Open source web-application to predict air quality on a city-block basis in Oakland and San Leandro, CA using machine learning models, trained on publicly available datasets on local pollutant concentrations, local meteorological data, emissions from local industrial sources, and traffic emissions.
* Web-application
* Source code
* Blog post
Image classification of land use types using the DeepSat-6 satellite image dataset and convolution neural networks.
* Source code
Open source spatial inventory of air pollutant sequestration data by grassland and shrubland vegetation in the US. The dataset was developed using the daily leaf area index of vegetation classes, the National Land Cover dataset, and the i-Tree Eco model. This open source data is currently used by
Contributed to a white paper titled Why all cities should have ‘Clean Air as a City Service’ that makes the case for cities to make air quality data readily available and to provide clean air as a service. The paper was written as a part of the Urban Futures Workshop 2020 hosted by Microsoft.