Database Reference
In-Depth Information
Multi-resolution and multi-scale.
Naturally occurring global
phenomena occur at different scales. For example, events such as ur-
banization, fires and deforestation tend to impact smaller areas than
droughts. The degree of spatial heterogeneity of each dataset deter-
mines the necessary grid size to resolve important characteristics. Some
datasets, e.g., population and political borders (important to connect
events to political decision making) are usually available for predefined
regions and need to be interpolated to the gridded space. Weighted aver-
age values can attributed to grid cells that cover multiple spatial regions.
One common approach is to build a bridge between these disparate scales
and develop algorithms that can identify patterns at multiple resolutions
without upsampling all data to the highest resolution.
Spatial autocorrelation.
Tobler's first law of geography states
that “Everything is related to everything else, but near things are more
related than distant things” [69]. Thus, the spatial dependence of earth
science data needs to be incorporated into data mining algorithms.
In the following two sections, we discuss two broad applications of
earth science sensor datasets - (i) event detection, and (ii) relationship
mining. We further supplement the discussion by providing illustrative
examples of problems and methods developed in each of the aforemen-
tioned earth science applications.
4. Event Detection
Identifying different kinds of occurrences of ecosystem events such
as forest disturbances, agricultural intensifications, urban expansions,
ocean eddies and high aerosol concentrations can provide earth scien-
tists and policy-makers with timely information to mitigate and adapt
to critical environmental pressures. Event detection aims at detecting
anomalous and/or change behavior across multiple spatio-temporal vari-
ables spanning multiple facets of information about the earth. Spatio-
temporal datasets such as vegetation indices, night-time lights, sea sur-
face height, land surface temperature, precipitation, aerosol concentra-
tion, and population can be used for identification of these events, as
they often exhibit a change or a characteristic pattern in one or more of
these types of data.
Since different sensor datasets capture unique (and often complemen-
tary) information about events of interest, we can leverage multiple
sources of information in earth science domain to (i) detect and charac-
terize events that exhibit different event characteristics in different vari-
ables (ii) improve confidence in the significance of a detected event and
