IEEE Geoscience and Remote Sensing Magazine - June 2017 - 90

and recognition within the games to attract good classifiers
(e.g., becoming the coauthor of a publication and contributing to address big issues). The tool is under evolution to
include more piles (e.g., water bodies, biodiversity, disaster
impact) and more features (e.g., pattern delineation). One
challenge is to check the quality of the classifications by
developing consensus and crowd-ranking algorithms.

backs processes, feedback, changes of regimes, ecological
transitions, and the impact of some climate changes and
land-use practices.

Parameters

OPEN INNOVATION IN THE DIGITAL MARKET PLACE
Open innovation is about involving many and diverse
actors in the innovation process, ranging from researchers
to entrepreneurs and civil society. Open science provides
the foundation for open innovation by providing the tools
DATA-INTENSIVE SCIENCE
and knowledge upon which innovation can build.
The challenges of exploiting big data are not only accessing
In this context, ESA is supporting a series of Open Inand processing large amounts of data but also integrating
novation activities, including app camps and Copernicus
them into a more complete picture of a multivariate system,
masters. ESA is also now supporting the development of a
taking account of their different nature and uncertainty, and
continuous pipeline for innovation aiming to transform inmost importantly, mining them for information and insight.
novative EO ideas into market ventures. This so-called Earth
One way to do that is through gridded data cubes,
Lab Accelerator will provide developers, entrepreneurs, and
organizing different data sets on the same geotempodata scientists with the EO data, infrastructure, access to
ral grid. This approach, with roots in the business secnetworks, and coaching necessary to harness the power of
tor, is now becoming widely used in the earth sciences
Earth data and science to rapidly prototype and benchmark
through applications such as the Australian Geosciences
innovative EO-based information services. At the beginData Cube [54]. The idea is to apply AI tools to discover the
ning of the pipeline, the accelerators will be continuously
structure of the data cube, find patterns of changes, and
fed through a series of open competitions for entrepreneurquantify extremes. Such an approach is ver y powerful,
ial ideas that can turn the grand challenges of our society,
as it provides a consistent framework that enables a truly
emerging business needs, and advances in digital technolomultivariate analysis across communities. A deep learngies into new growth opportunities for Europe. At the end of
ing approach also brings the challenges of interpretation
the pipeline, the entrepreneurs will display their new service
of correlations in terms of physical underpinning mechaon a digital marketplace for data and services.
nisms, as correlations do not imply causation (one of the
The service industry is currently developing a strategy
major pitfalls in big data mining).
to build a digital marketplace for these data and services.
ESA is now exploring the development of a network of
The idea is to have a platform (Figure 5) where data and
distributed data cubes dedicated to specific sensors (e.g.,
software users and providers can easily discover, comLandsat, Sentinels), themes (e.g., climate) or geographipare, and buy data (open and commercial), products,
cal coverage (e.g., regions) or specific communities (e.g., Futools, and processing services online, thereby doing for
ture Earth), together with the AI tec hnique s necessar y
data and services what Amazon is doing for books. In the
to iden tif y transient patterns of change and correladigital marketplace, the broker is connected to a type of
tions [30]-[32]. One of the goals is to develop the new generapp store, which transforms the data
ation of value c hains where human
into an information service that can
computations and AI are combined
be delivered to the user or made availto work in cooperation. The effective
able to the community. Such a platcombination of AI and human intelform would also enable direct links
ligence could unleash new possibilbetween scientists, innovators, and
ities to explore changes in very large
entrepreneurs, which in the end will
data sets, with multiple applications
most likely be the driving force to lefor climate, health, forecast and Earth
verage the geomarket's potential and
system science.
EO data utilization. This approach has
One example of a cube is the Earth
multiple benefits for all actors along
System Data Cube (Figure 4), develthe value range, including additional
oped by ESA to combine a wealth of
Space
revenues from unused resources, cost
obser vations on land-atmosphere
saving by pay-per-use models, no owninteractions ranging from soil moisership costs for users, and increased
ture, vegetation structure, vegetation FIGURE 4. A schematic of a multivariate
ownership of the data and data value
activity, and physical land surface Earth System Data Cube. The surfaces of
(e.g., the so-called IKEA effect). Such
properties up to atmospheric condi- extreme events identified by the machine
an approach would not only benefit
tions. The idea of the cube is to pro- learning algorithm are represented in red.
from open science but also by providvide a new framework for addressing a The blue arrows highlight their relationship
ing stringent tests on products feeding
variety of questions framed by Future with other parameters. (Image courtesy of
back into the science.
Earth about land-atmosphere feed- ESA/ppm.)
90

ieee Geoscience and remote sensing magazine

june 2017

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2017