Modelling and forecasting the fortnightly cladoceran abundance in the Saronikos Gulf (Aegean Sea)

Abstract

In the present study, we evaluated the ability of dynamic regression on the basis of its efficiency to model and forecast the fortnightly abundance of cladocerans during 1989-1993 in a coastal region of the eastern Saronikos Gulf (Aegean Sea). The following dynamic regression model explained 79% of the variability of the transformed cladoceran abundance during the fitting period (January 1989-December 1992): ln (cladocerans)(t) = 0.4314 ln (chlorophyll a)-5.9013 ln (salinity) + 8.7760 ln (temperature) + 0.2371 ln (cladocerans)(t-20). The model always predicted the start and the end of the cladoceran season, and forecasted the cumulative cladoceran abundance during January-December 1993 with an absolute percentage error of 14.5%. The positive effect of sea temperature and chlorophyll a, and the negative effect of salinity, on cladoceran abundance are all consistent with previous studies. The positive relationship between cladoceran abundance at times t and t-20 approximates the seasonal cycle of their abundance.