On the assessment of citric acid in aqueous solution and fruit titratable acidity levels using SWNIRS : (a) citric acid solutions

Shortwave (700 to 1100 nm) near infrared spectroscopy using diode array instrumentation suited to in-line or field portable instrumentation was considered in context of the assessment of titratable acidity (TTA) in intact fruit. A searching window method was used to optimise the wavelength window for model development (around 750 – 950 nm). In pure citric acid – water mixtures (mean + SD of 5.05 + 6.65 g citric acid /100 ml), a RMSECV = 0.34 and R = 0.999 was achieved, while if the matrix was varied with 0, 10 and 13% sucrose, a RMSECV = 0.59 and a R = 0.996 was achieved. Models developed on spectra collected of the cut surface of a lime fruit (mean + SD of 7.3 + 0.51 g citric acid equivalents/100 ml), possessed a RMSECV = 0.17, R = 0.93, and a SDR = 3.0 (where SDR = SD/RMSECV), comparable with prediction results (RMSEP = 0.16, R = 0.89, bias = -0.03, and SDR = 2.4). For intact lime fruit, model calibration results (RMSECV = 0.16, R = 0.92, and SDR = 3.1) were markedly better than prediction results (RMSEP = 0.30, R = 0.70, bias = -0.07, SDR = 2.1). For a low TTA product, peach (with spectra collected across fruit maturity stages; mean + SD of 0.88 + 0.17), model calibration results were relatively poor (RMSECV = 0.09, R = 0.83, SDR = 1.8), while in prediction the model failed (RMSEP = 0.104, R = 0.05, bias = 0.02, SDR = 0.9). We conclude that SWNIR is not appropriate for assessment of the acidity of intact low TTA fruit, and has limited use for high TTA fruit. The method has value for rapid assessment of the TTA of juice extracted from moderate and high TTA fruit.

本研究针对完整果实的可滴定酸度（titratable acidity, TTA）评估场景，采用适配在线或现场便携式设备的二极管阵列仪器（diode array instrumentation），开展了700~1100 nm波段短波近红外光谱（Shortwave near infrared spectroscopy, SWNIR）的相关研究。本研究采用搜索窗口法优化模型开发所需的波长窗口，最终选定约750~950 nm波段。在纯柠檬酸-水混合体系（柠檬酸含量均值±标准差（Standard Deviation, SD）为5.05±6.65 g/100 ml）中，模型的交叉验证均方根误差（Root Mean Square Error of Cross Validation, RMSECV）为0.34，相关系数（correlation coefficient, R）达0.999；当体系基质添加0、10和13%蔗糖以改变基质组成时，模型的RMSECV为0.59，R为0.996。以青柠果实切面的光谱数据建模时，样品的柠檬酸当量含量均值±SD为7.3±0.51 g/100 ml，所得模型的RMSECV为0.17，R为0.93，标准差比（Standard Deviation Ratio, SDR）为3.0（SDR定义为SD与RMSECV的比值），该结果与预测集表现相当，其中预测集均方根误差（Root Mean Square Error of Prediction, RMSEP）为0.16，R为0.89，偏差（bias）为-0.03，SDR为2.4。针对完整青柠果实的建模结果显示，其校正集RMSECV为0.16，R为0.92，SDR为3.1，显著优于预测集表现（RMSEP=0.30，R=0.70，bias=-0.07，SDR=2.1）。针对低TTA样品（桃果实，采集了不同成熟阶段的光谱数据，其柠檬酸当量含量均值±SD为0.88±0.17 g/100 ml），模型的校正集表现相对较差（RMSECV=0.09，R=0.83，SDR=1.8），且预测集模型完全失效（RMSEP=0.104，R=0.05，bias=0.02，SDR=0.9）。本研究结论表明，SWNIR不适用于完整低TTA果实的酸度评估，对高TTA果实的酸度评估应用也较为有限；但该方法可用于快速检测中、高TTA果实压榨果汁的可滴定酸度。