Counterfactual auditing of heat-early-warning dispatch under compound-hazard forecast ambiguity: a climate-analytics causal-inference estimation of false-alert exposure in informal settlements

Heat early-warning systems (HEWS) in West African cities must decide whether to dispatch alerts when forecasts are uncertain and when heat interacts with other hazards. The framework formalises dispatch as a decision under a joint distribution of heat and co-occurring hazards, defines the counterfactual false-alert rate as the probability of alert conditional on a non-event, and derives bounds on that rate from forecast ensemble spread and settlement-level exposure heterogeneity. The method is applied to a reanalysis record of monthly air temperature at 2 m for Accra, Ghana (5.6°N, 0.19°W), 2000–2023, comprising 312 monthly observations with a mean of 26.521°C and a warming trend of 0.1622°C per decade. The record is used to characterise the climatological baseline against which dispatch thresholds are set, not to estimate alert outcomes directly. The analysis yields three results: a formal identification condition for counterfactual false-alert exposure; a bound that widens with compound-hazard correlation and with settlement-level exposure variance; and a demonstration that threshold-based dispatch rules calibrated to temperature alone systematically understate false-alert exposure in informal settlements. The framework provides an audit procedure that can be executed with forecast ensembles and settlement exposure maps, and it clarifies why compound-hazard ambiguity is a structural feature of heat-early-warning dispatch rather than a data-quality problem.