Water-Level Fluctuations Can Confound the Source Interpretation of Principal Components: Concentrations and In-Lake Masses over 19 Years in Bosten Lake, Arid Northwestern China

Lake-water-quality assessment and multivariate statistical analysis are usually based on concentrations. In lakes where water level and storage fluctuate widely, however, hydrological change can produce a common concentration response across many solutes. Whether this shared signal becomes a dominant axis of the concentration matrix, and thereby affects the environmental or source interpretation of principal components, has not been directly tested. Using 19 years (2006–2024) of monitoring data from 17 sites in Bosten Lake, Northwestern China, we examined how water-level fluctuations relate to the dominant axis of the concentration matrix and compared long-term changes in concentration and estimated in-lake mass. Concentrations of all solutes except dissolved oxygen, including the conservative tracers chloride and electrical conductivity (EC), decreased as water level rose, whereas the corresponding estimated in-lake masses showed no negative relationship with water level; chemical oxygen demand (CODCr) concentration declined, while its mass showed no trend. The first principal component, loaded on both conservative and reactive solutes, explained 34% of the variance and was strongly associated with water level (r = −0.82 to −0.89 across three temporal scales). Controlling for water level reduced the shared variance between CODCr and EC from 58% to 4%. Water-level fluctuations can therefore contribute substantially to a dominant common-variation structure in concentration data and may confound the source interpretation of principal components; component scores should be tested against water level before components are interpreted, with volume-normalised measures reported alongside concentrations.