基于随机森林的香港海域海表盐度遥感反演模型

Remote sensing retrieval model of sea surface salinity in Hong Kong waters based on the random forest

提出了一种基于随机森林反演海盐的算法模型,基于研究海域的实测数据,分析并筛选出与海表盐度敏感性较高的影响因子(总氮、悬浮固体、温度),利用2003-2008年共6期ASTER影像数据,从中提取、计算敏感因子的光谱参数,结合相应实测盐度,作为模型的原始数据集,运用R语言构建随机森林算法对数据进行训练,将训练得到的随机森林用于海表盐度的预测。结果显示,预测值与实测值之间平均相对误差较小,吻合度高,R2均在0.85以上,多数达0.95以上。研究表明,基于多因子参数的随机森林反演海表盐度是可行且高效的。

A method of Sea Surface Salinity （SSS） inversion model based on random forest was proposed. Based on the data of the Hong Kong waters, the three factors （total nitrogen, suspended solid, temperature） with higher sensitivity were chosen. By analyzing the ASTER data from 2003 to 2008, the spectral parameters of the sensitive factors were taken as the model data set. The random forest obtained through the training by using R was applied to the sea surface salinity forecast. The result showed that the mean squared error （MSE） between predicted value and measured value was small, and R2 was all more than 0.85, mostly reached over 0.95. It was concluded that the retrieval model of SSS based on Random Forest with multi-factor was feasible and efficient.