SUMMARY
In the dissertation joint effects of large scale circulation and climatic
oscillation phenomena (ENSO: El Niño / Southern Oscillation,
NAO: North Atlantic Oscillation) on the Northern mid-latitudes have
been analysed. Two selected regions have been investigated: (1) the
Carpathian Basin in the Atlantic-European area, (2) the Midwest in North
America. Our main goal was to determine the statistical relationships,
and then considering these findings, to build models that are able to
estimate regional climate parameters.
In the presented studies traditional mathematical statistical techniques
(teleconnection analysis, EOF analysis, multivariate linear regression)
and an alternate technique (fuzzy rule-based models) have been applied.
Fuzzy logic is considered a new approach in the meteorology literature
in Hungary. In order to reveal statistical linkages between distant
regions teleconnection analysis is used, namely, anomaly maps of height
and temperature of several geopotential levels are compared during different
phases of atmospheric oscillation phenomena. Furthermore, monthly relative
frequency of large scale macrocirculation patterns and regional climate
parameters is evaluated in case of these phases. EOF analysis uses empirical
orthogonal functions to map spatial patterns of a given field possessing
the largest variances during different ENSO phases. Regions with the
largest positive and negative values in the spatial patterns of EOF
modes show the action centers of the given field. Multivariate linear
regression (MLR) and fuzzy rule-based (FRB) models determine the regional
climate parameters using predictor variables. Fuzzy logic is able to
consider several contradictory responses, which may be true to varying
degrees. In the dissertation sensitivity analysis of FRB models is carried
out, furthermore, model outputs are compared to results from the MLR
models using the same conditions (i.e., input variables, datasets).
Based on the results besides the more direct effect of large scale
macrocirculation, teleconnection of climatic oscillation phenomena (ENSO
and NAO) is considerably present in the selected regions of the Northern
mid-latitudes. Regional climate information obtained from the FRB models
using both the relative frequency of MCP types and the climatic oscillation
phenomena as input variables are able to reproduce the statistical characteristics
of the observed regional climate parameters. The best results can be
achieved if (1) both zonality and cyclonic/anticyclonic dominancy are
considered at MCP classification, (2) time lag of the climatic oscillation
is taken into account. Simulated time series much better represent the
observed time series if FRB models presented in the dissertation are
used instead of MLR models. Based on the comparison of different model
error terms the same conclusion can be drawn: errors of the FRB models
are smaller than those of the MLR models.
Main advantage of applying fuzzy rule-based models includes long range
estimation of climate parameters when regional climatological information
is needed for a long period.