[M-GI30] Near Surface Investigation and Modeling for Groundwater Resources Assessment and Conservation

[MGI30-06] An approach for identifying outliers of groundwater observations

*Shih-Yao Lee¹, Shang-Chen Ku¹, Hwa-Lung Yu¹ (1.National Taiwan University)

Keywords:Outliers detection, Data imputation, Bayesian maximum entropy

The groundwater head observations have been extensively collected in Taiwan since the groundwater monitoring network was established. The data quality varies depending on the maintenance of the observation wells. Therefore, the outlier and missing data should be detected and removed before further application. This study proposes an approach for identifying outliers, which examine whether the fluctuation of the time series from an observation well meets the proposed criteria. If the first-order differences of head observations meet the proposed criteria, the data points will be marked as outliers and removed. The Bayesian maximum entropy approach is applied to missing data estimation after outlier detection.