New York City Mayor's Office of Sustainability
At a Glance
Industry
Government/Public Administration
Project Types
Data Analysis, Engagement and Behavior Change, Sustainability and Energy Management Strategy
Year
2020
Location
New York, NY
Summary
The summer fellowship was dedicated to building data pipelines and data science models to aid energy efficiency, environmental justice, and decarbonization policy formation.
Goals
New York City has laid out broad ambitions to reach zero net emissions and buildings represent 70% of New York’s emissions; the vast majority of those are under 25k square feet. Policy is in development to decarbonize the existing small building stock. Several technical roadblocks were impinging on the Sustainability Office’s progress on several policy fronts, and the fellowship was dedicated to solving them.
Solutions
The solution was to use broadly available data science tools -- Python notebooks and Tableau Dashboards -- to bring relevant data, both raw and analyzed, to policymakers. The first project took census data from the American Community Survey (ACS) and connected it to city data related to buildings. These ~15m rows of detailed city demographic information were then placed into a freely accessible Tableau so that all stakeholders can breakdown boroughs, neighborhoods, or even census tracks across whatever variables are pertinent to their particular policy. The second project first found several data sources related to buildings in New York -- assessments, 2D footprint maps, 3D renderings -- and created a congealed dataset of ~1M buildings in New York. This dataset was run through a clustering algorithm, to find associations across buildings, assisting any broad retrofit policy.
Impact
The net impact is the broad democratization of complex data to the entire mayor’s office. The Tableau from the ACS Project has some 4000 variables across 200 dimensions that can be compared and contrasted in myriad ways. The clustering project, while still incomplete, provides base code for incorporating further data points, using different clustering algos, and/or using different mathematical principles.