厦门地区地面气温的多模式集成统计降尺度预报研究

The multi-model super-ensemble statistical downscaling forecast of the air temperature of Xiamen District

利用双线性插值与线性回归方法、消除偏差集合平均（bias-removedensemblemean，BREM）和多模式超级集合预报（Super-ensemblePrediction，SUP）方法对厦门地区的地面气温进行统计降尺度分析，结果表明：在2013年夏季的3个月中，降尺度后三个单模式对厦门地面气温的预报效果显著改善。使用多模式集成预报方法（BREM和SUP）后．预报误差进一步减小。对比整体预报效果最好的单模式ECMWF，降尺度后3-96h预报误差均在3℃以下。此外，结合SUP方法的降尺度预报能最大程度的改善地面气温的预报误差。

Based on the ensemble mean outcomes from forecasts of the surface temperature 2 m over the ground in Xiamen, which were provided by ECMWF, GFS and T639 data archive, a statistical downsealing forecast was studied by using the interpolation, linear regression in conjunction with bias- removed ensemble mean （BREM） and multi-model super ensemble （SUP） . The results showed that the statistical downscaling technique significantly improved the forecast skill of four single models during three months of 2013 summer. The SUP and BREM methods further reduced the errors of the single model downscaling forecast.The improvement percentage of the 3-96h forecast error of the downscaling forecast with BREM and SUP forecast schemes of the best single model ECMWF was below 3℃. in addition, the forecast skill of the statistical downscaling with SUP forecast was superior to that with BREM forecast.