Diatoms are important primary producers in lake ecosystems and, as a sink for dissolved (DSi) and biogenic silica (BSi) originating from land, can significantly impact the global Si cycle. After burial in lake sediments, resistant diatom frustules can also be used for reconstructions of past ecosystem change. The BSi content of lake sediments is thus often used as a proxy for past diatom productivity, and measured using a time-step analysis of Si extracted in 0.1 M Na2CO3. However, studies in soils and ocean sediments have shown that also certain non-biogenic Si fractions are prone to dissolve in alkaline solutions, contributing to a potential overestimate of sedimentary BSi concentration. In lakes, volcanic and terrestrial Si compounds reactive at high pH are likely to interfere with the analysis. In this study we used a continuous analysis of Si and Al extracted in 0.5 M NaOH and a new mathematical model improved from Koning et al. (Aquat Geochem 8:37-67, 2002) to distinguish between biogenic and non-biogenic Si fractions in lake-sediment extracts. We tested this approach in 43 samples of a 19,000-year sediment sequence recovered from Lake Rutundu, a volcanic crater lake on Mt. Kenya. Our results show that a significant fraction of the extractable Si is of non-biogenic origin, especially in the part of the sequence deposited during the glacial period. We conclude that this technique is essential for the characterization of different Si phases, and in particular the diatom-derived BSi, in the sediment of lakes situated in volcanic catchments. It allows calculating a correction, based on the distinct Si:Al ratio of each of those Si phases, that eliminates the contribution from non-BSi fractions