针对全国自动土壤水分实时小时观测资料,结合仪器标定方法、土壤水文物理常数等因素,研究土壤水分固有变化特征,分析异常数据误差来源和阈值判定,利用频率检测、界限值检测、无降水突变检测、异常峰值检测、恒值检测五类方法对土壤水分观测资料进行质量控制实验和检验。结果表明：（1） 自动土壤水分异常数据主要分为：粗大值、突变、异常峰值和僵值,主要由仪器失灵、土壤水文物理常数测定不准、传感器标校不合理等原因引起。（2） 频率检测可有效检出因仪器故障引起的错误数据。目前该方法已应用于综合气象观测数据质量控制业务系统,用于开展全国实时自动土壤水分小时数据的质量控制和质量评估。

Soil moisture plays a crucial role in the study of agricultural drought monitoring, yield prediction, and soil erosion, which are of great significance for agriculture, drought, and climate studies. Automatic soil moisture observation instruments have become an important component of the automatic soil moisture observation stations run by the meteorological department in China given their high precision, high spatial and temporal resolution, and nondestructive capabilities. The accuracy of data obtained through the process of observing soil moisture is seriously affected by the calibration methods used, the stability of the equipment, and the texture of the soil. Thus, it is especially important to establish a quality control method for automatic soil moisture observation data that are free from these error sources that affect the observation quality. In attempting to solve the outstanding quality problems in automatic soil moisture observation data, this paper studies the inherent variation characteristics of soil moisture on the basis of the data obtained by the automatic soil moisture observation stations in China between 2013 and 2015. Combined with the instrument observation principle and the data characteristics and error sources of abnormal data, this study classifies and statistically analyzes the form of the abnormal data, and given the threshold, it puts forward a preliminary practical set of quality control methods for the hourly soil moisture observation data, which includes frequency detection (FD), threshold detection for volume moisture content and relative humidity, break drop detection, sudden change detection, and constant detection. The application effect of this quality control method has been verified using data obtained in China in 2019. The results show that (1) the four main categories of abnormal data are gross errors, mutations, abnormally high values, and stiffness; these are mainly caused by instrument failure, abnormal soil hydrological constants, and unreasonable calibration of sensors. (2) Based on the FD method, which analyzes the change characteristics through the frequency values measured by the sensors in the air and water, this quality control method can effectively detect the abnormal data caused by instrument failure, with the accuracy rate reaching 100%. (3) The evaluation results from the hourly data of the national automatic soil moisture stations from June to September 2019 show that the accuracy rate of data is 93.8%, the abnormal rate is 1.7%, and the missing rate is 4.5%. The abnormal data are mainly distributed in the Qinghai, Sichuan, Shandong, Hebei, and Guangdong Provinces. The soil surface layer (0-20 cm) has a higher proportion of anomalous data than that of other layers, and only a few stations have an anomaly within the whole observation layer. Although the abnormal phenomena can be observed at most sites across the country, the primary problem is that a small number of sites have long-term abnormalities, which have been caused either by the calibration of the sensor or the inaccurate determination of the soil’s hydrological and physical constants. (4) This quality control method can effectively detect abnormal data in China. Presently, it has been applied to the Integrated Meteorology Observation Data Quality Control System.