Abstract This perspective traces the emergence and evolution of early warning systems (EWS) and anticipatory action (AA). It revisits how major disasters, advances in science and forecasting, and global policy frameworks progressively expanded the ambition of EWS—from hazard-centered monitoring to more impact-based, people-centered approaches—while exposing persistent gaps in communication, local capacity, financing, and political will that prevented warnings from reliably triggering timely action. We examine how AA emerged in the humanitarian sector from these shortcomings as an operational bridge between long-term risk reduction and disaster response using risk information to enable pre-agreed, financed early actions, reviewing evidence that anticipatory interventions can improve food security, protect assets and livelihoods, enhance dignity and agency, and strengthen coordination in humanitarian contexts. Despite having the same goal of protecting the lives and livelihoods of at-risk populations, AA and EWS investments remain fragmented, and they struggle in fragile, data-poor, and conflict-affected settings. Drawing on this history and evidence, we argue that efforts to strengthen EWS and AA must be coherent and mutually reinforcing, building on the broader work on disaster risk reduction and climate-resilient development. The paper outlines concrete action points: align investments across EWS and AA; link government-led and humanitarian mechanisms wherever possible; strengthen inclusive, locally grounded partnerships; level up regional hydrometeorological and climate service capabilities; and institutionalize accountability and learning at scale. By acting on these priorities, practitioners and policymakers can move from pilots and parallel initiatives toward integrated systems that consistently enable people to act ahead of crises. The paper aims to address the following research question: From a humanitarian perspective, how have early warning systems and anticipatory action evolved in theory and praxis over the last three decades?
Jaime et al. (2026) studied this question.