July 27, 2021

Measuring visits to parks and open spaces with mobility intelligence

Measuring visits to parks and open spaces with mobility intelligence
This blog post dives deep into how local government can leverage high-quality mobility data and machine learning for inferring visits to parks and open spaces and the dispersion movement patterns
Sign up for our webinars to learn from our data scientists and geospatial experts. They will show you how mobility intelligence can be used in the most cost-effective way to measure and analyze movement patterns.

CITYDATA.ai specializes in crowdsourcing anonymized mobility data and aggregating the privacy-compliant datasets to infer the footfall visits, the average dwell time, the dispersion or movement patterns, and the changes over time both at the micro and macro level. In doing so, CITYDATA.ai builds mobility replicas or mobility digital twins that represent the chronological historical patterns for people-presence, density, and movement for POI places and geofenced polygons.

Parks & Open Spaces

Urban-adjacent parks offer endless opportunities for recreational activities in proximity of cities and towns. With the COVID-19 variants still prevalent, people eager to vacation safely are turning to urban-adjacent parks to enjoy large outdoor areas that would allow for physical distancing. Per the CDC, being physically active is one of the best ways to keep the mind and body healthy. In many areas, people can visit parks, trails, and open spaces as a way to relieve stress, get some fresh air and vitamin D, stay active, and safely connect with members of their community. While the number of people visiting parks dropped significantly during the first wave of the pandemic; the visits started to increase as the country begins to recover from COVID-19.

Problem Statement

Most cities need to quantify their population and movement trends to make informed data-driven decisions for planning and budgeting. Satellite feeds are sporadic and cannot be relied on for continuous data. Drones and cameras pose surveillance-related privacy problems. Sensors can be expensive to deploy and onerous to maintain.

Google had previously published their Community Mobility Index to measure the high-level aggregated activity at parks and open spaces across the United States. While such reports were useful to understand the overall trend within a county, the data was not available for individual parks. The high-level data has since been discontinued.

Mobility intelligence companies like CITYDATA.ai are publishing their open datasets and footfall visit insights for each neighborhood park, urban-adjacent park, regional park, and national park in the country. The crowdsourced data samples are statistically relevant and invaluable for the Recreation and Parks departments. Such studies are also essential for public health officials, economic developers, urban planners, and local administrators for their policy decisioning.

Method and System

CITYDATA.ai measured the density, activity, and movement of people at urban-adjacent parks, open spaces, camp grounds and hiking trails that are adjacent to urban areas to study and understand the changes in footfall visit patterns and activity levels. The typical data processing and machine learning pipeline consists of the steps outlined below:

  • Collect crowdsourced mobility data for the entire city or county through a collaborative network of mobile app publishers, open WiFi hot spots, and Bluetooth beacon sensors
  • Check to ensure that 100% of the mobility identifiers in the dataset are hashed and anonymized for compliance with regional and national privacy regulation
  • Filter the dataset to only retain data rows with a horizontal accuracy of at least 20 meters or better
  • Filter the dataset to only retain latitude and longitude data points with at least 4 decimals representing a precision of 10 meters or better
  • Calculate the geohash grid cell identifier and the h3 grid cell identifier for each mobility data point
  • Cleanse, de-dupe, normalize and structure the filtered data to produce a high-quality and high-resolution mobility dataset for machine learning
  • Partner with the city's GIS team to collect the geofences or polygon shapes for relevant parks, trails, camp grounds, open spaces, downtown zones, business improvement districts, equity zones, transit zones, and study areas
  • Ingest the above geojson shapes into CITYDATA's cloud to create FENCES APIs for searching and querying the shapes
  • Optionally, partner with the city to import ground truth data stream from privacy-compliant sensors and cameras, where available
  • Ingest the ground truth data into CITYDATA's cloud to create TRUTH APIs for searching and querying the data
  • Optionally, partner with the city to also collect and import the daily 311 data set, daily crime reports data set, weekly building permits dataset, hourly air pollution, temperature, precipitation, visibility, and other relevant datasets, and create APIs for searching and querying such data
  • Combine the normalized mobility dataset, places POI data, local events and activities, census demographic data, the geofences, and optional datasets like ground truth data, 311 data, crime incidents, permits data, and weather data into CITYDATA's proprietary knowledge graph
  • Apply graph ML to spatiotemporally cluster the mobility data points to infer the significant patterns for each geofenced park, trail, zones, districts, and other places of interest
  • Apply scale-up models and filter outcomes based on 70% confidence threshold
  • For the specified date and time range for each geofence, infer:
  • Incorporate ground truth data if available, to improve the scale-up models for geofence visit inferences
  • Correlate geofence visits with 311 data, crime incidents, and building permits within the geofence and in the vicinity of the geofence
  • Index and publish results through the VISITS APIs and the MOVEMENT APIs in CITYDATA cloud
  • Upload the results data files to cloud storage through Amazon AWS s3 and to BigQuery in Google Cloud Platform
  • Pipe the results to the CITYDASH visualization platform and to third-party tools like Tableau, Carto, Data Studio, and Kepler
  • Integrate the results for each geofence into the overall mobility digital twin knowledge graph for the city or county in CITYDATA cloud
  • Share results with the local government stakeholders and decision makers

Important Definitions

The link below provides a comprehensive list of defined terms related to mobility data and geospatial intelligence: https://univercity.ai/mobility-intelligence-glossary/

GovTech Experts

The team of GovTech experts at CITYDATA.ai are passionate about geospatial intelligence, mobility patterns, machine learning and data science for civic use cases. We sincerely believe that our technology platform and our domain expertise can make our cities smarter, safer, more equitable and live-able. We invite you to partner with us in our quest to build mobility replicas for improving parks and open spaces around the world.


Sign up for our webinars to learn from our data scientists and geospatial experts. They will show you how geospatial intelligence can be used in the most cost-effective way to measure, analyze and advocate local parks and open spaces.