间隔偏最小二乘-紫外光谱法海水硝酸盐最佳建模波长区间选取

利用紫外光谱结合偏最小二乘法（PLS）建立了海水硝酸盐浓度计算模型，并利用间隔偏最小二乘法（interval PLS，iPLS）对建模区间进行了优化。相比全波段建模，优化后模型校正集均方根误差（RMSECV）从39.1降低到9.83。同时利用35个预测集样本对iPLS优化模型进行验证，iPLS优化模型将预测样本平均相对偏差从3.23%降低到1.81%。研究结果表明，相比全波段建模，利用iPLS进行建模波长区间优化，不仅减少了自变量个数，简化了运算，同时提高了模型预测准确度。

Optimal wavelength selection in determination of nitrate based on iPLS-UV spectrophotometry

A quantitative model for detecting the concentration of dissolved nitrate in seawater was established based on Ultraviolet spectrophotometry combined with partial least squares (PLS) method.The optimal wavelength range was determined using interval PLS (iPLS) method.Compared with the model used full spectrum, the model used the optimal wavelength range determined by iPLS method lowered the RMSECV of the training set from 39.1 to 9.83.The test set including 35 samples were used to verify the effect of iPLS method, the results showed that the iPLS method lowered the average relative deviation of the test set from 3.23% to 1.81%.The study showed that the optimal wavelength range determined by iPLS not only reduced the number of independent variables, simplified computing, but also improved the prediction accuracy.