A compositional framework for modelling post-earthquake damage, economic loss and recovery needs based on Post-Disaster Needs Assessments

Abstract. Earthquakes generate multisectoral impacts that Post-Disaster Needs Assessments (PDNAs) evaluate comprehensively to guide long-term recovery. This study develops a compositional framework to estimate post-earthquake damage, economic losses, and recovery needs from a multisectoral perspective, based on data from 17 PDNAs. The method combined aggregate and compositional estimates. The relative distribution of damage, losses, and needs across housing, transport, energy, WASH, health, education, and the productive sector was modeled using isometric log-ratio (ilr) transformations, while their total magnitudes were estimated in millions of USD. Disaster characteristics, severity indicators, and vulnerability dimensions were used as explanatory variables. Ordinary least squares (OLS) and generalized additive models (GAMs) were compared using leave-one-out cross-validation and MAE, MSE, and RMSE metrics. The results showed that the magnitude in USD and the composition of recovery follow different underlying logics: while economic damage provides an approximation of the total volume of losses and recovery needs, vulnerability dimensions play a central role in explaining how these needs are distributed across sectors. Our findings highlight the importance of incorporating general vulnerability into transport recovery planning; adaptive capacities in energy; epidemiological vulnerability in WASH; socioeconomic disparities in education; and socioeconomic development in the productive sector. To our knowledge, this is one of the first studies to apply compositional data analysis (CoDA) to the modeling of damage, economic losses, and recovery needs within the PDNA framework.