Multiscale periodicities in aerosol optical depth over India
S. Ramachandran, Sayantan Ghosh, Amit Verma, Prasanta K. Panigrahi
aa r X i v : . [ phy s i c s . a o - ph ] M a r Multiscale periodicities in aerosol optical depth overIndia
S. Ramachandran , Sayantan Ghosh , , † , Amit Verma and P.K. Panigrahi Physical Research Laboratory, Navrangpura, Ahmedabad 380009, India School of Chemistry and Physics, University of KwaZulu-Natal, Durban 4000,South Africa Dept. of Physical Sciences, Indian Institute of Science Education and ResearchKolkata, Mohanpur 741252, India † Currently at SUPA, School of Physics and Astronomy, University of St Andrews,North Haugh, St Andrews KY16 9SS, Fife, UK Department of Computer and Information Science and Engineering, University ofFlorida, Gainesville, Florida 32611, USAE-mail: [email protected]
Abstract.
Aerosols exhibit periodic or cyclic variations depending on natural andanthropogenic sources over a region, which can get modulated by synopticmeteorological parameters such as winds, rainfall and relative humidity, and long-range transport. Information on periodicity and phase in aerosol properties assumessignificance in prediction as well as to examine the radiative and climate effectsof aerosols including its association with changes in cloud properties and rainfall.Periodicity in aerosol optical depth, which is a columnar measure of aerosoldistribution, is determined using continuous wavelet transform over 35 locations(capitals of states and union territories) in India. Continous wavelet transform is usedin the study because it is better suited to extract the periodic and local modulationspresent at various frequency ranges, as these features are invisible in conventionalmethods such as Fourier Transform. Monthly mean aerosol optical depths (AODs)from MODerate Resolution Imaging Spectroradiometer (MODIS) on board the Terrasatellite at 1 o × o resolution from January 2001 to December 2012 are used. Annualand quasi-biennial oscillations (QBO) in AOD are evident in addition to the weaksemi-annual (5-6 months) and quasi-triennial oscillations ( ∼
40 months). The semi-annual and annual oscillations are consistent with the seasonal and yearly cycle ofvariations in AODs. QBO type periodicity in AOD is found to be non-stationary whilethe annual period is stationary. The 40-month periodicity indicates the presence oflong term correlations in AOD. The observed periodicities in MODIS Terra AODs arealso evident in the ground-based AOD measurements made over Kanpur in the Indo-Gangetic Plain. The phase of the periodicity in AOD is stable in the mid-frequencyrange, while local disturbances in the high-frequency range and long term changes inthe atmospheric composition give rise to unstable phases in low-frequency range. Thatmodulations in AOD over one location/region can influence the other is revealed bythe presence of phase relation among different locations. ultiscale periodicities in aerosol optical depth over India Keywords:
Aerosol, Periodicities, Phase, India, Region, Climate
1. Introduction
Atmospheric aerosols exert a cooling effect on the Earth’s climate through direct andindirect effects which partially offset the warming caused due to greenhouse gases.The sources of aerosols can be natural (dust, sea salt, biogenic and volcanic) andanthropogenic (combustion of fossil fuel from urban/industrial processes and biomassburning). Dust, sea salt and sulfate produced over the ocean surfaces dominate thenatural global aerosol abundance, however, a fraction of dust in the atmosphere couldbe due to anthropogenic activities (Habib et al., 2006; Prospero et al., 2002). Similarly,smoke from natural burning such as due to forest fires is treated as natural component ofbiomass burning; while the burning of fuel wood, dung cake and crop waste burning areanthropogenic processes. Atmospheric aerosols modify the Earth-atmosphere radiationbudget by scattering and absorbing the incoming solar radiation (direct effect), andthe processes of formation of clouds and precipitation (indirect effect). The directand indirect aerosol radiative effects remain a significant uncertainty in climate studies(Solomon et al 2007).Scattering (sulfate) and absorbing (black carbon) particles cool the Earth’s surface,however, their radiative effects in the atmosphere vary with altitude. For scatteringparticles, the top of the atmosphere forcing is almost the same as the surface forcing;while for absorbing aerosol species the surface forcing is about 2 to 3 times larger thanthe top of the atmosphere forcing, which gives rise to a large atmospheric warming.The greenhouse gases are longer lived, globally well mixed and their radiative effectsare homogeneous and one of warming throughout the atmosphere starting from thesurface. In contrast, aerosols reside in the atmosphere for about a week, exhibit regionalsignatures and can either warm or cool the atmosphere. Although aerosols are abundantnear source regions, they impact global climate as aerosols and their radiative influencecan be transported to other regions due to atmospheric circulation. On temporal scales,the forcing due to aerosols is greatest during daytime and in summer. In contrast, thegreenhouse gas forcing acts over the full diurnal and seasonal cycles. Thus, aerosolsperturb the Earth-atmosphere radiation budget differently than greenhouse gases.The most important characteristics required to estimate the radiative influence ofaerosols are aerosol optical depth (AOD), single scattering albedo (SSA) and asymmetryparameter. Aerosol optical depth is the most crucial parameter to study aerosol-climateinteraction among the three, because aerosol radiative forcing changes due to increasein AODs overwhelm the forcing changes due to the increases in single scattering albedoand the asymmetry parameter values (Ramachandran 2005). SSA (ratio of scatteringto extinction) values can range from 0 (absorber) to 1 (scatterer). The aerosol radiativeforcing at the surface is nearly linearly related to AODs. For the same AOD value whenSSA is lower, the surface forcing is higher and the atmospheric forcing also becomeslarger. In addition, the radiative forcing at the top of the atmosphere due to lower SSA ultiscale periodicities in aerosol optical depth over India
2. Study region
Asia accounts for about 60% of the world’s population, and faces serious environmentalthreats in terms of air pollution, monsoon floods, droughts and associated climatechange. Increases in aerosol loading due to growing population and industrializationin recent decades have resulted in an increase in health-related problems, and impactedair quality, agriculture and water resources in Asia (Lau et al 2008). Anthropogenic ultiscale periodicities in aerosol optical depth over India >
3. Data
The MODerate Resolution Imaging Spectroradiometer (MODIS) is a remote sensor onboard the Earth Observing System (EOS) Terra and Aqua satellites. MODIS Terra andAqua satellites operate at an altitude of 705 km with the Terra spacecraft crossing theequator at about 1030 LST (ascending northward) while the Aqua spacecraft crosses theequator at around 1330 LST (descending southward) (King et al 2003; Remer et al 2008).AOD data are available from Terra since March 2000, while AOD data are available fromAqua starting from July 2002. Level 3 MODIS Terra Collection 5.1 quality assured (QA)monthly average 0.55 µ m AOD at 1 o × o data from January 2001 to December 2012 areutilized in this study, to maintain uniformity and because more number of data pointson temporal scale is available from Terra. Validation and comparison of AODs retrievedfrom Terra and Aqua, with ground-based AErosol RObotic NETwork (AERONET) ultiscale periodicities in aerosol optical depth over India o × o cells on an equal-angle global grid from level 2 atmospheric products that span over 24hperiod (King et al 2003). MODIS retrieval algorithms attempt to match the MODISobserved surface reflectances to a look-up table of precomputed reflectances for a widevariety of commonly observed aerosol conditions (King et al 1999) over land and ocean.The predicted retrieval uncertainty of MODIS derived AODs over land is ± (0.05 +0.15AOD) (Remer et al 2008).MODIS AOD data corresponding to the capitals of 28 states and 7 union territoriesin India divided into seven regions are analyzed and discussed (Table 1, Ramachandranand Cherian, 2008). The division is based on geography and meteorological conditions,and within the context of urban and rural development patterns. The capitals of thestates and union territories are chosen because most of these locations are urban centers,have medium to dense population (based on 2001 Indian census). The capitals acrossdifferent regions are governed by different aerosol sources. For example, Delhi andMumbai in addition to being densely populated, they are also largest commercial centersin India. The metro cities (population >
10 million) such as Delhi, Mumbai, Kolkataand Chennai are sources of urban/industrial and automobile emissions. In contrast,northeast India is sparsely populated and is rich in natural resources of oil and gas.Seasonal mean climatology of aerosol optical depth distribution over Indiacalculated from the 12-year (2000-2012) MODIS Terra AOD data, and synoptic windpatterns are plotted in Figure 1. The mean synoptic surface winds vary as a function ofseason (Figure 1). During winter (December-January-February) the winds are calm,north/northeasterly and are from the northern hemisphere. During the southwestsummer monsoon (June-July-August-September), the winds are stronger, moist andcome from the marine and western regions surrounding India (Figure 1). The windsare in a transitory phase and start shifting in direction during post-monsoon (October-November) from southwest to northeast. During the pre-monsoon season (March-April-May) the winds originate and travel from the west of Indian subcontinent. AOD exhibitssignificant regional and seasonal variations across India (Figure 1). AODs are higher inmonsoon when compared to post-monsoon. AODs are higher over the Indo-GangeticPlain throughout the year when compared to the rest of India. AODs are higher overthe Indo-Gangetic Plain during winter and post-monsoon because of the dominance offine mode aerosols from fossil fuel and biomass burning, while the higher AODs duringpre-monsoon and monsoon are attributed to the dominance of coarse mode dust andsea salt particles.Seasonal variations in AOD are found to be significant over northwest, north andeast India (Ramachandran and Cherian 2008). Annual mean AOD over northeast Indiawas lowest (0.3) and the contribution of fine mode aerosols (aerosols of size < µ m) to theoptical depth was higher (0.95) (Ramachandran and Cherian 2008). Northeast Indiais meagerly populated, and aerosols from natural burning of forest fires and biomass ultiscale periodicities in aerosol optical depth over India ∼
4. Methodology
In this section details about continuous wavelet transform (CWT) and the phaseinformation obtained from the complex continuous wavelet transform are outlined.Wavelet transform, since its advent has been a valuable tool for signal processing(Daubechies 1992; Torrence and Compo 1998). It has been applied to analyze signals indifferent fields of science including geophysics, for example, to study El Ni˜no SouthernOscillations (Torrence and Compo 1998), tropical convection over the western Pacific(Weng and Lau 1994) among others. Detailed description on the use of wavelettransforms in geophysics can be found in Foufoula-Georgiou and Kumar (1995). TheCWT of a data set X = { x i } , i ∈ Z + is given by, W i ( s ) = N − X j =1 x j ψ ∗ (cid:18) i − js (cid:19) (1)where s is the scale and N is the data length. ψ ( s ) is a well localized (in bothphysical and Fourier domains), zero mean and integrable function and is called the mother wavelet . Eqn. 1 represents a convolution equation, where the wavelet coefficientsare calculated by convolving the scaled and translated versions of ψ ( n ) with x i . Thus, itis clear that s is the scaling parameter and j the translation parameter. In this analysis,Morlet wavelet is utilized, whose real form is given by, ψ ( n ) = C cos(5 n ) e − n (2)where C is a normalization constant. This function has a wide support and allowsto get more accurate results from the computationally performed convolution. Sincethis function is real it can be used to retrieve periodic structures from the AOD data.In order to obtain the phase relationships between various stations the complex Morletfunction is used. The complex Morlet function is expressed as ψ ( n ) = 1 √ πF b exp (cid:18) ıπF c n − n F b (cid:19) (3) ultiscale periodicities in aerosol optical depth over India F b = 1 and F c = 1 . φ ( n ) = tan − (cid:18) Im[ ψ ( n )]Re[ ψ ( n )] (cid:19) (4)Since the data length is 12 years (144 months), the analysis is limited to a scaleof 32. As shown in Figure 2, the cone of influence is big enough to make the waveletcoefficients significant and reliable at this scale. The results obtained using the realMorlet wavelet (Eqn. 2) are used to extract the multiscale periodicities in aerosol opticaldepths, while the complex Morlet function is used to establish the phase relations, ifany, between the study locations. Periodicity in atmospheric/geophysical parameterscan provide information on the cyclic/periodical behavior and local modulations of theparameter, while the phase relation can be used to understand the influence aerosols onthe local scale exert on a regional scale and/or surrounding locations.
5. Results and Discussion
Results pertaining to four locations in India for a comparative study are highlighted inthe study. Mumbai and Chennai represent the western and eastern coastal regions withBengaluru lying between them (Figure 1). These three locations are interesting becauseof high urbanisation and industrialization, which allows one to observe the anthropogenicinfluence in aerosol characteristics, in conjunction with their vastly different ambientenvironments (Mumbai and Chennai are urban cities located near the coast, whileBengaluru is a continental location) and climatic conditions (Figure 1). Kohima ischosen because it is less industrialized and its atmosphere is dominated by fine modeaerosols produced by natural biomass burning (forest fires). Kohima and Chennai AODsare lower than Bengaluru and Mumbai (Figure 3). Three dominant periods in thesignal can be seen, one corresponding to 12 months and the others corresponding toapproximately 24 and 40 months. A period of approximately 5-6 months is observed atall the locations, though it is not significant as the other periodicities. This is indicativeof an external influence (over the natural cycle) in AODs. The 5-6 and 12-month periodis consistent with the seasonal (summer high, winter low) and annual patterns seen inAODs (Jin et al 2005; Ramachandran and Cherian 2008). These dominant periods areobserved in all the other study locations also (not shown).Figure 3 shows the monthly averaged AOD data and the local variation in variousgeographical regions, obtained using Morlet wavelet technique for Chennai (south India),Mumbai (west India), Bengaluru (south India) and Kohima (northeast India). Thoughthe 12 month period is a stationary period (i.e., it does not vary over time), the 24 monthperiod is non-stationary. The 24-month non-stationary periodicity can be associatedwith quasi-biennial oscillation (QBO). Quasi-biennial oscillation refers to downwardpropagating easterly or westerly winds in the equatorial stratosphere ( ∼ ultiscale periodicities in aerosol optical depth over India ∼
12% to the columnar AOD (Kulkarni et al2008). It has also been found that the contribution of aerosols in the 10-30 km exhibitseasonal variations, and their contribution to the columnar AODs can vary between10 and 20% (Kulkarni et al 2008), consistent with the finding on increase/decrease inAODs during the different phases of QBO. A 40 month cycle in AOD is also seen, whichappears only after sufficient averaging (Figure 3). This implies the presence of long termcorrelations in AOD over the study locations. The results from the present study onthe QBO and 40-month periodicities in AODs agree with Beegum et al (2009).An analysis of the monthly mean 0.50 µ m AODs during 2001-2012 from Kanpur(26.5 o N, 80.2 o E, 123m above mean sea level (AMSL)), measured using ground-basedAERONET sun/sky radiometers (Holben et al 2001), is undertaken in order toexamine whether the observed periodicities are evident in ground-based measurementsof columnar AOD as well. Kanpur is an urban, industrial and densely populated citywith a population of more than 4 million and located ∼
250 km east of the mega city,New Delhi (Figure 1). Kanpur AODs are mostly in the range of 0.5 to 1 (Figure 4). Theanalysis reveals that AODs obtained between 2001 and 2012 over Kanpur also exhibit ∼ ∼ ∼
24 and ∼
36 month periodicities (Figure 4) consistent with periodicitiesderived using remote sensing AODs over different locations in India.In Figure 5, the phase relations obtained using Eqn. 4 are plotted. It is interestingto see a stationarity in the phases at scale 20. In contrast, at scale 10, which is the high-frequency range, non-stationarity in the phases is observed. All the stations undergoa phase variation around the month 35 (at scale 30). The phase relations also providea measure of the recovery time to their original phases at various locations; Chennairecovers the fastest while the other stations take more longer to recover. The low-frequency range is also non-stationary. This makes the analysis of their phase differencesimportant. In Figure 6 the phase differences between the various stations are depicted.At scale 20 (mid-frequency range) AODs in the study locations are periodically in andout of phase, while Chennai stays in phase with Mumbai and Bengaluru most of thetime. It should be noted that ∆ φ = 0 means zero phase lag, while ∆ φ > φ < ultiscale periodicities in aerosol optical depth over India
6. Conclusions
Multiscale periodicities and phase in aerosol optical depths are determined byperforming an analysis over 35 locations (capitals of states and union territories) in India.Continuous wavelet transform is chosen to extract the periodic and local modulationspresent at various frequency ranges which are invisible in the conventional methods suchas Fourier Transform. Monthly mean aerosol optical depths from MODIS on board theTerra satellite at 1 o × o resolution from January 2001 to December 2012 are used.The major findings of the study are:The study reveals the presence of 5-6 and 40 month periods in AOD, in additionto the annual and quasi-biennial oscillations over the study locations. The semi-annualand annual oscillations are consistent with the seasonal and yearly cycle of variationsin AODs. QBO type periodicity in AOD is non-stationary consistent with the variationin the temporal occurrence of QBO. The 40-month periodicity suggests the presence oflong term correlations in AOD.AODs obtained from ground-based measurements over Kanpur, an urban, denselypopulated location in the Indo-Gangetic Plain, also showed semi-annual, annual, QBOand 40-month periodicities consistent with the periodicities observed in MODIS TerraAODs.The phase of periodical behavior is stable in the mid-frequency range, while in thehigh- and low-frequency ranges the phases are unstable implying time-local disturbancesin the high-frequency range. The instability in the low-frequency range could beattributed to long term changes in the atmospheric composition and structure overthe locations due to various factors such as industrial emissions and changes in forestcover for example.The presence of phase relation among different locations signifies that themodulations in AOD over one location/region can affect the other.Results from this study on the quantitative determination of the periodic natureof aerosol characteristics over a large spatial domain governed by a variety of aerosolsources will be useful while conducting visibility and air quality studies, and in globalclimate simulations and/or predictions of aerosols and their radiative effects.This study should be extended to the other regions of the globe to ascertain andcorrelate the natural and anthropogenic sources/processes contributing to the cyclic be-havior in aerosol characteristics. Acknowledgments ultiscale periodicities in aerosol optical depth over India µ m are used in the study. References
Beegum S N, Moorthy K K, Babu S S, Reddy R R, Gopal K R and Ahmed Y N 2009 Quasi-biennialoscillations in spectral aerosol optical depth
Atmos. Sci. Lett. Ten lectures on wavelets
61 CBMS-NSF Regional Conference Series in AppliedMathematics (Society for Industrial and Applied Mathematics) p 377Foufoula-Georgiou E and Kumar P (Eds.) 1995
Wavelets in Geophysics (Academic Press) p 373Habib G, Venkataraman C, Chiapello I, Ramachandran S, Boucher O and Reddy M S 2006 Seasonaland interannual variability in absorbing aerosols over India derived from TOMS: Relationship toregional meteorology and emissions
Atmos. Env. J. Geophys.Res.
J. Geophys. Res.
D10S20 doi:10.1029/2004JD005081Kaufman Y, Tanr´e D and Boucher O 2002 A satellite view of aerosols in the climate system
Nature
Bull. Amer. Meteorol. Soc. IEEE Trans. Geosci. Rem. Sens. J. Geophys. Res. , D17207, doi:10.1029/2007JD009411Lau K-M and Kim K-M 2006 Observational relationships between aerosol and Asian monsoon rainfall,and circulation
Geophys. Res. Lett. L21810 doi:10.1029/2006GL027546Lau K-M, Ramanathan V, Wu G-X, Li Z, Tsay S C, Hsu C, Sikka D R, Holben B, Lu D, TartariG, Chin M, Koudelova P, Chen H, Ma Y, Juang J, Taniguchi K and Zhang R 2008 The jointaerosol-monsoon experiment
Bull. Amer. Meteorol. Soc. A Wavelet Tour of Signal Processing (San Diego: Academic Press) p 637Moberg A, Sonechkin D, Holmgren K, Datsenko N and Karlen W 2005 Highly variable NorthernHemisphere temperatures reconstructed from low- and high-resolution proxy data
Nature
Rev. Geophys. J. Geophys. Res.
D07207doi:10.1029/2004JD005563Ramachandran S and Cherian R 2008 Regional and seasonal variations in aerosol opticalcharacteristics and their frequency distributions over India during 2001-2005
J. Geophys. Res.
D08207 doi:10.1029/2007JD008560Remer L A, Kleidman R G, Levy R C, Kaufman Y J, Tanr´e D, Mattoo S, Martins J V, Ichoku C, ultiscale periodicities in aerosol optical depth over India Koren I, Yu H and Holben B N 2008 Global aerosol climatology from MODIS satellite sensors
J.Geophys. Res.
D14S07 doi:10.1029/2007JD009661Solomon S et al (ed) 2007 Climate change 2007
Fourth Assessment Report of the IntergovernmentalPanel on Climate Change (New York: Cambridge University Press) p 996Teolis A 1998
Computational signal processing with wavelets (Boston: Birkhauser) p 342Torrence C and Compo G 1998 A practical guide to wavelet analysis
Bull. Amer. Meteorol. Soc. J. Atmos. Sci. ultiscale periodicities in aerosol optical depth over India Figure 1.
Seasonal mean climatology (2000-2012) of aerosol optical depths and surfacewinds over India during (a) winter (December-January-February, DJF), pre-monsoon(March-April-May, MAM), (c) monsoon (June-July-August-September, JJAS) and(d) post-monsoon (October-November, ON). The shaded contours correspond to 0.55 µ m aerosol optical depths, on which surface winds (ms − ) represented by arrows areoverlaid. ultiscale periodicities in aerosol optical depth over India Time (in months) S c a l e
12 24 36 48 60 72 84 96 108 120 132 14451015202530 −0.3−0.2−0.100.10.20.3
Figure 2.
The scalogram for Kohima upto scale 32. The cone of influence (blackdotted line) is shown. The oscillations (drawn as blue dotted lines) at semi-annual (atscale 5), annual (at scale 10), QBO (at scale 20) and 40 month (at scale 30) are clearlyvisible. ultiscale periodicities in aerosol optical depth over India A O D (a) AOD−101 W i ( s ) (b) Scale 5−202 W i ( s ) (c) Scale 10−101 W i ( s ) (d) Scale 2012 24 36 48 60 72 84 96 108 120 132 144−101 W i ( s ) (e) Scale 30 Time (in months) Bengaluru Chennai Mumbai Kohima
Figure 3. (a) Monthly mean aerosol optical depths over Bengaluru, Chennai,Mumbai and Kohima from January 2001 to December 2012. Wavelet coefficients overBengaluru, Chennai, Mumbai and Kohima at (b) scale 5, (c) scale 10, (d) scale 20 and(e) scale 30. ultiscale periodicities in aerosol optical depth over India A O D (a)−0.500.5 W i ( s ) (b) scale 5 scale 10 scale 25 scale 3012 24 36 48 60 72 84 96 108 120 132 144−101 Time (in months) φ i ( s ) (c) Figure 4. (a) Monthly mean aerosol optical depths over Kanpur during 2001-2012.(b) Wavelet coefficients of aerosol optical depths obtained over Kanpur at differentscales. (c) Phase relation of aerosol optical depths over Kanpur. ultiscale periodicities in aerosol optical depth over India −202 φ i ( s ) (a) Scale 5−202 φ i ( s ) (b) Scale 10−202 φ i ( s ) (c) Scale 2012 24 36 48 60 72 84 96 108 120 132 144−202 φ i ( s ) (d) Scale 30 Time (in months) Bengaluru Chennai Mumbai Kohima
Figure 5.
The phase relation at different scales in Bengaluru, Chennai, Mumbai andKohima. ultiscale periodicities in aerosol optical depth over India −505 ∆ φ i ( s ) (a) Bengaluru − Chennai−505 ∆ φ i ( s ) (b) Chennai − Kohima−505 ∆ φ i ( s ) (c) Chennai − Mumbai−505 ∆ φ i ( s ) (d) Bengaluru − Kohima−505 ∆ φ i ( s ) (e) Bengaluru − Mumbai12 24 36 48 60 72 84 96 108 120 132 144−505 ∆ φ i ( s ) (f) Kohima − Mumbai Time (in months) scale 5 scale 10 scale 20 scale 30 Figure 6.