下载文档原格式
《大气物理与大气探测学》知识点1.熟悉大气物理与大气探测学研究的内容，也要明白大气物理与大气探测的区别。
大气物理学是研究大气的物理现象（声光电等）、物理过程及其演变规律的学科，是大气科学的一个分支。
大气探测学是大气科学的另一个基础性学科分支，主要研究大气状态和过程的信息探测技术、观测方法和信息处理技术。
探测的对象包括地面和高空的大气状态和过程参数。
2.基本名词的理解，从大气科学的角度解释，温室效应，温室气体，阳伞效应，ENSo，酸雨，大气污染，雾，露点（霜点），沙尘暴，极光，臭氧空洞，湖陆风（焚风），城市热岛，大气中的光现象解释（如海市蜃楼，虹，天空蓝色，海洋蓝色等），平流层急剧增温（SSw）1）温室效应：太阳（短波）辐射通过大气层到达地面并被其吸收，地面（长波）辐射则几乎全部被大气所吸收，大气向外太空和地面发出长波辐射，后者称为大气逆辐射，使地面升温。
2）温室气体：指二氧化碳、甲烷、一氧化二氮及水汽等。
其中co2是最主要的温室气体，主要来自火山喷发、有机物的燃烧、腐烂及动植物的呼吸等。
3）阳伞效应：由于排入空气的烟尘不断增加，使到悬浮在大气中的气溶胶颗粒就象地球的遮阳伞一样，反射和吸收太阳辐射，引起地面降温。
4）ENSo：ENSo循环：ENSocirculation赤道太平洋海面水温的变化与全球大气环流尤其是热带大气环流紧密相关。
其中最直接的联系就是日界线以东的东南太平洋与日界线以西的西太平洋—印度洋之间海平面气压的反相关关系，即南方涛动现象（So）。
在拉尼娜期间，东南太平洋气压明显升高，印度尼西亚和澳大利亚的气压减弱。
厄尔尼诺期间的情况正好相反。
鉴于厄尔尼诺与南方涛动之间的密切关系，气象上把两者合称为ENSo（音“恩索”）。
这种全球尺度的气候振荡被称为ENSo循环。
厄尔尼诺和拉尼娜则是ENSo 循环过程中冷暖两种不同位相的异常状态。
因此厄尔尼诺也称ENSo暖事件，拉尼娜也称ENSo冷事件。
5）酸雨：大气中含有的二氧化硫在常温下溶解于雨水中并达到气液相平衡后，雨水之pH值约为5.6。
Retrieval of seasonal dynamics of forest understory re flectance in a Northern European boreal forest from MODIS BRDF dataJan Pisek a ,⁎,Miina Rautiainen b ,Janne Heiskanen b ,Matti Mõttus ca Tartu Observatory,61602Tõravere,Tartumaa,Estoniab Department of Forest Sciences,PO Box 27,FI-00014University of Helsinki,FinlandcDepartment of Geosciences and Geography,PO Box 64,FI-00014University of Helsinki,Finlanda b s t r a c ta r t i c l e i n f o Article history:Received 24May 2011Received in revised form 12September 2011Accepted 16September 2011Available online 26October 2011Keywords:BorealUnderstory MODIS BRDFMulti-angle remote sensingThe spatial and temporal patterns of the forest background re flectance are critically important for retrieving the biophysical parameters of the forest canopy (overstory)and for ecosystem modeling.In this short com-munication paper,we retrieve the re flectance and seasonal changes of the forest background at 500m reso-lution with the 8-day MODIS bidirectional re flectance distribution function (BRDF)model parameters product.For the first time,the satellite data derived results are directly validated with in situ measured sea-sonal re flectance trajectories of boreal forest understory layers.Our results illustrate the importance of taking into account the documented quality and limitations of the MODIS BRDF product.©2011Elsevier Inc.All rights reserved.1.IntroductionRemote sensing of biophysical properties,such as leaf area index (LAI)or fraction of absorbed photosynthetically active ra-diation (fAPAR),of the forest canopy layer is often confounded by the dominating effect of the forest background re flectance on the total spectral signal (Spanner et al.,1990;Chen &Cihlar,1996;Eklundh et al.,2001;Rautiainen,2005;Eriksson et al.,2006;Garrigues et al.,2008).The separate spatial and temporal patterns of the forest background are also critically important for ecosystem modeling (Waring &Law,1994).For example in carbon cycle modeling,overstory and background vegetation need to be treated differently because carbon fixed through net primary productivity (NPP)has different residence times for these dif-ferent vegetation components in forest ecosystems (Vogel &Gower,1998;Rentch et al.,2003).Despite its importance,the predictions re-garding forest background spectral variations have remained inherently dif ficult (McDonald et al.,1998;Gemmell,2000).The evaluation of spectral properties and effect of understory veg-etation on re flectance observed by a sensor is particularly important in the sub-boreal and boreal zone due to the low canopy cover of the tree layer (Rautiainen et al.,2007;Pocewicz et al.,2007).In previ-ous campaigns,a few efforts were undertaken to collect various un-derstory components and/or creating limited spectral databases (ler et al.,1997;Lang et al.,2002;Rees et al.,2004;Peltoniemiet al.,2005;Rautiainen et al.,2007;Hallik et al.,2009).However,the site type and understory vegetation of the boreal forests can be quite variable with very different bidirectional properties (e.g.Rees et al.,2004;Kaasalainen &Rautiainen,2005;Peltoniemi et al.,2005),and remote sensing is the only technology to provide consistent data at the required spatially extensive scales (Pinty et al.,2008).Recently,Pisek and Chen (2009)produced a monthly forest back-ground brightness data set over North America with a spatial resolu-tion of one degree using Multiangle Imaging Spectroradiometer (MISR)data (Diner et al.,1998).While the dataset helped to reduce the forest background effect on canopy LAI retrievals (Pisek et al.,2010),the associated drawbacks of (1)very low spatial and temporal resolution of the dataset and (2)only indirect validation of the de-rived re flectances required further investigation.In this short communication paper,we address both drawbacks.We document that using the same multi-angle approach,we can retrieve the re flectance and seasonal changes of the forest background at 500m resolution with the 8-day MODIS bidirectional re flectance distri-bution function (BRDF)model parameters product.Second,the results are directly validated with in situ measured seasonal re flectance trajec-tories of boreal forest understory layers for Hyytiälä(Finland)forest site.2.Materials 2.1.Study areaOur study sites are located close to the Hyytiäläforest station (Finland,61°50′N,24°17′E)in the southern boreal zone.The sites comprise fourRemote Sensing of Environment 117(2012)464–468⁎Corresponding author.E-mail address:jan.pisek@utoronto.ca (J.Pisek).0034-4257/$–see front matter ©2011Elsevier Inc.All rights reserved.doi:10.1016/j.rse.2011.09.012Contents lists available at SciVerse ScienceDirectRemote Sensing of Environmentj o u r n a l h om e p a g e :ww w.e l s e v i e r.c o m /l o c a t e /r s estands representing different forest fertility site types:1)a herb-rich Silver birch(Betula pendula Roth)forest understory dominated by herbaceous species and graminoids,2)a mesic Norway spruce (Picea abies(L.)Karst)forest understory dominated by mosses and abundant dwarf shrubs,3)a sub-xeric Scots pine(Pinus sylvestris L.) forest understory dominated by mosses and dwarf shrubs,and4)a xeric Scots pine forest understory dominated by lichens and heather.A more detailed description of the sites is available in Rautiainen et al. (2011).2.2.Field measurements of reflectance spectra and LAIThe reflectance spectra for the understory layers of study sites were measured during the growing period of2010.Thefield cam-paign started on4May2010(day of year(DOY)124)and ended on22September2010(DOY265).During thefield campaign,the understory spectra for each study site were measured ten times, i.e.every two to three weeks.A28-meter long permanent tran-sect was established at each study site,and the spectrum of reflected radiation was measured approximately every70cm on the transect with a FieldSpec Hand-Held UV/VNIR Spectroradi-ometer(325–1075nm,spectral resolution3nm,field of view of about25°)manufactured by Analytical Spectral Devices(ASD). Only diffuse light conditions were used:measurements were car-ried out when the sun was completely blocked by clouds or completely attenuated by the long path length in tree crown layer at low solar elevation near sunrise or sunset.A leveled Spec-tralon reflectance panel was measured at the beginning and the end of transect as well as at every third measurement point along transect,i.e.approximately every2m.All the measure-ments for one transect were averaged to form a mean spectrumrepresenting the study site.The spectral measurements corre-spond to hemispherical–directional reflectance factors(HDRF), and MODIS'spectral response functions were used to compute understory HDRF's corresponding to the MODIS red(620–670nm)and NIR(841–876nm)bands for each of the four under-story types.A more detailed description of the spectral measure-ments and data processing is available in Rautiainen et al.(2011).Next,the understory spectra were upscaled by overlaying a forest site type map(produced by Metsähallitus,a state-owned enterprise managing the local forests)with the MODIS500m grid(Fig.1).The red and NIR understory reflectance factors corresponding to each MODIS pixel were computed as area-weighted mean values of the HDRF's representing each forest site type within the pixel.In addition to understory spectra,the effective leaf area index (LAI eff)of20forest stands representing a wide range of forest structures(and corresponding to the forest site types previously described)in Hyytiäläwas measured with the LAI-2000Plant+ Canopy Analyzer(Li-Cor Inc.)approximately at the same time with the understory spectra measurements.Below-canopy mea-suring height was1m above the ground in order to exclude leaf area coverage of the understory.Above-canopy measurements were collected by automatic logging every15s at a tower located in the study I of each of the20stands was calculated from canopy gap fraction values averaged over twelve measurements made according to the measurement protocol of the VALERI net-work(i.e.a cross with measurement points placed at4m inter-vals on a South–north transect(six points)and on a West–east transect(six points);http://w3.avignon.inra.fr/valeri/).Next,the stand-level LAI measurements were upscaled to corre-spond to the spatial resolution of MODIS pixels.The study stands were divided intofive groups:pure Scots pine stands(n=3),pure Norway spruce stands(n=2),pure birch stands(n=3),mixed conif-erous stands(n=8)and mixed coniferous-birch stands(n=4).A corresponding forest type classification of the study area was done by supervised classification of a SPOT5HRG image(31May2009)and forest inventory sample plots from2008(n=72).The training areas were manually delineated around the plots to cover a homoge-neous area of several pixels.The overall classification accuracy for the training sites was90.3%.Finally,the LAI eff values for each MODIS pixel were computed as area-weighted averages of the LAI eff values repre-senting each forest type within the pixel.2.3.MODIS dataThe MODIS BRDF/Albedo Product(MCD43A1,version5)(Lucht et al.,2000;Schaaf et al.,2002)is a MODIS standard product that provides the weighting parameters associated with the RossThick-LiSparse BRDF model that describes the reflectance anisotropy at 500m resolution.The BRDF parameters are produced every eight days with16days of acquisition using both Terra and Aqua data (Schaaf et al.,2002).We tested the forest background signal re-trieval with the BRDF model parameters(isotropic,volumetric, and geometric kernel weights(Roujean et al.,1992))for MODIS band1(red,620–670nm)and band2(NIR,841–876nm). MODIS data employed in this study were acquired for all dates and tiles covering Hyytiäläin2009–2010(33pixels;Fig.1). First,a40-day moving window was run on all three BRDF model parameters with the largest and the smallest value dropped be-fore averaging.This was done in order to limit previously noted noise in the seasonal trajectories of the linear BRDF parameters (Vermote et al.,2009;Quaife&Lewis,2010).Next,for each date,we reconstructed two BRF values:(1)view zenith angle (VZA)=0°;and(2)VZA=45°.The view azimuth angle(VAA) was140°with solar zenith angle(SZA)corresponding for both view angles to the Sun's position at10:00local time.Following Wang et al.(2010),we additionally corrected the bias in the orig-inal reconstructed MODIS BRF values in the red band before their input into the understory calculation algorithm.All the MCD43A1data were downloaded from the Data Pool of Land Processes Distributed Active Archive Center(LPDAAC).The Fig.1.Forest site type map of Hyytiäläwith the layout of the MODIS500m grid.465J.Pisek et al./Remote Sensing of Environment117(2012)464–468associated data quality(MCD43A2)product was also downloaded from the same source and was used to analyze the effect of re-trieval quality on the accuracy of calculating understory reflectance.3.Retrieval methodThe total spectral reflectance of a pixel(R)can be expressed as a linear combination of the contributions from the sunlit and viewed, and shaded and viewed components(Li&Strahler,1985;Chen et al., 2000;Bacour&Bréon,2005;Chopping et al.,2008):R¼R T·k TþR G·k GþR ZT·k ZTþR ZG·k ZGð1Þwhere R T,R G,R ZT,and R ZG are the reflectance factors of the sunlit crown, sunlit background,shaded crown,and shaded background,respectively. Here the term background refers to all the materials below the forest canopy such as understory,leaf litter,grass,lichen,moss,rock,soil, snow,or their mixtures(Kuusk,2001;Pisek&Chen,2009).The k j are the proportions of these components at the chosen view angle or in the instantaneousfield-of-view(IFOV)of the sensor at given irradiation geometry.Based on the assumption that the reflectance factors of the overstory and the understory at the given illumination angle differ little between chosen view angles,one can derive the background reflectance factor(R G).Forward-scattering reflectance factors of various targets off the principal plane were previously shown to be fairly constant within the angular range used in optical satellite RS(Peltoniemi et al.,2005; Bacour&Bréon,2005;Deering et al.,1999).The reflectance at nadir (R n)and another zenith angle(R a)can be then expressed by the Eqs.(2)and(3)(Pisek&Chen,2009):R n¼R T·k TnþR G·k GnþR ZT·k ZTnþR ZG·k ZGnð2ÞR a¼R T·k TaþR G·k GaþR ZT·k ZTaþR ZG·k ZGað3ÞThe shaded fractions of tree crowns(R ZT)and ground(R ZG)can be expressed dynamically as functions of their sunlit fractions and the mul-tiple scattering factor M(White et al.,2001,2002a,b),giving R ZT= M·R T and R ZG=M·R G,where M=Rz/R for a reference target.M is predetermined by the4-Scale geometric-optical model inversion (Chen&Leblanc,1997).The proportions of the components(k j)were retrieved from pre-calculated look-up tables generated with the4-Scale model(Chen&Leblanc,1997).The model inputs are provided in Table1;for more details see Pisek and Chen(2009).Combining and solving Eqs.(2)and(3),and inserting the MODIS-derived R n and R a, the background reflectance R G can be calculated.4.Results and discussionGiven the possible large variations of understory spectra even among the same species found few meters apart(Bubier et al., 1997;Peltoniemi et al.,2005),the MODIS BRDF understory signal re-trievals provide reasonable estimates in Hyytiälä.The mean abso-lute error(MAE)of understory retrievals from MODIS BRDF over the whole April–October period(DOY100–300)was within one standard deviation of thefield measurements(0.0118in red and 0.0339in NIR)with an exception of the NIR band when the MODIS BRDF parameters wereflagged to be of lower quality (Table2).The better performance in the red over NIR domain is linked to the overall better contrast between vertically clumped elements and the background because of weaker multiple scattering in the red wavelength(Pinty et al.,2002).Schaaf et al.(2011)stated that the lower-quality BRDF parame-ters(QA=1)may also include acceptable values,although they are more likely to be affected by residual cloud contamination and atmospheric turbidity and are thus less consistent.Our results confirm these statements(Fig.2A–C;Table2).The MODIS BRDF parameters of lower quality can occasionally produce forest background reflectances on par with the best quality retrievals,especially during thefirst half of the growing period(April–mid July;Fig.2A–C; Table2).However,the qualityflags also correctly pick up most of the retrievals that deviated from thefield measurements,either in the red(Fig.2C)or NIR(Fig.2B)band.The MODIS retrievals differed markedly from thefield estimates in few locations during the whole season(Fig.3).These pixels contain large proportions of non-vegetated areas.The upscaled in situ mea-surements represent the vegetated fraction only,while the MODIS signal is influenced by the areas with no vegetation as well.Chen (1999)reported that the bias can be up to45%of the correct value depending on the non-vegetated area fraction in the pixel;our re-sults as illustrated in Fig.3are in agreement with thesefindings. Besides the MODIS data quality,the amount of bias in the re-trieved forest understory signal is thus confirmed to depend on the surface heterogeneity as well(Fig.3).It should be also re-membered that MODIS does not sample from the same footprint every time it passes over,which further increases the uncertainty (Tan et al.,2006).In fact,a single pixel value may have been obtained from up to4different pixels.Except few locations men-tioned above,our Hyytiälästudy area is a rather homogeneous forest.However,this uncertainty can be a limitation for using the retrieval technique in a heterogeneous forest and making reli-able estimates on the change in background reflectance over time.The background forest retrievals from MODIS match closely the field values particularly up to mid-July(DOY100–200;Fig.4A; Table2).The best quality red band retrievals are still moderately suc-cessful after mid-July(DOY201–300;Table2).However,in the NIR band,even the best quality retrievals do not always match thefield measurements after mid-July(Fig.4B),especially after DOY250 (Fig.2B,C).One possible strategy to alleviate this issue can be to look at the retrievals from another year(s)with higher share of the best quality retrievals during the period in question.Indeed,the for-est background retrievals from2009have much improved seasonal trajectories after DOY250(Fig.2A–C),which are closer to thefield measurements as well.Previously,Kuusk et al.(2004)looked at the sensitivity of their hybrid type forest reflectance model to input understory parameters and they recommended using typical(average)parameter values while representing understory.Our results indicate that averaging the best quality retrievals from multiple years might be the mostTable1Input parameters to4-Scale model.The modeled stand density for both forest types was2000trees/I varied from0.1to10with a step of0.1.Tree shape H H b VCCΩEConiferous Cone+cylinder16469.50.7 Deciduous Spheroid20582.50.8H—tree height,m.H b—crown base height,m.VCC—vertical canopy cover,%.ΩE—element clumping index.Table2Mean absolute errors(MAE)of the forest background reflectance retrievals using MODIS BRDF product over Hyytiäläin2010.The MAE values smaller than the corre-sponding mean standard deviations of thefield measurements are in bold.QA=0re-fers to the best quality MODIS BRDF parameters and QA=1to the lower quality ones.Red band NIR bandQA=0QA=1QA=0QA=1DOY100–200(April–mid July)0.0070.0060.0190.022 DOY201–300(mid July–October)0.0090.0140.0450.058 DOY100–300(April–October)0.0080.0110.0310.045466J.Pisek et al./Remote Sensing of Environment117(2012)464–468suitable approach to obtain robust estimates of forest background re flectance from the MODIS BRDF product.5.ConclusionOur results demonstrate the capability of the MODIS BRDF param-eters and our multi-angle-based approach to retrieve meaningfulinformation about forest background at 500m resolution for Northern European forests.Our results also illustrate the importance of taking into account the documented quality and limitations of the MODIS BRDF product.A B CRRDOYRDOYDOYFig.2.Three examples of the retrieved seasonal courses of forest background re flec-tance in red (filled blue squares)and NIR band (empty blue squares)and their compar-ison with upscaled in situ measurements (red band-filled purple diamonds;NIR band-empty purple diamonds)at Hyytiäläin 2010.The seasonal trajectories for 2009are also included (red band-filled green triangles;NIR band-empty green triangles).Blue bars indicate MODIS BRDF parameters with lower quality flags in 2010;black bar indicates no data available.00.10.20.30.40.50.60.70.80.91DOYN D V IFig.3.Seasonal courses of forest background NDVI (blue crosses —2010;green crosses —2009)and their comparison with in situ measurements (empty purple circles —2010).The pixel represents a spatially heterogeneous case with a large fraction of open water.Blue bars indicate MODIS BRDF parameters with lower quality flags in 2010.0.10.20.30.4M O D I S0.10.20.30.40fieldM O D I SBAHyytiälä, DOY 100-200Hyytiälä, DOY 201-3000.10.20.30.40field0.10.20.30.4Fig.4.Scatter-plots of MODIS forest background re flectance retrievals versus field mea-surements in the red band (purple crosses)and NIR band (blue circles)over Hyytiäläin 2010.A.Period DOY 100–200.B.Period DOY 201–300.467J.Pisek et al./Remote Sensing of Environment 117(2012)464–468AcknowledgmentsThe authors wish to thank Anu Akujärvi and Titta Majasalmi for field work at the Hyytiälätest site.MODIS MCD43BRDF product tiles were acquired from Land Processes Distributed Active Archive Center(LP DAAC).Funding was provided from the Academy of Finland,Emil Aaltonen Foundation,University of Helsinki Research and Postdoctoral Funds.We are grateful to three anonymous reviewers for helping to improve the manuscript.The detailed suggestions by the third reviewer were particularly appreciated.ReferencesBacour,C.,&Bréon,F.M.(2005).Variability of biome reflectance directional signatures as seen by POLDER.Remote Sensing of Environment,98,80–95.Bubier,J.L.,Barrett,N.R.,&Crill,P.M.(1997).Spectral reflectance measurements of boreal wetland and forest mosses.Journal of Geophysical Research,102(D24), 29483–29494.Chen,J.M.(1999).Spatial scaling of a remotely sensed surface parameter by contex-ture.Remote Sensing of Environment,69,30–42.Chen,J.M.,&Cihlar,J.(1996).Retrieving leaf area index of boreal conifer forests using Landsat TM images.Remote Sensing of Environment,55,153–162.Chen,J.M.,&Leblanc,S.G.(1997).A four-scale bidirectional reflectance model based on canopy architecture.IEEE Transactions on Geoscience and Remote Sensing,35, 1316–1337.Chen,J.M.,Li,X.,Nilson,T.,&Strahler,A.(2000).Recent advances in geometrical opti-cal modeling and its applications.Remote Sensing Reviews,18,227–262. Chopping,M.,Moisen,G.G.,Su,L.H.,Laliberte,A.,Rango,A.,Martonchik,J.V.,et al.(2008).Large area mapping of southwestern crown cover,canopy height,and bio-mass using the NASA Multiangle Imaging Spectro-Radiometer.Remote Sensing of Environment,112(5),2051–2063.Deering,D.W.,Eck,T.F.,&Banerjee,B.(1999).Characterization of the reflectance anisotropy of three boreal forest canopies in spring–summer.Remote Sensing of Environment,67,205–229.Diner,D.J.,Beckert,J.C.,Reilly,T.H.,Bruegge,C.J.,Conel,J.E.,Kahn,R.A.,et al.(1998).Multiangle Imaging SpectroRadiometer(MISR)instrument description and experi-ment overview.IEEE Transactions on Geoscience and Remote Sensing,36,1072–1087. Eklundh,L.,Harrie,L.,&Kuusk,A.(2001).Investigating relationships between Landsat ETM+sensor data and leaf area index in a boreal conifer forest.Remote Sensing of Environment,78,239–251.Eriksson,H.,Eklundh,L.,Kuusk,A.,&Nilson,T.(2006).Impact of understory vegetation on forest canopy reflectance and remotely sensed LAI estimates.Remote Sensing of Environment,103,408–418.Garrigues,S.,Lacaze,R.,Baret,F.,Morisette,J.T.,Weiss,M.,et al.(2008).Validation and intercomparison of global leaf area index products derived from remote sensing data.Journal of Geophysical Research,113.doi:10.1029/2007JG000635. Gemmell,F.(2000).Testing the utility of multi-angle spectral data for reducing the effects of background spectral variations in forest reflectance model inversion.Remote Sensing of Environment,72,46–63.Hallik,L.,Kull,O.,Nilson,T.,&Penuelas,J.(2009).Spectral reflectance of multispecies herbaceous and moss canopies in the boreal forest understory and openfield.Canadian Journal of Remote Sensing,35,474–485.Kaasalainen,S.,&Rautiainen,M.(2005).Hot spot reflectance signatures of common boreal lichens.Journal of Geophysical Research,110,D20102.doi:10.1029/2005JD005834. Kuusk,A.(2001).A two-layer canopy reflectance model.Journal of Quantitative Spec-troscopy and Radiative Transfer,71,1–9.Kuusk,A.,Lang,M.,&Nilson,T.(2004).Simulation of the reflectance of ground vegeta-tion in sub-boreal forests.Agricultural and Forest Meteorology,126,33–46. Lang,M.,Kuusk,A.,Nilson,T.,Lükk,T.,Pehk,M.,&Alm,G.(2002).Reflectance spectra of ground vegetation in sub-boreal forests http://www.aai.ee/bgf/ger2600/index.html Accessed March,2011Li,X.,&Strahler,A.(1985).Geometric-Optical modelling of a conifer forest canopy.IEEE Transactions on Geoscience and Remote Sensing,23,207–221.Lucht,W.,Schaaf,C.B.,&Strahler,A.H.(2000).An algorithm for the retrieval of albedo from space using semiempirical BRDF models.IEEE Transactions on Geoscience and Remote Sensing,38,977–998.McDonald,A.J.,Gemmell,F.M.,&Lewis,P.E.(1998).Investigation of the utility of spectral vegetation indices for determining information on coniferous forests.Remote Sensing of Environment,66,250–272.Miller,J.,White,P.,Chen,J.,Peddle,D.,McDemid,G.,Fournier,R.,et al.(1997).Seasonal change in the understory reflectance of boreal forests and influence on canopy vegetation indices.Journal of Geophysical Research,102,29475–29482.Peltoniemi,J.I.,Kaasalainen,S.,Näränen,J.,Rautiainen,M.,Stenberg,P.,Smolander,H., et al.(2005).BRDF measurement of understory vegetation in pine forests:Dwarf shrubs,lichen,and moss.Remote Sensing of Environment,94,343–354.Pinty,B.,Lavergne,T.,Kaminski,T.,Aussedat,O.,Giering,R.,Gobron,N.,et al.(2008).Partitioning the solar radiantfluxes in forest canopies in the presence of snow.Journal of Geophysical Research,113.doi:10.1029/2007JD009096D04104.Pinty,B.,Widlowski,J.L.,Gobron,N.,Verstraete,M.M.,&Diner,D.J.(2002).Unique-ness of multiangular measurements—Part1:An indicator of subpixel surface heterogeneity from MISR.IEEE Transactions on Geoscience and Remote Sensing, 40,1560–1573.Pisek,J.,&Chen,J.M.(2009).Mapping forest background reflectivity over North Amer-ica with Multi-angle Imaging SpectroRadiometer(MISR)data.Remote Sensing of Environment,113,2412–2423.Pisek,J.,Chen,J.M.,Miller,J.R.,Freemantle,J.R.,Peltoniemi,J.I.,&Simic,A.(2010).Mapping forest background reflectance in a boreal region using multi-angle Compact Airborne Spectrographic Imager(CASI)data.IEEE Transactions on Geoscience and Remote Sensing,48,499–510.Pocewicz,A.,Vierling,L.,Lentile,L.,&Smith,R.(2007).View angle effects on relation-ships between MISR vegetation indices and leaf area index in a recently burned ponderosa pine forest.Remote Sensing of Environment,107,322–333.Quaife,T.,&Lewis,P.(2010).Temporal constraints on linear BRDF model parameters.IEEE Transactions on Geoscience and Remote Sensing,48,2445–2450. Rautiainen,M.(2005).Retrieval of leaf area index for a coniferous forest by inverting a forest reflectance model.Remote Sensing of Environment,99,295–303. Rautiainen,M.,Suomalainen,J.,Mõttus,M.,Stenberg,P.,Voipio,P.,Peltoniemi,J.,et al.(2007).Coupling forest canopy and understory reflectance in the Arctic latitudes of Finland.Remote Sensing of Environment,110,332–343.Rautiainen,M.,Mõttus,M.,Heiskanen,J.,Akujärvi,A.,Majasalmi,T.,&Stenberg,P.(2011).Seasonal reflectance dynamics of common understory types in a Northern European boreal forest.Remote Sensing of Environment.doi:10.1016/j.rse.2011.06.005.Rees,W.,Tutubalina,O.,&Golubeva,E.(2004).Reflectance spectra of subarctic lichens between400and2400nm.Remote Sensing of Environment,90,281–292. Rentch,J.S.,Fajvan,M.A.,&Hicks,R.R.,Jr.(2003).Oak establishment and canopy ac-cession strategies infive old-growth stands in the central hardwood forest region.Forest Ecology and Management,184,285–297.Roujean,J.-L.,Leroy,M.,&Deschamps,P.Y.(1992).A bi-directional reflectance model of the Earth's surface for the correction of remote sensing data.Jorunal of Geophysical Research,D-97,20455–20468.Schaaf,C.B.,Gao,F.,Strahler,A.H.,Lucht,W.,Li,X.,Tsang,T.,et al.(2002).First oper-ational BRDF albedo,nadir reflectance products from MODIS.Remote Sensing of Environment,83,135–148.Schaaf,C.B.,Wang,Z.,&Strahler,A.H.(2011).Commentary on Wang and Zender–MODIS snow albedo bias at high solar zenith angles relative to theory and to in situ observations in Greenland.Remote Sensing of Environment,115,1296–1300. Spanner,M.,Pierce,L.,Peterson,D.,&Running,S.(1990).Remote sensing of temperate coniferous leaf area index:The influence of canopy closure,understory vegetation and background reflectance.International Journal of Remote Sensing,11,95–111. Tan,B.,Woodcock,C.E.,Hu,J.,Zhang,P.,Ozdogan,M.,Huang,D.,et al.(2006).The impact of gridding artifacts on the local spatial properties of MODIS data:Impli-cations for validation,compositing,and band-to-band registration across resolu-tions.Remote Sensing of Environment,105,98–114.Vermote,E.,Justice,C.,&Breon,F.(2009).Towards a generalized approach for correc-tion of the BRDF effect in MODIS directional reflectances.IEEE Transactions on Geoscience and Remote Sensing,47,898–908.Vogel,J.S.,&Gower,S.T.(1998).Carbon dynamics of boreal jack pine stands with and without a green adler understory.Ecosystems,1,386–400.Wang,Y.J.,Lyapustin,A.I.,Privette,J.L.,Cook,R.B.,SanthanaVannan,S.K.,Vermote,E.F., et al.(2010).Assessment of biases in MODIS surface reflectance due to Lambertian approximation.Remote Sensing of Environment,114,2791–2801.Waring,R.H.,&Law,B.E.(1994).Remote sensing of leaf area index and radiation intercepted by understory vegetation.Ecological Applications,4,272–279. White,H.P.,Miller,J.R.,&Chen,J.M.(2002).Four-scale linear model for anisotropic reflectance(FLAIR)for plant canopies—Part II:Validation and inversion with CASI,POLDER,and PARABOLA data at BOREAS.IEEE Transactions on Geoscience and Remote Sensing,40,1038–1046.White,P.H.,Deguise,J.,Schwartz,J.,Hitchcock,R.,&Staenz,K.(2002).Defining shaded spectra by model inversion for spectral unmixing of hyperspectral datasets—Theory and preliminary application.Proceedings of the International Geosci.and Remote Sens.Symposium(IGARSS)and24th Canadian Symposium on Remote Sensing (pp.989–991).White,P.H.,Miller,J.R.,&Chen,J.M.(2001).Four-scale linear model for aniso-tropic reflectance(FLAIR)for plant canopies—Part I:Model description and partial validation.IEEE Transactions on Geoscience and Remote Sensing,39, 1073–1083.468J.Pisek et al./Remote Sensing of Environment117(2012)464–468。
MISR和MODIS二向性反射数据产品的对比分析陈永梅;王锦地;梁顺林;王东伟;马斌【期刊名称】《遥感学报》【年(卷),期】2009(013)005【摘要】The Bidirectional Reflectance Distribution Function (BRDF) of land surfaces specifies the behavior of surface directional reflectance as a function of illumination and viewing angles. The bidirectional reflectance products of the Moderate resolution Imaging Spectroradiometer (MODIS) and the Multiangle Imaging Spectroradiometer (MISR) and their BRDF model parameters products have been available respectively. Since the BRDF model parameters are inverted from limited angular observations for a given pixel, it is necessary to evaluate whether they can effectively characterize the directional reflectance in other viewing directions. In this study, we choose four kinds of land cover types to analyze the representation of MISR and MODIS BRDF model parameters, by comparing the directional reflectance extrapolated using the BRDF model parameters with the observed directional reflectance. The results show that: (1) both MODIS and MISR BRDF model parameters have better representation on directional reflectance at some viewing directions. Especially MISR BRDF model parameters have preferable representations at MODIS viewing directions; (2) the representative ability of both BRDF model parameters tends to weaken when the viewing zenith angle increases; (3) from theanalysis of the observed data set shown in this paper, it seems that the representative ability of MODIS BRDF models is better near cross-principal plane than that near principal plane.%利用4种地表类型共16个地面站点的MISR 9个角度反射率数据产品和MODIS反射率数据产品(MOD09A1),对比两种传感器获取的地表方向反射观测数据和用BRDF模型参数产品计算的方向反射数据,分析了这两种BRDF模型参数的外摊能力.研究结果表明:(1)MISR和MODIS BRDF模型参数数据产品都具有一定的对观测数据以外方向反射的外推能力,相比用MISR BRDF模型参数数据外推到相应MODIS观测角度方向反射的结果一致性较强.(2)采用两种BRDF模型参数推算的方向反射率与观测数据的接近程度有随观测天顶角的增大而减弱的趋势,在观测天顶角较小时一致性较好.(3)所用数据处理结果显示,MODISBRDF模型参数数据产品在近垂直主平面方向反射率推算结果相比近主平面方向的推算结果更接近观测数据.【总页数】15页(P801-815)【作者】陈永梅;王锦地;梁顺林;王东伟;马斌【作者单位】北京师范大学,地理学与遥感科学学院,100875;解放军理工大学气象学院,江苏,南京,211101;北京师范大学,地理学与遥感科学学院,100875;美国马里兰大学,地理系,MD20742;北京师范大学,地理学与遥感科学学院,100875;北京师范大学,地理学与遥感科学学院,100875【正文语种】中文【中图分类】TP702【相关文献】1.偏振反射与二向性反射的关系 --以不同物候期杨树单叶的室内光谱测量为例 [J], 宋开山;赵云升;张柏2.花岗岩表面二向性镜面反射分量和漫反射分量的比较研究 [J], 赵乃卓;赵云升;晏磊;吴太夏;相云3.多角度偏振反射与二向性反射定量关系初探 [J], 赵云升;吴太夏;胡新礼;罗杨洁4.基于地面光谱数据的二向性反射模拟与分析 [J], CHEN Lei;QIAN Da;ZHANG Hu;CUI Tie-jun5.基于航天与航空多角度观测提取二向性反射系数与方向半球反射信息(英文) [J], JohnV.Martonchik因版权原因，仅展示原文概要，查看原文内容请购买。
NASA对于有这方面兴趣的人，我推荐一本书：《地球卫星遥感》共有两卷。
主要是有关中分辨串成像光谱仪（MODIS）产品的信息和应用，介绍了美国国家极轨环境卫系统（NPOESS）和NPOESS预备计划（NPP），还探讨了其他卫星遥感装备和应用，论及NASA 用于监测和探测地球变化的主要卫星系统——地球观测系统（EOS），EOS包括的卫星Terra、Aqua 和Aura及其装载的MODIS、AIRS、AMSU、AMSR-E、OMI等遥感仪器，并讨论NPP将携带的4个NPOESS系统重要部件：可见光红外成像辐射组件（VIIRS），航线交叉红外探测器（CrIS），先进技术微波探测器（ATMS）以及臭氧成图和廓线仪装置（OMPS）。
既包括现代遥感技术的基础知识，又涉及卫星遥感的领域。
其中负责观测陆地的Terra、负责观测地球水循环的Aqua和负责搜集大气数据的Aura共同组成了完整的eos地球观测系统，服务于nasa的地球科学计划（ese）。
1 GRACE10. Gravity Recovery and Climate Experiment (GRACE)重力恢复与气候实验The primary goal of the GRACE mission is to accurately map variations in the Earth's gravity field over its 5-year lifetime. The GRACE mission has two identical spacecrafts flying about 220 kilometers apart in a polar orbit 500 kilometers above the earth.It will map the Earth's gravity fields by making accurate measurements of the distance between the two satellites, using geodetic quality Global Positioning System (GPS) receivers and a microwave ranging system. This will provide scientists from all over the world with an efficient and cost-effective way to map the Earth's gravity fields with unprecedented accuracy. The results from this mission will yield crucial information about the distribution and flow of mass within the Earth and it's surroundings.The gravity variations that GRACE will study include: changes due to surface and deep currents in the ocean; runoff and ground water storage on land masses; exchanges between ice sheets or glaciers and the oceans; and variations of mass within the earth. Another goal of the mission is to create a better profile of the Earth's atmosphere. The results from GRACE will make a huge contribution to NASA's Earth science goals, Earth Observation System (EOS) and global climate change studies.GRACE is a joint partnership between the NASA in the United States and Deutsche Forschungsanstalt fur Luft und Raumfahrt (DLR) in Germany. Dr. Byron T apley of The University of Texas Center for Space Research (UTCSR) is the Principal Investigator (PI), and Dr. Christoph Reigber of the GeoForschungsZentrum (GFZ) Potsdam is theCo-Principal Investigator (Co-PI). Project management and systems engineering activities are carried out by the Jet Propulsion Laboratory.9. TerraTerra is a multi-national, multi-disciplinary mission involving partnerships with the aerospace agencies of Canada and Japan. Managed by NASA’s Goddard Space Flight Center, the mission also receives key contributions from the Jet Propulsion Laboratory and Langley Research Center. Terra is an important part of NASA’s Science Mission, helping us better understand and protect ourhome planet.NASA launched the Earth Observing System's flagship satellite "Terra," named for Earth, on December 18, 1999. Terra has been collecting data about Earth's changing climate. Terra carries five state-of-the-art sensors that have been studying the interactions among the Earth's atmosphere, lands, oceans, and radiant energy. Each sensor has unique design features that will enable scientists to meet a wide range of science objectives. The five Terra onboardsensors are: ASTER, or Advanced Spaceborne Thermal Emission and Reflection Radiometer (先进星载热发射和反射辐射仪)∙CERES, or Clouds and Earth's Radiant Energy System∙MISR, or Multi-angle Imaging Spectroradiometer∙MODIS, or Moderate-resolution Imaging Spectroradiometer(中分辨率成像光谱仪)∙MOPITT, or Measurements of Pollution in the Troposphere Corpus Christi, TexasThe city of Corpus Christi, Texas, is tucked against the southern shore of Corpus Christi Bay on the Gulf of Mexico. Inland, the city is surrounded by the large, green grid of croplands. To the south and east, the landscape is dominated by marshes, lagoons, and barrier islands, the longest of which is Padre Island. Although the part of Padre I sland visible in this scene is developed with roads, residences, and resorts, just south of the southern edge of the scene, Padre Island National Seashore begins. The seashore is the longest remaining undeveloped stretch of barrier island in the world.Upstream of Corpus Christi Bay is Nueces Bay, which takes its name from one of the two freshwater inputs to the bay system, the Nueces River. The other is Oso Creek, which flows into Corpus Christi Bay along the south shore. The Corpus Christi Bay estuary is located in a semi-arid region, and the total freshwater input into the system is naturally low. Flows are further diminishedby irrigation and urban water demands.These factors combine to make the system particularly sensitive to accumulation of water pollutants and salt, which compromises the health of the plants and animals that live in the estuary (including commercially and recreationally important species such as oysters and shrimp.) For these reasons, the Environmental Protection Agency has included the Corpus Christi Bay Estuary in its National Estuary Program. Their goal is to develop water re-use and conservation strategies that will meet urban, agricultural, and ecological needs as the city continues to grow.Satellite images such as this view from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) on NASA’s Terra satellite captured on June 29, 2007, can document land cover changes such as the conversion of natural landscapes to cropland, or cropland to urban development. Information on how fast and where changes are occurring can help scientists and urban planners predict future water supply and demand.2 AMSREA MSR - E通过测量来自地球表面的微波辐射来研究全球范围的水循环变化。
拉曼光谱仪的工作参数介绍概述拉曼光谱仪是一种利用拉曼散射效应分析样品的仪器。
拉曼散射现象是指，当激光照射样品时，样品中的分子或晶体因振动而引起散射光，且散射光的频率比激光光源的频率稍微降低，这种现象被称为拉曼散射效应。
拉曼光谱可以用于物质的分析、组成分析、结构分析等。
拉曼光谱仪是一种非常重要的分析仪器，可以用于化学、材料、地质、生物等领域。
本文将对拉曼光谱仪的工作参数进行介绍。
光源拉曼光谱仪的光源通常是激光，激光的波长是固定的。
常见的激光波长有532nm、785nm、1064nm等。
不同波长的激光适用于不同类型的样品。
特别地，可见激光主要适用于有机物和有机高分子的分析。
而近红外激光则适用于无机化合物和矿物质的分析。
红光、绿光和蓝光的激光波长分别为650nm、532nm、488nm。
分光镜拉曼光谱仪利用分光镜分离散射光和激光，并将散射光投射到探测器上。
拉曼光谱分光镜一般有两种，低通滤光片和金属滤光片。
低通滤光片的作用是将激光衰减，以避免光谱的偏移和饱和。
金属滤光片通常用于在样品被激光照射之前消除自发荧光信号。
反射镜反射镜是拉曼光谱仪的重要组成部分，可将激光引导到样品上，使激光与样品产生相互作用，然后将样品的信号收集到探测器上。
反射镜分为两种：倾斜反射镜和平面反射镜。
平面反射镜是将激光与样品垂直以实现散射，而倾斜反射镜则在一定角度下将激光引向样品。
探测器探测器是测量拉曼光谱的关键组件。
常见的探测器有CCD探测器和光电倍增管探测器。
CCD探测器具有高分辨率、低背景噪声等优点，适用于高精度的测量。
而光电倍增管探测器则具有高速度、高灵敏度等优点，适用于大量样品快速测量。
其他参数除了上述参数外，拉曼光谱仪还有一些其他的参数需要注意：•分辨率：分辨率是指一个光谱测量中重心波数和半高峰宽度的示例差异。
分辨能力越好，就可以得到更加精确的光谱数据。
•精度：精度是指测量结果与真实值的偏差，精度越高，测量结果越可靠。
•灵敏度：灵敏度是指在低样品浓度下，还能够检测到样品中的信号强度。
第38卷第8期 2017年8月哈尔滨工程大学学报Journal of Harbin Engineering UniversityVol.38 No. 8Aug. 2017光谱相似性度量方法研究进展赵春晖，田明华，李佳伟(哈尔滨工程大学信息与通信工程学院，黑龙江哈尔滨150001)摘要：为了进一步分析光谱相似性度量在高光谱图像处理中的应用，从距离、投影等角度充分归纳总结了现有二元光谱相似度量方法，并分析讨论了二元光谱相似度量存在的问题。
重点介绍了一种多元光谱相似性测量方法，也称N 维立体光谱角（N-dimensional solid spectral angle, NSSA)方法。
NSSA方法从本质上突破了传统的二元光谱角（spec- tral angle mapping, SAM)仅能计算两个光谱之间夹角的局限性，具备联合计算多元光谱欧氏空间夹角的能力，为评价多元光谱联合相似性提供了一种定量化的度量手段。
最后，对NSSA方法在高光谱波段选择及端元提取领域的潜在研究价值和应用现状进行了分析和展望。
通过分析表明NSSA方法所具备的特性可更好地实现光谱相似性度量，在 高光谱图像处理领域具有较高的研究价值。
关键词：高光谱图像；光谱相似性度量；二元光谱角；N维立体光谱角方法；多元光谱相似性度量；波段选择；端 元提取D O I: 10. 11990/jheu. 201612063网络出版地址：http://www. cnki. net/kcms/detail/23. 1390. u.20170428. 1622. 064. html中图分类号:TN911. 73文献标志码:A文章编号= 1006-7043(2017)08-1179-11Research progress on spectral similarity metricsZHAO Chunhui,TIAN Minghua,LI Jiawei(College of Information and Communication Engineering, Harbin Engineering University, Harbin 150001, China) Abstract ：Spectral similarity metrics are important in the field of hyperspectral data analysis.To further analyze their application in hyperspectral image processing,the current binary spectral similarity metrics method based on distance or projection was summarized.The problems in the binary spectral similarity metrics were analyzed and discussed.Then,this study chiefly introduced a multiple spectral similarity metric called N-dimensional solid spectral angle (NSSA).The NSSA method breaks through the limitation of the traditional binary spectral angle mapping in essence,which can not only calculate the angle between two spectra but also the angle constructed by multiple spectra jointly in Euclidean space.The method provides a quantitative measure to evaluate the joint similarity of multivariate spectra.The potential research value and applications of the NSSA method in hyperspectral band selection and endmembers extraction were analyzed and forecasted.The analysis indicates that the NSSA method can better realize the spectral similarity measure and has high research value in the field of hyperspectral imaging process.Keywords：hyperspectral imagery；spectral similarity metrics;binary spectral angle mapping；N dimensional solid spectral angle；multiple spectra similarity metric；band selection；endmember extraction随着成像光谱仪的发展，高光谱遥感成像技术 得到了广泛的关注。
Meris 介绍The MERIS Product Handbook can be downloaded from http://envisat.esa.int/dataproducts/. MERIS sample data are freely available at http://envisat.esa.int/services/sample_products/meris/. For a full list of the reports, see http://earth.esa.int/pcs/envisat/meris/reports/cyclic.Access_to_MERIS_data document (seehttp://earth.esa.int/pcs/envisat/meris/documentation/Access_to_MERIS_data.pdf)or directly contacting the EOHelp@esa.int.Image of the dayhttp://www.brockmann-consult.de/ImageOfTheDay/index.htmMERIS stands for “MEdium Resolution Imaging Spectrometer”.MERIS – (A)ATSR Workshop (26-30 September 2005)http://envisat.esa.int/workshops/meris_aatsr2005/ENVISAT/ERS Symposium (6-10 September 2004)http://earth.esa.int/salzburg04/ENVISAT MERIS and AATSR Validation Team Workshop - MAVT-2003 (20 - 24 October 2003) http://envisat.esa.int/workshops/mavt_2003/Envisat Validation Workshop (9-13 December 2002)http://envisat.esa.int/pub/ESA_DOC/envisat_val_1202/proceedings/Envisat Calibration Review (9-13 September 2002)http://envisat.esa.int/calval/proceedings/ERS-Envisat Symposium "Looking down to Earth in the New Millennium" (16-20 October 2000) http://earth.esa.int/pub/ESA_DOC/gothenburg/start.pdfESA Data User Element (DUE) Projectshttp://dup.esrin.esa.it/index.aspESA EO PI Projectshttp://eopi.esa.int/esa/esa工具软件1.The Basic ERS & Envisat (A)ATSR and Meris Toolbox (BEAM) is a collection of executable tools and an application programming interface (API) which has been developed to facilitatethe utilisation, viewing and processing of ESA MERIS, (A)ATSR and ASAR data. Seehttp://www.brockmann-consult.de/beam/ and in particular the BEAM FAQ at addresshttp://www.brockmann-consult.de/BEAMWiki/Wiki.jsp?page=FAQ .2.The EnviView tool provided by ESA (see http://earth.esa.int/services/tools/enviview/) can alsobe used to convert ENVISAT data into hdf file format. This allows the data to be automaticallyread by any third-party software supporting this format.All the Envisat tools are available from: http://www.envisat.esa.int/services/tools_table.html.ENVISAT is an advanced polar-orbiting Earth observation satellite providing measurements of the atmosphere, ocean, land, and ice (see http://envisat.esa.int/). It was launched on 1st March 2002 by ESA.Envisatsun-synchronous polar orbitabout 800-km altitude.The repeat cycle of thereference orbit is 35 days, although most sensors, being wide swath, providecomplete coverage of the globe within one to three days.The MERIS level 2 products provide geophysical information ready to be used for various applications. The primary mission of MERIS is tomonitor the ocean colour including chlorophyll concentrations for open oceans and coastal areas, yellow total suspended matter.In addition, MERIS provides with land parameter measurements like vegetation indices and atmospheric parameters like water vapour, cloudtop pressure, cloud types, aerosol optical thickness, and Angstr?m coefficients.MERIS is a push-broom instrument composed of five cameras (also called modules). Swath of each camera overlaps with the successive one. Each camera is composed of one CCD array for each oneof the 15 bands. The ground sampling for one CCD (resolution) represents 260 metres across track by 290 metres along track in a “full resolution” mode.A on-board electronic unit computes a combination of four adjacent samples across-track over four successive lines leading to a “reduced resolution” of 1040 metres across track by 1160 metres alongtrack.The full resolution is processed on-demand while the reduced resolution is processed systematically;The instrument's 68.5° field of view around nadir covers a swath width of 1150 km.MERIS product processing levelslevel 1B – are images resampled on a path-oriented grid, with pixel values having been calibrated to match the Top Of Atmosphere (TOA) radiance.level 2 – are images deriving form the level1B products, with pixel values having beenprocessed to get geophysical mesurements.level 3 – are synthesis of more than one MERIS products (and possibly external data) to display geophysical measurements for a time period.Disclaimers addressing issues affecting MERIS product quality are published athttp://envisat.esa.int/dataproducts/availability/.Full Resolution260 m across track 290 m along trackReduced Resolution 1040 m across track 1160 m along trackAlgal Pigment Index IThe MERIS algal pigment index is a measurement of the concentration expressed in Log10(mg/m3)of phytoplankton (algae) in the water.The concentration is derived by the direct relationship between the ratio of the blue and green signa l leaving the water surface and the concentration of algal pigments.The relationship, based on published data,is valid over clear waters and spans a concentration range from mg/m3 to tens of g/m3 .Note:The Algal Pigment Index I is not applicable in water with significant amounts of suspended matter or yellow substance. In such cases, corresponding product confidence flag is raised.For more information, see ATBD 2.9Algal Pigment Index IIThe second MERIS algal pigment index is also a measurement of the concentration expressed inLog10(mg/m3) of phytoplankton (algae) in the water but, is part of a suite of oceanic products derived by inverting a model of the optical properties of the ocean by the use of a neural network. The otheroceanic products are suspended matter and yellow substance.In clear waters the Algal Pigment Index II product is more noisy than the Algal Pigment Index I. For more information, see ATBD 2.12the “smile effect ?http://earth.esa.int/pcs/envisat/meris/documentation/MERIS_Smile_Effect.pdfThe spectral measurements of each pixel along an image line are made by its own set of CCD sensors: this causes small variations of the spectral wavelength of each pixel along the image that constitute the so-called "smile effect". The variation of the wavelength per pixel is in order of 1nm from one camera to another, while they are in the order of 0.1nm within one camera.Even though this variation is small compared to the spectral bandwidth of a band, which is typically10nm, and can hardly be seen in an image, it can cause disturbances in processing algorithms, which require very precise measurements, for example the retrieval of chlorophyll in the ocean. These disturbances could result in a visual artefact, "camera borders.The smile effect is corrected for level 2 products.。
Mapping annual mean ground-level PM2.5concentrations usingMultiangle Imaging Spectroradiometer aerosol optical thicknessover the contiguous United StatesYang Liu1,2,Rokjin J.Park,3Daniel J.Jacob,3Qinbin Li,3Vasu Kilaru,4and Jeremy A.Sarnat5Received15May2004;revised25July2004;accepted3August2004;published24November2004.[1]We present a simple approach to estimating ground-level fine particulate matter(PM2.5,particles smaller than2.5m m in diameter)concentrations by applying local scalingfactors from a global atmospheric chemistry model(GEOS-CHEM with GOCARTdust and sea salt data)to aerosol optical thickness(AOT)retrieved by the MultiangleImaging Spectroradiometer(MISR).The resulting MISR PM2.5concentrations arecompared with measurements from the U.S.Environmental Protection Agency’s(EPA)PM2.5compliance network for the year2001.Regression analyses show that the annualmean MISR PM2.5concentration is strongly correlated with EPA PM2.5concentration(correlation coefficient r=0.81),with an estimated slope of1.00and an insignificantintercept,when three potential outliers from Southern California are excluded.The MISRPM2.5concentrations have a root mean square error(RMSE)of2.20m g/m3,whichcorresponds to a relative error(RMSE over mean EPA PM2.5concentration)ofapproximately20%.Using simulated aerosol vertical profiles generated by the globalmodels helps to reduce the uncertainty in estimated PM2.5concentrations due to thechanging correlation between lower and upper tropospheric aerosols and therefore toimprove the capability of MISR AOT in estimating surface-level PM2.5concentrations.The estimated seasonal mean PM2.5concentrations exhibited substantial uncertainty,particularly in the west.With improved MISR cloud screening algorithms and the dustsimulation of global models,as well as a higher model spatial resolution,we expect thatthis approach will be able to make reliable estimation of seasonal average surface-levelPM2.5concentration at higher temporal and spatial resolution.I NDEX T ERMS:0305Atmospheric Composition and Structure:Aerosols and particles(0345,4801);0345Atmospheric Compositionand Structure:Pollution—urban and regional(0305);0394Atmospheric Composition and Structure:Instruments and techniques;K EYWORDS:MISR AOT,GEOS-CHEM,PM2.5Citation:Liu,Y.,R.J.Park,D.J.Jacob,Q.Li,V.Kilaru,and J.A.Sarnat(2004),Mapping annual mean ground-level PM2.5 concentrations using Multiangle Imaging Spectroradiometer aerosol optical thickness over the contiguous United States,J.Geophys. Res.,109,D22206,doi:10.1029/2004JD005025.1.Introduction[2]Epidemiological studies around the world have shown strong and consistent correlations between adverse health effects and outdoor fine particle matter(PM2.5,particles with diameters less than2.5m m)[Pope,2000].Some of the mortality studies as reviewed by Wallace[2000]have consistently shown an increase of1–8%in deaths per 50m g/m3increase in outdoor air particle concentrations without apparent threshold.The importance of long-term PM2.5monitoring has recently been emphasized in order to evaluate the health effects of low or moderate exposure as well as repeated exposure to elevated pollution levels [Samet et al.,2000;Schwartz et al.,1996].However,for many epidemiology studies,particle measurements from stationary ambient monitoring(SAM)sites have been used as surrogates of exposures for individuals living substantial distances(20–100miles)from the sites[Ito et al.,2001].[3]Since aerosol optical properties such as aerosol opti-cal thickness(AOT,a dimensionless measure of aerosol abundance and its light extinction capability in the entire air column)derived from satellite observations is directly related to particle mass loading[Chow et al.,2002;Malm et al.,1994],studying its association with surface-level PM2.5mass concentration may provide a cost effectiveJOURNAL OF GEOPHYSICAL RESEARCH,VOL.109,D22206,doi:10.1029/2004JD005025,2004 1Division of Engineering and Applied Sciences,Harvard University,Cambridge,Massachusetts,USA.2Now at ENVIRON,Arlington,Virginia,USA.3Department of Earth and Planetary Sciences,Harvard University,Cambridge,Massachusetts,USA.4National Exposure Research Laboratory,U.S.Environmental Protec-tion Agency,Research Triangle Park,North Carolina,USA.5Department of Environmental Health,Harvard School of PublicHealth,Boston,Massachusetts,USA.Copyright2004by the American Geophysical Union.0148-0227/04/2004JD005025$09.00way for PM2.5pollution monitoring.To date,satellite remote sensing has been applied to monitor long-range transport of Asian and Saharan dust[Chiapello and Moulin, 2002;Husar et al.,2001;Wang et al.,2003]and charac-terize ambient particulate pollution[Falke et al.,2001;Liu et al.,2002;Sifakis and Deschamps,1992].[4]The launch of NASA EOS satellite Terra in December 1999provided a new opportunity for monitoring particle pollution from space.Two instruments aboard Terra,MISR and the Moderate-Resolution Imaging Spectroradiometer (MODIS),were specially designed to retrieve aerosol opti-cal properties including AOT over most of the land surface and the oceans[Diner et al.,1998;Kaufman et al.,1998, 2002;Martonchik et al.,1998].Preliminary studies have shown that MODIS AOT data can be used to detect and track the transport of pollutants and extreme pollution episodes[Engel-Cox et al.,2004;Hutchison,2003].In addition,a strong linear relationship(correlation coefficient r=0.70)was found between MODIS AOT measurements and24-hour PM2.5concentrations from seven sites in Alabama,indicating a good potential for satellite derived aerosol optical properties to be used in air quality studies [Wang and Christopher,2003].[5]We showed in a previous study that an empirical regression model using MISR AOT and a few geographical and meteorological parameters is able to estimate surface-level24-hour average PM2.5concentrations within approx-imately45%,with a correlation coefficient of approximately 0.7between observed and predicted PM2.5concentrations (Y.Liu et al.,Estimating ground level PM2.5over the eastern United States using satellite remote sensing, submitted to Environmental Science and Technology, 2004,hereinafter referred to as Liu et al.,submitted manuscript,2004).This paper extends our previous study by developing a simple approach that establishes a predictive relationship between surface PM2.5concentra-tions and AOT.This simple approach relies on a global chemistry and transport model(CTM)to provide a better physical basis for relating satellite AOT measurements to the spatial and temporal pattern of surface PM2.5 concentrations.[6]In section2,we describe this approach,which involves analysis of data from the CTMs,PM2.5data from the U.S.Environmental Protection Agency’s(EPA)PM2.5 compliance network and speciation and trend network (STN),and AOT retrievals from MISR for the year2001. In section3,the summary statistics of the different data sets are presented.The agreement in geographical patterns and seasonal variations between model and observations is discussed.PM2.5concentrations derived from the MISR AOT are compared with EPA PM2.5data and sources of uncertainties are discussed in detail.Finally,major findings and potential future improvements to the current analysis are summarized in section4.2.Description of Data and Method2.1.Simulated Aerosol Data by GEOS-CHEMand GOCART[7]The GEOS-CHEM model is a global3-D tropospheric chemistry and transport model driven by assimilated meteo-rological observations from the Goddard Earth Observing System(GEOS)of the NASA Global Modeling and Assim-ilation Office.The fully coupled oxidants-aerosol simulation by GEOS-CHEM provides sulfate(SO42À),nitrate(NO3À), ammonium(NH4+),elemental carbon(EC),and organic carbon(OC)aerosol concentrations for the period of2001 at3-hour temporal resolution,2°latitudeÂ2.5°longitude horizontal resolution,and30sigma vertical layers.When calculating AOT using aerosol dry mass concentrations, particle growth with increased relative humidity is taken into account by applying different hydroscopic growth factors to all hydrophilic species using local relative humidity condi-tions[Martin et al.,2003].Detailed descriptions of GEOS-CHEM as well as its aerosol simulations can be found elsewhere[Bey et al.,2001;Park et al.,2003,2004].The lowest model levels are centered at approximately10,50, 100,200,and400m above the surface.[8]Since EPA’s compliance network measures24-hour average PM2.5concentration according to a fixed sampling schedule(every third or sixth day)regardless of weather conditions,seasonal mean GEOS-CHEM surface PM2.5 concentrations were calculated using all eight3-hour out-puts per day for each species.The calculation of columnar AOT values followed the methodology given by Chin et al. [2002].The3-hour values were first interpolated to10a.m. local time values(MISR measurement time window),sam-pled on the dates when MISR had valid AOT retrievals, then integrated into seasonal averages in order to be compared with seasonal mean MISR AOT values.[9]Monthly mean dust and sea salt concentrations for 2001from the Georgia Tech/Goddard Global Ozone Chem-istry Aerosol Radiation and Transport(GOCART)model were used to complement GEOS-CHEM aerosol fields. General descriptions of the GOCART model simulation of dust and sea salt are provided elsewhere[Chin et al.,2002; Ginoux et al.,2001].Particle concentrations and the asso-ciated AOT values for fine sea salt(effective radius of 0.80m m),and fine dust(sum of effective radius ranges0.1–0.18m m,0.18–0.30m m,0.3–0.6m m,0.6–1m m,and part of 1–1.8m m)provided by the GOCART model were com-bined with the particle concentrations for SO42À,NO3À,NH4+, EC,and OC and their associated AOT values provided by GEOS-CHEM.From this point on,we refer to the total aerosol AOT and total surface aerosol concentration pre-dicted by combining GEOS-CHEM and GOCART outputs as simulated AOT and simulated PM2.5concentration.2.2.EPA24-Hour Average PM2.5Mass Concentration and Speciation Data[10]The EPA’s PM2.5compliance network was initiated in 1997and designed to measure compliance of both the annual and24-hour PM2.5National Ambient Air Quality Standard (NAAQS).Daily average PM2.5concentrations measured by gravimetric methods(Federal Register40CFR part50, 5Feb.1998)in2001from1137sites of EPA’s compliance network,primarily located in urban areas and surrounding suburbs,were collected and integrated into seasonal averages in each GEOS-CHEM model grid cell(Figure1).Validated daily average mass concentrations of SO42À,EC,OC,and mineral dust were collected from131sites of EPA’s PM2.5 speciation trends network(STN)and integrated into seasonal averages in2°Â2.5°model grid in order to analyze the difference between simulated PM2.5concentrations and EPAPM 2.5measurements for individual aerosol components [Rao et al.,2002](Figure 1).To reduce the influence of potential outliers,only the grid cells with more than 30EPA measure-ments per season were included in the analysis.2.3.MISR AOT Retrievals[11]MISR AOT data have a spatial resolution of 17.6km and achieve global coverage in nine days [Diner et al.,1998;Martonchik et al.,2002].It is most sensitive to particles in the diameter range from 0.05to 2.0m m [Kahn et al.,1998],corresponding to the size range of PM 2.5.All MISR AOT data (mostly version 12)that covered the contiguous United States for 2001were obtained from the NASA Langley Research Center Atmospheric Sciences Data Center (/$imswww/imswelcome/index.html),and then integrated into seasonal averages in 2°Â2.5°model grid cells.We excluded AOT values greater than 1.5from our analysis because they are probably contaminated by clouds (D.Diner,personal communication,2003).In addition,the data from North Dakota,South Dakota and Minnesota in the winter and spring is excluded because of a potential cloud contamination in MISR AOT.This issue will be further discussed in the following analysis.We previously showed that the overall retrieval error of MISR AOT is D AOT =±0.04±0.18ÂAOT over the contig-uous United States [Liu et al.,2004].The seasonal and geographical variability of MISR AOT errors was partly corrected in this analysis by applying linear regressions between MISR and AERONET AOT values presented by Liu et al.[2004].Finally,only those grid cells with more than 30AOT measurements per season were included in the analysis in order to reduce the influence of potential outliers.2.4.Coupling of the Global Models With MISR[12]We previously showed that an empirical regression model is able to predict surface PM 2.5concentrations byusing MISR AOT data and simple meteorological and geographical predictors with a relative error of approxi-mately 45%(r %0.7)and with no significant biases compared to observations (Liu et al.,submitted manuscript,2004).However,half of the variability in PM 2.5concen-trations cannot be explained probably because of the lack of information on aerosol vertical profile and long-range aerosol transport events.In addition,empirical models must be calibrated before transferring to other regions.We here use the simulated AOT and PM 2.5concentrations from GEOS-CHEM and GOCART model to define a physically consistent relationship between AOT and surface-level PM 2.5concentration:MISR PM 2:5Concentration¼Simulated Surface Level PM 2:5Concentration Simulated Column AOTÂMISR AOTð1ÞThis relationship,as defined in equation (1),is then applied to MISR AOTs to infer PM 2.5distributions.We refer to the PM 2.5concentrations derived from this simple model as MISR PM 2.5concentrations hereinafter.The terms of particle mass concentrations and optical properties used in this analysis is summarized in Table 1.[13]The MISR PM 2.5concentrations differ from the simulated PM 2.5concentrations in three ways.First,the MISR PM 2.5concentrations are less likely to be affected by possible biases in the aerosol vertical distribution estimated by the global models because the biases are attenuated by the ratio of simulated PM 2.5concentrations over simulated AOTs.For example,if both simulated PM 2.5concentrations and AOT have consistent biases (i.e.,both high or both low),the uncertainty of MISR AOT measurements will be more influential in determining the uncertainty in MISR PM 2.5than either simulated PM 2.5concentration orsimu-Figure 1.Spatial distribution of the EPA PM 2.5compliance monitoring sites (FRM sites,circles)and speciation and trend sites (STN sites,triangles)in the contiguous United States.Data in this study are collected from 1137FRM sites and 131STN sites for the year 2001.lated AOT.Therefore the MISR PM2.5is likely to have less discrepancy in spatial and seasonal patterns than simulated PM2.5concentrations when compared with EPA measure-ments.Second,because MISR AOT has a much higher spatial resolution than the global model simulations,MISR PM2.5concentrations would be able to reflect the impact of subgrid variation of particle properties.Finally,it has been shown that the discrepancy between gravimetric PM2.5 concentrations and the sum of all measured particle com-ponents can be as large as28–42%[Andrews et al.,2000; Turpin and Lim,2001].This discrepancy is likely due to uncertainties in organic carbon and dust measurements. These differences between global model predictions and EPA PM2.5measurements are likely to be reduced with the calibration of MISR AOT.It should be noted that equation (1)assumes that the atmospheric column is dominated by one aerosol component.When two or more important aerosol components with different optical properties and vertical distributions are present,AOT and PM2.5concen-trations are likely to have a nonlinear relationship.Under such circumstances,MISR PM2.5derived from equation(1) would only be a first-order approximation of EPA PM2.5 measurements.3.Results and Discussion[14]The final data set consists of totally577seasonal data records,each containing the seasonal average EPA PM2.5measurement,simulated PM2.5concentration and AOT,MISR AOT,and MISR PM2.5concentrations in each GEOS-CHEM model grid cell,and159annual data records after averaging all seasonal records.The relationships among three PM2.5concentration parameters and two AOT parameters are studied using scatterplots,and Spearman’s correlation coefficients.In addition,reduced major axis lines are used to characterize the overall relationship between simulated PM2.5concentrations and EPA PM2.5 measurements,as well as the agreement between simulated AOT and MISR AOT retrievals[Hirsch and Gilroy,1984]. When comparing the MISR PM2.5concentrations with EPA PM2.5concentrations,simple linear regression is used because we are interested in examining the capability of equation(1)in estimating individual PM2.5concentration over a given grid cell.parison Between Simulated PM2.5andEPA PM2.5Measurements[15]The annual average simulated PM2.5concentration (8.36(mean)±3.28(standard deviation)m g/m3)is approx-imately20%lower than EPA PM2.5concentrations(10.76±3.14m g/m3)with a similar dynamic range(Table2).Figure2 compares annual average simulated PM2.5and observed PM2.5concentrations in the contiguous United States.The EPA PM2.5measurements are plotted on a0.5°Â0.5°grid. The annual mean simulated PM2.5concentrations capture the geographic characteristics of EPA PM2.5measurements very well nationwide with the exception of the San Joaquin Valley and Southern California where the models substan-tially underestimate PM2.5concentrations.A scatterplot shows that annual average simulated PM2.5concentrations have a good linear relationship with EPA measurements(r= 0.74,reduced major axis line slope=1.04)but with a negative offset of2.88m g/m3(Figure3).The three data points that apparently deviate from the general pattern of the data set are all from Southern California.Excluding the three potential outliers does not have a significant impact on the parameter estimates of the reduced major axis line. [16]Seasonally,the difference between simulated PM2.5 concentrations and EPA measurement is the largest in the winter(12.17m g/m3versus8.26m g/m3,32%difference)Table1.Definitions of Terms Used in This AnalysisTerm Unit DefinitionSimulated PM2.5concentration m g/m3sea salt and dust mass concentrations are derived from GOCART,and mass concentrationsfor SO42À,NO3À,NH4+,EC,and OC are derived from GEOS-CHEM;these two sets ofconcentration data are combined to form the total simulated PM2.5concentrations Simulated AOT unitless AOT values for sea salt and dust are from GOCART,and AOT values for the remainingparticulate species are from GEOS-CHEM;these two sets of AOT data are combined toform the total simulated AOT estimatesEPA PM2.5concentration m g/m3daily average PM2.5mass concentrations measured by gravimetric methods collected fromthe EPA PM2.5Monitoring and Compliance NetworkMISR AOT unitless total column AOT retrieved by MISR instrument aboard the Terra satelliteMISR PM2.5concentration m g/m3daily average PM2.5concentrations estimated by equation(1)Table2.Summary Statistics of Annual and Seasonal AverageMISR PM2.5Concentrations,EPA PM2.5Concentrations,ModelColumnar AOT,MISR AOT,and MISR PM2.5Concentrations inthe United States for the Year2001Season Variable Unit N Mean SD a Min MaxAnnual EPA PM2.5m g/m315910.76 3.14 4.7220.51simulated PM2.5m g/m31598.36 3.28 3.7017.17MISR PM2.5m g/m31599.68 3.68 3.2218.24MISR AOT unitless1590.130.030.050.22simulated AOT unitless1590.100.030.060.22Winter EPA PM2.512912.17 3.62 5.2124.77simulated PM2.51298.27 3.93 2.3219.69MISR PM2.51298.56 3.20 2.4317.99MISR AOT1290.080.020.030.11simulated AOT1290.070.030.030.16Spring EPA PM2.513510.23 3.34 3.8520.48simulated PM2.51358.50 3.46 2.9017.29MISR PM2.513511.51 5.00 4.2926.92MISR AOT1350.160.040.080.32simulated AOT1350.110.030.070.25Summer EPA PM2.515511.32 4.91 4.1522.98simulated PM2.51558.39 3.34 2.7417.39MISR PM2.515511.68 5.81 4.0829.23MISR AOT1550.180.080.050.38simulated AOT1550.130.050.060.29Fall EPA PM2.51589.98 3.18 3.2824.05simulated PM2.51588.32 3.18 3.7915.05MISR PM2.51587.71 3.37 2.0719.46MISR AOT1580.090.030.040.19simulated AOT1580.100.030.040.20a SD,standard deviation.and smallest in the fall (9.98m g/m 3versus 8.32m g/m 3,17%difference).The global models underestimate PM 2.5con-centrations by approximately 30%during the summer.Simulated PM 2.5concentrations are significantly correlated with EPA measurements in all seasons with the exception of the western United States where correlation coefficients are insignificant in the winter and summer (p >0.05)(Table 3).EPA measurements are substantially higher than simulated PM 2.5concentrations in California during the winter and the fall.In addition,during the winter,EPA measurements show strong spatial variation in the northwest region with a number of stations observing much higher PM 2.5concen-trations than other stations in the same region.Scatterplots of simulated PM 2.5concentrations versus EPA PM 2.5mea-surements show that simulated PM 2.5concentrations agree with observations better in the east than in the west(Figure 4).Larger scatter is found in the west especially in the winter and summer as compared to in the east.[17]The overall underestimation of PM 2.5concentrations might be attributed to the discrepancy between chemical and gravimetric measurements found in surface-level mon-itoring campaigns,with the sum of all component concen-trations often smaller than the gravimetric measurements of PM 2.5concentrations,as previously mentioned.The possi-ble reason for the summer bias in the east is described by Park et al.[2004].The weak correlation found in the west in the summer might be due to the high bias in GOCART monthly mean dust concentrations.Along the coast from Washington State to central California,GOCART dust concentrations are approximately 5–7m g/m 3.However,dust concentrations measured by STN sites are generally below 1.5m g/m 3.In addition,in the inland northwest region,GOCART also overestimates dust concentrations by a factor of 2–3.This high bias and disagreement in spatial pattern between simulated and observed dust con-centrations likely cause the insignificant correlation.Since in the eastern United States,PM 2.5concentrationsareFigure 3.Scatterplot of annual average simulated PM 2.5concentration versus EPA PM 2.5measurements.The reduced major axis line is shown as the solid line in the plot.The 1:1line is shown as the short-dashed line for reference.Three potential outliers pointed out by arrows are all Southern California gridcells.Figure 2.Two-dimensional plot of annual (left)simulated PM 2.5concentrations integrated in 2°Â2.5°grid cells versus (right)EPA PM 2.5measurements integrated in 0.5°Â0.5°grid cells.The scale saturates at 18m g/m 3to best display the color contrast in the plot (99th percentile of EPA PM 2.5measurement =17.05m g/m 3).Table 3.Spearman’s Correlation Coefficients Between EPA PM 2.5and MISR AOT,Simulated PM 2.5and MISR PM 2.5Concentrations in Each Season in the Eastern and Western United StatesSeason Region Correlation Coefficient (p Value)aSimulated PM 2.5MISR AOT MISR PM 2.5Winter east b0.73(<0.0001)À0.03(0.78)0.49(<0.0001)west c À0.15(0.23)0.26(0.04)0.09(0.51)Spring east 0.58(<0.0001)À0.04(0.77)0.19(0.10)west 0.58(<0.0001)0.09(0.51)0.55(<0.0001)Summereast 0.82(<0.0001)0.50(<0.0001)0.67(<0.0001)west À0.02(0.88)0.55(<0.0001)0.39(0.0007)Falleast 0.68(<0.0001)0.09(0.40)0.38(0.0004)west0.65(<0.0001)0.43(<0.0001)0.64(<0.0001)aThe p values reflect the significance level of the correlation coefficients.The correlation is not significant if p is greater than 0.05.bRefers to regions to the east of 95°longitude.cRefers to regions to the west of 95°longitude.generally dominated by sulfate that are transported from regional sources of SO 2,the impact of urban excess con-centrations is relatively small [West et al.,1999].Therefore good agreement between model and observations is noted.[18]The large discrepancy and weak correlation in the northwest region in the winter may be because there is only a limited number of EPA monitoring sites in this region.The EPA PM 2.5concentrations from some of the urban sites are heavily influenced by strong carbonaceous aerosols emitted from local sources such as automobiles and wood fires and can be substantially higher than surrounding suburban and rural areas [Rao et al.,2002].Since EPA sites are relatively sparse in the northwest,a small number of urban sites can greatly influence the mean PM 2.5concentration in a 2°Â2.5°grid cell.Therefore it is not surprising to see that global models substantially underestimate PM 2.5concentrations at current spatial resolution.parison Between Simulated and MISR AOTs [19]Annual average simulated AOT (0.10±0.03)is approximately 20%lower than MISR AOT (0.13±0.06)with a similar dynamic range (Table 2).The largest seasonal difference is found in the spring when simulated AOT (0.11±0.03)is approximately 30%lower than MISR retrievals (0.16±0.04).Annual average simulated AOT generally captures the spatial pattern of MISR AOT mea-surements,with higher values in the east and lower values in the west (Figure 5).A scatterplot shows that simulated AOT has a good linear relationship (r =0.80)with MISRAOT with a small offset (intercept =À0.007)although simulated AOT shows a low bias of 17%(reduced major axis line slope =0.83)(Figure 6).[20]A seasonal comparison shows that although signifi-cant in all four seasons,the correlations between simulated and MISR AOT values are substantially stronger during the summer (r =0.78)and the fall (r =0.62)than during the winter (r =0.39)and the spring (r =0.42)(Figure 7).The overall agreement between simulated AOT and MISR AOT retrievals does not vary significantly by geographical region except that MISR AOT is substantially higher than simulated AOT over coastal Washington State,North Dakota,South Dakota and Minnesota during the spring.[21]Current MISR AOT data includes the aerosol extinc-tion effect in the entire atmospheric column in both the troposphere and the stratosphere (C.Welch,personal com-munication,2003).Although stratospheric AOT is usually at least an order of magnitude smaller than tropospheric AOT [Kent et al.,1994],it likely contributes to the difference of approximately 0.03between the means of MISR AOT and simulated AOT.In addition,as previously mentioned,the sum of the known particle species concen-trations can be significantly smaller than PM 2.5concentra-tions measured by gravimetric methods.This deficit is also likely reflected in the underestimation of AOT by GEOS-CHEM and GOCART.[22]The discrepancies in spatial and seasonal patterns between simulated AOT and MISR AOT may be attributed to the uncertainties associated with both the globalmodelsFigure 4.Scatterplots of seasonal average (left)simulated PM 2.5concentrations with (right)EPA PM 2.5measurements over the United States in 2001.Data from the eastern United States are shown as circles,and data from the western United States are shown as crosses.The total number of data points in each season N ,the correlation coefficient r ,and the reduced major axis equation are presented in each plot.The reduced major axis line is shown as the thick solid line in each plot.The 2:1,1:1,and 1:2lines are shown as the thin dashed lines for reference.。
