湛江气温的多时间尺度特征及其变化趋势预测

利用MHAT小波函数分析了湛江市近50年来年平均气温的多时间尺度变化特征,并采用Kalman滤波模型对年平均温度的变化趋势进行了预测研究。结果表明,在半个多世纪的气候变化过程中包含了16 a4、a和准2 a的周期振荡,20世纪80年代中期以来湛江气温持续升高,且未来几年仍处于偏高阶段。卡尔曼滤波模型对湛江年平均气温的预测中,独立样本预测的平均绝对误差(MAE)等于0.30℃,相对误差为1.24%,对短期气候预测工作具有较高的参考价值。

Wavelet Analysis and Kalman Filter Forecasting Model on Annual Mean Temperature in Zhanjiang

By using Mexican-Hat wavelet transformation,the features of annual mean temperature in the past 50 years(from 1951 to 2000) in Zhanjiang region are discussed.And a new forecasting system is developed based on Kalman filter.Results show that the mean temperature in Zhanjiang is various periodically in 16a,4a and 2a,but it has become warmer gradually since the middle of 1980s.The prediction made by Kalman filter model from 2001 to 2003 shows that its mean absolute error(MAE) is only 0.30℃,and mean absolute percentage error(MAPE) is about 1.24%.Therefore,this model is able to provide a valuable result for forecasters in the short-term climate forecasting.