1 SIGNAL PROCESSING METHODS IN ANALYSING OF SPATIAL-TEMPORAL CLIMATE DATA Branimir Reljin 1 , Irini Reljin 1,2 , Gordana Jovanović 3 1 Faculty of Electrical Engineering, University of Belgrade, Bulevar Kralja Aleksandra 73 2 PTT College, Zdravka Čelara 16, 11000 Belgrade 3 Federal Hydro-Meteorological Institute, Birčaninova 6, 11000 Belgrade, Serbia and Montenegro ABSTRACT Climate data fields are characterized by a large amount of information (having, usually, many non-dominant patterns), multivariability, and non-linear nature. Several statistical and signal processing methods are applied in analyzing such of data – some of them are used here in analyzing of climate in the region of Serbia and Montenegro. The teleconnection with the two major generators of the global climate variability: the ENSO (El Nino Southern Oscillation) and the NAO (North Atlantic Oscillation) is approved. 1. INTRODUCTION The climate system is composed of all processes that control the atmospheric state of the Earth. The ultimate source of energy that drives the climate system is radiation from the Sun. The climate is highly dynamical system, responding to variations in external forcing on a wide range of space- and time scales. The response time of the various components of the climate system is very different: from days to weeks – as observed in the troposphere, to decades but up to centuries or millennia – for oceans. But, even without changes in external forcing, the climate may vary naturally, because, in a system of components with very different response times and non- linear interactions, the components are never in equilibrium and are constantly varying. A well-known example is the quasi-periodically varying ENSO (El Nino Southern Oscillation) phenomenon, caused by atmosphere-ocean interaction in the tropical Pacific. Although the central Pacific region Nino 3,4 (120-170W, 5N-5S) occupies only one-seventh of the equatorial Earth perimeter, and its sea- surface temperature (SST) variation is very small (during 50 years registered data [1] the maximal deviation of monthly temperatures lies within the boundaries of ±3 0 C), see Fig. 1, the temperature anomalies in this region transform weather around the globe. Another example is the North Atlantic Oscillation (NAO). The NAO index is defined as a normalized difference in sea surface pressures (SSP) between the center of high pressure above Azores and low pressure stationed above Iceland [2]. Pressure oscillation between these centers can be imagined as a great see saw of air masses in the North-South direction. The influence of NAO is predominant near the activity centers, but its consequences are felt in the wide area from North America to Siberia and from the Arctic to Northern Africa. Local climate is generally much more variable than the climate on a hemispheric or global scale because it is influenced by a number of additional events. But, even local climate is dictated by global changes, too; in other words, the climate exhibits teleconnection behavior. The impact of ENSO and NAO on climate in different regions on the Earth was approved by different methods and by many authors [2- 7]. Here, we will present some of our investigations in describing the influence of these two major generators of the climate variability on the climate in the region of Serbia and Montenegro. We used several linear analyzing methods, such as the Fourier analysis, the correlation and the EOF (Empirical Orthogonal Function) analysis [8], as well as some of non-linear methods: the fractal and multifractal analyses [9-10], and the SOM (Self-Organizing Map) neural network [11]. 2. ANALYZING OF CLIMATE DATA As a first example let us observe monthly averaged SST in the region Nino 3,4 and available monthly averaged temperatures in the region of Serbia, during the same 48-year period (from 1951 to 1998), as depicted in Fig. 1. At the first glance the two time series in Fig. 1, each having 576 samples (monthly values), exhibit no visible coincidence: SST series behave more fractal than periodical, while temperatures in Serbia have strong annual periodicity. 1951 1961 1971 1981 1991 year 22 24 26 28 30 Sea-surface temperatures of Niňo3,4 region from 1951 to 1998 yr. 0 C Average temperature 25.8 0 C Warmer than normal – El Niňo Cooler than normal – La Niňa 1951 1961 1971 1981 1991 year -10 0 10 20 30 Temperatures in Serbia from 1951 to 1998 0 C Average temperature 10.6 0 C Figure 1. Monthly averaged temperatures in the Nino 3,4 region and in Serbia