Bering Sea climate dynamics forecast by novel multivariate natural hazard assessment method, utilizing self-deconvolution scheme

This case study advocates a generic state-of-the-art multidimensional natural hazards evaluation methodology, applied to windspeeds and wave heights, measured in different offshore locations. Due to complex nonlinear spatiotemporal crosscorrelations between different environmental system components and covariates, it is challenging to assess associated environmental risks, utilizing existing reliability techniques. Hence, it is necessary to develop novel multimodal reliability and risk assessment methods for natural hazards prognostics further, given global climate variability. Advocated multivariate risk assessment methodology being particularly suitable for both environmental and offshore/ocean structural systems, which have been either physically measured or numerically simulated over a representative period. National Oceanic and Atmospheric Administration (NOAA) buoys, operating in the central Bering Sea, provided the raw in situ measurements of windspeeds and wave heights, utilized in this case study. A relatively limited amount of underlying data had been analyzed – only 4 months between June and September 2024. The presented multimodal natural hazards prognostics methodology has a generic nature, hence, large amounts of measured data can be analyzed if available. A novel non-parametric deconvolution extrapolation scheme has been utilized to accurately forecast in situ extreme environmental climate dynamics events. System’s quasi-stationarity was assumed; otherwise, for nonstationary multidimensional dynamic systems with underlying multivariate trend, this trend has to be identified first, before the advocated reliability methodology to be applied. Distinct advantage of presented multivariate reliability methodology versus existing ones lies within its ability to overcome “curse of dimensionality”, namely ability to treat systems with dimensionality above two.