===============
最优参数评估过程
===============


* 对于如何从批量试验结果中确定出最优FDDA参数方案，我们将敏感试验的24小时预报结果与NCEP GDAS 分析场进行比较，分别计算FDDA 试验6-, 12-, 18- 和 24-小时预报温度，湿度和风场和NCEP分析场之间的偏差（Bias)和均方根误差（stander deviation, STD)，选取偏差和均方差最小的参数方案作为最优选择。所用到的NCEP分析场资料如下所列::

    >tree -l gfs_00h/
    ├── gfs.0p25.2018060200.f000.grib2
    ├── gfs.0p25.2018060300.f000.grib2
    ├── gfs.0p25.2018060400.f000.grib2
    ├── gfs.0p25.2018060500.f000.grib2
    ├── gfs.0p25.2018060600.f000.grib2
    ├── gfs.0p25.2018060700.f000.grib2
    ├── gfs.0p25.2018060800.f000.grib2
    ├── gfs.0p25.2018060900.f000.grib2
    ├── gfs.0p25.2018061000.f000.grib2
    ├── gfs.0p25.2018061006.f000.grib2
    ├── gfs.0p25.2018061012.f000.grib2
    ├── gfs.0p25.2018061018.f000.grib2
    ├── gfs.0p25.2018061100.f000.grib2
    ├── gfs.0p25.2018061200.f000.grib2
    ├── gfs.0p25.2018061300.f000.grib2
    ├── gfs.0p25.2018061400.f000.grib2
    ├── gfs.0p25.2018061500.f000.grib2
    ├── gfs.0p25.2018061600.f000.grib2
    ├── gfs.0p25.2018061700.f000.grib2
    ├── gfs.0p25.2018061800.f000.grib2
    ├── gfs.0p25.2018061900.f000.grib2
    ├── gfs.0p25.2018062000.f000.grib2
    ├── gfs.0p25.2018062100.f000.grib2
    ├── gfs.0p25.2018062200.f000.grib2
    ├── gfs.0p25.2018062300.f000.grib2
    ├── gfs.0p25.2018062400.f000.grib2
    ├── gfs.0p25.2018062500.f000.grib2
    ├── gfs.0p25.2018062600.f000.grib2
    ├── gfs.0p25.2018062700.f000.grib2
    ├── gfs.0p25.2018062800.f000.grib2
    ├── gfs.0p25.2018062900.f000.grib2
    ├── gfs.0p25.2018063000.f000.grib2
    ├── gfs.0p25.2018070100.f000.grib2
    ├── gfs.0p25.2018070200.f000.grib2
    ├── gfs.0p25.2018070300.f000.grib2
    ├── gfs.0p25.2018070400.f000.grib2
    ├── gfs.0p25.2018070500.f000.grib2
    ├── gfs.0p25.2018070600.f000.grib2
    ├── gfs.0p25.2018070700.f000.grib2
    ├── gfs.0p25.2018070800.f000.grib2
    ├── gfs.0p25.2018070900.f000.grib2
    ├── gfs.0p25.2018071000.f000.grib2
    ├── gfs.0p25.2018071100.f000.grib2
    ├── gfs.0p25.2018071200.f000.grib2
    ├── gfs.0p25.2018071300.f000.grib2
    ├── gfs.0p25.2018071400.f000.grib2
    ├── gfs.0p25.2018071500.f000.grib2
    ├── gfs.0p25.2018071600.f000.grib2
    ├── gfs.0p25.2018071700.f000.grib2
    ├── gfs.0p25.2018071800.f000.grib2
    └── gfs.0p25.2018071900.f000.grib2

* 下图就是所有敏感试验与控制试验（福建省气象局业务系统RAP/HRRR的FDDA参数配置）平均所得到的偏差和均方根误差。可以看到调整后的FDDA参数在一定程度上改进了预报的准确性::

* 注：蓝线代表控制试验，红线代表最优参数试验。

* 6小时预报偏差和均方根误差::
.. figure:: ./../Figure_fdda/results/rmse/wrf_tpw_dfprof_f06.000009.png
.. figure:: ./../Figure_fdda/results/rmse/wrf_tpw_dfprof_f06.000010.png

* 12小时预报偏差和均方根误差::
.. figure:: ./../Figure_fdda/results/rmse/wrf_tpw_dfprof_f12.000009.png
.. figure:: ./../Figure_fdda/results/rmse/wrf_tpw_dfprof_f12.000010.png

* 18小时预报偏差和均方根误差::
.. figure:: ./../Figure_fdda/results/rmse/wrf_tpw_dfprof_f18.000009.png
.. figure:: ./../Figure_fdda/results/rmse/wrf_tpw_dfprof_f18.000010.png

* 24小时预报偏差和均方根误差::
.. figure:: ./../Figure_fdda/results/rmse/wrf_tpw_dfprof_f24.000009.png
.. figure:: ./../Figure_fdda/results/rmse/wrf_tpw_dfprof_f24.000010.png