracting the modeled concentrations of in-situ HCHO and CH 3 CHO from the factor 3, the levels of transported photochemically-formed HCHO and CH 3 CHO were estimated.The mean concentrations of the transported photochemically-formed HCHO and CH 3 CHO at TMS were 0.67 ± 0.02 and 0.31 ± 0. 01 ppbv, respectively, accounting for about 25 ± 1% and 21 ± 1% of the observed HCHO and CH 3 CHO at TMS, respectively, much higher than those at TW (0.13 ± 0.01 (5 ± 1%) and 0.22 ± 0.05 ppbv (10 ± 2%) for HCHO and CH 3 CHO, respectively).This is not surprising because air masses at TMS were photochemically-aged and air pollutants could be brought to the mountainous site from the urban areas by mesoscale circulation and regional transport from the PRD (Ling et al., 2015).It is noteworthy that there were uncertainties for the above estimated contributions of photochemical formation of HCHO and CH 3 CHO, which were mainly generated from the uncertainties in the PMF-derived secondarily-formed HCHO and CH 3 CHO values and in the PBM-MCM simulated in-situ phtotochemically-formed HCHO and CH 3 CHO values.The uncertainty of the PMF model was mainly caused by data input and model itself, which was estimated through the bootstrapping analysis in the model.The uncertainty for factor 3, i.e., secondarily-formed HCHO and CH 3 CHO was 4-8% for the two sites.On the other hand, the input data for the PBM-MCM model had uncertainty in measurements, i.e., the analysis accuracy of each species, which would subsequently generate uncertainty for the outcomes.By estimation, the uncertainty from PBM-MCM simulation was about 7-17%.Besides, the inherent uncertainties of the PBM-MCM could also be induced by one or more of the following reasons: 1) the uncertainty associated with the deposition of HCHO and CH 3 CHO.As the actual dry deposition rates of species were not measured in this study, the deposition rates of HCHO and CH 3 CHO, which could influence their sink, were adopted from the parameterization of previous studies (Zhang et al., 2003;Li et al., 2014); and 2) The uptake of HCHO by aerosols (MacDonald et al., 2012;Li et al., 2014) and the influence of physical processes, including the vertical and horizontal dispersions.These factors were not considered in the PBM-MCM model (Lam et al., 2013;Lyu et al., 2015).

CONCLUSIONS
In this study, simultaneous measurements of air pollutants were carried out at different elevations of the highest mountain in Hong Kong, i.e., Mt.Tai Mo Shan.The spatial patterns showed no difference of HCHO levels between the urban site and the mountain site, while the average CH 3 CHO levels were higher at the urban site.Analysis of diurnal patterns and the variations of HCHO and CH 3 CHO on O 3 and non-O 3 episode days suggested the influence of secondary formation on the increment of HCHO and CH 3 CHO, while the correlation of HCHO with CH 3 CHO and the ratios of HCHO/CH 3 CHO revealed the different impacts of primary emissions and secondary formation on ambient HCHO and CH 3 CHO at the two sites.The PMF and PBM-MCM results showed that in-situ formation and vehicular emissions were the major contributors to both HCHO and CH 3 CHO at the urban site, while the in-situ formation and the transport of photochemically formed HCHO and CH 3 CHO were the most important source of HCHO and CH 3 CHO at TMS, followed by vehicular emissions, biogenic emissions, biomass burning and solvent usage.Overall, this study quantified the contributions of different sources to ambient HCHO and CH 3 CHO in Hong Kong, which could assist environmental policy makers in Hong Kong to devise appropriate control measures of HCHO, CH 3 CHO and their precursors.

Fig. 2 .
Fig. 2. Average diurnal patterns of HCHO and CH 3 CHO during O 3 and non-O 3 episode days at TMS and TW.

Fig. 3 .
Fig. 3. Correlation of HCHO with CH 3 CHO on O 3 and non-O 3 episode days at TMS and TW.

Fig. 4 .
Fig. 4. Explained variations of species in the identified sources extracted by the PMF model.The blue and green bars represent values at TMS and TW, respectively.

Table 1 .
Comparison of HCHO and CH 3 CHO with other studies.