CAJ | 학술논문

基于42组不同年份不同季节获得的遥感反射率、水体各组分吸收系数的实测数据,对QAA(QuasiAnalytical Algorithm)和GSM(Garver-Siegel-Maritorena)算法在寡营养的南海和富营养的福建沿岸两种不同类型水体的吸收系数反演进行了检验,以水样测量值为参考,两种算法在本研究水体中的反演成效与他人在其它水域的研究结果相当.QAA算法在南海的反演成效高于福建沿岸水体.对于443nm的总吸收系数a(443).南海的对数均方根误差RMSE为0.044,平均相对误差ε为7.9%.对数甲均偏差δ为0;福建沿岸水体RMSE=0.190,ε=30.6%,δ=-0.167.GSM算法在两类水体的反演成效类似.a(443)之RMSE和ε,南海分别为0.156和27.7%.福建沿岸分别为0.146和32.1%;但从a(443)的δ值看.其在南海反演值低于实测值(8=-0.142).在福建沿岸则略呈高于实测值(δ=0.016).两种算法中的部分经验参数与实测值之间的差异是产生反演误差的主要原因.为了提高反演精度,对算法中经验参数的更进一步区域化调整可能是必要的.
기우42조불동년빈불동계절획득적요감반사솔、수체각조분흡수계수적실측수거,대QAA(QuasiAnalytical Algorithm)화GSM(Garver-Siegel-Maritorena)산법재과영양적남해화부영양적복건연안량충불동류형수체적흡수계수반연진행료검험,이수양측량치위삼고,량충산법재본연구수체중적반연성효여타인재기타수역적연구결과상당.QAA산법재남해적반연성효고우복건연안수체.대우443nm적총흡수계수a(443).남해적대수균방근오차RMSE위0.044,평균상대오차ε위7.9%.대수갑균편차δ위0;복건연안수체RMSE=0.190,ε=30.6%,δ=-0.167.GSM산법재량류수체적반연성효유사.a(443)지RMSE화ε,남해분별위0.156화27.7%.복건연안분별위0.146화32.1%;단종a(443)적δ치간.기재남해반연치저우실측치(8=-0.142).재복건연안칙략정고우실측치(δ=0.016).량충산법중적부분경험삼수여실측치지간적차이시산생반연오차적주요원인.위료제고반연정도,대산법중경험삼수적경진일보구역화조정가능시필요적.
With 42 in situ measurements of remote sensing reflectance and component absorption coemcients taken in the South China Sea and coastal waters off Fujian,China during different seasons,the authors evaluate the performance of the Quasi-Analytical Algorithm(QAA)and the Garver-Siegel-Maritorena(GSM)algorithm for water absorption coefficients.It is found that the retrieval performances of the two algorithms are similar to those of the algorithms by other researchers conducted in other regions.In this study,QAA performs better in the South China Sea than in the waters off coastal Fujian.For the total absorption coefficient at 443 nm a(443),the rootmean-square error(RMSE)is 0.046 in the South China Sea,with an averaged percentage error(ε)of 7.9%,and averaged error in log scale(δ)close to 0.For the waters off coastal Fujian,RMSE,ε and δ are 0.194,30.6%,and -0.167,respectively.The performance of the GSM is similar for the two waters.For a(443),RMSE and ε are 0.161 and 27.7% in the South China Sea and 0.149 and 32.1% in the waters off coastal Fuiian,respectively;their δ values,however,are negative(-0.1 42;indicating underestimation)in the South China Sea and positive (0.016;indicating slight overestimation)in the waters off coastal Fujian.Further analysis indicates that the differences between the empirical parameters employed in the algorithms and actual values of the studied waters are the main reasons causing the errors in remote sensing retrievals,therefore,it is necessary to regionally refine those parameters in order to improve the algorithm performance.