DevMachTech/Daily-air-quality-estimates-for-urban-centers-in-Africa

Implementation of an air quality model adapted from a Zindi winning solution # Daily-air-quality-estimates-for-urban-centers-in-Africa Implementation of an air quality model based on satellite data adapted from a Zindi winning solution In this project, Jonathan Whitaker and Yasin Ayami present a method for predicting historical air quality (as measured by daily median PM2.5 concentration) for locations where no ground-based sensors are present, by using weather data and remote sensing data from sources like the Sentinel-5P satellite. Air quality data is obtained for 555 cities and supplemented by satellite and weather data. This is then used to build a model to predict the air quality for a given date and location. A competition hosted by Zindi was used to crowd-source the creation of the model used, with the winning code forming the basis of our modelling approach. We used the trained model to create a new dataset of historical air quality predictions for cities across Africa, available at ​ github.com You can view winner, Nikhil Mishra's, solution here. ## Implementing the model View the model implentation on South African Spatial Data Infrastructure (SASDI) Report: docs.google.com or as PDF ('AirQ Report.pdf') Dashboard to view results: datasciencecastnet.com ### Accessing Predictions The predictions can be found in the Data/results folder. - za_cities_predictions contains the predictions for cities across South Africa - pcs_predictions contains predictions for some additional population centers in South Africa - af_cities_predicitons contains the predictions for major cities across Africa ### Replicating the project: - Download the historical air quality measurements from aqicn.org and store the files in Data/waqi_downloads - Run '1. WAQI Data Prep.ipynb' to merge the relevant data into a single csv file, which is stored in Data/intermediate/citypm25.csv - R …