AN ANALYSIS OF THE FISH POPULATIONS BY USING ANN AND WAVELET TECHNIQUES

摘要

Air – sea climate, environmental and biological conditions show various differences on several spatio-temporalscales. Climate change associated with anthropogenic activity and natural global multi-decadal climate variations effects onair-sea interactions and water surface–atmosphere–biosphere climate system. In the first part of this paper is related withArtificial Neuro Network analyses for prediction of fish stocks in Marmara and Black Sea. The second part of this study isbased on wavelet analyses and, the results were compared with former wavelet and harmonic analyses to explain seasonaleffects of NAO and ENSO on fish population. The influence of climatic oscillations (based on NAO and ENSO) on monthlycatch rates of fish population such as sea bass, Atlantic bonito,blue fish sea (pomatomus population between 1991-2012) inBlack Sea and Marmara have been analyzed by discrete wavelet transform (DWT) with Meyer and Daubechie's. Waveletanalysis is an efficient method of time series analysis to study non-stationary data. Wavelet analyses allowed us to quantifyboth the pattern of variability in the time series and non-stationary associations between fish population and climatic signals.Phase analyses were carried out to investigate dependency between the two signals. We reported strong relations betweenfish stock and climate series for the 4- and 5-yr periodic modes, i.e. the periodic band of the El Ni?o Southern Oscillationsignal propagation in the Black Sea and Marmara Sea. These associations were non-stationary, evidenced from 1995 to2012. It is recognized that other factors in small, meso and large scales may modulate fish stocks beginning from 1995 andmore clearly from 2005.