Лëд и снег (Feb 2022)
Stochastic modeling of sea ice concentration fields for assessment of navigation conditions along the Northern Sea Route
Abstract
Article describes a probabilistic model (stochastic generator) of spatial-temporal variability of sea ice concentration. Values of the ice concentration are generated at the nodes of the spatial grid with 10 km resolution; the model time step is one day. The change in ice concentration with time (temporal variability) is modeled on the basis of a matrix of transient probabilities (discrete Markov chain), each row of which is a distribution function of the conditional probability of changes in the ice concentration. Spatial variability is determined by empirical probability fields, with which the observed changes in fields of the ice concentration are associated with known conditional probability distribution functions. To identify the parameters of the stochastic generator, satellite data from the OSI SAF project for the period 1987–2019 were used. The generator takes into account seasonal, interannual and climatic variability. Interannual and climatic variability are determined on the basis of a stochastic model of changes in the types of ice coverage. In order to verify the developed stochastic generator, we compared the statistical indicators of observed and calculated ice fields. The results showed that the fieldaverage absolute error of statistical characteristics of the ice concentration (mean and standard deviation) does not exceed 3.3%. The discrepancy between the correlation intervals of ice coverage calculated from the model and measured ice concentration fields does not exceed 2 days. The variograms of the modeled and observed fields have a similar form and close values. As an example, we determined the duration of navigation of Arc4 ice class ships between the Barents and Kara Seas using synthetic fields of the ice concentration reproduced by the stochastic generator.
Keywords
