Bringing the Geospatial Perspective to Data Science Degrees and Curricula
Summer 2018
Virtually every sector of industry, business, government, and science is awash in data of great volume, variety, and velocity. In light of calls for fairness, accountability, transparency, and reproducibility, data accuracy and authority are also highly relevant. As an interdisciplinary field, there are high expectations for the capabilities of data science1 to address myriad demands for innovative breakthroughs. “Data Scientist” has become an in-demand job title, though the nature of the positions varies widely. The most common skill sets required are analytical and quantitative in nature: to be able to manage and help others interpret large and diverse data sets.
xxxxxxxxxxxxxxxxxxxxxxxxxxx
A 2011 McKinsey Global Institute report estimates a value of “about $600 billion annually by 2020” from leveraging personal location data to reduce fuel waste, improve health outcomes, and better match products to consumer needs. Spatial data are critical for societal priorities such as national security, public health & safety, food, energy, water, smart cities, transportation, climate, weather, and the environment. For example, remotely-sensed satellite imagery is used to monitor not only weather and climate but also global crops3 for early warnings and planning to avoid food shortages.
xxxxxxxxxxxxxxxxxxxxxxxxx
The fundamental geographic qualities of spatial autocorrelation, which assumes properties of geographically proximate places to be similar, and geographic heterogeneity, where no two places on Earth are exactly alike, violate assumptions of sample independence and randomness that underlie many conventional statistical methods. Other spatial challenges include how to choose between a plurality of projections and coordinate systems and how to deal with the imprecision, inaccuracy, and uncertainty of location measurements. To deal with such challenges, practitioners in many fields including agriculture, weather forecast, mining, and environmental science incorporate geospatial data science4 methods such as spatially-explicit models, spatial statistics5, geo-statistics, geographic data mining6, spatial databases7, etc.
Xxxxxxxxxxxxxxxxxxxxxxxx
..... robust computational and statistical skills are necessary to achieve the types of analytical results that decision makers have been promised from data science, and have come to expect. At the same time, only spatial skills allow for the holistic analysis of the local and neighboring environmental context (e.g., climate, infrastructure, hazards), and social data (e.g., laws, demographics, culture) at multiple scales. Integrating this geographic contextual perspective significantly advances students’ skill sets to ). The bottom line: spatial is key within the “variety” dimension of big data and data science, and represents a vital variety of methods beyond the initial “one-size-fits-all” toolkit.
Spatial data, methods, and use cases may be leveraged to integrate ethically-based decision-making in an engaging manner via current societal debates such as those in election gerrymandering court cases or the negative impact of social feedback loops13 in crime hotspot analyses.