华东沿海ASCAT反演风速的检验和订正

基于2010—2014年ASCAT反演风速、华东沿海14个浮标站和浙江沿海249个自动气象站资料，对华东沿海ASCAT反演风速进行检验和订正。研究表明:站点ASCAT风速误差不仅与离岸距离相关，而且与站点周围地形有关，误差较大的5个浮标站均位于舟山群岛附近海区，平均偏大4.79 m·s-1，其他海区浮标站的ASCAT反演风速平均偏差仅为0.46 m·s-1。ASCAT反演风速与浮标站风速的线性回归可有效减小反演风速误差，订正后误差大幅减小，误差越大的站点订正效果越好。相距160 km内的浮标站点间风速误差呈正相关，且站点间距越小，误差正相关越明显。考虑带影响半径的反距离权重，采用邻站方程订正法和邻站误差订正法分别对华东沿海ASCAT反演风速进行订正，均能明显减小平均偏差和均方根误差，两种方法订正效果接近，即两种方法均有较好的订正效果，可用于实际业务。

Verification and Correction on ASCAT Wind Velocities Within the Offshore East China Sea

Based on ASCAT wind velocities, observations of 14 meteorological buoys in the offshore East China Sea, and 249 automatic weather stations (AWS) along coastal Zhejiang Province from 2010 to 2014, verification and correction methods are implemented on ASCAT wind velocities and buoy observations. The analysis indicates ASCAT wind velocities are overestimated for all the 14 buoys in comparison with observations, but only 5 of them, all located off Zhoushan Archipelago, hold deviations greater than 2 m·s-1 with mean bias of 4.79 m·s-1, and the mean bias for the rest buoys is only 0.46 m·s-1. Results also imply ASCAT wind velocities are not only related to distances away from the coastal line, but also to the local terrains. Regression methods are applied to investigate relations between ASCAT wind velocities and observations at all the buoys with regression and independent test samples ratio of 70% to 30%. It shows that linear regression can help reduce ASCAT wind deviations at all the buoys, decreasing the mean bias from 2.02 m·s-1 down to 0.14 m·s-1, especially at those stations with big errors. The relation of ASCAT deviations among buoys is also studied, indicating there is a positive correlation between the ASCAT wind errors and distances for buoys within 160 km, the closer the distances between buoys are, the bigger the coefficients are, with the logarithmic fitting taking advantages of the linear fitting. Two methods, namely regression and deviation, are carried out to make corrections on ASCAT wind velocities, with effective radius taken into account while doing inverse distance weighing interpolations. Results show the mean deviations and root mean square errors decrease obviously after revision, two methods reduce the mean biases by 1.86 m·s-1 (67.9%) and 1.74 m·s-1 (64.2%), and reduce the root mean square errors by 1.19 m·s-1 (29.2%) and 0.89 m·s-1 (29.6%), repectively. Case study on the regression method is carried out with corrected ASCAT wind velocities compared with the 10 m wind fields at lead time 0 h of European Centre for Medium-Range Weather Forecasts (ECMWF) fine model (resolution of 0.25°×0.25°). It shows that two methods are proved positive and can help decrease mean wind deviation. Further analysis shows that the deviation method gets the least mean deviation when AWS observations are taken into account, implying that the enhancement of station resolution can help increase the correction result.