The methods of aquaphotomics were explored as an aid to improve near infrared spectroscopic predictive modelling of the soluble solids content of pure apple juice at different temperatures. The study focussed on the first overtone region of the O–H stretching vibration of water (1300–1600 nm). A transmission-based FT-NIR (Fourier transform near infrared) spectrometer was used to acquire 103 spectra of freshly expressed juice samples from individual ‘Braeburn’ apples over the wavelength range of 870–1800 nm with a 1 mm cuvette at three temperatures, 20, 25 and 30°C. The aquagram of the first overtone water region showed a trend of increasing bound water absorption with rising soluble solids content, from 7.3 to 13.7°Brix, and increasing free water absorption with rising temperature from 20 to 30°C. Predictive models for apple juice soluble solids content at 25°C were developed using partial least squares regression with spectral pre-processing by standard normal variate (SNV) followed by second derivative transformation (SNV + 2D) or no pre-processing on absorbance spectra at all. The best result, with lowest standard error of prediction of 0.38°Brix, was obtained using the first overtone water region with partial least squares regression on the SNV + 2D spectra. The method of extended multiplicative scatter correction was used, as an additional pre-processing step, to improve apple juice soluble solids content prediction at different temperatures. The interference component selected for the extended multiplicative scatter correction method was the first principal component loading measured using pure water samples taken at the same three temperatures (20, 25 and 30°C). Such extended multiplicative scatter correction pre-processing greatly reduced the soluble solids content prediction bias, when applying the partial least squares regression model developed at 20°C to samples measured at 25 and 30°C, from 0.23 to 0.08 and 0.36 to 0.13°Brix, respectively. Model precision (in terms of standard error of prediction) was also slightly improved by 0.02°Brix in each case, from 0.40 to 0.38 and 0.46 to 0.44°Brix at 25 and 30°C respectively.

中文翻译：

---

研究水光组学对纯苹果汁可溶性固形物含量的独立温度预测

探索了水光组学方法，以帮助改进不同温度下纯苹果汁可溶性固形物含量的近红外光谱预测模型。该研究集中在水的 O-H 伸缩振动（1300-1600 nm）的第一个泛音区域。使用基于透射的 FT-NIR（傅里叶变换近红外）光谱仪在 870–1800 nm 的波长范围内，使用 1 mm 比色皿在三个温度、20 , 25 和 30°C。第一泛音水区的aquagram显示出随着可溶性固形物含量的增加，结合吸水率增加的趋势，从7.3到13.7°Brix，随着温度从20到30°C的升高，自由吸水率增加。25°C 下苹果汁可溶性固形物含量的预测模型是使用偏最小二乘回归通过标准正态变量 (SNV) 进行光谱预处理，然后进行二阶导数转换 (SNV + 2D) 或不对吸收光谱进行预处理而开发的全部。使用第一个泛音水区域在 SNV + 2D 光谱上进行偏最小二乘回归，获得了预测的最低标准误差 0.38°Brix 的最佳结果。使用扩展乘法散射校正方法作为额外的预处理步骤，以改进不同温度下苹果汁可溶性固形物含量的预测。为扩展乘法散射校正方法选择的干涉分量是使用在相同三个温度（20、25 和 30°C）下采集的纯水样品测量的第一主分量载荷。当将在 20°C 开发的偏最小二乘回归模型应用于在 25 和 30°C 下测量的样品时，这种扩展的乘法散射校正预处理大大降低了可溶性固形物含量预测偏差，从 0.23 到 0.08 和 0.36 到 0.13°Brix ， 分别。在每种情况下，模型精度（就预测的标准误差而言）也略微提高了 0.02°Brix，在 25°C 和 30°C 下分别从 0.40 到 0.38 和 0.46 到 0.44°Brix。将在 20°C 下开发的偏最小二乘回归模型应用于在 25°C 和 30°C 下测量的样品时，分别为 0.23 至 0.08 和 0.36 至 0.13°白利糖度。在每种情况下，模型精度（就预测的标准误差而言）也略微提高了 0.02°Brix，在 25°C 和 30°C 下分别从 0.40 到 0.38 和 0.46 到 0.44°Brix。将在 20°C 下开发的偏最小二乘回归模型应用于在 25°C 和 30°C 下测量的样品时，分别为 0.23 至 0.08 和 0.36 至 0.13°白利糖度。在每种情况下，模型精度（就预测的标准误差而言）也略微提高了 0.02°Brix，在 25°C 和 30°C 下分别从 0.40 到 0.38 和 0.46 到 0.44°Brix。