EIE 2020 mobility data now available to cities as they continue sustainability efforts
By Christopher Bian, Sr. Software Engineer, Environmental Insights Explorer
In 2018, Environmental Insights Explorer (EIE) set out to achieve what had not been done before: a globally consistent, freely available dataset to measure buildings and transportation emissions as well as solar offset potential. Over the last two years, we’ve learned a lot from our users and we are continuously working to refine EIE; this includes investments in expanding outreach, improving technical infrastructure, and continuous data analysis/modeling — the work that we consider to be never-ending. Following the release of 2019 transport data last year, our team is excited to share mobility data for 2020. Through this EIE release, cities and local governments will now have the ability to perform year-over-year analysis for three consecutive years dating back to 2018 to track activity changes over time for their climate action planning efforts.
A quick refresher on how EIE calculates transportation emissions: we are able to characterize and aggregate trips taken within the city boundaries and the trips that crossed the city boundaries, using proprietary data from Google Maps. Similar to Agent Based Activity (ABA) models used by transportation planning departments, we classify trips based on origins and destinations to determine travel direction (inbound, outbound and in-boundary) data for each administrative boundary. We compute annual aggregations by city, mode, and travel direction by applying scaling factors for general population and average vehicle occupancies. We then apply emission factors that are available in the Climate Action for Urban Sustainability tool (CURB) that contain fleet mix and fuel type distributions by country/region. Finally, we sum overall carbon emissions by mode as metric tons of CO₂ equivalents.
What’s new in the 2020 release?
A unique and challenging year
The occurrence of a global pandemic impacted travel behavior deeply, in ways that the world is just beginning to fully understand. EIE captured many of these dynamic behaviors, such as overall decreased travel in the early phase of the pandemic, then the subsequent rise during popular travel and holiday time periods. Depending on the region, some mode shift behaviors became evident: increased/decreased bicycling in cities such as New York City and Copenhagen; decrease in public transit usage in urban areas like Kyoto and San Francisco.
Updated population for 2020
A city’s population is a dynamic figure that changes over time. In the 2020 data release, a 2020-based set of population figures are used, which can result in emissions differences depending on whether a city has grown or contracted. These population data may vary somewhat from other sources, depending on boundaries and other factors. The population dataset is based on the work of WorldPop, which develops open, high-resolution spatial datasets based on peer-reviewed research and methods. Learn more about WorldPop here.
Improved year-on-year user experience
As EIE continues to publish data each year as part of Google’s third decade of climate action, we understand the need for better tools to measure and track sources of emissions over time. In this release, we have made improvements to the menu bar that allow for easier comparison and benchmarking from year-to-year.
- Forward/backward arrows toggle between years more easily, without needing to access the dropdown menu every time.
- A new multi-year bar chart enables side-by-side comparison of annual aggregate emissions.
Data availability and privacy
EIE transportation insights leverages inputs from Google Maps that include location information from user trips. We protect the privacy of our users by applying best-in-class techniques to aggregate and anonymize the activity data. You might notice that EIE focuses on locations with significant populations — we do this to ensure data quality and privacy. You also may find that certain subsets of data are not included due transportation attributes unique to your locale. This is intentional, and does not signify a lower quality dataset. Reasons that could contribute to this are:
- Trips of a specific mode into, out of, or staying within the administrative boundaries are too low a volume to be represented in EIE according to privacy protections.
- Trips of a specific mode into, out of, or staying within the administrative boundaries do not exist (such as light rail or bus stops outside of the city limits).
- Your locale encompasses a small area, or a population size that does not meet minimum levels for privacy protections.
- Data quality thresholds determined by EIE contain higher margins of error than are deemed to be useful, and so are excluded.
Frequently Asked Questions
For more information, please visit the EIE Methodology page here.
METHODOLOGY & DATA CONSIDERATIONS
How does EIE define inbound, outbound and in-boundary trips?
EIE transport data tracks trips based on origin and destination; it does not account for passthrough traffic to comply with Scope 1 and 3 emissions methodologies. For in-boundary travel, 100% of trips that start and end in the same city are accounted for. For inbound and outbound travel, EIE includes 100% of the trip between the origin and destination cities to provide the full picture of travel behavior. In order to be compliant with GPC’s induction (origin-destination) method, cities can simply take 50% of transboundary trips + 100% of in-boundary trips.
Which emissions factors does EIE use to estimate transportation emissions and why?
EIE uses default emissions factors pulled from the CURB tool, which accounts for vehicle fleet mix, fuel efficiencies and fuel carbon intensities by country. These contain mixes for passenger vehicles, and light and medium duty trucks. Some limitations include lack of plug-in electric (EVs) and heavy duty commercial freight emissions factors. These were chosen due to their applicability across the globe, as well as the sources/organizations that created the factors. In addition, users can change the default emissions factors based on local knowledge for more accurate emissions.
How does EIE determine the vehicle occupancy factor?
Vehicle occupancy is a highly challenging metric to capture consistently across the world due to a dizzying array of vehicle capacities, and transportation behaviors that determine average occupancy by mode. As such, EIE uses assumptions based on the US Dept. of Transportation averages for certain modes such as 1.7 persons per passenger vehicle and 9.7 persons per bus. For very high occupancy public transit modes such as trains and subways, AVO is not applied, and their data is noted as representing total passenger distance not vehicle distance. We are actively exploring ways to improve emissions data for these modes.
Why does EIE data differ from my city’s GHG inventory or local transit authority?
Different data sources and methodologies often result in different outcomes, despite attempting to measure the same behaviors. EIE is a bottom-up model based on measured trips (per GPC) that are scaled to the full population, before applying emissions factors. The resulting vehicle distance traveled or emissions can differ from your methodology. We recommend adjusting for the following:
- Apply the 50% GPC induction method of in/outbound trips.
- Consult with your local transit authority to determine if there is a reliable average vehicle occupancy (AVO) for each mode, particularly passenger vehicles, and buses. Adjust the vehicle distance traveled using the formula:
EIE VKT for mode * EIE AVO for mode / local AVO - Methodological differences between EIE and other methods such as fuel sales estimates, 4-step traffic demand models, etc. Some of these models may also include pass-through trips (such as those that include inputs from pneumatic tube vehicle counters), while EIE does not.
Are EVs included in the automobile mode?
EVs or plug-in electric vehicles are included in the distance traveled. However, their tailpipe emissions are not accounted for (Scope 2, GPC), as they come from upstream stationary energy generation (typically associated with building energy).
Are commercial or heavy duty freight travel included in EIE emissions?
EIE leverages data from location services, which commercial/freight operators typically rely on. Travel from these types of trips will be included in EIE but they are not broken out into their own category. In addition, the emissions factors from CURB do not contain heavy duty/freight fleet and fuel mixes so emissions from this sector will be underestimated. Whether this has a substantial impact on emissions will depend on the local freight characteristics of your region or jurisdiction. Users may find that adjustment of emissions factors to weight more heavily towards freight fuel types is appropriate.
Why doesn’t EIE show emissions from trains, trams or ferries?
Currently, these public transit modes in EIE do not have emissions calculations applied, because there is no globally available, consistent, and reliable source for vehicle occupancy data at the local level. As such, emissions from these modes will be overestimated and therefore emissions are not given. However, the trip and distances traveled are still provided by EIE so that local knowledge of your municipality can be applied to estimate emissions from these modes.
Why am I missing data for certain years or modes?
Depending on factors unique to your region’s surrounding topography, and its transportation attributes (or usage of transportation modes), you may find that certain subsets of data are not included. For example, you may find that public transit trips (trains, buses, etc.) into and out of your region are not included. This could be a result of low ridership volume of these types of trips because your transit system is largely confined to your administrative boundary, or that there has been a large reduction in ridership or operating schedule for a given mode. This is particularly likely in 2020 during Covid response for public transit. The population in your boundary may also play a role as EIE only exposes data for a minimum population threshold where privacy protections can be applied.
What should I do if I am missing data for certain years or trips?
We recommend that you consult with a local expert or authority for your administrative area to determine whether the combination of the mode of travel and direction exists and would be expected to comprise a significant (>1%) of overall trips. These criteria may not be met, resulting in EIE not publishing this data. If based on your local knowledge, EIE should indeed be capturing this information, please let us know so that we can do a deeper investigation and improve our models where necessary.
DATA QUALITY
How does EIE anonymize its transportation data?
At Google, protecting the privacy of user data is of utmost importance. We apply a variety of techniques to prevent reverse-engineering of individual users from the transportation dataset, including:
- Aggregating data to appropriate spatial and temporal resolutions
- Laplace noise
- K-anonymity
- Limiting user contributions in each cell
These collective techniques add significant privacy protections while retaining the value and insights in the underlying data.
How does the QA process work and what does EIE look for?
EIE data undergoes a rigorous and constantly evolving data quality process that involves manual and automated checks. At the high level, we look to see that EIE broadly agrees with externally available sources of general travel behavior, such as gasoline sales and airport TSA checkpoint volumes. We also use a variety of heuristics and statistical techniques to detect if there are any significant anomalies in the data, including symmetry between inbound and outbound travel directions, and analyses based on standard deviation and outliers.
How does EIE data compare to ground truth data (like from utility companies or road sensors) or other inventories?
To date, we have compared EIE data with ground truth data for three cities: Mountain View, CA and Boulder, CO in the US, and Dublin, Ireland at multiple points on the M50 toll road in Ireland. While we will continue to validate EIE data against ground truth, these studies have shown that EIE has a range of 0.91 to 0.99 Pearson’s correlation with ground truth sources. Learn more here.
We want to hear from you!
Once you have the opportunity to access and review your transportation data for your administrative area(s), you may have feedback. If you have further questions or, please see additional information in our Methodology section. If you’d like to inquire about the data being presented, you can contact our team using the feedback links in the footer of the EIE website.
