Fig 2 : Boxplots of the accuracy of one-point, two-point, and
multipoint normalizations that are matrix matched and bounded (A) and
matrix mixed/both mixed and extrapolated (B). Significance is shown in
the compact letter display, where any two methods with the same letter
are not significantly different as per Dunn’s post-hoc testing. Nitrogen
(uppercase) and carbon (lowercase) letters are not compared with each
other.
Effect of matrix and isotope range on normalization
accuracy
To isolate the effect of matrix-matching on normalization errors, we
present results from only three-point normalizations. When the results
are constrained to bounded normalizations, the matrix of the standards
relative to the quality controls has a significant effect for both N and
C (Fig. 3). Nitrogen normalizations that have calibration standards
composed of a different matrix than the quality controls (median error =
0.104‰, n = 688) exhibit significantly higher errors than
matrixed-matched normalizations (median error = 0.064‰, n = 102,p < 0.0001) and normalizations where the standards and
quality controls are both composed of a combination of matrixes (median
error = 0.058‰, n = 190, p < 0.0001). Similarly,
mixed-matrix C normalizations (median error = 0.165‰) exhibit
significantly greater errors than normalizations composed of matched
matrixes (median error = 0.056‰, p < 0.0001) and
both-mixed matrixes (median error = 0.121‰, p <
0.0001). Furthermore, N normalizations with both-mixed matrixes have
significantly greater errors than matched-matrix normalizations
(p < 0.0001).