Graph-topology-aware conformal prediction for cross-border flood nowcasts under gauge-network reconfiguration: a climate-analytics evaluation of coverage-loss spillover to downstream warning services

Conformal prediction provides distribution-free coverage guarantees for flood nowcasts, but its exchangeability assumption is violated when gauge networks are reconfigured across borders. The framework is formalised through graph Laplacian eigenmodes, a topology-conditioned nonconformity score, and a proposition establishing that spillover is bounded by the spectral gap of the reconfiguration operator. A climate-analytics evaluation is designed around a Lagos, Nigeria case study using NASA POWER monthly reanalysis air temperature at 2 m for 2000-2023 (312 monthly observations; mean 26.414 C; standard deviation 0.943 C; warming trend 0.1553 C per decade), which serves as a proxy for the hydroclimatic forcing that drives gauge-network stress. The analysis is methodological and theoretical: no empirical coverage experiment is reported. The contribution is a formal characterisation of how topological reconfiguration propagates coverage loss, together with a set of testable propositions and a proposed evaluation protocol for cross-border flood nowcast services.