Database Reference
In-Depth Information
ing of the interactions between different processes of the earth system
at large, and thereby improving predictive power. Together, these two
research tasks offer fertile grounds for developing novel knowledge dis-
covery approaches which focus on addressing a range of interconnected,
societally-relevant themes at the core of impending environmental con-
cerns cutting across diverse disciplines. They enable a wide community
to analyze changes in the earth's system, interactions between different
processes from local to global scales, and their impacts on the carbon
cycle, hydrology, air quality, biodiversity, and other research areas. In
particular, qualitative inferences about changes and relationships in the
earth's system and their impacts may be transformed into quantitative
historic and predictive insights based on a combination of hypothesis-
driven and data-guided discovery processes.
The remainder of the chapter is organized as follows. In Section 2, we
provide an overview of the types of sensor datasets used in earth science
research. Section 3 focuses on the data-centric challenges posed by earth
science applications. Sections 5 and 6 introduce two broad categories of
research problems using earth science data, namely event detection and
relationship mining, providing illustrative examples in each category.
Section 7 contains concluding remarks and directions for future work.
2. Overview of Earth Science Sensor Datasets
Earth science sensor datasets possess varying data characteristics, ac-
quisition methods and domains of coverage (both in space and time).
They either consist of local sensor recordings (
in situ
data) or are ob-
tained through instruments mounted on satellites or other remotely
based locations (remote sensing data).
In situ
sensors which are non-
uniformly distributed in space at local or regional scales can be processed
and made available at a fixed spatial grid using basic interpolation, ag-
gregation and sampling techniques so that the processed data is free
from missing values or non-uniformly spaced data. Further, interpo-
lation methods can range from simple linear interpolation to reanalysis
techniques using climate simulation models. We provide a brief overview
of the diverse types of datasets used in earth science research in the fol-
lowing subsections.
2.1 Observational Data
Observational datasets that are commonly used in earth science re-
search can be broadly classified into
station-based
and
gridded
data.
