A model for forecasting water temperature of a small bay at Sanriku coast

为了建立日本三陆沿岸小海湾--大槌湾的水温预报模式，首先利用国际海岸研究中心提供的天气和海况资料，研究了该海湾海面热量收支的年际变化，东京大学和附近的气象站，日本气象厅。结果表明，80'和90'之间的热收支存在明显差异。这种差异是由于东北地区向下的短波辐射和风速的差异。其次，我们进行了测量的湍流活动在湾口和内湾。一般情况下，能量耗散率估计在10^<-7>至10^<-8>W/kg的数量级，但有时<-6>在内湾15至20 m深度处高达10^ W/kg的数量级。第三，利用GETM（通用河口输运模式）模拟了风驱动的水循环。所得结果与观测结果所得到的环流形势相比较，吻合得很好。冬季西风造成的垂直环流表现为8- 10 m深度以上的外流和8- 10 m深度以下的入流，并试图建立一个预报海湾水温的统计模式。利用1978 ~ 2003年的数据，估计了自回归模型（AR）和自回归移动平均模型（阿尔马）的统计系数。利用这些系数，我们计算了2004年的水温变化。结果表明，该模型对该湾水温的预报偏差为0.58℃，标准差为0.88℃。

To develop a model for forecasting water temperature of a small bay, Otsuchi Bay, located at the Sanriku coast of Japan, we clarified physical process of the bay, such as heat balance, water circulation, turbulent activity and so on. We first investigated inter-annual variation of heat budget through the sea surface of the bay by using weather and sea condition data supplied by International Coastal Research Center, the University of Tokyo and neighborhood meteorological stations, the Japan meteorological agency. The results show a clear difference in heat budget between the 80' and 90'. This difference is attributed to differences in both downward shortwave radiation and wind speed in Tohoku area. Second, we have carried out a measurement of turbulent activity both at the bay mouth and inner bay. The energy dissipation rate is estimated at the order of 10^<-7> to 10^<-8> W/kg in general, but sometimes up to the order of 10^<-6> W/kg at the depth of 15 to 20m of the inner bay. Third, we simulated wind driven water circulation by using GETM (General Estuarine Transport Model). The results are compared quite well with the circulation pattern obtained by the observational results. The vertical circulation caused by westerly in winter season shows the out flow above the depth of 8-10m and inflow below it. We also tried to investigate a statistical model for forecasting water temperature of the bay. On the basis of the data from 1978 to 2003, statistical coefficients of AR (Auto Regressive model) and ARMA (Moving Average of Auto Regressive model) are estimated. By using these coefficients, we calculated water temperature variation in 2004. As a result, this model forecasts water temperature of the bay with the bias of 0.58℃ and standard deviation of 0.88℃.