Analysis of Wind Characteristics of Wind-Profiler Radars and Their Quality Control Methods for Data Assimilation

With the 0.5°×0.5° analysis fields of the National Centers for Environmental Prediction/Global Forecast System (NCEP/GFS) as the numerical forecast background and using surface precipitation data aimed at data assimilation, in this study, we first analyzed the quality of wind products from 12 L-band wind-profiler radars in Fujian Province, including three CFL-03 radars and nine CFL-06 radars obtained at 0000 UTC, 0600 UTC, 1200 UTC, and 1800 UTC from January to December 2017. Then, we considered different quality control (QC) schemes and their effects. The results indicate that: (1) Winds detected by CFL-06 radars are obviously better than those by CFL-03 radars with respect to the maximum detection height, effective data availability, and horizontal wind quality at low levels. (2) Great differences exist in the horizontal winds detected by different wind-profiler radars in same series, including data availability, effective detection height, and the vertical distributions of the standard deviation, correlation coefficients, and bias. These differences have no direct relationship with the geographical locations of the wind-profiler radars, i.e., coastal area or inland, or their heights above sea level. (3) Wind-profiler radar products have obvious systematic negative biases relative to the GFS u-wind field, that is, the u winds detected by wind-profiler radars are lower than the GFS background field. This does not meet the no-bias requirement for data assimilation, so bias corrections are necessary in the data assimilation process. In contrast, the v-wind data are relatively better than u-wind data. (4) Precipitation has a great impact on the detection capability of wind-profiler radars. On precipitation days, the data availability is decreased in the middle-low levels but greatly enhanced in the middle-high levels. The standard deviations of u and v winds both increase in the middle-low levels, whereas the standard deviation of the v winds increase and those of u winds greatly decrease in the middle-high levels. (5) Two QC schemes, i.e., a scheme using different high-confidence ranges and a scheme using different effective detection heights are introduced to different wind-profiler radars to compare their results with those of the fixed effective detection height scheme. The results show that the two QC schemes both have obvious advantages. The QC effect of the different high-confidence range scheme is much more obvious, with the horizontal wind data of different radars more fully and effectively identified. This scheme reduces unnecessary loss of radar data, eliminates poor quality data, and achieves good results regarding the precipitation conditions.