Abstract: In order to improve the daily management capabilities of lobster pond in Jiangsu Province and reduce lobster losses caused by the high temperature disasters, using hourly observational data of dissolved oxygen, pH, water temperature of different layers and the corresponding meteorological station data of Xuyi lobster pond in summer in 2014, hourly change and daily variation characteristics for dissolved oxygen, pH and water temperature were analyzed. Based on BP neural network, prediction model of water temperature of different layers and dissolved oxygen was built. The results showed that water temperature of the deeper layers were higher than air temperature in summer and hourly variation amplitude of 10 cm water temperature was most dramatic. The highest water temperature of 10 cm appeared first, followed by 30 cm, 60 cm, 100 cm. The fluctuations of air temperature and water temperature were consistent. The fluctuation of dissolved oxygen negatively correlated with temperature, and pH was relative stable. Compared with air temperature, water temperature of different layers fluctuated in small amplitude with a certain hysteresis effect. Correlation analysis showed that the maximum water temperature of different layers correlated with maximum air temperature and maximum temperature in the lasted 1, 2, 3, 4 days and the correlation coefficient was above 0.8. Average daily dissolved oxygen negatively correlated with maximum temperature of that day and the day before, 100 cm maximum temperature. Determination coefficient of measured and simulated values was above 0.98 and the relative error was less than 1%, indicating that the prediction model possess high simulation accuracy. Prediction model constructed in this study about the water temperature and dissolved oxygen in Xuyi lobster pond can provide a theoretical basis for lobster farming and meteorological disaster prevention.