Abstract: Water temperature had significant impacts on safe wintering of tropical and subtropical fishes whose cold enduring are relatively weak.In this paper,change characteristics of water temperature of fishing pond and its forecasting model in winter in Guangzhou area were researched in accoring with the observing data of water temperature in fishing pond and air temperature nearby meteorological station.The results showed that daily amplitude of water temperature of fishing pond was relatively small and occuring time of daily maximum water temperature lagged compared with air temperature.This character was very obvious in hourly temperature and daily mean temperature.Daily change of water temperature under clear weather was far greater than that under overcast.Daily amplitude of water temperature gradually decreased and phase lagged more obviously with the increase of water depth.Cross correlation analysis between water temperature and air temperature showed that water temperature had distinct correlations with air temperatures of present day,former one day,former two day and former three day,nevertheless had no distinct correlations with air temperatures of former four day and former five day.Based on air temperature,forecasting modes of water temperature in different depths were established by stepwise regression analysis.Verifications of these models were statistically significant and their average absolute and relative errors were not bigger than 0.5 ℃ and 10% respectiviely.These high precise models could be used to forecast water temperature of fishing pond.

Key words: Even-aged pure stands, Self-thinning rule, Density change, Genetic algorithms