北京地区加密自动气象站数据的质量分析
Evaluating the Quality of Meteorological Data Measured at Automatic Weather Stations in Beijing during 1998－2010
中文摘要：
在逐小时气温资料质量评估流程的基础上，着重对1998—2010年北京地区降水、相对湿度、气压、风速、风向5个基本气象要素逐时观测资料的总体运行情况、缺测情况、错误和可疑情况进行了分析。结果表明：降水和风速（风向）的空间分布整体上具有较强的一致性；运行时长较长的站点主要集中在主城区和近郊，此外，西南的房山区和北部的怀柔区也有部分运行时间较长的代表站点。错误资料的检查结果显示，不同要素的观测资料发生错误的特点不尽相同，降水数据主要为极值错误，是在各要素中极值错误中发生率最高的，而相对湿度、气压两气象要素的错误数据的发生时间具有较好的一致性。可疑数据判识结果表明，相对湿度可疑数据在2003年和1999年甄别为错误数据的比例最高，而气压资料的可疑数据在2004年以后未通过空间一致性检查的数据占据较大比例，其原因有待进一步分析。通过对加密自动站数据的质量评估，可以帮助人们更好地把握北京地区高密度自动站观测资料的质量特性，为资料的合理利用提供参考。
中文关键词：
质量评估，数据缺测，自动气象站
KeyWords：
quality evaluation, data-lacking, automatic weather stations
Abstract：
Data quality control is a basic assurance for meteorological researches and data applications. In order to efficiently use the data sets from AWS (Automatic Weather Station) constructed in 1998, hourly data (including precipitation, relative humidity, air pressure, wind speed and direction) measured by 187 AWS over Beijing from 1998 to 2010 were evaluated by its integrality, veracity and confidence in the paper. The approach included the definition and principle applied in the temperature data assessment, in which the AWS data can be divided into correct, severe missing, missing, discrete and slight continuous missing, and dubious data. The spatial and temporal consistent detections are employed in the data quality control flow. The results show as follows: The AWS net of Beijing was set up following a fine layout. During the AWS construction process, the regional character was considered for the representation of AWS. Even in the beginning period, the AWS were set up in urban areas and mountainous areas as well according to the need of different region representation. It is beneficial to urban and rural climate comparison, data sequence reconstruction and regional climate research. The collectivity of the five kinds AWS data is fine. The severe and bad missing data was few. The dispersing and lightly consecutive missing data were concentrated with regional consistency. It is shown from the error test that the incidence of mistakes were different among the five kinds data. Although the number of extreme errors in precipitation is higher than other error styles, the highest number of cumulative errors of precipitation data is not beyond 43 (2005). Relative humidity and air pressure, had errors almost simultaneously. The results of suspicious data discriminating revealed that the suspicious data of relative humidity had the highest proportion in 2003 and 1999, after 2004, while the suspicious data of pressure, which failed, in space consistency check occupy a larger proportion in 2004. The cause remains for further study.