Remote sensing of water optical property for China's inland Lake Taihu

Remote Sensing of Water Optical Property for China’s Inland Lake Taihu Using the SWIR Atmospheric Correction With1640and2130nm BandsMenghua Wang,SeungHyun Son,Yunlin Zhang,and Wei ShiAbstract—Using the shortwave infrared(SWIR)atmospheric correction algorithm with1640and2130nm bands,water optical property data for China’s inland freshwater Lake Taihu have been derived from measurements of the Moderate Resolution Imaging Spectroradiometer(MODIS)on the satellite ing MODIS-Aqua measurements from2002to2010,seasonal and spatial distributions of the normalized water-leaving radiance spectra from visible to the near-infrared(NIR)and nm)are derived and used for study and charac-terization of Lake Taihu water optical properties.In particular, for thefirst time,spatial and seasonal variations ofat the SWIR1240nm are derived for theand show some different those of the red and NIRand.Time series of monthly MODIS-derived spectra for Taihu are obtained and analyzed,showing seasonal and interannual variations. The results indicate that the SWIR contributions in the lake are mainly due to the presenceficant amounts of algae,while the red and NIR variations result from changes of total suspended sediment amounts in the water column.Furthermore,this study shows that all three SWIR bands at1240,1640,and2130nm are useful and required for satellite water quality remote sensing for extremely turbid near-shore and inland waters.Index Terms—Atmospheric correction,lakes,remote sensing, water pollution,water quality.I.I NTRODUCTIONL AKE Taihu is the third-largest inland freshwater lake in China with an areal coverage of2300km and a mean water depth of2m.China’s Lake Taihu is located in the Yangtze River delta(Fig.1),which is one of the world’s most heavily populated regions with the highest rate of economic Manuscript received November08,2012;revised December18,2012;ac-cepted January22,2013.Date of publication March12,2013;date of cur-rent version November21,2013.The in situ data collection in Lake Taihu(Y. Zhang)was jointly supported by the Knowledge Innovation Project of the Chi-nese Academy of Sciences(KZCX2-YW-QN312),the Major Projects on Con-trol and Rectification of Water Body Pollution(No.2012ZX07101-010),and the National Natural Science Foundation of China(41271355,40825004,and 41230744).(Corresponding author:M.Wang.)M.Wang,S.Son,and W.Shi are with the NOAA National Environmental Satellite,Data,and Information Service,Center for Satellite Applications and Research,College Park,MD20740USA(e-mail:Menghua.Wang@). S.Son and W.Shi are with CIRA,Colorado State University,Fort Collins, CO80523USA.Y.Zhang is with the State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology,Chinese Academy of Sciences, Nanjing,China.Color versions of one or more of thefigures in this paper are available online at .Digital Object Identifier10.1109/JSTARS.2013.2243820development in the last ke Taihu provides normal water usage for several million residents in nearby Wuxi City (Fig.1).Waters in Lake Taihu are consistently highly turbid [1],with the exception of East Taihu Bay and some of the East Lake regions(Fig.1),where waters are often clear,with aquatic vegetation in the bottom[2].In addition,Lake Taihu has frequent algae blooms in the spring-summer,polluting the lake water[1],[3]–[5].In particular,in the last decade or so, algae blooms happen more often with largescale effects and disastrous impacts regionally,e.g.,the2007blue-green algae (Microcystis)bloom event in the lake[3],[4].Algae-polluted waters in the lake have affected and interrupted the normal life of the several million residents nearby.Therefore,there is an urgent need for effectively monitoring and managing water quality in inland freshwater lakes such as Lake Taihu,and for a broader understanding of the optical,biological,and ecological processes and phenomena in all fresh inland waters. Recently,Wang et al.(2012)[1]examined and characterized Lake Taihu water optical,biological,and biogeochemical properties using measurements from the Moderate Resolution Imaging Spectroradiometer(MODIS)onboard the Aqua satel-lite.In the study,the shortwave infrared(SWIR)atmospheric correction algorithm[6]–[8]using MODIS SWIR bands of 1240nm and2130nm in addition to an iterative approach, has been developed specifically for Lake Taihu in order to account for some extremely turbid water cases,where the SWIR1240nm band has important water-leaving radiance contributions[1].To accurately derive satellite-measured nor-malized water-leaving radiance spectra,for which the definition can be found in various references[9]–[13], sensor-measured radiances at the two near-infrared(NIR)or SWIR bands are used for atmospheric correction[6],[9],[10] with the assumption of known water-leaving radiance contri-butions for these two bands,e.g.,the black ocean assumption [9]or modeling for the NIR radiance contributions[14]–[17]. There are also some other attempts with non-ocean color satel-lite sensor for deriving water optical property of Lake Taihu [18].In a more recent study,the SWIR atmospheric correction algorithm was used to derive and characterize water turbidity for a freshwater Lake Okeechobee in Florida from long-term MODIS-Aqua measurements[19].However,with the SWIR atmospheric correction algorithm using MODIS-Aqua bands at 1240and2130nm for deriving water property data for Lake Taihu,Wang et al.(2012)[1]shows that the SWIR1240nm band is not always black for the entire lake,i.e.,the normalized water-leaving radiance at1240nm is not always1939-1404©2013IEEEFig.1.Maps of China’s inland freshwater Lake Taihu with indications of four in situ data locations(a)–(d),which correspond to latitudes and longitudes of (31.26N,119.98E),(31.01N,120.24E),(31.14N,120.16E),and(31.16N,120.19E),respectively.negligible for waters in some regions[1],in particular,for regions with very significant amounts of algae(algae blooms), such as those with surface-floating algae[20].Thus,an iterative approach was developed in the previous study[1],accounting for non-negligible water-leaving radiance contribution at the SWIR1240nm band.It should be noted that satellite-measured spectra data are essential for remote sensing of water optical,biological,and biogeochemical properties,e.g.,chloro-phyll-a concentration,the diffuse attenuation coefficient,total suspended sediment(TSS),as well as water inherent optical properties(IOPs)[21]–[27].In this paper,we further our efforts in deriving water optical properties for China’s Lake Taihu using the SWIR-based atmo-spheric correction algorithm with MODIS SWIR bands1640 and2130nm.In particular,at the SWIR band1240nm can be derived routinely for Lake Taihu.Both MODIS SWIR bands at wavelengths of1640and2130nm are generally black(i.e.,negligible radiance contributions at these two SWIR bands)for Lake Taihu and can be used for carrying out atmospheric correction[28].Although detailed simulations[6] and some MODIS case studies[28]have been done,it is for the first time that long-term satellite measurements are extensively used to characterize the water optical property for the highly turbid inland Lake Taihu using the SWIR atmospheric correction with MODIS1640and2130nm bands.Water optical property results for the lake from this study are compared with those from the previous study,showing consistent results from both methods,although there are some differences,particularly for spectra at the shorter wavelengths.In addition,spatial and seasonal variability of spectra from visible to the NIR and SWIR(1240nm)for Lake Taihu is examined,including climatology spatial distributions of(1240)for the lake derived from MODIS-Aqua measurements from2002to2010.II.T HE MODIS-A QUA-D ERIVED W ATER O PTICALP ROPERTY D ATAA.The SWIR Atmospheric Correction Algorithm Using1640 and2130nm BandsSince Lake Taihu contains consistently highly turbid waters [29]with significant NIR water-leaving radiance contributions [1],the SWIR-based atmospheric correction algorithm[6]–[8] is required for deriving lake water optical property data from MODIS-Aqua measurements.The methodology of the SWIR atmospheric correction algorithm[6]is essentially the same as the Gordon and Wang[9]NIR atmospheric correction algo-rithm.Instead of using two NIR bands for atmospheric cor-rection[9],the SWIR atmospheric correction algorithm uses a combination of two SWIR bands,e.g.,1240and2130nm,or 1640and2130nm[6].However,it has been further demon-strated that the SWIR1240nm band is not always black for the entire Lake Taihu,i.e.,at the wavelength of1240nm is not always negligible for waters in some regions[1].Thus, in this study,the SWIR-based atmospheric correction algorithm with MODIS SWIR bands at1640and2130nm is used[6].The black ocean pixel assumption for the SWIR bands at1640and 2130nm is generally valid[28]even for the most turbid waters because of considerably large water absorptions at these wave-lengths[30],[31].The performance of the SWIR atmospheric correction algorithm including some error analysis has been dis-cussed in detail in[6].For MODIS-Aqua,however,four out of ten detectors for the SWIR1640nm band are inoperable and cannot be used effectively.For this reason,the MODIS-Aqua SWIR1240 and2130nm bands have been typically used for carrying out atmospheric correction,e.g.,in the previous study[1].On the other hand,the SWIR1640and2130nm bands have beenWANG et al.:REMOTE SENSING OF WATER OPTICAL PROPERTY FOR CHINA’S INLAND LAKE TAIHU2507used previously to assess the black ocean pixel assumption for three MODIS SWIR bands[28],and thus the SWIR1640 and2130nm algorithm is readily available to apply to Lake Taihu.In addition,the SWIR1640and2130nm atmospheric correction algorithm has also been used to derive sea ice optical properties in the Bohai Sea[32],[33].In fact,using the SWIR atmospheric correction algorithm with1640and2130nm bands,satellite ocean(water)property data can be derived directly and more efficiently without the iteration process.It should be noted that,with only six working detectors(four dysfunctional detectors)for the MODIS-Aqua SWIR1640nm band,satellite-derived ocean(water)color products will miss a significant amount of data,in particular,for the individual Level-2file[28].Due to significantly less amount of Level-2 data,in this study,we focus on deriving climatology seasonal results(composite images)for Lake Taihu using the SWIR 1640and2130nm atmospheric correction algorithm.It should also be noted that the implementation of the NOAA satellite ocean(water)color data processing system—NOAA Multi-Sensor Level-1to Level-2(NOAA-MSL12)—is different from the current NASA data processing system;particularly for the SWIR-based data processing[8],[34],e.g.,cloud masking [35],an ice detection algorithm[36],the lookup tables[6], [37]–[39],detection of turbid waters and absorbing aerosols [40],and some codes in data processing,as well as vicarious gains,are different.The vicarious gains for MODIS-Aqua visible to NIR bands are derived consistently for carrying out the NIR-SWIR combined atmospheric correction for deriving ocean(water)color products in open oceans and coastal regions [8],[34].Thus,the NOAA-MSL12data processing system used in this study is different from the NASA standard data processing system.Also,for NOAA operational ocean color data processing(near-real time),ancillary data(the total column ozone amount,sea surface wind speed,atmospheric pressure, and total column water-vapor amount)from the Global Forecast System(GFS)model are used[41].B.MODIS-Aqua Data Compared With In Situ Measurements In addition to the in situ data used in the previous study for Lake Taihu[1],there are some other in situ data that are now available to use in this study.It should be noted,however,that these new in situ data were only available after the previous work[1]was published,although these data were acquired much earlier[42].Zhang et al.[42],[43]provide some detailed description and discussions about in situ data.Briefly,five seasonal cruises in Lake Taihu were carried out in January 7–9,2006,July29–August1,2006,October12–15,2006, January7–9,2007,and April25–27,2007for collecting in situ physical,optical,and biological lake data[42].In particular, in situ water-leaving radiance spectra data from350–2500 nm were measured using an Analytical Spectral Devices,Inc. (ASD)FieldSpec Dual UV/VNIR spectrometer,for which the instrument spectral sampling interval is about1nm for its entire spectral coverage(350–2500nm).In situdata in the SWIR wavelengths,however,are quite noisy as radiance values in the SWIR bands are very low.It should be noted that,for near-zero values such as for the SWIR data,it is quite reasonable one gets both positive and sometime slightly negative in situ values due to instrument noise and noise from the in situ data processing.The in situ data collection and processing were carried out following the procedures outlined in the NASA ocean optics protocols[44]. Specifically,in the in situ data processing,the radiance that is contributed by the water surface reflection from the sky radiance to the instrument detector has been calculated and removed[42]as outlined in[45].Thus,MODIS-Aqua-derived spectra in Lake Taihu can be compared with those from the in situ measurements.It is important to note that we use the same in situ matchup procedure as described in[8]from the MODIS-Aqua-derived daily data,in particular, a valid satellite/in situ matchup requires50%valid pixels from satellite measurements,i.e.,13pixels in a55box surrounding the in situ data,and the recalculation of satellite data for satisfying a uniformity criterion[8].Fig.2provides four examples of spectra data for these validation comparisons.Fig.2(a)–(d)are spectra data that were collected at the locations indicated in Fig.1 marked as(a)–(d),i.e.,the corresponding latitudes and longi-tudes for the data obtained in Fig.2(a)–(d)are at(31.26N, 119.98E)for January9,2006,(31.01N,120.24E)for July 30,2006,(31.14N,120.16E)for July30,2006,and(31.16 N,120.19E)for October15,2006,respectively.In Fig.2, there are two sets of MODIS-Aqua-measured spectra data,i.e.,from the SWIR1240and2130nm method with an iteration approach[1]and the SWIR1640and2130nm method (this study).In addition,spectra results with the SWIR 1640and2130nm method also include MODIS-measured data in Fig.2.The in situ data in Fig.2 are shown from400to1240nm.Results in Fig.2show that both SWIR-based atmospheric correction algorithms(i.e.,SWIR1240and2130nm with an iterative approach and the SWIR1640and2130nm method) performed reasonably well for these highly turbid inland lake waters.Indeed,both in situ[42],[46]and MODIS-Aqua-mea-sured data show significantly high NIR water-leaving radiance values,i.e.,values at the wavelength of859nm range from0.3to 1.5mW cm m sr (see also results in Fig.3).In particular,in Fig.2(a)–(d),the corresponding MODIS-derived values using the SWIR1640and2130nm method are0.038,0.008,0.002, and0.01mW cm m sr,respectively.Significant contribution in Fig.2(a)leads to biased low MODIS-Aqua-derived in the visible bands with the SWIR1240and2130nm method.For cases with negligible (1240)contributions,on the other hand,both SWIR-based atmospheric correction algorithms produce consistent spectra results.Nevertheless,results in Fig.2show that the MODIS-derived at the visible bands compare reason-ably well with the in situ measurements.Fig.3further shows comparisons between MODIS-Aqua and in situ-measured values at the selected six(Fig.3(b)) or seven(Fig.3(a))MODIS-Aqua bands(412,443,488,555, 645,859,and1240nm).It is noted that for extremely turbid waters some of MODIS ocean color bands are often saturated. Fig.3(a)shows results(including SWIR from the SWIR1640and2130nm atmospheric correction algorithm,2508IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERV ATIONS AND REMOTE SENSING,VOL.6,NO.6,DECEMBER2013Fig.2.MODIS-Aqua-derived normalized water-leaving radiance spectra data using the modi fied SWIR 1240and 2130nm method [1]and the SWIR 1640and 2130nm algorithm in Lake Taihu compared with in situ data that were acquired on (a)January 9,2006,(b)July 30,2006,(c)July 30,2006,and (d)October 15,2006.The locations in latitude and longitude are indicated in each plot and also shown in Fig.1.while Fig.3(b)demonstrates cases using the SWIR 1240and 2130nm with an iteration approach [1].For a fair comparison,the same in situ data for the first six MODIS bands (visible to NIR)are used for both Fig.3(a)and (b).It should be noted that there were signi ficant reductions in the number of matchup data in order to satisfy the criterion of 50%of valid pixels in a 55box surrounding the in situ data.For example,there are 40–60satellite/in situ matchup comparisons with a criterion of 5valid pixels (20%)in a 55box,while with the re-quired 13valid pixels (50%)and the spatial uniformity cri-teria in a 55box the valid number of comparisons reduces to 12–24.It was further con firmed from in situ observations that these matchup requirements [8]are necessary,particularly for such a highly turbid and often spatially non-uniform lake.With a relaxed requirement,e.g.,a criterion of 5valid pixels in a 55box,we found some in situ data even larger than MODIS-Aqua-measured top-of-atmosphere (TOA)radi-ances due to highly spatial non-uniform distribution of water property in Lake Taihu.In Fig.3,the corresponding number of matchup data forat MODIS wavelengths of 412,443,488,555,645,and 859nm are 12,12,14,18,18,and 24,respectively.The number of matchup data for the SWIRin Fig.3(a)is 21.Although there are some variations (noise errors)due to both sensor (e.g.,poor MODIS SWIR bands performance)and algorithm performance errors [6],[47]results in Fig.3show reasonable matchup comparisons between MODIS-Aqua and in situ measurements.In Fig.3mean ratio values inbetween MODIS-Aqua-derived and in situ-measureddata for wavelengths of 412,443,488,555,645,859,and 1240nm with the corresponding standard deviation (STD)values are provided and also summarized in Table I.Results in Fig.3and Table I show that the SWIR 1640and 2130nm method performed slightly better with reduced bias errors in,compared with those from the modi fied SWIR 1240and 2130nm method [1].However,noise errors (i.e.,STD values)are generally lower with the SWIR 1240and 2130nm method as also demonstrated in [6].It is particularly noted that comparisons of the SWIR are very noisy with a very large STD value (as expected).In summary,satellite/in situ comparison results show that the SWIR 1640and 2130nm algorithm can produce reasonable MODIS-Aqua-measured spectra data for Lake Taihu,particularly for spectra at the visible wavelengths.WANG et al.:REMOTE SENSING OF WATER OPTICAL PROPERTY FOR CHINA’S INLAND LAKE TAIHU2509Fig.3.MODIS-Aqua-derived spectra data using(a)the SWIR1640and2130nm method and(b)the SWIR1240and2130method with an iteration approach[1]in Lake with in situ measurements for wavelengths of412,443,488,555,645,859,and1240nm.The SWIR comparison results are only shown in plot(a).The mean ratio(MODIS-Aqua vs.in situ)and STD values at each wavelength are listed in both also shown in Table I.TABLE IM ATCHUP C OMPARISON R ESULTS(F IG.3)OF THE M EAN R ATIO(MODIS VS.I N S ITU)ANDTHE C ORRESPONDING STD V ALUES FOR QUA V ARIOUS B ANDS2510IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERV ATIONS AND REMOTE SENSING,VOL.6,NO.6,DECEMBER2013Fig.4.MODIS-Aqua-measured (2002–2010)seasonal climatology (panels a–d),(panels e–h),(panels i–l),(panelsm–p),and(panels q–t)in Lake Taihu using the SWIR 1640nm method.data are cm m sr.III.S ATELLITE -M EASURED S PECTRA F ROM V ISIBLE TO THE NIR AND SWIR B ANDSMODIS-Aqua measurements from 2002to 2010in Lake Taihu were acquired to produce water optical property data using the SWIR-based (1640and 2130nm)atmospheric correction algorithm.These satellite remote sensing data are used for char-acterizing water optical properties for Lake Taihu;in particular,for the first time,satellite-measured at the SWIR band 1240nm are derived for Lake Taihu.Detailed results and discussions are provided in the following sections.WANG et al.:REMOTE SENSING OF WATER OPTICAL PROPERTY FOR CHINA’S INLAND LAKE TAIHU2511A.Spectra From Visible to the NIR Wavelengths Fig.4provides seasonal climatology spectra images at wavelengths of443,555,645,859,and1240nm for the entire Lake Taihu.These images were derived from MODIS-Aqua measurements from2002–2010using the SWIR-based at-mospheric correction algorithm with bands1640and2130nm [6].Thus,spectra results in Fig.4provides overall characteristics of optical property both spatially and seasonally for Lake Taihu.It is noted that color scales in Fig.4are dif-ferent(indicated in the right side)for images in various MODIS-Aqua bands.Fig.4(a)–(d)show seasonal climatology(2002–2010)of at the wavelength of443nm for spring (March,April,and May),summer(June,July,and August),fall (September,October,and November),and winter(December, January,and February),respectively,representing character-ization of algae absorption distributions for Lake Taihu[48]. Spatially,low values are observed in various bay regions,e.g.,the Meiliang Bay,Gonghu Bay,and Zhushan Bay, compared with other lake regions,particularly in the central lake region.Overall,for most of the lake,high values occur in the spring season,while lows are shown in the fall season. On the other hand,seasonal climatology images (Fig.4(e)–(h))show relatively stable values throughout the year for Meiliang Bay,Gonghu Bay,and Zhushan Bay,while there are significant seasonal changes in for most of the other lake regions.For most of the lake regions,highvalues are observed in the winter-spring seasons and low values are shown in the summer-fall seasons.These varia-tions are mainly influenced by changes of the TSS concentra-tion in the lake.The TSS spatial and seasonal variations in Lake Taihu are also represented through seasonal climatology results in(Fig.4(i)–(l)),which can be related to TSS con-centration[26],[27],[49],[50],with highs in the winter-spring seasons and lows in the summer-fall seasons.The TSS seasonal variation in Lake Taihu has been observed from in situ measure-ments[43].Furthermore,results of the NIR water-leaving radiance con-tributions for Lake Taihu(Fig.4(m)–(p))show that the seasonal pattern in is similar to that of, i.e.,highs in the winter-spring and lows in the summer-fall sea-sons for most of the lake regions.Clearly,variation in Lake Taihu is also mainly controlled and influenced by the TSS variation in the lake.Fig.4(p)shows that the winter season includes the most turbid water for Lake Taihu with the largest located in the lower western part of the lake.It is important to note that for Lake Taihu,contribution is significant over most of the lake and has to be accurately ac-counted for in satellite remote sensing of the lake water property with the NIR atmospheric correction algorithm[9].These spatial and seasonal distribution results from the visible to the NIR bands are consistent with those derived from the previous study using the SWIR-based(1240and 2130nm)bands with an iterative approach for atmospheric correction in Lake Taihu[1].However,compared with the results from the previous study,at the blue bands are notably larger with the SWIR1640and2130nm method(e.g.,Figs.2(a)Fig.5.Histogram results of for Lake Taihu from(a)MODIS-Aqua-measured(2002–2010)data climatological four seasons(spring, summer,fall,and winter)and(b)overall distributions from all MODIS-Aqua-measured(2002–2010)and in situ data.and3).In addition,more data noises are notable from the SWIR 1640and2130nm method mainly due to four dysfunctional de-tectors in MODIS-Aqua SWIR band1640nm,i.e.,considerably less data for deriving the seasonal composite images.B.Data in the SWIR1240nm BandFig.4(q)–(t)provide seasonal climatology of at the SWIR wavelength of1240nm for Lake Taihu. Note that the color scale for is from0to0.15mW cm m sr.For a reference,contribution ofwith a value of0.1mW cm m sr is very significant. In fact,the TOA radiance at the SWIR1240nm for a clear sky over ocean is generally0.1mW cm m sr.For most part of the lake,contribution is negligible.How-ever,there are some noticeable contributions in var-ious bay regions,e.g.,Zhushan Bay,Meiliang Bay,etc.Par-ticularly,it is important to note that both spatial and seasonal distributions in(Fig.4(q)–(t))are different from those of(Fig.4(m)–(p))or(Fig.4(i)–(l)), indicating that the reason and mechanism for(or and variations are different for Lake Taihu.Some significant contributions in Zhushan Bay and Meiliang Bay are in the summer-fall seasons,likely due to the presence of significant amounts of algae,e.g.,sur-facefloating algae[20].In the winter,while has2512IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERV ATIONS AND REMOTE SENSING,VOL.6,NO.6,DECEMBER2013Fig.6.MODIS-Aqua-derived(from2002–2010)density-scattering plots of(a)vs.,(b)vs.,(c)vs.,and(d)vs.for China’s Lake Taihu.the highest contributions in Lake Taihu(Fig.4(p))due to high loading of TSS[51],is actually negligible for al-most entire Lake Taihu(Fig.4(t)).Fig.5further quantifies the contributions in Lake Taihu,showing histogram results of for the four climatological seasons(Fig.5(a))and overall distributions from satellite and in situ measurements(Fig.5(b)).Results in Fig.5(a)show that mode values in histograms for spring,summer,fall,and winter seasons are quite small with the corresponding values of0.02,0.005,0.01,and 0.01mW cm m sr,respectively.Again,a very small negative value(near zero)is due to data noise.Some notice-able cases with large(e.g.,larger than0.1mW cm m sr show up seasonally in the order of the most to the least as summer,fall,spring,and winter,respectively, showing that the most significant SWIR contribu-tions are in the summer-fall season in Lake Taihu(Fig.5(a)). On the other hand,Fig.5(b)compares overall histogram results in from MODIS-Aqua and in situ measure-ments for the lake.In fact,because of data noise in, comparisons of histogram results are more meaningful.The total numbers of data contributed to the results in Fig.5(b)from MODIS-Aqua and in situ measurements are58134and135, respectively.In Fig.5(b),the mode,mean,median,and STD values for the MODIS-Aqua results are0.010,0.023,0.016, and0.030mW cm m sr,respectively,compared with the corresponding values from in situ measurements of0.010, 0.079,0.021,and0.369mW cm m sr,respectively. Both mode and median values from satellite and in situ data are comparable,with significantly high mean and STD values from in situ data.Again,these values should be compared with the TOA radiance of0.1mW cm m sr at the SWIR 1240nm.Fig.5shows clearly that there are some significant contributions in Lake Taihu,and con-tributions need to be accurately accounted for in atmospheric correction for deriving accurate satellite spectra data.C.Relationships Among Various MODIS-MeasuredDataUsing all MODIS-Aqua measurements from2002 to2010in Lake Taihu,relationships(or correlations)between at various bands vs.NIR have been inves-tigated.Fig.6provides results of density-scattering plots for vs.at wavelengths of555,645,748,and 1240nm.It is noted that MODIS-Aqua-measured data in Fig.6are from Level-2results.Results in Fig.6show that for Lake Taihu there are strong correlations betweenvs.(Fig.6(a)),vs.(Fig.6(b)), and vs.(Fig.6(c)),while correlations between vs.are quite weak(Fig.6(d)). The weak correlation has also been shown and discussed in the previous results(Fig.4)due to different mechanisms forWANG et al.:REMOTE SENSING OF WATER OPTICAL PROPERTY FOR CHINA’S INLAND LAKE TAIHU2513Fig.7.MODIS-Aqua-derived monthly mean values ofover entire Lake Taihu as a function of time covering from 2002–2010for parameters of (a)and ,(b)and,(c),and (d),respectively.driving and changes.As also found fromShi and Wang (2009)study [28],at first,both andincrease almost linearly as increase of(Fig.6(a)and (b)).Then,with further increase of ,the rate of increase in (or is reduced.Finally,with more increase of and reach their corresponding maximum values (first and then ,respectively.This phenomenon [52],[53]is attributed to the dominance of for extremely turbid waters in a function of and are total absorption coef ficient and backscattering coef ficient,respec-tively),of which can be expressed as a function [54].When reaches its maximum value.In addition,with all MODIS-Aqua-measured data in Fig.6,the best fitting curves for vs.(Fig.6(a)),vs.(Fig.6(b)),and vs.(Fig.6(c))for Lake Taihu have been derived and shown in each plot in Fig.6,i.e.,(1)where fitting coef ficients of (for wave-lengths of 555nm in Fig.6(a),645nmin Fig.6(b),and 748nm in Fig.6(c)are (0.7517,0.2181,0.1455),(0.6727,0.4475,0.1510),and (0.2864,0.8675,0.1498),respectively.These derived rela-tionships (Fig.6(a)–(c))from MODIS-Aqua measurements can be used for a regional (Lake Taihu)atmospheric correction for satellite remote sensing of water optical properties,e.g.,with measurements from the Korean Geostationary Ocean Color Imager (GOCI)[17].Furthermore,we found from the results in Fig.6(c)that the NIR normalized water-leaving re flectance ratio [55],[56]between bands at 748and 859nm at low re flectance value is about 1.73,compared to the value of 1.84—the ratio of water absorption coef ficients between bands at 859and 748nm [55]D.Time Series of MODIS-Aqua-Measured for LakeTaihuTime series of monthly averages of MODIS-Aqua-derived,anddata are obtained for the entire Lake Taihu.Fig.7provides the monthly mean values of over entire Lake Taihu at spectral bands of the blue and 488nm (Fig.7(a)),green and red(Fig.7(b)),NIR (Fig.7(c)),and。