Analysis of Windows of Opportunity for Weather-Sensitive Operations

Abstract

Summary This paper presents an approach to the statistical analysis of weather windows of opportunity, which are defined as the time span over which the stringent, multiparametric conditions required by weather-sensitive marine operations (such as heavy lift, topside float over, pipeline tie in) are met. For this paper, the topside float over has been used as a case study and application of the statistical analysis. The basic data for the analysis are the numerically generated long-time-series wind, wave, and sea currents, which are becoming increasingly available from weather-forecasting models, global reanalysis studies, and project-specific metocean assessment. These long time series allow identification of climatically significant weather windows at the seasonal, and often at the monthly, time scale. Appropriate statistical distributions to fit the weather windows are required to interpolate within and extrapolate from windows durations, smooth the sample data, and obtain objective assessment of weather-window statistics. For this task, two possible distributions are proposed - the Johnson set of transformations (Johnson 1949) and the Kappa distribution (Hosking 1994) – that, although well-documented in the statistical literature, have seldom, if ever, been applied in offshore analysis. Fits on sample data are performed with the L-moment method (Hosking 1990, 1996, 2000) for the Kappa distribution and with both moment (Draper 1952; Hill et al. 1976) and quantile (Wheeler 1980) methods for the Johnson distribution. The entire procedure (definition of limit conditions, identification of weather windows, and statistical analysis of weather windows) is exemplified with reference to a hypothetical float-over operation in the Andaman Sea (Myanmar). The simulation carried out indicates that the use of statistical distributions can enhance the reliability of the weather-window analysis significantly because of smoother description of weather-window durations, improved interpolation and extrapolation capabilities, and higher discriminating power of alternative design solutions. Both the Johnson distribution and the Kappa distribution provide an overall good performance in fitting the sample data, either when subjectively assessed by visual inspection and in terms of the objective analyses of the resulting mean-square errors; hence, they are suggested as promising candidates for routine analysis of weather windows.