Increasing national annual rainfall alongside trends in Kenya’s equatorial bimodal rainy seasons and recurrent lower-tail years, 2000–2024: dependence-aware analysis with exploratory 1991–1999 context

Solomon Musau Nzioka , Lolem Bosco Lokolile, Emmanuel Wamalwa , Jeremiah Mushosho

National annual rainfall trends can conceal seasonal changes relevant to water availability and climate-sensitive health risks. This study examined whether increasing national rainfall in Kenya was reflected consistently across its two principal rainy seasons, March–May (MAM) and October–December (OND), and whether seasonal lower-tail years coincided. The analysis covered 2000–2024, with 1991–1999 examined separately as exploratory historical context. Trends were estimated using Theil–Sen slopes, Newey–West heteroskedasticity- and autocorrelation-consistent (HAC) regression, and circular block bootstrapping. Complementary one-sided tests assessed directional rainfall increases, while season-specific empirical twentieth-percentile thresholds identified lower-tail seasons.
National annual rainfall increased by 78.25 mm per decade (95% block-bootstrap confidence interval [CI]: 43.74–102.91; HAC p=0.0103). Three annual totals were at or below the empirical 10th-percentile threshold and five at or below the 20th percentile. MAM rainfall increased by 28.66 mm per decade (95% block-bootstrap CI: 7.55–40.78; two-sided HAC p=0.0266), with directional support across tested HAC lags and bootstrap block lengths (one-sided p=0.0164), although the Holm-adjusted two-sided result was just above the 0.05 threshold (p=0.0532). OND had a smaller positive trend estimate of 14.65 mm per decade (95% block-bootstrap CI: −2.28 to 31.03; two-sided p=0.0707). Its directional result was supported under primary settings (p=0.0411), but was not robust across sensitivity settings. Five MAM (20%) and five OND (20%) seasons fell below their respective empirical twentieth-percentile thresholds, with no year experiencing deficits in both seasons.
The findings reveal that increasing national annual rainfall does not translate into uniform seasonal rainfall increases or eliminate recurrent seasonal deficits. Evidence of an increase was stronger for MAM than OND. This seasonally differentiated evidence provides a basis for climate–health vulnerability assessments that integrate rainfall dynamics with the spatial distribution of populations exposed to climate hazards; health outcomes were not directly assessed.