Research And Realization Of Analogue Prediction Method For Seawater Temperature

The seawater temperature forecast is very important for the oceanography element forecast.To understand the dynamic fields in the ocean,the temperature field is a prerequisite,at the same time,it is an input parameter for atmosphere and ocean forecast system.An analogue method for predicting the mid long term point-by-point refinement forecasting of seawater temperature is presented in this paper.In order to predict the seawater temperature for the future month,the present method searches for a similar time evolution of seawater temperature in the historical data set.The present algorithm shows reasonable skill in predicting seawater temperature for a lead-time up to one month.First of all,this paper expounds the basic principles and methods of statistical forecast.The similarity criteria is the key to analogue method and plays a decisive role in the forecast results,therefore,this paper focuses on the analysis of common similarity criterion—Hamming distance,analogue quantity and analogue deviation.Through the sample data can improved that the analogue deviation better than others.Secondly,the characteristics of seawater temperature are studied in detail.The similarity of seawater temperature is analyzed based on the optimal interpolation sea surface temperature data set of NOAA,by compare the temperature data with the curve,the results show that the variation of seawater temperature has similarity and can predict by analogue method.At the same time,it can be obtained by the calculation of the temperature data that the analogue deviation has better selection ability than Hamming distance and analogue quantity.So,in this paper,analogue deviation is used to construct the analogue method model.Thirdly,a prediction model of seawater temperature lattice based on analogue deviation is designed.The forecast result with the resolution of 0.25° latitude by 0.25° longitude,and correct them with the bias correction technique and the weighted average integrated forecast technique.Taking the prediction result of seawater temperature in the South China Sea in 2016 as an example,the influence of different factors on the forecast result and different forecast results for different months are analyzed.It is concluded that the forecast result is best when select 10 days data before the forecast month as the predictor and all years data use the method of weighted mean to forecast.At the same time,the correlation coefficients and errors of forecast data and actual data in March,April,December,and October are all optimal.In the case of maintaining the same error in other months,the correlation coefficients is poor,but all the forecast results are better than climatic results(that is,the statistical day average of data for all years).Finally,by forecast the seawater temperature for each month of 2017 in the same region,we have obtained the experimental results which the same to the 2016 seawater temperature forecast results,verified the correctness of the analysis results,The forecast results show that the feasibility of analogue method to forecast seawater temperature of mid long term.