Application of near infrared reflectance (NIR) and fluorescence spectroscopy to analysis of microbiological and chemical properties of arctic soil

Applicability of near infrared reflectance (NIR) and fluorescence spectroscopic techniques was tested on highly organic arctic soil. Soil samples were obtained at a long-term climate change manipulation experiment at a subarctic fell heath in Abisko, northern Sweden. The ecosystem had been exposed to treatments simulating increasing temperature (open-top greenhouses), higher nutrient availability (NPK fertilization) and increasing cloudiness (shading cloths) for 15 years prior to the sampling. For each of the 72 samples from the 0 to 5 cm soil depth and 36 samples from the 5 to 10cm depth, the wavelength range of 400-2500nm (visible and near infrared spectrum) was scanned with a NIR spectrophotometer and fluorescence excitation-emission matrices (EEMs) were recorded with a spectrofluorometer. Principal component analyses of the visible, NIR and fluorescence spectra clearly separated the treatments, which indicates that the chemical composition of the soil and its spectral properties had changed during the climate change simulation. Similarly to the results from the conventional analyses of soil chemical and microbiological properties, fertilization treatment posed strongest effects on the spectra. Partial least-squares (PLS) regression methods with cross-validation were used to analyse relationships between the spectroscopic data and the chemical and microbiological data derived from the conventional analyses. The fluorescence EEMs of the dried solid soil samples were moderately related to soil ergosterol content (correlation coefficient r = 0.84), bacterial activity analysed by leucine incorporation technique (r = 0.78) and total phospholipid fatty acid (PLFA) content (r = 0.74), but in general fluorescence provided inferior predictions of the chemical and microbiological variables to NIR. NIR was highly related to soil organic matter content (r > 0.9) and showed promising predictions of soil ergosterol content (r > 0.9), microbial biomass C, microbial biomass P, and total PLFA contents (r = 0.78-0.79). These results suggest that especially NIR could be used to predict soil organic matter and fungal biomass. Since it is rapid and inexpensive, and requires little sample mass, it could be used as a 'quick and dirty' technique to estimate progression of the treatment responses in long-term ecosystem experiments, where extensive soil sampling is to be avoided. (c) 2007 Elsevier Ltd. All rights reserved.