Article Factors Contributing to Haze Pollution: Evidence from Macao, China Wai-Ming To 1 ID , Peter K. C. Lee 2,* ID and Chi To Ng 2 1 School of Business, Macao Polytechnic Institute, Macao, China; Energies 2017, 10, 1352 2 of 12 haze adversely affected people’s health physiologically and psychologically. Haze consists mostly of particulate matter and aerosols [8–12]. Jansen et al. [8] collected air samples in Hangzhou, China and found that the contribution of secondary inorganic aerosols to PM2.5 mass increased by around 50 percent during haze episodes. Li et al. [10] collected PM2.5 samples in Nanjing, China, during a severe haze-fog event. They reported that the mass concentrations of metals including Zn, As, Pb, Cd, Mo, and Cu were higher during haze-fog events than those in non-haze fog days. Some metals such as Zn, As, Pb, Cd, and Cu are found in fossil fuels [13] and traffic-related emissions [14]. Guo et al. [12] indicated that non-methane hydrocarbons are amongst the key ozone precursors, which can cause haze episodes in urban areas. Guo et al. [12] also noted that non-methane hydrocarbons may come from gasoline evaporation, solvent use, vehicular emission, and natural gas leakage, which are likely the same sources as the metals mentioned by Jansen et al. [8] and Li et al. [10]. Haze is defined as “an aggregation in the atmosphere of very fine, widely dispersed, solid or liquid particles, or both, giving the air an opalescent appearance that subdues colors” [15]. In cities, air pollutants from a diverse range of sources disperse in the atmosphere and merge into a featureless, uniform layer of haze that adversely affects visibility [15] and causes mild to moderate psychological stress [16]. The American Meteorological Organization describes haze as a mixture of aerosols and particles suspended in air, reducing visibility by scattering light while the World Meteorological Organization has yet to establish a classification of haze densities [17]. The Macao Meteorological and Geophysical Bureau (the Bureau) defines ‘haze’ as the condition under which visibility is equal to, or less than 5 km and relative humidity is less than 80 percent. These parameters are chosen because 2 Macao is a small coastal city with a land area of 30.4 km and the distance between the monitoring station and Macao’s border is only slightly larger than 5 km. The Bureau keeps daily records of the numbers of hours with haze. In the 1980s and 1990s, haze hours were recorded by professional meteorologists in the Bureau. Since the Millennium, the number of haze hours was determined by hourly readings from a forward scatter visibility meter and a relative humidity meter. The figures were confirmed by professional meteorologists in the Bureau. Specifically, it has been found that Macao was generally covered by haze almost continuously between September and March, year after year over the past decade [18]. In contrast to past research that had focused on the characterization of haze based on the chemical analysis of samples collected during haze events and the association of haze with the physical and psychological reactions of residents [7,10,11,16], this paper explores the extent of haze in Macao by using long-term time-series data. It also investigates the association between energy use, meteorological conditions, and the numbers of hours with haze in Macao. Understanding the direction and strength of association between energy use, meteorological conditions, and haze should shed light on possible solutions to environmental degradation. Macao’s Social and Economic Development 2 Macao was a small tourism city in Asia in the 1980s. It had a land area of about 16 km and a population of 0.3 million. After returning to China at the end of 1999, the Government of Macao decided to liberalize the gaming industry in 2002. Since then, casino operators have invested heavily to build mega-casinos, resorts, convention and exhibition centers, shopping malls, and theatres. In recent years, Macao has attracted over 30 million visitors a year. Macao’s population also increased from 2 0.43 million in 2000 to 0.64 million in 2016. Meanwhile, Macao’s area increased from 25.4 to 30.4 km 2 by land reclamation projects. Currently, Macao’s urban density is about 21,000 people per km (one of the highest in the world), and this figure increases by 50 percent due to the surge of visitors during weekends and holidays. Energies 2017, 10, 1352 3 of 12 2. Method 2.1. Data Sources All data of this study were official data obtained from government sources. Specifically, the annual energy consumption data were obtained from the Macao Statistics and Census Service [19,20]. The Macao Meteorological and Geophysical Bureau performs real-time measurements of air temperature, atmospheric pressure, and rainfall in its meteorological observation station located at the eastern part of the Grand Taipa Hill (i.e., the central location of Macao). Air temperature was recorded using a calibrated liquid-in-glass thermometer, atmospheric pressure was recorded using a calibrated mercury barometer, and rainfall was recorded using a high-precision rain gauge [21]. As mentioned above, the number of haze hours was determined by hourly readings from a forward scatter visibility meter and a relative humidity meter, and was confirmed by professional meteorologists in the Bureau. Daily records of the number of hours with haze, monthly mean atmospheric pressures, monthly mean air temperatures, and monthly total rainfall figures were obtained from the Macao Meteorological and Geophysical Bureau [22]. To explore whether haze is a cross-border environmental issue from the north, annual haze days in Guangzhou were collected from the online network of Guangzhou Meteorological Bureau [23]. 2.2. Stationarity of Time-Series A time-series can be modeled by an autoregressive (AR) model or its derivative such as an autoregressive moving average (ARMA) model when it is a stationary process. It is known that a stationary time series has a constant mean, a constant variance, and the autocovariance function between two points in the series only depending on the time interval. A typical time-series can be expressed as: yt = ρyt−1 + εt (1) where yt is the variable of interest, ρ is a coefficient, εt is the error, and t is the time index. When ρ = 1, unit root is present and the time-series is not stationary. When |ρ| < 1, the time-series is stationary. The Dickey-Fuller test is a very popular method to determine whether a time-series has a unit root. Sometimes, a time-series can have a trend and it is stationary after removing the trend. Hence, there are three main versions of the Dickey-Fuller test: i. Test for a unit root: ∆yt = βyt−1 + εt (2) where ∆ is the delta operator defined by ∆yt = yt − yt−1 and β is (ρ − 1). ii. Test for a unit root with a drift: ∆yt = a0 + βyt−1 + εt (3) where a0 is the drift. iii. Test for a unit root with a drift and a deterministic time trend: ∆yt = a0 + a1t + βyt−1 + εt (4) where a1 is coefficient of the time trend. Each version of the Dickey-Fuller test has its own critical tau value [24], which depends on the sample size. In each case, the null hypothesis is that there is a unit root and the time series is not stationary, while the alternative hypothesis is that the time series is stationary. In this paper, only the versions (ii) and (iii) of the Dickey-Fuller test were performed because the test version (ii) covered the test version (i). Energies 2017, 10, 1352 4 of 12 2.3. Association between Variables 2.3. Association between Variables The association between two variables was characterized by Pearson’s correlation coefficient. The cToheeffiacsiseonct,i art, iwonasb ceatwlcueelantetdw aos variables was characterized by Pearson’s correlation coefficient. The coefficient, r, was calculated as n n � xi � x� � yi � y� r r =� √ n i∑=i� 1(xi −2x√)(yni − y) 2 (5(5) )  i�∑=n11� (xxii �−xx� )2 i∑=� n11(� yi −� yy� )2 where x and y are values of the variables, and n is the size of dataset. where x and y are values of the variables, and n is the size of dataset. 2.4. Regression Analysis Multiple regression analysis was employed to assess the relationships between the independent variables and the dependent variable. The accuracy of the identiffiied model was determined by the coeffificient of determination, r2. TThhee ccooeeffffiicient provides a measure of how well the collected data are predicted by the modeell [[2255]].. It was calculated by: r r22=� S SSSSRTTR=� 11−� SSTE (6(6) ) where SSR is the sum of squared regression, SST is the sum of squared total, and SSE is the sum of squared error.. 3. Results and Analysis 3.1. Haze and Meteorological Conditions Visibiliittyy iiss aann inintteeggrraatitviveep paararammeteetrert hthatart erfeefresrtso toth tehaeb ailbitilyittyo tsoe eseaed ai sdtaisnttaonbt joecbtje[c2t5 ][.25W].h Wenhaeinr paiorl lpuotallnuttsaanrtes eamreit teemdiftrtoemd farnotmhr oapnothgreonpicosgoeunrice so, huarczesi,s hfoarzme eisd fuonrdmeerda sutnabdleera atm stoasbplhee articminosvpehrseiroinc lianyveerr,saiodnv elarsyeelry, adffveectrisneglyt hafefevcitsiinbgil itthyeo vviesirbtihlietyc iotyv.eHr tahzee chitays. bHeaeznea hcaosm bemenon a pcohmenmomone npohneninomweintoenr in Cwhinintear’s inci tCiehsin[3a,’8s, 9c,i1t2ie,1s 8[,32,68],.9,I1n2M,18a,c2a6o]., Ihna zMeaiscadoe,fi hnaezde ais tdheeficnoendd aitsi otnheu cnodnedriwtiohnic hunvdiseirb iwlihtyicihs evqisuibailltioty, oisr eleqsusatlh taon, o5rk lmessa nthdarne 5la ktimve ahnudm riedlaittyivies hleusms tihdaitny8 i0s pleesrsc ethnat.nT 8h0e pceorncdenitti.o Tnhies rceocnodrditeiodnb iys trreacionreddedm ebtye otraoilnoegdis tms aetnedorcoolnofigrimstse danbdy caofnofriwrmaredd sbcyat ate frovriwsiabridlit sycmateteter rv. iTshibeilMitya cmaoetMere. tTeohreo Mlogaiccaaol aMnedteGoerolpohgyicsaicl aalnBdu Greeaouphkyeespicsald Bauilryeraeuc okredeps so df tahileyn ruecmorbdesr of thhoeu nrusmwbitehr hoaf zheo.uFrisg wuriteh1 haaszheo. wFisguthre r1eac sohrdows fso trhteh erepceoridosd f1o9r 8t6heto p2e0ri1o6d. 1986 to 2016. 800 600 400 200 0 1986 1991 1996 2001 2006 2011 2016 Year (a) (b) Figure 1. (a) The number of hours with haze per year for the period 1986 to 2016; (b) The number of Figure 1. (a) The number of hours with haze per year for the period 1986 to 2016; (b) The number of hours with haze per month for the period 19986 ttoo 22001166.. Figure 1a shows that the number of hours with haze increased steadily from 0 in 1990 to 297 in 2002. However, it sharply increased to 694 in 2003. Between 2003 and 2008, the number of hours with Haze (hours/year) Energies 2017, 10, 1352 5 of 12 Energies 2017, 10, 1352 5 of 12 haze ranged from 536 and 766. The figure has decreased since 2008. Figure 1b shows the monthly numbFeigrsu roef 1haosuhrosw wsitha thtahze.n Numotbee trhoaft hhoauzres hwaisth bheaezne ai nccormeamseodns tpehaedniloymfreonmon0 in 1a9u9tu0mton2s9 a7nidn 2w0i0n2t.erHs,o iw.ee.,v beer,twit esehna rSpelpytienmcrbeears eadndto M6a9r4cihn, 2d0u0r3in. gB etthwe epeanst2 0fo0u3ratneedn2 y0e0a8r,st.h e number of hours with hazeFriagnugred 2far–ocm sh5o3w6 a mndon7t6h6l.y Tmheafing uatrme ohsapshdeercicre pasresdsusirnec, em2o0n0t8h. lFyi gmueraen1 baisrh toewmspethraetumroen, tahnldy nmuomntbhelrys toofthaol urarisnwfailtlh fhigauzree.sN fotre ththea pt heraizoedh 1a9s8b6e teon 2a0c1o6m. Tmhoensep fhigenuorems esnhoonwi nthaautt uMmancasoa nhdasw hiingtheresr, ia.tem., obseptwhereinc Sperpestesumrbee, rloawnedrM aiarr tcehm, dpueraintugrteh, eapnads ltefsosuert reaeinnfyaelal rbse.tween October and March i.e., in the aFuitguumren a2nad– cwsinhtoewr mmonotnhtsh.l y mean atmospheric pressure, monthly mean air temperature, and monthly total rainfall figures for the period 1986 to 2016. These figures show that Macao has higher atmospheric pressure, lower air temperature, and lesser rainfall between October and March i.e., in the autumn and winter months. (a) (b) (c) Figure 2. (a) Mean atmospheric pressure; (b) Mean air temperature from; and (c) Total rainfall figures over the period 1986 to 2016. Figure 2. (a) Mean atmospheric pressure; (b) Mean air temperature from; and (c) Total rainfall figures 3.2. Energy Use in Macao over the period 1986 to 2016. Macao’s social and economic development had remained relatively steady before the 3li.b2e. rEanliezragtyioUns oefi nthMe gacaamoing industry in 2002. Table 1 summarizes the consumption of primary energy sources in 1986, 1991, 1996, 2001, 2006, 2011, and 2016, respectively. It also shows the consumption of Macao’s social and economic development had remained relatively steady before the liberalization electricity (1 GWh is equivalent to 3.6 × 10 TJ)—a secondary energy in these years. of the gaming industry in 2002. Table 1 summarizes the consumption of primary energy sources in Figure 3 shows the primary energy use figures for Macao over the period 1986 to 2016. This 1986, 1991, 1996, 2001, 2006, 2011, and 2016, respectively. It also shows the consumption of electricity figure shows that Macao’s total primary energy use had been increasing continuously from 1986 to (1 GWh is equivalent to 3.6 × 10 TJ)—a secondary energy in these years. 2005. The consumption of oil products such as gas oil/diesel and fuel oil has decreased quite Figure 3 shows the primary energy use figures for Macao over the period 1986 to 2016. This figure significantly since 2005. This was because Macao’s power companies had imported more electricity shows that Macao’s total primary energy use had been increasing continuously from 1986 to 2005. from mainland China, thus burning less fuel oil and gas oil/diesel since 2005. The percentage of The consumption of oil products such as gas oil/diesel and fuel oil has decreased quite significantly imported electricity increased from 15 percent in 2005 to 81 percent in 2016 [20]. since 2005. This was because Macao’s power companies had imported more electricity from mainland China, thus burning less fuel oil and gas oil/diesel since 2005. The percentage of imported electricity increased from 15 percent in 2005 to 81 percent in 2016 [20]. Energies 2017, 10, 1352 6 of 12 Energies 2017, 10, 1352 6 of 12 Table 1. Import of fossil fuels (primary source of energy) in 1986, 1991, 1996, 2001, 2006, 2011, and 2016. Table 1. Import of fossil fuels (primary source of energy) in 1986, 1991, 1996, 2001, 2006, 2011, and 2016. Primary Energy in TJ Year Motor Gas oil & Liquefied Natural Aviation Fueloil PrimKaerryoEsenneerg y in TJ 1 Gasoline Diesel Petroleum Gas Gas Kerosene Year Motor Gas oil & Liquefied Natural Aviation 1986 6 4878 560 53 358 0 0 Fueloil Kerosene 1 Gasoline Diesel Petroleum Gas Gas Kerosene 191 748 4240 8438 417 561 0 0 19896 16263 348978 105,686402 5363 39582 0 20507 1290901 173480 34826490 128,453089 4147 151617 0 3803 1996 1263 3498 10,842 363 952 0 2057 206 17 6195 12,427 295 148 0 8769 2001 1340 3869 12,509 447 1117 0 3803 201 2659 5787 3864 190 1954 2873 603 2006 1777 6195 12,427 295 1488 0 8769 2016 329 4526 6270 95 2032 393 10,258 2011 2659 5787 3864 190 1954 2873 6003 Secondary Energy (i.e., Electricity in GWh) 2Y0e1a6r 3329 4526 6270 95 2032 393 10,258 2 3 Net Imported Electricity Net Electricity Produced Locally Electricity Consumed Secondary Energy (i.e., Electricity in GWh) 1986 42 489 531 Year 2 3 1991 Net Import9e5d Electricity Net Electricity P75ro6 duced Locally Electricit8y6C0 onsumed 196 2 126 1348 1986 42 489 531 1290901 2495 173563 18602 1290906 924262 1415296 2143248 201 3126459 1639513 31865062 2016 439046 1948589 524924 2011 3165 691 3856 1 2 The Macao International Airport was opened for commercial operations in November 1995. Macao 2016 4306 988 5294 3 has started to rely more heavily on imported electricity since 2007. Macao’s oil-fired power plants 1 2 The Macao International Airport was opened for commercial operations in November 1995. Macao has started primarily used fuel oil (93 percent) and diesel o3il (7 percent) to generate electricity before 2007. to rely more heavily on imported electricity since 2007. Macao’s oil-fired power plants primarily used fuel oil (M93apcearoc’esn pt)oawnderd iceoseml opial n(7yp setracerntet)dt ougsiennger afuteeel loecitlr aicnitdy bneaftourera2l0 0g7a.sM taoc agoe’nseproawtee recloemctpriacniytys tianr t2e0d0u8s ibnugtf uitesl oil and natural gas to generate electricity in 2008 but its natural gas supplier was not able to supply natural gas to natural gas supplier was not able to supply natural gas to Macao’s power plants in 2012, 2013, 2015, Macao’s power plants in 2012, 2013, 2015, and 2016. Macao’s oil-fired power plant efficiency was about 41 percent and 2016. Macao’s oil-fired power plant efficiency was about 41 percent while Zhuhai’s coal-fired while Zhuhai’s coal-fired power plant efficiency was about 40 percent [27]. power plant efficiency was about 40 percent [27]. 35000 30000 25000 Aviation Kero Natural Gas 20000 LPG 15000 Kerosene (local) 10000 Fuel Oil 5000 Gas oil/diesel Gasoline 0 1986 1991 1996 2001 2006 2011 2016 Year Figure 3. The consumption of primary energy in Macao from 1986 to 2016. Figure 3. The consumption of primary energy in Macao from 1986 to 2016. More specifically, gasoline is consumed primarily by private vehicles, gas oil/diesel by buses, coaches, trucks, industrial boilers and Macao’s power plants, fuel oil by jetfoils and Macao’s power More specifically, gasoline is consumed primarily by private vehicles, gas oil/diesel by buses, plants, kerosene by industrial companies, liquefied petroleum gas by domestic cookers and hot water coaches, trucks, industrial boilers and Macao’s power plants, fuel oil by jetfoils and Macao’s power burners, and aviation kerosene by aircrafts (after the opening of Macao International Airport in 1995). plants, kerosene by industrial companies, liquefied petroleum gas by domestic cookers and hot water Macao’s power company has used natural gas as a fuel since 2008. burners, and aviation kerosene by aircrafts (after the opening of Macao International Airport in 1995). Macao’s power company has used natural gas as a fuel since 2008. 3.3. Stationarity of the Annual Haze Time-Series Data and the Annaul Energy Consumption Data 3.3. STtahteioannanriutya lofh tahzee Atinmneu-asle Hriaezsea Tnidmteh-Seearinens uDaalttao atnald pthriem Aanrnyaeunl eErngeyrgcyo nCsounmsupmtipotniotni mDea-tsae ries were examined using the Dickey-Fuller tests based on Equations (3) and (4). Table 2 shows the results of The annual haze time-series and the annual total primary energy consumption time-series were the Dickey-Fuller tests on the null hypothesis that the annual haze time-series and the annual energy examined using the Dickey-Fuller tests based on Equations (3) and (4). Table 2 shows the results of consumption time-series had unit roots. As each of the calculated tau values was greater than the the Dickey-Fuller tests on the null hypothesis that the annual haze time-series and the annual energy critical value in the table of critical values [24], the null hypothesis was rejected, and two time-series consumption time-series had unit roots. As each of the calculated tau values was greater than the Energy consumption (TJ) Energies 2017, 10, 1352 7 of 12 were found to be non-stationary. As the annual haze time-series and the annual total primary energy consumption time-series were found to be non-stationary, AR or ARMA models cannot be reliably used for these non-stationary time series. Table 2. Results of Dickey-Fuller tests. Annual Total Primary Energy Annual Haze Time-Series Consumption Time-Series Critical/Tau Values Constant, Constant, Constant, No Constant, No Trend with Trend Trend with Trend Calculated tau value −1.693 −1.256 −1.351 −1.081 Significant at −3.75 −4.38 −3.75 −4.38 0.01 level For a sample size of 25 Significant at −3.00 −3.60 −3.00 −3.60 Critical 0.05 level value [24]: Significant at −3.58 −4.15 −3.58 −4.15 0.01 level For a sample size of 50 Significant at −2.93 −3.50 −2.93 −3.50 0.05 level Null hypothesis-stationary rejected rejected rejected rejected Note: n = 31 in these cases. 3.4. Correlation and Multiple Regression Analysis Bivariate correlations were performed between the number of hours with haze, the consumption of motor gasoline, gas oil/diesel, fuel oil, kerosene, liquefied petroleum gas, natural gas, and aviation kerosene, and meteorological variables including annual mean atmospheric pressure in hPa, annual ◦ mean air temperature in C, and annual total rainfall in mm using the annual data from 1986 to 2016. The calculated Pearson’s correlation coefficients are shown in Table 3. Although one might expect years with more rainfall days or more total rainfall to be cleaner years due to scavenging effects, the results show that the number of haze hours per year was not significantly correlated with annual total rainfall (r = 0.08, n.s.). It could be explained by the emissions of pollutants that had a much greater significant effect on haze than that of the cleansing mechanism on an annual basis. Table 3. Correlations between haze, energy consumption, and meteorological conditions. Energy Use (in TJ) Motor Gas Liquefied Natural Aviation Fuel Oil Kerosene Gasoline Oil/Diesel Petroleum Gas Gas Gasoline Haze 0.38 * 0.73 ** 0.17 −0.33 0.59 ** 0.34 0.66 ** Annual Meteorological Conditions Annual Mean Annual Mean Air Temperature Annual Total Rainfall Atmospheric Pressure Haze −0.10 −0.35 0.08 Note: n = 31 (i.e., 31 pairs). * p < 0.05. ** p < 0.001. The correlation values show that there were significant positive relationships between haze and some oil products such as motor gasoline, gas oil/diesel, and liquefied petroleum gas consumed locally, and between haze and aviation kerosene consumption that was partly consumed in Macao during aircrafts’ taking off and landing. A multiple regression analysis was then employed to determine the effects of primary energy consumption on the number of hours with haze. All fuel consumptions were first input as the independent variables, while the numbers of hours with haze was selected as the Energies 2017, 10, 1352 8 of 12 Energies 2017, 10, 1352 8 of 12 Haze = 0.043×GasOil + 0.042 × FuelOil + 0.076 × NG + 0.033× Kerosene − 497 aviation dependent variable. A backward procedure was used in which variables with a p value less than 0.05 5.49 5.48 3.81 4.73 − 4.47 (7) were retained. After several iterations, the following multiple regression equation was obtained: [0.000] [0.000] [0.000] [0.000] [0.000] Haze = 0.043 × GasOil + 0.042 × FuelOil + 0.076 × NG + 0.033 × Keroseneaviation − 497 where Haze is the number of hours with haze, GasOil is the consumption of gas oil/diesel in TJ, FuelOil 5.49 5.48 3.81 4.73 − 4.47 (7) is the consumption of fuel oil in TJ, NG is the consumption of natural gas in TJ, and Keroseneaviation is [0.000] [0.000] [0.000] [0.000] [0.000] the consumption of aviation kerosene in TJ. Figures underneath the coefficients are values for t- wstahteirsetiHcsa azendis ftihgeunreusm inb ebrroafckheotus rasrwe pit-hvahlauzees,.G asOil is the consumption of gas oil/diesel in TJ, FuelOil 2 is theTchoen sru-mvaplutieo nofo fEfquueal toioilnin (7T)J ,wNaGs 0is.7t4h1e, cio.en.,s u7m4.1p tpioenrcoefnnt aotfu trhael gvaasriannTcJe, ainn dthKee rnousemnbeaevria toiofn hisouthres cwointhsu hmazpeti oconuolfda bveia etixopnlakinereods ebnye tihneT cJo. nFsiguumrepstiuond oefr ngeaast hoitlh/deiecsoeelf,fi fcuieln tosila, rneavtualruael sgfaosr, ta-s twateisllt iacs anvidatfiiognu rkeserionsbernaec.k Tethsea relpa-tvioanlusehsip. s between these three variables and haze were significant at 2 p < 0.T00h1e. rTh-vea cloueffoifciEeqnutsa itniodnic(a7t)ewd athse0 r.7e4la1t,ivi.e .i,n7f4lu.1epnceerc oefn tthoef itnhdeevpaernidanencet vinartihaeblneus monb edreopfenhdouenrst wvairthiabhlaez. eInc oouthlderb eweoxrdpsla, ian eudnibt yinthcreeacsoen sinu mthpet icoonnosuf mgapstioinl/ odfi egsaesl ,ofiul,e lfuoeil, onial,t unraatul rgaals ,gas, wanedll avs iaavtiioanti oknerkoesreonsee nine .TTJ hinecrelaasteiodn tshhei pnsubmebtwere oefn htohuesres wthirtehe hvaazreia bbyle 0s.0a4n3d, 0h.a0z4e2,w 0e.0r7e6s, iagndifi 0c.a0n3t3 aht pe<r 0y.e0a0r1,. rTehspeeccoteivffieclyie. nTths ein 9d5i cpaetercdetnhte croenlafitdiveencine flinuteenrcvealosf ftohre tihnedseep ceonedfefinctievnatrsi awbelerse o[0n.0d2e7p, e0n.0d5e8n]t, v[0a.r0i2a6b,l e0.0I5n8o]t, h[0e.r0w35o,r 0d.s1,1a7]u, naint din [c0r.e0a1s9e, i0n.0t4h8e]c, ornespuemctpitvieolny.o f gas oil, fuel oil, natural gas, and aviation kerosFeingeurien 4T cJoinmcpreaareses dthteh neunmumbebresr oof fhhoouursr swwitihth hahzaez eusbiyng0 .E0q4u3,a0ti.o0n42 (,70) .a0g7a6i,nasnt dac0tu.0a3l3 dhatpae. rIty aelsaor, rpersopveicdteivse tlhy.e Tbhaend95 opf eprrceednitccteodn fivdaelunecse uinstienrgv athlsef 9o5r tpheersceencot ecfofincfiiednetns cwe eirnete[0rv.0a2l7s, o0f. 0t5h8e] ,co[0e.f0f2ic6i,e0n.t0s5.8 ], [0.035, 0.117], and [0.019, 0.048], respectively. 1000 Figure 4 compares the numbers of hours with haze using Equation (7) against actual data. It also provides the ba90n0d of predicted values using the 95 percent confidence intervals of the coefficients. 800 700 Upper bound 600 500 Predicted 400 300 Actual 200 100 Lower bound 0 1986 1991 1996 2001 2006 2011 2016 Year Figure 4. Variations in the actual and predicted number of hours of haze in Macao over the period 1986–2016. 3F.5i.g Aurseso4c. iVatairoinatsi obnestwineetnh eHaacztue aalnadn dMpertedoricotleodgincaulm Vbaerriaobflheso uonrs ao fMhoanzethilny MBacsiaso over the period 1986–2016. Figure 2a–c shows monthly mean atmospheric pressure, monthly mean air temperature, and 3.5. Associations between Haze and Meteorological Variables on a Monthly Basis monthly total rainfall figures between 1986 and 2016. Pair-wise correlations were performed between the nFuimgubreer o2fa h–cousrhso wwisth mhaoznet halnyd meaecahn ofa tmheo mspehteorircolporgeicssaul rvea,rimabolenst holny am meoanthaliyr btaesmisp feorra etuacreh, ayneadr mforn ththlye tpoetarilorda in1f9a8l6l –fi2g0u1r6e. s Rbestweaerecnhe1r9s 86havned u20se1d6. pPair-wise correlations wtoe reidpeenrtfiofyrm tehde breeltawtieoennshthipesn buemtwbeerno fphoour rvsiswibitihlithya azneda nmdoenathchlyo mf tehteomroeltoegoircoallo cgoicnadlivtiaorniasb [l2e5s].o Tnaablme o4n sthholywbs atshies fvoarlueeasc hofy peairr-fworisthe ecoprerreiloadtio1n9 8b6e–t2w0e1e6n. Rhaeszea arnchde eraschha mveetuesoerdolpoagiirc-awl ivsaerciaobrlre.l aIt iwonas tfouidnedn tthifayt the rneulamtiboenrs ohfi phsouberstw weiethn hpaozoer wviassib piloistyitiavnedlym, sotrnotnhglylym, aentedo srioglnoigficalnctloyn adsistioocniast[e2d5 ]w. iTtahb mleo4nsthloyw msetahne vatamluoesspohferpiac ipr-rwesisuerceo irnr ethlaet ipoanstb deetwcaedeen whhaizle iat nwdaes ancehgamtievteeloyr aonlodg wicaeal kvlayr iaasbsolec.iaIttewd awsitfho umnodnthlayt tmheannu amirb termopfehraotuurrse wanitdh mhoaznethwlya tsoptaols riatiivnefalyll, sitnrocen g1l9y9,6a. nItd iss ibgenciafiucsaen tthlye caosrsroecliaattieodn,w r-ivthalmueo, nmthulsyt mbee garneatemr othspanh e0r.i4c9p7 rfeosrs pu r<e 0i.n05t hfoerp 1a0s dt edgerceaedse owf fhreileediot mw.a Isnn oetghaetri vweolyrdasn, dthwe enaukmlybears soof chiaotuerds with mhaoznet hmlyighmte baen aasisrotceimatpeder watiuthre haingdh mlevoenltsh olyf mtoetalnr atinmfoasllpshienrciec 1p9r9e6ss.uIrtei,s wbhecicahu soefttehne lceoardrse ltaot itohne, rlo-vwaelur em, imxeudst labyeegrr heeaitgehr t h[2a8n].0 R.4e9s7eaforcrhpe<rs0 [.2095,3f0o]r h1a0vdee aglrseoe rseopfofrrteeedd tohmat. lIonwoetrh merewano radirs ,tethmepneuramtubreer omfahyo ruerdsuwceit hthhea azme bmieignht tvbeentaislastoicoina.t eHdewncieth, hhaizgeh sleevmels otof mbee manoaretm porospnhouernicepdr edsusurirneg, wthhei cahutoufmten and winter months from September to March in Macao. Haze (hours/year) Energies 2017, 10, 1352 9 of 12 Energies 2017, 10, 1352 9 of 12 Table 4. Pair-wise correlations between haze and meteorological conditions. leads to the lower mixed layer height [28]. Researchers [29,30] have also reported that lower mean air Correlation between Haze Correlation between Correlation between temperature may reduce the ambient ventilation. Hence, haze seems to be more pronounced during Year and Monthly Mean Haze and Monthly Haze and Monthly the autumn and winter months from September to March in Macao. Atmospheric Pressure Mean Air Temperature Total Rainfall 1986 −0.47 −0.47 −0.11 Table 4. Pair-wise correlations between haze and meteorological conditions. 191 −0.24 0.2 0.24 1996 0.10 −0.33 −0.35 2001 Correlation0b.3e1tw een Haze Correlat−i0o.n15b etween Correl−a0t.i3o1n between Year and Monthly Mean Haze and Monthly Haze and Monthly 2006 0.70 ** −0.43 −0.75 ** Atmospheric Pressure Mean Air Temperature Total Rainfall 2011 0.79 ** −0.82 ** −0.63 * 19826016 −0.05.14 7* −−00..4437 −0−.302. 11 1991 −0.24 0.22 0.24 Note: n = 12. * p< 0.05. ** p< 0.01. 1996 0.10 −0.33 −0.35 2001 0.31 −0.15 −0.31 3.6. Association between Haze in Macao and Haze Days in Guangzhou 2006 0.70 ** −0.43 −0.75 ** 2011 0.79 ** −0.82 ** −0.63 * Macao is located on the southwestern side of the Pearl River Delta while Hong Kong is at the 2016 0.51 * −0.43 −0.32 southeastern side. Whenever haze is reported, Macao (or Hong Kong) people claim that most of the Note: n = 12. * p < 0.05. ** p < 0.01. air pollutants come from the north across the border [31]. Although Guangzhou, Zhongshan, and Zhuhai are located on the north of Macao, only the Meteorological Bureau in Guangzhou has been 3.6. Association between Haze in Macao and Haze Days in Guangzhou recording the annual numbers of haze days for more than three decades, while there were no official data on haze days/hours published by the local meteorological stations in Zhongshan and Zhuhai. Macao is located on the southwestern side of the Pearl River Delta while Hong Kong is at the Figure 5 shows the number of haze days per year in Guangzhou for the period 1986 to 2016. A southeastern side. Whenever haze is reported, Macao (or Hong Kong) people claim that most of the air comparison of the total number of hours with haze in Macao, shown in Figure 1a, and the number of pollutants come from the north across the border [31]. Although Guangzhou, Zhongshan, and Zhuhai haze days in Guangzhou shown in Figure 5 demonstrates that changes in haze conditions in these are located on the north of Macao, only the Meteorological Bureau in Guangzhou has been recording two cities are very different. A correlation analysis was performed to test whether the haze condition the annual numbers of haze days for more than three decades, while there were no official data on haze in Macao was associated with haze condition in Guangzhou. The result showed that there was no days/hours published by the local meteorological stations in Zhongshan and Zhuhai. Figure 5 shows significant relationship (r = −0.06, p > 0.05) between haze in Guangzhou and haze in Macao in the long the number of haze days per year in Guangzhou for the period 1986 to 2016. A comparison of the total run, implying that haze in Macao is primarily caused by local emissions rather than cross-border air number of hours with haze in Macao, shown in Figure 1a, and the number of haze days in Guangzhou pollution from the north. Besides, haze needs relatively still air—so haze is not likely to be shown in Figure 5 demonstrates that changes in haze conditions in these two cities are very different. transported from one place to another. A correlation analysis was performed to test whether the haze condition in Macao was associated with haze condition in Guangzhou. The result showed that there was no significant relationship (r = −0.06, 250 p > 0.05) between haze in Guangzhou and haze in Macao in the long run, implying that haze in Macao is primarily caused by local emissions rather than cross-border air pollution from the north. Besides, 200 haze needs relatively still air—so haze is not likely to be transported from one place to another. 150 100 50 0 1986 1991 1996 2001 2006 2011 2016 Year Figure 5. The numbers of haze days in Guangzhou over the period 1986–2016. 4. Discussion Haze often appears in media headlines. This paper is one of the first publications to employ long-term timeF siegruirees 5d. aTthae inu imnvbersstiogfahtianzge dthayes rienlaGtuioannsghzihpo ubeotvweretehne tpheer iondum19b86e–r2 o0f1 6h.ours with haze (i.e., the integral form of air pollutants discharged from a wide range of sources), the consumption of 4fo. sDsiils fcuuesls einoenrgy, and the meteorological variables in Macao on an annual basis. The number of hours with haze increased continuously from 0 to 766 h a year between 1990 and 2007 while the Haze often appears in media headlines. This paper is one of the first publications to employ consumption of primary energy (including oil products, natural gas, and aviation kerosene) had long-term time series data in investigating the relationship between the number of hours with haze increased by 130 percent. Multiple regression analyses have shown that gas oil/diesel, fuel oil, and Haze (days/year) Energies 2017, 10, 1352 10 of 12 (i.e., the integral form of air pollutants discharged from a wide range of sources), the consumption of fossil fuel energy, and the meteorological variables in Macao on an annual basis. The number of hours with haze increased continuously from 0 to 766 h a year between 1990 and 2007 while the consumption of primary energy (including oil products, natural gas, and aviation kerosene) had increased by 130 percent. Multiple regression analyses have shown that gas oil/diesel, fuel oil, and natural gas consumed locally and the burning of aviation kerosene were significantly associated with the number of hours per year with haze in Macao. A correlation analysis between monthly haze data and monthly records of meteorological variables has indicated that the number of hours with haze was strongly, significantly associated with high atmospheric pressure that leads to the lower mixed layer height in the past decade. The combustion of oil products produces particulate matter, sulfur oxides, nitrogen oxides, and volatile organic compounds including non-methane hydrocarbons. All of these air pollutants are well-known sources of haze. Besides, around 150 commercial aircrafts, which emit a significant amount of particulate matter in low troposphere, take off and land at the Macao International Airport each day [32]. It is possible that there is a significant correlation between the number of humid hours (the hours with a relative humidity between 75 and 80 percent) each year/month and the number of haze hours per year/month. However, the hourly relative humidity data were not available from the Bureau’s publications. Since 2005, Macao’s power company has been importing more and more electricity from Zhuhai, and has reduced the burning of fuel oil and gas oil/diesel at its power plants significantly. As a result, the amount of oil products consumed locally decreased continuously from 2005 to 2014. The number of haze hours also changed from 567 to 346 per year during the same period. Macao’s power company has announced that it would import more electricity from Zhuhai and reduce the operational capability of its local power plants further, implying the haze condition of Macao will continue to improve over the near future. Nevertheless, the imported electricity from Zhuhai is produced by a coal-fired power plant at Jinwan—located at 50 km southwest of Macao. Since Zhuhai Jinwan Power Plant is on the Chinese mainland where the pollution regulations, compliance, and monitoring are not likely to be as stringent as in Macao, the increases in its power production will emit larger quantities of carbon dioxide, sulfur dioxides, nitrogen oxides, and particulate matter affecting its vicinity. Moreover, the number of aircrafts departing from and arriving at the Macao International Airport will very likely increase, leading to an increase in the consumption of aviation kerosene. As a result, it is very likely that the number of hours with haze will decrease in the short-run due to the continual decrease in the burning of oil products in Macao’s power plants. The anticipated future increase in local pollutant emissions and the haze will most likely be due to the Zhuhai Jinwan Power Plant. The implications are twofold. First, the Macao Government, or governments sharing a similar context, should take a proactive approach to consider the impact of energy mix on environmental conditions. Second, the transfer of environmental burdens (e.g., power generation) to neighboring areas needs to be re-examined as an environmental policy because it runs against environmental ethics and does not help in reducing the actual pollution. 5. Conclusions Haze is a major concern in many cities. In Macao, the number of hours with haze increased from 0 to 766 hours a year between 1990 and 2007 and decreased quite dramatically to 57 h in 2016. That provided an opportunity to explore what caused the change in hazy weather condition. During the same period of time, Macao’s energy mix also changed quite dramatically. Specifically, Macao started importing a significant percentage of electricity from its neighboring city and cut down the burning of fuel oil and gas oil/diesel substantially in its power plant. The results of multiple regression analysis indicated that gas oil/diesel, fuel oil, and natural gas consumed locally and the burning of aviation kerosene were significantly associated with the number of hours per year with haze in Macao.