An emerging GHG estimation approach can help cities achieve their climate and sustainability goals

A credible assessment of a city’s greenhouse gas (GHG) mitigation policies requires a valid account of a city’s emissions. However, questions persist as to whether cities’ ‘self-reported inventories’ (SRIs) are accurate, precise, and consistent enough to track progress toward city mitigation goals. Although useful for broad policy initiatives, city SRIs provide annual snapshots that may have limited use to city managers looking to develop targeted mitigation policies that overlap with other issues like equity, air quality, and human health. An emerging approach from the research community that integrates ‘bottom-up’ hourly, street-level emission data products with ‘top-down’ GHG atmospheric observations have begun to yield production-based (scope 1) GHG estimates that can track changes in emissions at annual and sub-annual timeframes. The use of this integrated approach offers a much-needed assessment of SRIs: the atmospheric observations are tied to international standards and the bottom-up information incorporates multiple overlapping socio-economic data. The emissions are mapped at fine scales which helps link them to attribute information (e.g. fuel types) that can further facilitate mitigation actions. Here, we describe this approach and compare results to the SRI from the City of Indianapolis which shows a yearly difference of 35% in scope 1 emissions. In the City of Baltimore, we show that granular emission information can help address multiple issues, e.g. GHG emissions, air pollution, and inequity, at the sub-zip code scale where many roots and causes for each issue exist. Finally, we show that the incorporation of atmospheric concentrations within an integrated system provides rapid, near-real-time feedback on CO2 emissions anomalies that can uncover important behavioral and economic relationships. An integrated approach to GHG monitoring, reporting and verification can ensure uniformity, and provide accuracy to city-scale GHG emissions, scalable to states and the nation—ultimately helping cities meet stated ambitions.


Introduction
Many cities across the globe recognize their impact on climate change and have committed to longterm, ambitious greenhouse gas (GHG) emission mitigation targets. The fact that cities are pledging to reduce their GHG emissions suggests that selforganized, city-scale actions might compensate for the shortcomings of international climate treaties, regulation, and climate finance/carbon markets (Seto et al 2014, IPCC 2018. Typical near-term city mitigation targets range between 30% and 50% reductions by 2030 compared to emissions estimated in a city's chosen baseline year. More ambitious cities aim to achieve carbon neutrality or be net-zero emitters by 2050 (ARUP, C40 2014).
A comprehensive accounting, or inventory, of GHG emissions not only establishes a baseline from which to prioritize actions but also helps a city monitor progress if regularly updated. Although there are different ways to account for emissions, all approaches aim to yield estimates that are accurate, comparable, comprehensive, and complete (Ibrahim et al 2012, Ramaswami et al 2012. Several guidelines have been developed, each having a slightly different perspective on how a city should account for their emissions (Arioli et al 2020).
The original city-scale guidelines borrowed much from the IPCC framework for nations (IPCC 2006, ICLEI 2009). Using this framework, emissions are estimated by combining activity data with sectoral emission factors obtained from, for instance, the IPCC emission factor database (www.ipccnggip.iges.or.jp/EFDB/main.php), to provide emissions by economic sector. From this perspective, only scope 1 GHG emissions are inventoried (i.e. those emissions directly resulting from activities that take place within a city's jurisdiction). For example, electricity generation-related emissions are tied to those from powerplants that are physically located within a city as opposed to those from any powerplant (inside and outside a city) caused by the consumption of electricity by city residents (Pichler et al 2017).
Since then, work on corporate supply-chain (Chen et al 2017) and life-cycle analysis (Ramaswami et al 2008, Kennedy et al 2010, and trade (Lin et al 2015), provided additional perspectives on carbon accounting. These perspectives included inventorying any emissions associated with the consumption of electricity (scope 2) and emissions associated with the complete supply-chain of goods and services (scope 3). A scope 3 approach allows a city to allocate emissions from factories and commercial business worldwide to residents that consume products. A 'consumption-based' approach generally involves all three scopes' emissions, while a 'production-based' perspective considers only scope 1 emissions.
Numerous mixtures and hybrids of consumptionand production-based accounting have been developed and applied at the city-scale (Ramaswami and Chavez 2013, Chen et al 2016, Seto et al 2016, Lombardi et al 2017, Jones et al 2018. For example, many incorporate full transboundary activities to avoid truncation errors at physical boundaries, e.g. for aviation or marine sectors, or ways to avoid the double counting of emissions (Creutzig et al 2015). Most recently, some yield spatially disaggregated emission information , Gurney et al 2019a, Han et al 2020, Gurney et al 2020b or the use there of (Lin et al 2014).
While the academic literature has presented differing accounting perspectives, cities and non-government organizational (NGO) networks have taken up the practical task of building inventories to support target-setting and mitigation policy. Generally, a city itself accounts for its own emissions following one of several protocol guidance documents/tools e.g. ICLEI (2012). Once developed, cities can report their self-reported inventory (SRI) publicly, e.g. through the CDP (formally the Carbon Disclosure Project) (https://data.cdp.net/). In 2017, over 229 cities worldwide reported emissions on the CDP (https://data.cdp.net/widgets/kyi6-dk5h, accessed March 2021) including 45 of the 100 most populated cities in the US (Markolf et al 2020).
Although mitigation targets set by cities are impressive, questions have been raised about the accuracy and numerical integrity of reported emissions (Satterthwaite 2008, Deetjen et al 2018. This is due, in part, to the city-centric nature of the guideline approach. Urban practitioners make multiple subjective decisions based on political, demographic, and socioeconomic circumstances and goals (Kramers et al 2013). They collect and quality control data that may come from different years, national statistics, assumptions, and difficult to find sources if local data is not available (Bader and Bleischwitz 2009, Nangini et al 2019. Ideally, default emission factors should be refined (but are not always) to represent local processes (Shan et al 2019). Compiling an SRI takes time, resources, and expertise: maintaining consistent upto-date inventories is challenging (Nangini et al 2019, Markolf et al 2020. Many of these logistical and practical issues are likely the cause of discrepancies noted in  rather than fundamental flaws within the protocol guidelines themselves. It is unclear whether progress can be independently evaluated since there are no universal standards to chart progress (Markolf et al 2020). Indeed, the decentralized nature of climate action largely puts the onus on a city itself to independently evaluate progress. Protocol guidelines recommend that cities choose the verification that meets their needs and capacity (ICLEI 2012) with only a few having the budget and staff to do so (Blackhurst et al 2011, Markolf et al 2018. Uncertainties associated with SRIs can be >50%, larger than many reduction goals (Blackhurst et al 2011). Reporting provides a measure of transparency but choices made by practitioners may not be documented or publicly accessible (Markolf et al 2020). But, details of SRIs and thirdparty scrutiny are critical to developing/monitoring a mitigation strategy (Hoornweg et al 2011). The extent to which an SRI is accurate, comparable, comprehensive, and complete is largely city dependent. Comparing one city's SRI to another's is extraordinarily difficult.
Recent peer-reviewed literature raises questions about the accuracy of SRIs (specifically CO 2 ) due to large differences between reported city emissions and those published by academics. These differences cast doubt on whether cities are reducing their emissions as planned or reporting accurate totals. For example, Gurney et al (2019a) provides on-road 2010 emissions that are 10.7% larger than those reported by the local metropolitan planning agency in Los Angeles. Chen et al (2020) reported differences that range from −62% to +148% when comparing SRIs to a downscaled emission product for 12 cities across the globe. In an analysis of SRIs from 48 U.S. cities,  argues that cities under-report their scope 1 emissions by 18.3% on average with a range of −145.5% to 63.5% when compared to a finescale emission product (aka Vulcan). Note, the Vulcan emission product is consistent with atmospheric radiocarbon (14 • C) measurements at the continental scale (Basu et al 2020, Gurney et al 2020a.
These reported discrepancies suggest that SRIs are inconsistent with one another and may contain systematic biases or omissions. The comparisons also raise numerous questions such as: how well can we determine whether a city is moving toward meeting its targets? Can we assess whether activities at the city scale, when aggregated, have a national/global impact on emissions? Could new perspectives and methods help check the accuracy of reported emissions while providing an additional level of consistency?
To this end, the geoscience research community has developed approaches to quantify city and regional CO 2 and methane (CH 4 ) emissions (Hutyra et al 2014). These methods integrate 'topdown' atmospheric GHG observations with granular 'bottom-up' emissions data products using atmospheric inversion techniques (Tarantola 2004, Enting 2018. Atmospheric inversions have a rich history in carbon cycling science (Law 1999, Gurney et al 2002, Peters et al 2007, Ogle et al 2015. Their application over the last decade at urban scales have begun to yield estimates that can track changes in GHG emissions at annual and sub-annual timeframes (Lauvaux et al 2016, Sargent et al 2018, Turnbull et al 2019, Yadav et al 2019, Lauvaux et al 2020, Yadav et al 2021. Each component of the integrated approach, i.e. atmospheric observations, the inversion process, and the granular data products, brings unique value to the estimation of emissions. Atmospheric observations are key since they root the granular emission data to an atmospheric measurement of GHG tied to international standards (Tans et al 1990, Tsutsumi et al 2009. Thus, the atmospheric observations provide a measure of accuracy to the estimates missing from other carbon accounting techniques. Observations also contain the integral of all sources in a defined area including anthropogenic emissions and biological fluxes. The granular emissions, whose development has accelerated in the last decade, provide detailed information built from multiple data sources, including directly observed emission quantities (e.g. from continuous emission monitoring systems, aka CEMS), providing added specificity (at hourly/building-level scales) (Gurney et al 2012, 2019b. The inversion process provides the translation between atmospheric concentrations and surface emissions (Staufer et al 2016, Nickless et al 2018, Lauvaux et al 2020, Yadav et al 2021; it is in this translation where a fair amount of uncertainty arises at sub-city spatial scales (refer to the implication section).
Here, we present three case studies that brings together data from the published literature. This work focuses on scope 1 GHG emissions because they (a) can be linked with atmospheric observations (Nangini et (Dodman 2009 and thus, should be easier to estimate (and more well-known) than other types of emissions. Thus, any discrepancies in scope 1 estimates between reported emissions (by the city and other groups) suggest similar inconsistencies in emissions associated with other scopes that require more assumptions, even more data inconsistencies, etc.
The first case study compares the SRI (scope 1 emissions only) for Indianapolis, Indiana with wholecity emissions estimated with an integrated approach. We use Baltimore, Maryland in a second case study to show how spatially and temporally explicit emissions can point to specific places where the city can achieve co-benefits. Identifying underlying spatial processes not only helps city planners contextualize aggregated annual city-wide totals but provides support for targeted action. Finally, our third case study, focused on the City of Baltimore, demonstrates that relationships between citizen behavior, market forces, and stressors (like weather shocks), are best uncovered using emission information at sub-annual scales. Understanding relationships can explain year to year variability in annual emissions and allow cities to concentrate on levers to reduce emissions that are within their control. Such case studies have not been previously presented in the literature, largely because there are only a few applications of an integrated approach at the urban scale. A summary of SRIs and the integrated approach is provided in table 1.

Data and methods
In our whole-city case study (figure 1), we extract scope 1 CO 2 emissions and associated uncertainties from those reported in Lauvaux et al (2020), for the city of Indianapolis' jurisdictional boundary (∼950 km 2 ) and sum them annually for 2013 and 2014. Lauvaux et al (2020) employed an integrated approach to estimate 1 km 2 /5 day emissions for a nine county domain for 2013 and 2014 using (a) high-accuracy atmospheric CO 2 observations from Table 1   year-specific scope 1 CO 2 Hestia-Indianapolis granular emissions (1 km 2 hr −1 , figure S1 (available online at stacks.iop.org/ERL/16/084003/mmedia)) (Gurney et al 2012) linked through an inversion. We also separately show Hestia-Indianapolis CO 2 emissions and uncertainties (Gurney et al 2012) for the city domain since they cover a longer timespan (2010)(2011)(2012)(2013)(2014)(2015)(2016). A summary of the model used in Lauvaux et al (2020), the Hestia-Indianapolis method, means of extracting city emissions and uncertainties, and conversion of CO 2 eq to CO 2 are provided in the SI along with annual totals and sectoral emission sums (table S1). We compare the extracted and aggregated city-wide emissions from Lauvaux et al (2020) with those of Hestia-Indianapolis. We derive scope 1 CO 2 emission from the City of Indianapolis' Thrive report and GHG Inventory (2018, 2021) (detailed in table S2). We directly compare the (a) city-wide totals (2010,2013,2016), (b) city reported scope 1 sectoral emissions (2013) to (c) Hestia-Indianapolis emissions. In our sub-city case study, we use the annual sum of the on-road sector CO 2 emissions from the granular Hestia-Baltimore data product (for the city domain only; Roest et al 2020). A city-scale integrated approach cannot be used to estimate emissions for 2014 given atmospheric observation limitations for that year. We assume that Hestia-Baltimore provides a realistic spatial representation of on-road emissions that would result from an application of an integrated approach. These emissions have been compared to the City of Baltimore SRI and differences were diagnosed by Roest et al (2020). Hestia on-road emissions are estimated for each road segment as detailed in Gurney et al (2020a). Briefly, the developers used an inverse distance weighting of average annual daily traffic (AADT) from the Federal Highway Administration's vehicle counting station data, which were disaggregated by road type. The AADT were then used with the road segment length to estimate the vehicle miles travelled for each road segment, which provides a relative metric to distribute county-level FFCO 2 emissions from the 2011 EPA National Emissions Inventory (NEI). Local gap-filled traffic data were used to distribute emissions in time to the hourly scale, though only annual emissions are used in this study. More details on the on-road emission methods can be found in the Vulcan (Gurney et al 2020a) and Hestia-Baltimore (Roest et al 2020). We also use a housing market typology (HMT) map employed by the city to target interventions and guide municipal investments based on neighborhood conditions (https://planning.baltimorecity.gov/mapsdata/housing-market-typology). Additionally, we utilize the Environmental Protection Agency's (EPA) National Air Toxic Assessment (2014) (www.epa.gov/ national-air-toxics-assessment) which lists zip-codes where air toxins and emission source types pose respiratory health risks for each state. Zip codes used in this study were highlighted as having poor air quality Figure 2. Flowchart for analysis for the seasonal/whole-city case study for the City of Baltimore. The orange box represents the integrated approach used in Yadav et al (2021) to estimate emissions in January-May 2018, 2019, and 2020. Emissions were extracted for the City of Baltimore jurisdictional domain from the wider estimation extent used in Yadav et al (2021). Relative reductions were estimated and compared to those reported in Yadav et al (2021) and evaluated using proxy data. due primarily to on-road emissions. More information on data is provided in the SI.
For our seasonal/whole-city case study (figure 2), we extract and aggregate monthly scope 1 CO 2 emissions and uncertainties for the City of Baltimore (∼240 km 2 ) from those reported in Yadav et al (2021). Yadav et al (2021) used an integrated approach, for the Baltimore/Washington DC region, to discern the impact of COVID-19 on CO 2 emissions. Yadav et al (2021) estimated daily emissions at a native resolution of 2 km 2 /daily for these metropolitan regions (∼18 000 km 2 for both domains). Emissions were estimated for January-May 2020 along with coincident time-periods for 2019 and 2018 (the averages providing baseline months). The integrated approach used high accuracy atmospheric observations from in-situ tower networks (figure 4(a)) and 2015 granular emissions from Vulcan (Gurney et al 2020a). A summary of the method is provided in the SI. To diagnose cause and effect relationships at the wholecity scale, we compare features in emission proxy data with the extracted and summed CO 2 emission estimates. The emission proxy data includes natural gas consumption and gasoline data for the State of Maryland from the Energy Information Administration (EIA; www.eia.gov/state/?sid=MD); (www.eia.gov/dnav/pet/hist). We also use monthly emission data from EPA's Clean Air Market Database (CAMD; www.epa.gov/airmarkets) for a powerplant within the Baltimore city domain.

Annual/whole-city comparison: Indianapolis
After extracting and summing the Lauvaux et al (2020) emissions for the City of Indianapolis (as estimated using an integrated approach), we find them statistically consistent with those from Hestia-Indianapolis for 2013 and 2014, with less than 3% difference. This mirrors the results in Lauvaux et al (2020). Note, that the reported agreement is not happenstance. Lauvaux et al (2020) biased Hestia-Indianapolis emissions by up to 10% and achieved almost identical results at the nine-county scale. The convergence in estimates demonstrates that highaccuracy atmospheric CO 2 observations within an integrated system provide the necessary constraint to adjust granular emissions (if biased) to be consistent with atmospheric CO 2 -which accounts for all possible city sectoral CO 2 emissions. We do not have emissions from the integrated approach for other years. However, we assume the consistency between Hestia-Indianapolis' city-wide emissions and those extracted/summed from Lauvaux et al (2020) allow us to confidently use Hestia-Indianapolis' emissions as an independent point of comparison to those reported by the city.
The Hestia-Indianapolis emissions and the city's scope 1 reported emissions trend well together for each of the three reporting years ( figure 3(b)). Note, this trend does not necessarily reflect reductions stemming from the implementation of city-policies but mainly reflects market forces in the electricity production sector. For example, the Hestia-Indianapolis' electricity sector emissions indicate a decrease in emissions from 2014 to 2016 due to fuel switching at the two largest in-domain power plants (figure S1), e.g. Harding St. (a 12 unit, 1196 MW capacity) and Perry K. (a small steam producing multi-fired power station). The city's 2013 and 2016 emissions also capture this change.
However, unlike the trend, the absolute magnitudes of both the annual, sectoral totals of Hestia-Indianapolis and the city's SRI are significantly different ( figure 3(b)). The city's reported emissions are consistently 35% lower for 2010, 2013 and 2016 compared to the Hestia-Indianapolis emissions. Further analysis suggests that almost all emission sectors are underreported in the city's SRI by different amounts for various reasons (City of Indianapolis 2018, tables 2 and S3). The city's SRI attributes the majority of GHG emissions to Indianapolis' indomain residential buildings and traffic. But the difference in the transportation sector is substantial (43%). Given the size of the sector's emissions, the discrepancy questions whether the city will be able to assess whether it can achieve its goal of reducing on-road emissions by ∼67% in 2025 from a 2016 baseline (refer to tables S4 and S5 for policies outlined in Thrive including public benchmarking).
As cities commit to becoming net-zero emitters by 2050, the differences in total emissions will increasingly be important to reconcile; assessing trends will not be enough. As a signatory to the UNFCCC Initiative 'Race to Net Zero' (https://unfccc.int/climateaction/race-to-zero-campaign), the City of Indianapolis will have to balance any of its 2050 emissions with carbon offsets. Currently, the absolute difference  between Hestia-Indianapolis emissions and the city's reported totals is significant (2-sigma) with an average annual discrepancy of ∼4 MMTCO 2 -raising questions as to whether the city will be able to confidently assess their emissions for proper offsetting. For example, as an interim step to 2050, the city aims to reduce ∼5.86 MMTCO 2 by 2025, with 2.2 MMTCO 2 from the on-road sector (Scope 1) alone. Since annual/whole-city emissions estimated from an integrated approach are consistent with atmospheric levels of CO 2 , they could provide a credible point of comparison to SRIs to ensure proper accounting.

Sub-city case study: Baltimore
Mitigating GHG emissions can have co-benefits that improve quality of life by addressing air pollution and equity issues which occur at the local scale. In the City of Baltimore, the overlay of zip codes associated with very high respiratory risk from transportation (aka on-road) emissions specified in EPA's National Air Toxic Assessment with Hestia-Baltimore and the city's HMT map enables us to identify neighborhoods and roads that have large associated CO 2 emissions and air quality criteria pollutants like PM 2.5 . As in other studies (Levy et al 2007), figure 4(b) highlights economically disadvantaged communities that are disproportionately impacted by air pollution. Note that at an aggregate city-scale, emissions along interstates and arterial roadways make up the largest percentage of on-road emissions. However, exposure to vehicle exhaust that causes increased respiratory risk are generally associated with proximity to local roadways (Zwack et al 2011). Indeed, for these zip codes, those streets that contribute 50% or more to overall GHG emissions are smaller arterial roads (table 3) which suggests that congestion and 'street canyons' may cause significant air pollution levels at these localities rather than major thoroughfares . The city aims to use planning activities outlined in their CAP (e.g. City of Baltimore, 2014; activity LUT1. A further outlined in table S7) to decrease transportation related GHG emissions and improve residents' quality of life overall. Figure 4(b) implies that information at the zip-code scale may be too coarse to effectively achieve such co-benefits, since hotspots and intersections and specific roads within these areas are responsible for most emissions and air pollutants. Beyond achieving co-benefits, sectoral emissions resolved at granular scales can enable city planners detect the largest emitters in the wider Figure 4. The Washington DC/Baltimore region with urban areas defined by the 2016 census, the City of Baltimore (green polygon) and the locations (black circles) of sites that have CO2 observations are shown in (a). The four zip codes outlined in (b) are in the 90-100th percentile of those in the State of Maryland with respiratory risk from PM2.5 due to on-road diesel fuel combustion. They are also within the 85-100th percentile in terms of respiratory risk from all criteria air pollutants (www.epa.gov/ national-air-toxics-assessment). The housing marketplace typology (HMT) map shown in (b) combines sale price, vacancy, foreclosures, and other variables to indicate stressed areas along with specific streets and their associated Hestia emissions within these four zip-code. The roads with the most substantial contribution to the zip codes' total CO2 emissions are shown in table 3. urban extent-helping cities achieve their climate goals. Within its CAP, the city has targeted improving the energy efficiency of buildings. The city intends to audit commercial and industrial buildings over 10 000 sqft to identify simple and low-cost energy efficiency measures (City of Baltimore, 2014; activity ESS.1.C). This amounts to ∼3800 buildings with 76% being commercial (according to Hestia-Baltimore). Roest et al (2020) implied that a large fraction of commercial building emissions (∼76%) are from a small number of structures. Hestia-Baltimore could help the city prioritize which buildings should be audited first based on their emissions profiles (table  S7). In doing so, Hestia-Baltimore may also be able to provide a more realistic GHG reduction estimate based on efficiency measures implemented at these structures compared to the reduction numbers provided in Baltimore's CAP.

Seasonal analysis/whole-city analysis: Baltimore
Sub-annual emission information can provide valuable insights into the causes of emissions changes (e.g. weather events, sudden shifts in behavior, and abrupt market forces) that might be obscured in annual totals (Perugini et al 2021). However, estimates often have latency that lags real-time by 3-5 years ( figure S2). This may change in the future as more city-specific data becomes available without too much delay that can be ingested in granular emission products. In contrast, atmospheric observations are available near real-time (figure S2). When using an integrated approach, the atmospheric observations use 'older' granular emissions (for spatial and temporal patterns) but 'pull' emissions to capture abnormal events for more recent years (Yadav et al 2019, Turner et al 2020. Note, the specificity of granular information within an integrated approach, even for a prior year, is important to achieve citytotals consistent with atmospheric CO 2 (Oda et al 2017).
We show this with our extracted Yadav et al (2021) emissions and associative uncertainties for the City of Baltimore. Our extracted monthly summed CO 2 emissions show nearly identical relative reductions (31% in April 2020; figure 5(a)) in Baltimore as Yadav et al (2021) (33%) but with much wider uncertainty bounds (1-sigma 30% compared to 11% respectively). These large uncertainties reflect, in part, that some high-accuracy observations in the city were not available during the height of the change in emissions (April 2020).
For the City of Baltimore, variations in emission proxy data (e.g. gasoline fuel sales and natural gas consumption) explain much (but not all) of the relative drop in emissions during this time, particularly gasoline fuel sales (figure S3). These relationships are explored in Yadav et al (2021). Herein, we look at electricity output since, at the national scale, emissions from this sector have been shown to correlate with pandemic shifts (−15.1% April-May relative decline; Gurney et al 2021) when compared to emission baseline years. There are several powerplants within the city's boundary. The large Wheelabrator waste-to-energy facility (partially fueled by petroleum waste products) experienced a steep decline in February/March 2020 ( figure 5(b)). Global petroleum coke (petcoke) supply was affected by OPEC+ negotiation fallout and events associated with COVID-19 (Deloitte 2020) which likely explains this jump. The sudden drop in Wheelabrator emissions may have contributed to the 30% relative reduction in the Baltimore emissions.
The density of people and activities in cities make their citizens, economies, and carbon emissions vulnerable to various stressors-both natural and man-made. Understanding the relationships between CO 2 emissions and shocks can help cities (a) design policies to adapt to unexpected events while mitigating emissions and (b) tease apart those emissions that can be influenced by polies (e.g. building codes) compared to others outside of the city's control (e.g. market forces).

Implications
Cities have emerged as vanguards of climate leadership and have a stronger relationship with their citizens. But they have a limited set of policy levers to alter their emissions alone. Coordination and consistency with state, regional, and national emissions is crucial; e.g. cities like Baltimore are counting on state and national measures to help cut their emissions (20% from the state's renewable portfolio in 2022, and 11% from EPA's passenger vehicle and light duty fuel efficiency standards) but little coordination is evident. If estimated for a large enough area, emissions from an integrated approach are completely consistent across metropolitan, state, regional, and national domains while being tied to atmospheric observations whose accuracy is ensured by international standards. Nested national to regional systems could (a) use various levels of data more quickly, (b) enable policy levers at different scales of government, (c) foster private-public relationships, (d) avoid truncation issues associated with transportation (e.g. on-road, aviation, and commercial marine shipping), and (e) allow for an assessment of the progress of action across the county. In this manner, they complement existing tools/methods.
To provide the type of emission data shown herein, continued progress is needed. This involves grappling with observational constraints (e.g. in-situ, flask, low-cost, and aircraft GHG measurements, along with satellite retrievals), improving latency issues, enhancing workforce capabilities, tackling costs, and engaging stakeholders. Presently, implementation is prohibitively expensive for most city governments. Organizations like the World Meteorological Organization's Integrated Global GHG Information System initiative (WMO-IG 3 IS; https:// ig3is.wmo.int/en/welcome), NGOs, documentary standards organizations, etc must also help. These organizations can develop a scientifically-based and internationally recognized framework that would ensure that methods are tied to verifiable standards.
Researchers are also actively improving elements of the inversion process that lead to a fair amount of uncertainty on estimated emissions. These include: improving transport models which link observations to emissions in specific locations (Deng et al 2017), separating atmospheric CO 2 entering a city's domain , distinguishing anthropogenic sources from biogenic fluxes , and expanding the suite of measurements to constrain specific sectors (Nathan et al 2018), etc. Confidently estimating biological contributions using the observations could also help with accounting for programs like composting, increasing green areas, etc.-but more work is needed to do so.

Conclusions
In this paper, we show that combining atmospheric GHG observations with granular emissions data products in an integrated approach could help assess the uncertainty of SRIs, support city climate and sustainability goals, and uncover relationships between GHG emissions and their drivers. Near realtime feedback offers the potential for the 'coursecorrection' of policies if needed.
The United Nations Environment Programme (UNEP) estimates that the world would need to cut carbon emissions by 7.6% per year for the next decade to prevent the globe from warming more than 1.5 • C above pre-industrial levels (UNEP 2020). To do so, most cities must achieve net-zero emissions by midcentury. But without atmospherically checked emission data that speaks to scales of human behavior, this will be difficult. The discrepancies between SRI and our emission estimates shown herein are too large to have assurance in reported emissions. Methods like the integrated approach, as demonstrated within this work, can help cities achieve multiple goals and point global climate efforts in a new and more effective direction-all of which are needed to drive down emissions locally and globally.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).

Acknowledgments
Support for K Gurney, S Ghosh, G Roest, and P DeCola was provided by NIST grants 70NANB10H245, 70NANB19H132, 70NANB19H129, and 70NANB19H131 respectively. T Lauvaux was supported by the French program Make Our Planet Great Again (CIUDAD).
The opinions/recommendations/ findings/conclusions expressed do not necessarily reflect the views/policies of NIST/US Government.