Relationship Modeling between Nitrogen Phosphorus Content and Near-infrared Spectrum of Offshore Sediments

Abstract

Rapid in-situ determination of total nitrogen and total phosphorus in offshore sediments is an urgent need for real-time monitoring of the marine environment. Spectral analysis technology is a commonly used non-chemical method for the determination of total nitrogen and total phosphorus in soil. Establishing a relationship model between total nitrogen (TN) and total phosphorus (TP) in offshore sediments and its spectra can effectively support its rapid in-situ determination. Therefore, this paper proposes a method to model the relationship between TN and TP in offshore seabed sediments and their visible near-infrared spectra. The modeling process includes exception sample removal, sample spectral preprocessing, sample set partition and relationship modeling. Mahalanobis distance (MD) method is used to eliminate exception samples; Savitsky-Golay (S-G) filtering, multiplicative scatter correction (MSC) and standard normal variate (SNV) are used to pre-process sample spectra; the Kennard-Stone (K-S) selection method is used for sample set partition; partial least squares regression (PLSR) is used for relationship modeling; determination coefficient ( $R$ 2) and root mean square error (RMSE) are used to evaluate the modeling. Based on the 177 offshore sediments sample data collected in Qingdao offshore, using the visible near-infrared (250-950nm) spectra, the test results show that the method based on SNV and PLSR has the best performance. Based on the proposed relationship modeling process, the software for rapid determination of TN and TP in offshore seabed sediments is presented, which integrates sample spectral acquisition, sample relationship modeling and sample testing.