final_program_abstracts[1]

11 IMSC Session Program
Space-time stochastic modeling of climate changes
Tuesday - Poster Session 10
Vladimir A. Lobanov and Svetlana A. Gorlova
Russian State Hydrometeorological University, St.Petersburg, Russia
Time series of any climate characteristic in present and past can be represented as a
statistical ensemble of different time scales fluctuations. The general space-time
modeling includes a generalization of information in three main directions: intra-year,
inter-year and space. Description of all fluctuations includes the following steps:
- description of intra-annual fluctuations in form of the parameters of seasonal
function or averaged values;
- extraction of interannual, decadal, centural, millennia and other time-scale
components, connected with different factors of climate change, and their presentation
in the form of stochastic or deterministic-stochastic model;
- spatial classification and determination of homogeneous regions;
- development of spatial statistical models of different forms for homogeneous
regions.
Statistical methods and the particular tools have been developed for realization of
each step of development of such joint model. Among them are:
- two parameters linear model of seasonal function as a number of relationships
between averaged data for long-term period and data of each year;
- robust statistical methods of decomposition and smoothing for extraction of
interannual, decadal, centural and other time-scale components from observed time
series, which do not misrepresent the properties of fluctuations;
- application of new dynamic characteristics, which characterize abrupt climate
change and other dynamic properties (period and amplitude of cycles, speed of
increasing and decreasing of fluctuations and their durations, etc.);
- identification of random events and their generalization into the time model as
probable distribution function (pdf) or time function for modeling and forecast;
- tool for classification and regionalization of climate change components, which
takes into account a spatial correlation and threshold index of intersection of the same
sites in each class;
- linear spatial model, which connects mean historical climatic field with each year
field and includes two main coefficient: gradient and level of the field (external
properties) and a parameter of space-time non- homogeneity (internal properties).
Application of each technique is shown in form of the particular case studies and they
are:
- processing of time series of palaeodata for the last 17-20 thousand years;
- time model for more than 3 centural instrumental record of air temperature for the
Central England and precipitation records in the same region and a couple model of
extracted climate change components;
Abstracts 117