Application of near infrared reflectance spectroscopy (NIRS) to assess some properties of a sub-arctic ecosystem

Abstract

Summary Investigations of temporal and spatial variation of chemical properties of ecosystem components on a landscape level are hampered by the need to analyze large numbers of samples. Near infrared reflectance spectroscopy (NIRS) might provide a useful tool to overcome this problem. Here we investigated the possibilities and limitations to quantify the chemical composition of different plant parts and ecological properties with the help of NIRS. For this, we addressed the following questions: (1) Can NIRS-models be used to quantify different primary compounds (nitrogen, fibre), groups of secondary compounds (condensed tannins, total phenolics) as well as specific phenolic components (e.g., salicin) in leaves, twigs and litter of Salix phylicifolia? (2) Can NIRS be used to predict ecological properties such as moose browsing on willow or the decomposition rate of leaf litter? NIRS predicted the different primary compounds and grouped secondary compounds in different plant material with high accuracy. Results were inconsistent for specific phenolics. For ecological properties (moose browsing, litter decomposition rate) NIRS-models had high coefficients of determination. But tests of the models with a second independent set of samples (independent-data-set-test: IDS-test) showed that the predicted values were too low even though they were ranked correctly. Based on these results, the application of a second independent test is recommended. In the present study this second validation indicated inconsistencies in the NIRS-models that had not been revealed by the conventional validation procedures used to develop the models (test-set and cross-validation). According to the present results NIRS represents a suitable and cost-effective tool to measure primary and groups of secondary components of plant material. Application to specific phenolic compounds requires more elaborate testing. The successful reconstruction of moose browsing and the prediction of litter decomposition rate show that NIRS offers new possibilities for ecological applications.