Selecting temperature-dependent variables in near-infrared spectra for aquaphotomics

Temperature-dependent near-infrared (NIR) spectroscopy has been developed and taken as a new technique for aquaphotomic studies to analyze water structures in aqueous systems. However, due to the overlapping, it is difficult to obtain the information from the spectra to understand the temperature dependency. In this work, a method was proposed for selection of the temperature-dependent variables (wavenumbers) from the NIR spectra measured at different temperature. Continuous wavelet transform (CWT) was used to decompose the spectra into the spectral components with different frequencies, and then Monte-Carlo uninformative variable elimination (MC-UVE) was employed to evaluate the dependency of the variables with temperature. The feasibility of the method was tested by simulated datasets and the applicability was proved by the temperature-dependent NIR spectra of water and aqueous solutions. Several variables were selected from the spectra of water, indicating the complexity of water structure. The variables are located at similar but not identical wavenumbers for different solutions and they have a good relationship with the concentration of the solute, demonstrating that water spectrum can be a mirror to reflect the difference of the aqueous solutions. The method may provide a tool to identify the characteristic variables in NIR spectra for understanding the function of water in aqueous systems.