Database Reference
In-Depth Information
over North America at the same time is largely due to a combination of
NAO and ENSO [27]. Further, the ability to address important ques-
tions like the degree of climate change and its potential impacts requires
a deeper understanding of the behavior and interactions of these atmo-
spheric processes as well as to capture them precisely. Discovery of rela-
tionships or dependencies among climate variables involved is extremely
challenging due to the nature and massive size of the data. Data-guided
approaches thus offer a huge potential for characterizing and discovering
unknown relationships along with advancing climate science.
Figure 15.8.
Southern Oscillation Time Series at Tahiti and Darwin.
A novel graph based approach was recently proposed in [41, 42] where
the nodes of the graph were represented by regions on the Earth and the
edges were represented by the correlation between the anomaly time
series of two regions. It was shown that the negative correlations are
key for detecting dipoles, and thus need be preserved in both sign and
magnitude. The approach discovered dipoles using a Shared Recipro-
cal Nearest Neighbor (SRNN) algorithm and even enabled tracking the
movements of these dipoles and studying their interactions in a princi-
pled fashion. Data-guided dipole discovery techniques thus offer better
predictive ability of temperature and precipitation anomalies and can
be used for understanding various General Circulation Models (GCMs).
6. Concluding Remarks
Sensor datasets that are used in earth science research provide vast
amounts of accurate, timely and reliable information about earth's com-
