439
16.5 Challenges and Opportunities for Computer Science
The FEW Nexus poses tough challenges as well as exciting opportunities for transformative research in computer science. The topics discussed in this section are
inspired by the contributions of participants at the INFEWS Data Science Workshop
held in 2015.
Although spatial computing scientists expect to tackle issues on many fronts
in the coming years, this section will concentrate on challenges relevant to the
FEW Nexus. We begin by discussing the challenges of coupling and fusing multiple
model types (Webber 2015) for the FEW Nexus, followed by a review of key issues
that arise from spatiotemporal modeling. Next is a discussion about how methodologies for guiding stakeholders to reach consensus on FEW issues. A proposal is
made to develop a life cycle thinking methodology for understanding the FEW
Nexus. The final section presents novel computing research questions that address
issues of data uncertainty, incompleteness, and bias in spatiotemporal Nexus data.
16.5.1 Coupling and Fusing Multiple Models
As described above, causal and mechanistic models often start with model building
and theories. Moreover, such models are critical to understand and evaluate the
potential of implementing FEW nexus approaches. For example, the climate model
that coupled ocean and atmosphere models created a breakthrough in El Nino prediction. Similarly, models are beneficial for understanding the role of “forces” such
as climate change, socioeconomic and technological change, and policy interventions. These examples show that models serve to understand and quantify the risks.
This subsection presents the key issues, science questions, and challenges that arise
when modeling for the FEW Nexus. Note that in Chap. 15, a variety of models are
discussed through a sector-centric categorization. In this chapter, we used another
taxonomy of models where these are categorized as causal and mechanistic models
(e.g., process models) and data-driven models.
(a) Mechanistic and causal models versus data-driven models.
Traditionally, FEW models have been based on mechanistic process models
(e.g., see Sect. 1.4). For example, a model of passive hyperspectral sensing
might consider incident solar radiation, re-radiation at various wavelengths,
attenuation and refraction by the atmosphere, and noise in a satellite-borne
CCD (“charge-coupled device”). A model of hydrology might encompass precipitation, solar radiation, and wind inputs, canopy capture, through-fall, surface runoff, soil infiltration, evapotranspiration (Elshorbagy et al. 2010;
Solomatine 2005), and groundwater storage and release into streams (Fatichi
et al. 2016). Each of the individual processes involved could, in turn, be based
on appropriate mechanisms (plant physiology, percolation rates, etc.).
16 Computing
16.5 Challenges and Opportunities for Computer Science
The FEW Nexus poses tough challenges as well as exciting opportunities for transformative research in computer science. The topics discussed in this section are
inspired by the contributions of participants at the INFEWS Data Science Workshop
held in 2015.
Although spatial computing scientists expect to tackle issues on many fronts
in the coming years, this section will concentrate on challenges relevant to the
FEW Nexus. We begin by discussing the challenges of coupling and fusing multiple
model types (Webber 2015) for the FEW Nexus, followed by a review of key issues
that arise from spatiotemporal modeling. Next is a discussion about how methodologies for guiding stakeholders to reach consensus on FEW issues. A proposal is
made to develop a life cycle thinking methodology for understanding the FEW
Nexus. The final section presents novel computing research questions that address
issues of data uncertainty, incompleteness, and bias in spatiotemporal Nexus data.
16.5.1 Coupling and Fusing Multiple Models
As described above, causal and mechanistic models often start with model building
and theories. Moreover, such models are critical to understand and evaluate the
potential of implementing FEW nexus approaches. For example, the climate model
that coupled ocean and atmosphere models created a breakthrough in El Nino prediction. Similarly, models are beneficial for understanding the role of “forces” such
as climate change, socioeconomic and technological change, and policy interventions. These examples show that models serve to understand and quantify the risks.
This subsection presents the key issues, science questions, and challenges that arise
when modeling for the FEW Nexus. Note that in Chap. 15, a variety of models are
discussed through a sector-centric categorization. In this chapter, we used another
taxonomy of models where these are categorized as causal and mechanistic models
(e.g., process models) and data-driven models.
(a) Mechanistic and causal models versus data-driven models.
Traditionally, FEW models have been based on mechanistic process models
(e.g., see Sect. 1.4). For example, a model of passive hyperspectral sensing
might consider incident solar radiation, re-radiation at various wavelengths,
attenuation and refraction by the atmosphere, and noise in a satellite-borne
CCD (“charge-coupled device”). A model of hydrology might encompass precipitation, solar radiation, and wind inputs, canopy capture, through-fall, surface runoff, soil infiltration, evapotranspiration (Elshorbagy et al. 2010;
Solomatine 2005), and groundwater storage and release into streams (Fatichi
et al. 2016). Each of the individual processes involved could, in turn, be based
on appropriate mechanisms (plant physiology, percolation rates, etc.).
16 Computing