Tracing biogeochemical processes in stream water and groundwater using non-linear statistics

Stream water and groundwater solute concentration are subject to a multitude of biogeochemical processes that act at different scales and are often characterized by non-linear relationships and feedback loops. Different multivariate statistical methods were applied to investigate the interplay of different processes. The data set from the Lehstenbach catchment in South Germany comprised 2641 stream water and groundwater samples from 38 different sites in the catchment, where 13 different solutes had been determined. According to the correlation dimension analysis, the number of dominant processes was four. The first four components determined via principal component analysis comprised 88% of the total variance, whereas the non-linear isometric feature mapping explained 92% with the first four components. These components were ascribed to prevailing biogeochemical processes and were used to investigate spatial and temporal patterns. Redox processes and contamination by road salt explained 35% of the variance each. Another 13% were ascribed to near-surface runoff in the acidified topsoil, and 9% to the impact of contaminated filter gravel in some of the groundwater wells.