空间变异对海洋叶绿素a质量浓度遥感产品精度验证的影响

Impact of spatial variability on the validation of ocean chlorophyll- a concentration remote sensing product

海洋叶绿素a质量浓度遥感产品是海洋初级生产力与海洋生态系统固碳能力研究的重要数据源,为了保证数据的可靠性,对遥感产品进行精度验证以及验证误差的成因分析尤为重要。遥感产品的验证过程中,由于空间变异的存在,使得遥感像元尺度内的实测数据具有不同的离散程度和统计分布特征,并由此产生了不同的误差统计结果。本文选择MODIS-Aqua、MODIS-Terra、MERIS、SeaWiFS等卫星传感器叶绿素a质量浓度遥感产品为研究对象,统计分析了数据产品的空间变异与验证精度的关系。结果表明:空间变异是造成验证误差的直接原因之一,平均绝对百分比误差(Mean Absolute Percentage Error:MAPE)与空间变异系数(Coefficient of Variation:CV)呈幂指数模型关系;当CV<0.05时,MAPE随CV的增加明显;当CV>0.15时,MAPE的变化趋于平缓。不同卫星传感器叶绿素a质量浓度产品验证结果表明,SeaWiFS精度最高,MERIS次之,MODIS-Terra精度最低。

Ocean chlorophyll-a concentration remote sensing product is important data sources for the studies of ocean primary productivity and carbon sequestration capacity of ocean ecosystems.In order to ensure the data reliability,it is particularly important to verify the accuracy of remote sensing product and analyze the causes of validation errors.In the verification of remote sensing products,due to the existence of spatial variation,the measured data within the remote sensing pixel have different statistical distribution characteristics,and thus produce different statistical error results.In this work,chlorophyll-a concentration remote sensing products derived from MODIS-Aqua,MODIS-Terra,MERIS and SeaWiFS satellite sensors,were quantitatively analyzed on the relationship between spatial variability and validation accuracy.The results of statistical analysis show that spatial variation is one of the direct causes of validation error.The relationship between Mean Absolute Percentage Error(MAPE)and spatial Coefficient of Variation(CV)is in accordance with power exponential model.When CV<0.05,MAPE increases significantly with CV,and when CV>0.15,MAPE changes gently.The validation results of different chlorophyll-a concentration remote sensing products show that SeaWiFS has the highest accuracy,MERIS takes the second place,and MODIS-Terra has the worst accuracy.