Fig 1.
Conceptual overview of the fitting procedure.
A) Individual pigment spectra (here, chl a, chl b, ββ-Car, Lut, c-Neo), and B) background basis spectra are assigned weights (C, D) in the NNLS regression, where the weights depend on the composition of the pigment extract. The weighted component spectra are summed to obtain the total pigment spectrum (E) and the total background spectrum (F). Adding together the pigment and background spectra yields the total spectrum of the extract (G).
Table 1.
Pigment list.
Fig 2.
Results from the Monte-Carlo simulation of pigment detectability.
In A), the course of the false negative rate (FNR; black line) and the false positive rate (FPR; grey line) as the standard deviation of the random error is increased (see text), is shown for each pigment. The pigments were sorted from highest to lowest FNR, also shown in B). Here, we estimated how high the added random noise had to be for the FNR and FPR to exceed 0.05 (5%). Black dots represent the standard deviation of this random error for the FNR, the grey dots for the FNR. Asterisks mean that the FNR (black asterisks) or FPR (grey asterisks) never exceeded 5% for that pigment.
Fig 3.
Absorbance spectra measured on sediment extracts.
Data from two depths in the sediment column is shown: A) and B) are from the upper cm, C) and D) from 5 cm below the sediment surface. The difference in absorbance between the two samples can be attributed to lower pigment concentrations deeper in the sediment. A & C) The measured total absorbance spectrum (grey dots) overlaid with the fitted total absorbance spectrum obtained using the modified-GPS method (solid black line). Background spectra (scattering and/or absorbance by non-algal pigment) are represented by dotted lines; the thick line being the modelled background, while the thin line is the measured blank spectrum (absorbance of a well filled with 96% ethanol). Thin grey lines represent the different component spectra making up the modelled background spectrum. In B) and D), the modelled background spectrum is subtracted from the total spectrum, yielding only pigment absorbance. The spectra of the individual pigments recognized by the modified-GPS method are shown in colour.
Fig 4.
Total chlorophylls and total carotenoids from sediment.
A) Concentrations (μg pigment per gram sediment dry weight) of total chlorophylls (green) and total carotenoids (orange) as estimated by modified-GPS (y-axis) and HPLC (x-axis). The dotted line shows the 1:1 relationship. Note that both axes are log-transformed due to the large variation in concentrations. Pigment concentrations decreased with sediment depth, both total chlorophylls (B) and total carotenoids (C).
Fig 5.
Observed and fitted spectra from lake samples.
Absorbance spectra from the 75 natural lake samples (not normalized to unit path-length) plotted as grey dots. For clarity, spectra are ordered from lowest to highest maximum absorbance, and dots plotted only for every fifth nm. A vertical offset is added between each spectrum, therefore no units on the y-axis. Fitted spectra are added as black lines.
Fig 6.
Total chlorophylls and total carotenoids in lake samples.
A) Observed vs. fitted absorbance for all 75 spectra in Fig 5. r2 > 0.999. B) Concentrations of total chlorophylls (green) and total carotenoids (orange) estimated by modified-GPS (y-axis) and HPLC (x-axis). The dotted line shows the 1:1 relationship. Note that both axes are log-transformed due to the large variation in concentrations.
Fig 7.
Observed vs. fitted spectra from culture samples.
Observed absorbance spectra (grey dots) overlaid with fitted spectra (black lines) for four different phytoplankton cultures. Modelled background spectra are shown as dashed lines.
Fig 8.
Pigment concentrations in culture samples.
Concentrations of pigments in four phytoplankton culture samples presented as fractions of chl a. A) Chlamydomonas reinhardii, B) Cryptomonas ozolinii, C) Planktothrix 98, D) Planktothrix 68. Dark grey bars: modified-GPS, light grey bars: HPLC.
Fig 9.
Modelled background spectra (scattering, and/or absorption by non-algal components) estimated after subtracting the microplate blank spectrum from the raw absorbance spectra.
A) Sediment samples (n = 40) and B) natural lake samples (n = 75). Each grey line represents the modelled background spectrum from one sample. C) The fraction of the total spectrum constituted by the background spectrum. It was calculated by dividing the integral under each background spectrum by the integral under the corresponding total (pigment + background) spectrum. Horizontal line equals mean fractions. The background constituted a higher and more variable fraction in the sediment samples.