Database Reference
In-Depth Information
1. Introduction
Climate and earth sciences have recently experienced a rapid transfor-
mation from a data-poor to a data-rich environment. With the recent
advances in earth monitoring technologies, observations from remote
sensors on satellites and weather radars, or from
in situ
sensors and sen-
sor networks, as well as outputs of climate or earth system models from
large-scale computational platforms, provide terabytes of temporal, spa-
tial and spatio-temporal data about earth's complex processes. In addi-
tion, the increasing use of geographical information systems for decision-
making has provided an additional source of large spatial datasets. These
massive and information rich datasets offer a huge potential for advanc-
ing earth science research and its impacts on related domains. Examples
of earth science research tasks include developing robust, global-scale al-
gorithms that provide spatially explicit and regularly updated techniques
for global monitoring of the entire earth's land surface and oceans, de-
termining relationships between multivariate events and variables, etc.
The understanding and monitoring of earth science sensor datasets
offer unique data-centric and algorithmic challenges imposed by the vol-
ume, variety and richness. It is not only the massive size of datasets that
poses a challenge, but also complexities due to the unique data character-
istics such as spatial heterogeneity, temporal variability, and uncertainty.
An additional challenge arises from the broad range of questions posed
by diverse scientific disciplines covered in the broad purview of ecosys-
tem or environmental sciences. Specifically, the analysis and discovery
approaches need to be cognizant of climate and ecosystem data char-
acteristics, the value of physically-motivated conceptual understanding
and functional associations of the earth's system, as well as possible
thresholds and tipping points in the impacted natural, engineered, or
human systems. Thus, there is a strong need for understanding and
advancing the state of the art in computational algorithms and robust
data analysis methods which are tailored for applications in the earth
science domain, crossing the traditional boundaries between computer
science and earth science.
The analysis and discovery techniques for understanding and moni-
toring earth science sensor datasets can be broadly classified into the fol-
lowing two categories - (i) event detection, and (ii) relationship mining.
First, detecting events of interest over land, ocean and atmosphere using
multiple data sources enables earth scientists to monitor natural as well
as anthropogenic processes and to quantitatively assess their environ-
mental and socioeconomic impact. Second, finding relationships between
spatio-temporal events and variables is crucial for improved understand-
