and thus contribute more to the signals. In the case of malic and
lactic acids, most of the samples that contain a low concentration
of lactic acid had a high concentration of malic acid and vice
versa. One would therefore expect weaker signals from lactic acid
and these will be affected by stronger signals from malic acid for
wines before malo-lactic fermentation. After malo-lactic fermentation
the intense signals from lactic acid would be expected to
affect the weaker signals from malic acid. Due to the high concentration
of tartaric acid, the performance of succinic acid at lower
concentrations could be affected by the similar absorption band
of tartaric acid. Furthermore, concentration levels for succinic acid
were below 0.6 g L−1 and in some cases the values were close to
the LOQ. Another possible reason for the poor performance concerns
the accuracy of the reference method; i.e. a high relative
error was shown by the reference method in the low range. It is
also known that succinic and acetic acids do not have the alcohol
functional group and they therefore do not show the C–O
and O–H bands from the secondary alcohol that the other organic
acids have. However, acetic acid did show 84 and 74% acceptable
results in the lower range for wine and spirit drinks. For those samples,
the concentration range of acetic acid was very low. In the
full range calibration (Table 8, Fig. 3), the high error was due to
spirit and wine samples, which shows that when performing the
calibration the data distribution is critical to obtain a robust calibration.
In any mode the FT-IR measurement is based upon the
predictive ability of the reference method. In acetic acid measurements
the reference analysis was not associated with interferences
and a low relative error at low concentration levels was found.