IS THE PHOTOSYNTHETIC PROCESS NONLINEAR

We applied two non-linear time series analysis methods to photosynthet ic data obtained from a single-species population of Thalassiosira pse udonana incubated in situ in order to identify whether the time series were generated predominantly by linear or non-linear processes. The t ests used to make these distinctions involved a comparison of the pred ictability of the observed data under a non-linear hypothesis versus a linear hypothesis. Two records were analyzed. For the first data segm ent taken in the morning, the linear method performed as well or bette r than the nonlinear methods. Although a weak non-linearity was detect ed in the second data set observed in the afternoon, the time series i s dominantly linear. The best embedding dimension, whose value suggest s the number of participating independent parameters in the system, is 2 for the morning data and 7 for the afternoon data. These results ar e true for aggregate productivity measures (1.8 x 10(8) cells) on a ti me scale of 1-5 min.