Temporally annotated corpora about historic events can be crucial to digital humanities research: they allow to extract and date events as well as reactions to them, and to construct timelines of events and of language use, among other applications. However, producing a precise corpus of a particular event in history is very challenging due to the lack of noise-free digitalized data. This paper introduces RussianFlu-DE, a temporally annotated corpus of 639 articles extracted from noisy OCR text of newspaper issues in German. All articles are about the Russian flu epidemic that took place during 1889–1893. We describe the development of RussianFlu-DE, including methods to clean different types of noise in the OCR text, and our tool for extracting Russian flu related articles. In addition, the task of temporal annotation using the TIMEX2 schema is discussed and the characteristics of the corpus compared to other corpora are presented. To show how our contribution supports epidemiology, we present some preliminary yet interesting results obtained from analyzing the articles in RussianFlu-DE. The corpus and associated tools for exploration are publicly available.