No label defined (Q4258103)

We have demonstrated that the methods of symbolic dynamics are applicable in principle to ecosystem time series. They are leading to a quantification of the hypothesis that ecosystems act as filters in their abiotic environment by extracting information from irreversible input fluxes. Whether this picture is consistent with or complementary to a thermodynamically-based description (ecosystems as negentropy sinks) is an open question. The quantification of the filtering abilities opens the possibility of a comparison of ecosystems. The higher entropy difference for the Lange Bramke catchment as compared to Birkenes is consistent with the naive expectation from the different sizes and especially soil characteristics of the catchment. The impact of the biotic aspects of these ecosystems remains to be studied. The reconstruction of finite state machines from the observed time series failed in our case. So, strictly speaking, an evaluation of the metric complexity was not possible. This fact shows that a direct application of algorithms tested for artificial data series has its limits. On the other hand, one could envisage enhancements of these applications which might be successful, e.g. a systematic prescription for deleting a few ``nasty'' values from the original series by defining a threshold. Another interesting point is the connection between different time series, as also the processes observed are connected at least in one direction (we exclude systems so large that the feedback of evapotranspiration and runoff on the atmospheric processes may not be safely neglected). This seems to require another kind of machine which is externally controllable; finitary transducers are a promising candidate for that. For the entire collection see [Zbl 0913.00025].