Reference climate periods shape estimates of spring onset trends: evidence from Estonia

The timing of spring onset is a widely used indicator of climate change impacts, yet estimates of phenological trends depend critically on the choice of reference climate period. Using long-term daily air temperature records from 21 meteorological stations across Estonia, this study examines how different climatological normals influence estimates of spring onset timing and its change. Spring onset was defined as the sustained transition of mean daily air temperature above 5 °C and analysed across four overlapping reference periods (1965-1990, 1971-2000, 1981-2010, and 1991-2020). Mean spring onset dates vary substantially among reference periods, demonstrating that the concept of an "average" spring is sensitive to the selected normal. Cumulative changes derived from linear trends reveal a widespread advancement of spring onset over 1965-2020, with particularly strong shifts at several inland and coastal stations. When contrasting equal-length 30-year normals (1971-2000 vs. 1991-2020), trend magnitudes are generally weaker in the most recent period and, at some stations, near zero or positive. Comparison across reference periods indicates pronounced temporal heterogeneity in phenological responses. While some stations exhibit strong advances during the late twentieth century followed by weakening in the most recent climate normal, others show sustained strong change or weak and inconsistent responses. These patterns cannot be adequately described by a single linear trend and do not follow a simple geographic gradient. The results demonstrate that reference period selection can substantially affect both the magnitude and interpretation of phenological trends. It is therefore recommended that phenological studies explicitly report reference periods and, where possible, compare trend estimates across multiple climatological normals to improve the robustness and transparency of climate impact assessments.