Towards a global arctic-alpine model for Near-infrared reflectance spectroscopy (NIRS) predictions of foliar nitrogen, phosphorus and carbon content

Francisco Javier Ancin Murguzur,Adriaan Smis,Kari Anne Bråthen,Eric Struyf,Hanna Böhner,Patrick Meire,Marjorie Bison

doi:10.1038/s41598-019-44558-9

Abstract

Near-infrared spectroscopy (NIRS) is a high-throughput technology with potential to infer nitrogen (N), phosphorus (P) and carbon (C) content of all vascular plants based on empirical calibrations with chemical analysis, but is currently limited to the sample populations upon which it is based. Here we provide a first step towards a global arctic-alpine NIRS model of foliar N, P and C content. We found calibration models to perform well (R2validation = 0.94 and RMSEP = 0.20% for N, R2validation = 0.76 and RMSEP = 0.05% for P and R2validation = 0.82 and RMSEP = 1.16% for C), integrating 97 species, nine functional groups, three levels of phenology, a range of habitats and two biogeographic regions (the Alps and Fennoscandia). Furthermore, when applied for predicting foliar N, P and C content in samples from a new biogeographic region (Svalbard), our arctic-alpine NIRS model performed well. The precision of the resulting NIRS method meet international requirements, indicating one NIRS measurement scan of a foliar sample will predict its N, P and C content with precision according to standard method performance. The modelling scripts for the prediction of foliar N, P and C content using NIRS along with the calibration models upon which the predictions are based are provided. The modelling scripts can be applied in other labs, and can easily be expanded with data from new biogeographic regions of interest, building the global arctic-alpine model.

Highlights

The essential role of N and P in plants and ecosystem functioning has been emphasized over the last decades[1,2,3,4,5,6,7,8,9]
We hypothesised that NIR spectra can be used for modelling foliar N, P and C content across a range of functional types and ecological contexts and across a range of biogeographic regions
We developed Near-infrared reflectance spectroscopy (NIRS) calibration models and evaluated their capacity to accurately estimate foliar N, P and C content of a total of 552, 291 and 424 samples respectively

Summary

Introduction

The essential role of N and P in plants and ecosystem functioning has been emphasized over the last decades[1,2,3,4,5,6,7,8,9]. One of the challenges with NIRS methodology is that its application is limited to closed sample populations[11] This means spectral characteristics of sample types not included in the calibration model may interfere with model predictions and cause spurious results. This limitation restrains the application potential of NIRS for ecological studies, because the range of multiple ecological contexts is not accounted for when developing population-specific calibration models. A non-essential element occurring in inorganic form in leaves of several functional types of vascular plants, support the potential for a global NIRS model[21]. The magnitude of the imprecision can be reduced by using large sample sizes and thereby reduce the dependency on single, potential imprecise measurements

Methods

Results

Conclusion