We evaluated models predicting the spectral chlorophyll-a (Chl a)-specific absorption coefficient (a*ph (¿)) from Chl a concentration [Chl a] on the basis of 465 phytoplankton absorption spectra collected in estuarine, coastal and oceanic waters. A power model on ln-transformed data provided the best model fit compared to a power model on non-transformed data previously applied to parameterize the relationship between a*ph (¿) and [Chl a]. The variation in a*ph (¿) was parameterized over four orders of magnitude in [Chl a] (0.01-100 mg Chl a m-3) producing a 13-fold range in a*ph (0.19 to 0.015 m2 mg-1 Chl a) at 440 nm, the peak absorption of Chl a in the blue part of the spectrum. The variations in the modelled a*ph spectra were within realistic predictions of a*ph (¿) and the model satisfactorily reproduced the spectral flattening with increasing [Chl a]. The parameterization of a*ph (¿) confirmed the indirect dependency of a*ph (¿) on [Chl a] through co-variations between [Chl a] with pigment packaging and pigment composition. Although pigment packaging determined the spectral flattening, analysis of absorption ratios revealed a systematic change in pigment composition with profound influence on the variability of a*ph in the 440 to 495 nm region. Modelled spectra deviated by approximately 20% from the measured spectra on average and model accuracy was independent of [Chl a]. Although the model cannot fully replace spectral measurements of phytoplankton absorption, it does permit realistic reconstructions of a*ph (¿) from simple measurements of [Chl a] sampled in estuarine, coastal and oceanic waters.
International Journal of Remote Sensing, 2004, Vol 25, Issue 22, p. 5117-5130