Near-infrared spectrometry (NIRS) for the analysis of seston carbon, nitrogen, and phosphorus from diverse sources

Abstract

Seston carbon (C), nitrogen (N), and phosphorus (P) concentrations are routinely measured by limnologists and oceanographers. Unfortunately, common methods for the determination of seston C, N, and P are expensive and labor intensive and result in sample destruction. It has been suggested that near-infrared spectroscopy (NIRS) may be preferable to primary chemical analyses because NIRS is rapid and nondestructive. The ability of NIRS to measure seston element concentrations has been demonstrated, but an impediment to widespread adaptation of this technique relates to the generality of prediction (i.e., standard) equations. For diverse environmental samples, the ability to develop one global predictive equation is important, since the calibration process involves a considerable investment. It is unclear how often during routine use an investigator must standardize and calibrate NIRS equations. Here we explore the use of NIRS for analysis of seston C, N, and P in diverse aquatic samples, including those from monocultures, Lake Superior, and two sets of numerous lakes and ponds. Predictive equations were developed for specific datasets and for all the datasets combined (global equation). In most cases dataset-specific equations and the global equation accurately predicted (R2 > 0.90) concentrations of seston C and N but not P. Prediction error varied among seston types and increased when equations were tested on novel datasets. Our results indicate that NIRS analysis is an effective alternative to primary C and N chemistry, particularly for large aquatic monitoring programs. However, care must be taken during calibration and routine use, as accuracy depends on the types of seston in the calibration dataset.