Malaria seasonality — from a stale rainfall proxy to climate-driven emergence

The emergence / transmission shape feeds OpenMalaria’s EIR and should lag rainfall (aquatic development + habitat filling). The intervention onset drives product timing and sits at or before rainfall onset — chemoprevention needs drug onboard before biting ramps, and roads wash out once rains arrive.

They share rainfall but diverge on purpose, so the emergence-model lag can never corrupt deployment timing.

τ is a per-species persistence time-constant (in days) — roughly how long a larval habitat lingers after rain before it dries out. It is a model parameter, not a directly measured quantity, and it enters the kernel through a monthly water-balance recursion:

habitat_m = habitat_m−1 × decay_m + rain_m
decay_m = exp( −(30 / τ) × PET_m / PET_ref )

Each month, standing water is topped up by rain and drained by evaporative demand. A hot, high-PET month drains it faster; a larger τ makes it persist — carrying transmission into the dry season. So τ is exactly what separates a flashy, rain-tracking vector from one that breeds in water that lasts.

Peak-transmission month under the legacy country method vs the admin1 kernel, per admin, weighted two ways: by population and by estimated infections (population × MAP PfPR_2–10). The decision is not academic — it moves the recommended timing for a majority of people, and the burden-weighted share is higher still where disagreement is largest.