344
actions are not isotropic and vary across directions. Third, spatial dependency exists
in multiple spatial scales. Finally, spatial big data exhibits heterogeneity, that is,
identical feature values may correspond to distinct class labels in different regions.
Thus, learned predictive models may perform poorly in many local regions.
Depending on the scale, these effects may get exasperated. Moreover, scale may
affect the computational performances of algorithms that were used for FEW
research. Therefore, researchers are encouraged to find a correct balance between
performance and accuracy, depending on needs.
Key Points
• FEWS questions tend to be inherently multiscalar, but there are key scales at
which data, models, metrics, and processes are more readily available.
• Pasteur’s quadrant for use-inspired research is aspirational for FEWS practitioners and scholars.
• Scales can be important to determine in framing, researching, and implementing
key FEW challenges and solutions.
• Spatial scales include establishment, process, micro, meso, and macro.
• Temporal scales include interval, daily, monthly/seasonal, annual, and decadal,
although scales of thousands to millions of years to arise in climate modeling.
• Scale takes on a unique meaning for metrics, data, models, and computing.
• Metrics, data, models, and computing each answer important questions for
FEWS work.
Discussion Points and Exercises
1. What are the characteristics of a well-defined question or study in a FEWS
context?
2. Identify a FEWS question that you are very interested in.
(a) What is the best scale of space and time at which to study that question,
considering what you know about the physical and human processes involved
in this part of the system?
(b) What is the best scale of space and time at which to study that question,
based on what you know about the availability of data and tools?
(c) Do the answers to (a) and (b) above match, and if not what can you do to
close the gap?
3. If you are doing research work on FEWS, identify which part of Pasteur’s quadrant that work most nearly belongs to, and explain why. What could you do to
move your work into Pasteur’s quadrant?
4. Do you agree with our characterization and nomenclature for the primary scales
in a FEW system? Is something missing?
5. What questions are not being asked about FEWS right now?
6. Consider and discuss whether you think basic or applied research is more important and/or more attractive to you, personally—and why. Does your discipline,
occupation, or program provide a different answer to this question from your
personal answer?
M. Carbajales-Dale et al.
actions are not isotropic and vary across directions. Third, spatial dependency exists
in multiple spatial scales. Finally, spatial big data exhibits heterogeneity, that is,
identical feature values may correspond to distinct class labels in different regions.
Thus, learned predictive models may perform poorly in many local regions.
Depending on the scale, these effects may get exasperated. Moreover, scale may
affect the computational performances of algorithms that were used for FEW
research. Therefore, researchers are encouraged to find a correct balance between
performance and accuracy, depending on needs.
Key Points
• FEWS questions tend to be inherently multiscalar, but there are key scales at
which data, models, metrics, and processes are more readily available.
• Pasteur’s quadrant for use-inspired research is aspirational for FEWS practitioners and scholars.
• Scales can be important to determine in framing, researching, and implementing
key FEW challenges and solutions.
• Spatial scales include establishment, process, micro, meso, and macro.
• Temporal scales include interval, daily, monthly/seasonal, annual, and decadal,
although scales of thousands to millions of years to arise in climate modeling.
• Scale takes on a unique meaning for metrics, data, models, and computing.
• Metrics, data, models, and computing each answer important questions for
FEWS work.
Discussion Points and Exercises
1. What are the characteristics of a well-defined question or study in a FEWS
context?
2. Identify a FEWS question that you are very interested in.
(a) What is the best scale of space and time at which to study that question,
considering what you know about the physical and human processes involved
in this part of the system?
(b) What is the best scale of space and time at which to study that question,
based on what you know about the availability of data and tools?
(c) Do the answers to (a) and (b) above match, and if not what can you do to
close the gap?
3. If you are doing research work on FEWS, identify which part of Pasteur’s quadrant that work most nearly belongs to, and explain why. What could you do to
move your work into Pasteur’s quadrant?
4. Do you agree with our characterization and nomenclature for the primary scales
in a FEW system? Is something missing?
5. What questions are not being asked about FEWS right now?
6. Consider and discuss whether you think basic or applied research is more important and/or more attractive to you, personally—and why. Does your discipline,
occupation, or program provide a different answer to this question from your
personal answer?
M. Carbajales-Dale et al.