473
querying techniques. The physical coverage of a spatial range of interest may not be
uniform due to physical constraints (inaccessibility of terrain) or other operational
constraints. Worse, sensing may not provide sufficient density/coverage throughout
the entire region of interest. In such scenarios, one is compelled to resort to techniques for estimating the values of the physical phenomena throughout the geospatial domains, relying on techniques such as Kriging. The fidelity of the source data
and models of data must match the decision-maker’s need for information: imprecise matches introduce uncertainty.
The relevant data for FEWS science are heterogeneous: data obtained from different sensing devices vary in type and format, and in ontology (Chap. 14). For
instance, the water-resources data of the Food and Agriculture Organization
(AQUASTAT) summarized by continents and the major region is an Excel file; food
data (FASTAT) is maintained in a different software package. Similarly, the USGSproduced National Land Cover Data (NLCD) uses a 21-class classification scheme,
whereas the USDA-generated Crop Data Layer (CDL) has more than 100 classes.
Not only do the data come from multiple sources at different scales (see Chap. 12),
but the data are also subject to additional (implicit) uncertainty due to the different
levels of granularity of its storage in (spatial and/or spatiotemporal) data warehouses. Heterogeneous data not collected by a single agent or for a single purpose
cannot be combined without introducing various uncertainties. Finding a significant
variable that allows linkages between various models can be a critical challenge.
Thus, uncertainty and risk are part of all FEWS Nexus decisions: as such, it must
be explicitly addressed when making decisions or providing decision support. It is
important to know how much uncertainty is too much to make the decision and to
know whether the metrics, data, models, computing, and tools available can achieve
a tolerable uncertainty. It may be better to postpone decision making in the presence
of too much uncertainty, rather than making a serious mistake based on inadequate
information. However, paralysis by analysis can also be its own form of risk: waiting to make a decision until everything is absolutely certain can also lead to problematic outcomes.
17.5 Communities of Science and Practice
Bringing about the tools described above requires an interdisciplinary community
of scientists with the diverse skills needed to develop the Big Data, tools, and digital
platforms. These tools require knowledge sharing and communication with nonscientific stakeholders and decision-makers to put the knowledge into practice.
Community of Science and Practice is a common concept in social sciences,
private and public organizations, governments, education, and international development. Multi-stakeholder participation is an essential element for addressing
sustainability challenges and necessary for developing solutions palatable to civil
society (Kates et al. 2001).
17 Applying Science to Practice
querying techniques. The physical coverage of a spatial range of interest may not be
uniform due to physical constraints (inaccessibility of terrain) or other operational
constraints. Worse, sensing may not provide sufficient density/coverage throughout
the entire region of interest. In such scenarios, one is compelled to resort to techniques for estimating the values of the physical phenomena throughout the geospatial domains, relying on techniques such as Kriging. The fidelity of the source data
and models of data must match the decision-maker’s need for information: imprecise matches introduce uncertainty.
The relevant data for FEWS science are heterogeneous: data obtained from different sensing devices vary in type and format, and in ontology (Chap. 14). For
instance, the water-resources data of the Food and Agriculture Organization
(AQUASTAT) summarized by continents and the major region is an Excel file; food
data (FASTAT) is maintained in a different software package. Similarly, the USGSproduced National Land Cover Data (NLCD) uses a 21-class classification scheme,
whereas the USDA-generated Crop Data Layer (CDL) has more than 100 classes.
Not only do the data come from multiple sources at different scales (see Chap. 12),
but the data are also subject to additional (implicit) uncertainty due to the different
levels of granularity of its storage in (spatial and/or spatiotemporal) data warehouses. Heterogeneous data not collected by a single agent or for a single purpose
cannot be combined without introducing various uncertainties. Finding a significant
variable that allows linkages between various models can be a critical challenge.
Thus, uncertainty and risk are part of all FEWS Nexus decisions: as such, it must
be explicitly addressed when making decisions or providing decision support. It is
important to know how much uncertainty is too much to make the decision and to
know whether the metrics, data, models, computing, and tools available can achieve
a tolerable uncertainty. It may be better to postpone decision making in the presence
of too much uncertainty, rather than making a serious mistake based on inadequate
information. However, paralysis by analysis can also be its own form of risk: waiting to make a decision until everything is absolutely certain can also lead to problematic outcomes.
17.5 Communities of Science and Practice
Bringing about the tools described above requires an interdisciplinary community
of scientists with the diverse skills needed to develop the Big Data, tools, and digital
platforms. These tools require knowledge sharing and communication with nonscientific stakeholders and decision-makers to put the knowledge into practice.
Community of Science and Practice is a common concept in social sciences,
private and public organizations, governments, education, and international development. Multi-stakeholder participation is an essential element for addressing
sustainability challenges and necessary for developing solutions palatable to civil
society (Kates et al. 2001).
17 Applying Science to Practice