Diagnostic framework for causal inference in seasonal urban flooding: Precipitation-based control selection and synthetic difference-in-differences in Lagos, Nigeria

Establishing causal relationships between urban flooding and behavioral responses is challenging in tropical coastal cities experiencing seasonal flooding, where exposure often limits distinct control areas, rainy seasons with inundation episodes complicate discrete treatment timing, and satellite temporal resolution constrains flood tracking. We develop a framework that facilitates causal inference by shifting the unit of analysis from geographic locations to facility types. The framework uses two screening metrics to identify donor categories: Maximal Information Coefficient (MIC) for identifying facility types whose visitation patterns exhibit minimal sensitivity to precipitation variability, thereby screening for weather-resilient categories rather than direct flood impacts, and Coefficient of Variation (CV) for assessing temporal stability across flood phases. The framework then integrates these selected donors into a hybrid Synthetic Control-Difference-in-Differences estimator. In Lagos, Nigeria's June-July 2020 rainy season, the framework integrates Location-Based Services data, ERA5 precipitation reanalysis, Sentinel-1 SAR imagery, and OpenStreetMap infrastructure. Analysis reveals heterogeneity: healthcare visitation increased 40% during flooding and remained elevated at 51% above baseline through recovery; transportation declined 22% with no recovery; retail exhibited post-flood rebounds of 35%. Effect directions remained consistent across three control specifications (Religious-only, Residential-only, and optimized synthetic control), with the synthetic approach achieving 42-64% reductions in standard errors relative to fixed controls. The framework provides a systematic approach for impact assessment in data-constrained disaster contexts where spatial controls are limited, and discrete event isolation is constrained by monitoring infrastructure. By using precipitation as a temporally resolved proxy for flood exposure, the framework estimates compound flood-season effects using data increasingly accessible in tropical urban settings.