Earth Observation-Driven Spatial Disaggregation of Exposure Models for Seismic Risk Analysis

The spatial resolution of exposure models is a critical factor in probabilistic seismic risk assessments. Aggregating exposure data at a regional scale often leads to inaccuracies in risk estimates, underscoring the need for spatial disaggregation at finer resolutions. Traditional methods typically rely on readily available data, such as population density, while newer approaches utilize advancements in Earth Observation (EO) technologies from remote sensing. This study examines the sensitivity of seismic risk estimates to various EO-based disaggregation methods, incorporating population counts, built-up areas, and building heights.  These methods are tested in countries with high resolution exposure models: Chile, France, and Nepal. The analysis involves aggregating exposure data at the first administrative level, followed by spatial disaggregation and subsequent testing through risk calculations using the OpenQuake engine. A uniform spatial grid resolution of 0.01° decimal degrees (approximately 1km) is employed. The study evaluates the spatial distribution of key risk metrics, including the number of buildings, replacement costs, occupants, and average annual loss (AAL). Results show that disaggregating exposure using a combination of population and built-up area data produces estimates that more closely align with actual exposure distributions, reducing errors in AAL. Moreover, EO-derived methods combined with fine grid resolutions are promising for enhancing risk modeling, with potential applications to other hazards, such as floods.