Potential application of hyperspectral imaging and FT-NIR spectroscopy for discrimination of soilless tomato according to cultural practices and water use efficiency

This study was aimed to evaluate the suitability of hyperspectral imaging (HSI) and Fourier Transform (FT)-NIR spectroscopy for classifying sustainable-produced tomatoes according to a) cultural practices and b) levels of water use and fertilizer use efficiency (WUE and FUE). Three different cultivation strategies for water and fertilizer use were applied across two cultivation cycles for two varieties (cv ‘Carminio’, and cv ‘Mose’): i) free drain open cycle cultivation (OPEN); ii) open cycle cultivation with on-demand sensor-based fertigation (SMART); iii) closed cycle cultivation (CLOSED). Reflectance spectra were acquired using HSI in Vis-NIR and NIR ranges, and a FT-NIR spectrometer, for about 300 fully ripe tomatoes per variety. Partial least squares discriminant analysis (PLS-DA) was first aimed to discriminate the three cultivation systems, and then the levels of WUE and FUE. Model performances were higher when using FT-NIR and HSI in the Vis-NIR range, but the last one needed less latent variables. Good performance on external prediction were obtained for discriminating tomatoes from three cultural practices for each variety. In addition, an excellent performance was reached by classifying tomatoes in two different levels of WUE and FUE with accuracy, specificity and sensitivity higher than 95%. Finally, for the general models based on three levels of WUE, over the two experiments, using only 20 significant wavelengths, yielded accuracy and specificity of 89.8% and 91.7%, respectively. Results of this study indicate promising potential of these techniques for the authentication of agricultural crop grown with low inputs, which need further investigation.