Discovery and Analysis of Ocean Climate Indices Using DSNN Clustering Algorithm

Abstract

This Paper based on finding interesting spatio-tempral pattern from Earth Science data. The data consists
measurements of various Earth Science variables (include Temperature and pressure) which are related with time
series. Earth Science data has strong seasonal components that needs to be removed prior to pattern analysis, as the
Earth Scientist are primarily interested in pattern that represent deviation from normal seasonal variations such as
anomalous climate event (e.g. , E1 Nino) or tends (e.g., global warming). We used ―monthly‖ Z Score to remove
seasonality. After processing, we apply DSNN clustering algorithm to cluster the temperature time series associated
with point on the ocean, yielding clusters that represent ocean regions with relatively homogeneous behavior. The
centroids of these clusters are time series that summarize the behavior of these ocean areas and thus, represent
potential OCIs (Ocean climate indices).To evaluate cluster centroid for their usefulness, we must determine which
cluster centroids significantly influence the land area. For this task, we use variety approaches that analyze the
correlation between potential OCIs and time series.