HIGH-PERFORMANCE COMPUTING FRAMEWORKS FOR CLIMATE AND ENERGY INFRASTRUCTURE RISK ASSESSMENT

Abstract

This study addresses a persistent problem in climate and energy infrastructure risk assessment: despite growing adoption of cloud and enterprise high performance computing (HPC) environments, decision makers still lack clear quantitative evidence on which HPC framework capabilities most improve the quality and usefulness of risk assessment outputs. The purpose of the study was to test, in a quantitative cross sectional, case-based design, how key HPC capability dimensions predict Risk Assessment Effectiveness (RAE) in operational settings. Data were collected using a 5-point Likert scale survey from a case sample of cloud and enterprise HPC users involved in climate exposed energy infrastructure risk work (N = 168), including 41.7% operations and reliability, 27.4% risk and resilience planning or asset management, 18.5% data and analytics, and 12.5% IT or HPC administration. The key independent variables were Scalability or Throughput (SCAL), Computational Efficiency (EFF), Data Integration Readiness (DATA), Simulation or Model Execution Capability (SIM), and Reliability or Availability (REL), with RAE as the dependent variable. The analysis plan included internal consistency testing, descriptive statistics, Pearson correlations, and multiple regression with multicollinearity diagnostics. The measures demonstrated strong reliability (Cronbach’s alpha ranged from 0.81 to 0.90; overall alpha = 0.93). Descriptive results indicated above neutral capability levels (SCAL M = 3.74, SD = 0.71; EFF M = 3.61, SD = 0.76; DATA M = 3.92, SD = 0.69; SIM M = 3.79, SD = 0.73; REL M = 3.58, SD = 0.78) and moderately high effectiveness (RAE M = 3.83, SD = 0.68). RAE correlated positively with all capability dimensions, with the strongest associations observed for DATA (r = 0.69) and SIM (r = 0.61). The regression model was statistically significant and explained 58% of the variance in RAE (F(5,162) = 44.72, p < .001; R² = 0.58; Adj. R² = 0.56), with acceptable multicollinearity (VIF 1.42 to 2.18). Headline findings show that DATA (β = 0.31, p < .001), SIM (β = 0.24, p = 0.002), and REL (β = 0.19, p = 0.008) were significant predictors, while SCAL was marginal (β = 0.11, p = 0.087) and EFF was not significant (β = 0.07, p = 0.210). These results imply that organizations seeking stronger climate and energy risk assessments should prioritize data integration pipelines, reliable platform availability, and scenario execution capacity, and treat raw scaling and efficiency as secondary unless they directly enable those capabilities.