Impacts of a climate change initiative on air pollutant emissions: Insights from the Covenant of Mayors

Highlights • Evaluation of air pollutant emissions corresponding to locally reported CO2 emissions.• Methodology applied to over 1600 Covenant of Mayors signatories.• Most changes correspond to co-benefits for both climate and air pollution.• The role of technological improvement to limit emissions is highlighted.

reports the matching of the CoM sectors to the GAINS ones. Unfortunately, a one to one match is not possible. All CoM sectors referring to buildings are mapped to the domestic (DOM) sector of GAINS (including residential, commercial and services buildings), the ones relative to transport are mapped into several GAINS sector.
The matching of carriers and fuels is more straightforward and is reported in Table 2. Most fuels have a one to one match, the exceptions are biofuels and other biomass matched to the category BIO, other fossil fuels matched to the category ALL FF, (standing for 'all fossil fuels)', heating oil matched to the fuel HO (heating oil). BIO, ALL FF and HO are not GAINS fuels.
HO is associated for the domestic sector to the HF (heavy fuels) and MD (medium distillates) reported in GAINS. Some cities however report the use of HO in the transport sector, which could be an input error. Because there is not HO matched to transport in GAINS, in this case, the assumption is that they are actually referring to MD.
BIO, when used in the residential sector, corresponds to all biogenic fuels used in this sector in the selected scenario (denoted as OS1 or the specific fuel name in the GAINS database, see Table 3). On the contrary, when used in the transport sector it is matched to all fuels used in this sector (as the scenario does not report biofuels as a separate fuel). In this case, in the absence of more detailed data, the assumption is that the emission factors of air pollutants of biofuels and bio-fossil fuels mixtures are similar to those of the corresponding fuel mixture reported in the scenario. This is a limitation due to the emission factors we have available at the moment through the GAINS database. The matching could be refined and improved if more detailed emission factors were available, e.g. for biofuels or plant oil for the different sub-sectors of transport. However, it should be underlined that, in the current CoM database, the reporting of biogenic fuels in the transport sector is not always consistent and detailed, therefore the assumption used in this study can be considered reasonable.
According to the report Koffi et al. (2017), other fossil fuels should correspond to peat and the fossil fraction of municipal solid waste. However, looking at the data, this carrier is sometimes associated to the transport sector as well as the residential sector. In this analysis, 'other fossil fuels', is associated to ALL FF used in the corresponding sector under the assumption that the reported energy use is correct.
The GAINS on-line platform provides emissions for different pollutants and activities, for

CoM-EDGAR sector and carrier matching
The same analysis presented in this study can be carried out using the Emission Database for Global Atmospheric Research (EDGAR) database. In this case the emission factors are not the result of a scenario and a model, but rather the result of the careful compilation of emissions, activities as well as abatement technologies around the world (Crippa et al. (2018)).
The EDGAR database is currently being updated, therefore we preferred presenting results in the main text of this work using the GAINS model. However, a preliminary analysis on the same CoM dataset and using both inventories, shows comparable results and the same trends.
We provide therefore also the code for using EDGAR data. For completeness, Table a and Table   5 report CoM and EDGAR matching for sectors and carriers.    Further cleaning of the CoM data-set is carried out in order to limit, as much as possible, input errors for the specific analysis in this work. The following checks and changes are conducted: • CO 2 emissions associated with no energy consumption, or vice versa, energy consumption of a fossil fuel reported without CO 2 emissions (rare): the missing value (CO 2 emissions or energy consumption) is calculated using default emission factors.
• Coal used for transport: at the moment the corresponding emissions are not considered (even though they could correspond to rail transport).
• Heating oil used in transport: is assumed as an input mistake and is considered as diesel.
Internal consistency checks between the macro-sectors in the BEI and MEI data-sets are also performed. As we are looking at changes (between the BEI and the MEI years), the following outliers are removed for the display of data and analysis.
• Sectors (intended as total transport or domestic) for which energy consumption is reported to be 0 or almost 0 for the BEI or the MEI are removed.
• Sectors (intended as total transport or domestic) for which the relative energy consumption or CO 2 emissions varies between the BEI and the MEI by less than -90% and more than +500% are removed.
Finally, the data displayed in the figures corresponds to 1655 Signatories.

Trade-offs at city level
In order to show the variability between cities, Figure 1 presents CO 2 with respect to NOx and PM2.5 emission changes per city and per person. As in the main text, the data is presented with a count plot.  Overall results From Figure 2 t is evident that: • For the same level of CO 2 emission, transport displays the highest NOx emissions.
• There is a clear correlation between air pollutant emissions and CO 2 emissions (as expected), The correlation is stronger for the transport sector (circles) whereas, for the residential one (triangles) and especially for PM2.5 (red triangles) data-points are more dispersed. This dispersion is due to the larger variability of fuels/technologies and therefore of emission factors used in the domestic sector.
Co-benefits and trade-offs for the residential sector In the main text we presented the co-benefits and trade-offs for the transport sector, highlighting the effect of technological improvement. The corresponding results for the residential sector are presented in Figure 3. In this case the effect of technological improvement is less evident.
With Tec. Improvement Without Tec. Improvement  year) are reported in Figure 4. Some of the difficulties encountered during this comparison are, for example: • It is not always possible to find data for the years corresponding to the BEI and the MEI years.
• CoM inventories generally report the emissions that the city itself, with its actions, can reduce. Therefore a one to one matching is not always possible.
• Following from the previous point, even at macrosector level (e.g. transport), literature data may have a different aggregation than the one provided by the CoM. E.g. it may include shipping or heavy duty transport.
• The boundaries of the city may also be different.
Given these constraints and limitations, it was not possible to carry out, at the moment, a more thorough comparison. The few results obtained, however, generally show a promising picture.

Code
The code supporting this analysis is available at: https://github.com/esperluette/airpollutants_script_for_com At the moment, the entire CoM database is not yet publicly available (though it is planned to become accessible soon) so it was not possible to provide the entire dataset used for this work. However, a fictive input data file is provided for one example city, which could allow any city to test the data they submit to the CoM with the proposed approach. Minimum input data is also provided from GAINS and EDGAR. Additional data from GAINS and EDGAR (i.e. for different countries and different years) can be accessed through their respective contacts and on-line platforms (see the code for more details).