Clustering of groundwater hydrographs to reveal common patterns for the Baltic region

The aim of this study is to identify salient patterns of groundwater level dynamics in the Baltic region. The study investigates correspondence between (grouped) groundwater level dynamics and catchment, well and physiographic site characteristics. The analysis was carried out in five consecutive steps. Firstly, 1691 groundwater hydrographs were collected from Baltic surveys responsible for national groundwater level monitoring (Latvain Environment, Geology and Meteorology Centre; Estonian Environment Agency; Geological Survey of Lithuania). Observation wells represent both unconfined and confined aquifers. Groundwater level time series were checked for errors and treated according to an approach proposed by Retike et al., 2021. Secondly, the dataset was limited to the time period when daily groundwater level measurements were available in all three countries. The resulting time period covers 8.5 years at 689 wells. Then, the acceptable amount of missing values was defined as the balance between most complete hydrographs, the number of retrained hydrographs and spatial coverage of the wells. Gaps (if present) in the remaining 283 wells were filled with missForest (Stekhoven and Bühlmann, 2012). The results were visually inspected to identify groundwater hydrographs with suspicious modeling patterns and five wells were removed from further analysis based on expert judgment. Finally, 278 groundwater hydrographs (136 from Latvia, 58 from Lithuania and 86 from Estonia) were retained and clustered using Hierarchical Cluster Analysis. The identified clusters of groundwater level times series were then explained using descriptive geological (like aquifer lithology, thickness), hydrological (distance to the nearest stream), climatic (precipitation, seasonality) and anthropogenic (land use) characteristics. This research is funded by the Latvian Council of Science, project “Spatial and temporal prediction of groundwater drought with mixed models for multilayer sedimentary basin under climate change”, project No. lzp-2019/1-0165.

References:

Retike, I., Bikše, J., Kalvāns, A., Dēliņa, A, Avotniece, Z., Zaadnoordijk, W.J., Jemeljanova, M., Popovs, K., Babre, A., Zelenkevičs, A., Baikovs, A. (2022) Rescue of groundwater level time series: How to visually identify and treat errors. Journal of Hydrology, 605, 127294. https://doi.org/10.1016/j.jhydrol.2021.127294

Stekhoven, D.J., Bühlmann, P. (2012) Missforest-Non-parametric missing value imputation for mixed-type data. Bioinformatics, 28, 112–118. https://doi.org/10.1093/bioinformatics/btr597