nephatmwanza/zambia-rainfall-extremes

# Zambia rainy-season extremes: flood and dry-spell risk, 1981–2026 Have Zambia's rainy-season dry spells, wet spells and extreme-rainfall totals shifted over the past 45 seasons — and if so, where? Short answer: **almost nothing is statistically significant on its own, but the direction of change is remarkably consistent, and it points to a redistribution rather than a simple wetting or drying.** Rain is falling on *more* days and adding up to *more* total, while the heaviest single-day downpours *weaken* and the longest dry spell *within* the season gets *longer*. For rain-fed agriculture that combination matters more than the seasonal total does. > **Companion study — zambia-temperature-extremes →** > The same test, run on ERA5 temperature over the same country and the same 45 years, > gives the opposite answer: the hottest day of the year is rising in **all ten provinces**, > and in **this same ONDJFM window 969 grid cells survive false-discovery-rate control** > against **zero** for any rainfall index. That contrast is what makes the rainfall null > result below credible rather than merely inconclusive. *Trend per decade in each index, 1981/82–2025/26. Blue is wetter, red is drier, throughout — CDD uses a reversed scale because a **rising** dry spell means a **drier** season. Stippled cells are significant at p < 0.05 before multiple-testing control (see Finding 2).* --- ## The question, and why it is framed this way Zambia's rains fall in a single wet season, **ONDJFM** — 1 October to 31 March. Almost all national crop production depends on it. Two things can go wrong inside a season that a rainfall *total* completely hides: - a **dry spell** long enough to kill a crop mid-season, even in a wet year - a **cloudburst** heavy enough to flood, even in a dry year So the analysis is built on spell-length and percentile-exceedance indices rather than totals, and every index is computed over the ONDJFM window only. A season is labelled by its **start** yea …