82
Chlorophyll a "g!1
(observed data: solid line;
predicted data: dotted line)
3 •• ,.···· ...... · ....... ······ .................. · .. ·, ........ · .... · ................................... ,
, ..
100
F. Recknagel
Microcystis cellslml
(observed data: solid line;
predicted data: dotted line)
800000 , - - - - - , - - - - - ,
600000
400000
200000
800000
d
600000
400000
200000
600000
400000
200000
1993
Fig. 5.4. 7-days-ahead prediction of Chlorophyll-a and Microcystis for Lake
Kasumigaura (Japan) in 1986 and 1993 by ANN (a and b), EDE (c and d), and ER
(e and f)
During simulations only those agents will be fired at a certain time that best suit
occurring conditions but otherwise resting in the agent bank. Fired (emerging)
agents simultaneously evolve based on EC in order to reach their optima (see Fig.
5.2b). This agents bank will be the key to enable the adaptive agents model to
change the composition of agents during simulations by temporarily activating or
resting agents (e.g. algal species) depending on excitatory or inhibitory
environmental conditions.