After the adoption of the EU Climate and Energy Package in 2008, the European Commission launched the CoM to endorse and support the efforts of local authorities to implement sustainable energy policies. Today, the CoM is the main European movement dedicated to local and regional authorities who voluntarily commit to meet and exceed the European Union's objective of 20 % GHG emission reduction by 2020 (Covenant of Mayors, 2009). The number of towns participating in the CoM is steadily increasing.

Furthermore, the CoM has already been extended to Eastern and Southern European neighbouring countries. While keeping the 20 % emission reduction target, it was adapted to specific characteristics of the new countries. Former Soviet countries are allowed to set an emission reduction target with respect to the level of emissions estimated in a business-as-usual scenario for 2020. Thus, their emissions can grow from their baseline level. Towns from North-African countries are explicitly requested to address the problem of water and waste management.

After joining the CoM, signatories have to:

a.

Compute and submit a Baseline Emission Inventory (BEI) for their territories. This identifies the baseline level of GHG emissions to serve as reference for the reduction target for 2020. The BEI is based on city's energy consumption patterns (per key economic sector and energy carrier), and provides implicit evidence on the most suited actions for the city to reach its CO2 reduction target.

By 14 March 2013, 5049 municipalities joined the CoM, covering 187.5 million citizens (of which 96 % were from the EU-27). The highest number of signatories is in Italy (2582 municipalities, corresponding to 51 % of CoM signatories) and Spain (1323 municipalities, corresponding to 26 % of CoM signatories).

Christophoridis et al. (2013) further commented the diffusion of the CoM initiative in Europe. Most of the towns are located in Southern European countries where dedicated bodies, including Covenant Territorial Coordinators (CTCs), supported cities in the process of adhesion to the CoM. The CTCs are regional authorities which voluntarily join the movement committing to promote it within their respective territory and to offer technical and/or financial support to the signatories which choose to work under their coordination.

The peculiarity of the CoM movement, compared to other GHG mitigation networks, is the elicitation of small towns' interest and engagement in the effort to reduce greenhouse gas emissions. On the 14 March 2013, 4453 small and medium size towns (with a population of less than 50 000) joined the CoM. They are 88 % of total signatories and account for 16.5 % of the total population covered by the CoM.

This suggests that small cities can also play an important role for climate change mitigation. Based on three case studies, Melica et al. (2014) indicated that the multilevel governance approach adopted within the CoM has been a key determinant to get the involvement of small towns in the movement.

On the contrary, big cities (with more than 50 000 inhabitants) account for a relevant share of population in the CoM: 56.5 % of CoM population live in cities with 50 000 to 1 000 000 inhabitants, and the sole 24 cities with more than 1 million inhabitants in the CoM (e.g. London, Berlin, Madrid and Rome), represent 27 % of the total CoM population.

Given all the above mentioned, the CoM allows to collect a unique bottom up inventory of local greenhouse gas emissions and related emission reduction potentials, as estimated by local authorities. This can be used to enhance the precision of existing emission inventories and further explore the relative importance of small and big towns in the effort for climate change mitigation.

Alternative examples of international networks of municipalities, similar to the Covenant of Mayors, are the following:

a.

The C40 Cities Climate Leadership Group (C40) is an association of 58 megacities internationally (as of end-2013) which have the common purpose to locally implement sustainable climate-related policies. It covers almost 20 million tonnes of CO2 and involves 8 % of the world population and 18 % of the global GDP. It targets a 30 % reduction of CO2 between 2005 and 2025.

In line with the established framework of the UNFCCC, project guidelines for emission inventory within the CoM (Bertoldi et al., 2010) broadly follow the guidelines of the Intergovernmental Panel on Climate Change (IPCC). Similar to the UNFCCC, the recommended baseline year for reporting is 1990, or the closest subsequent year for which the most comprehensive and reliable data can be provided. Economies in transition within Annex I countries (e.g. East European countries and newly independent States) are expected to choose a year after 1990, but close to it and representative of the current situation, in order to avoid a bias related to the 1990–1991 economic breakdown. At the moment of the analysis, the baseline years which were chosen by the majority of the signatories were from 2005 to 2008.

Signatories are given various options to calculate emission inventories. They can choose the standard IPCC approach or the life cycle assessment (LCA) approach. In the standard IPCC approach, emission factors are based on the carbon content of fuels. Only CO2 reporting is mandatory, as it is the most important among all the GHGs when talking about emissions associated with fuel combustion. Nonetheless, signatories can report emissions of methane (CH4) and nitrous oxide (N2O), converted into CO2-equivalents (CO2eq.) according to their global warming potential (GWP). CoM guidelines for emission factors are based on IPCC 2006 Guidelines (IPCC, 2006), while CO2eq characterisation factors are based on the IPCC 4th Assessment Report (IPCC, 2007). However, local authorities can choose different emission factors, provided that they are in line with the IPCC approach. Finally, CO2 emissions from the sustainable use of biomass/biofuels, as well as emissions of certified green electricity, are considered carbon-neutral on an annual basis.

With the LCA approach, the overall life cycle of the energy carrier is calculated. This approach includes all emissions within the supply chain, from extraction of natural resources to processing (e.g. refinery) transport and final use (e.g. combustion). As a result, emissions generated beyond the administrative boundaries of a town are included if they can be associated to final consumption of the city. For example, this approach leads to associate positive GHG emissions to the use of carbon-neutral fuels because of the energy carrier of the supply chain.

Other non-energy related GHG emissions such as methane from landfills and waste water management could be included in the local authority's emission reports, converted into CO2eq.

Direct emissions in urban areas derive mainly from two macro-sectors: transport and buildings. Moreover, they can be directly influenced by local policies. Thus, mayors are recommended to design a strategy for emission reduction that includes both of them. In particular, these two macro-sectors have been disaggregated into four sectors and signatories are required to include in the BEI at least three of them: i.

municipal buildings, equipment and facilities (it can optionally include municipal public lighting);

As a consequence, the assessment of differences between CoM emission inventories and other inventories from the UNFCCC or from the European Commission requires careful identification and association of comparable sectors.

This section develops a comparison between the methodologies adopted in different international emission inventories to calculate the amount of greenhouse gases emitted by a region in a given time-scale. Different international emission inventories are available, including the UNFCCC National Inventories, the Global Carbon Budget, the Emissions Database for Global Atmospheric Research (EDGAR) and AR5 data (which includes the previous two and a third emission inventory based on IEA energy data). IEA and EUROSTAT provide data on both energy consumption and GHG emissions.

In order to validate the data in the “CoM sample 2013”, CoM baseline emissions per capita were compared with the EDGAR inventory for fossil fuel emissions in the buildings sector (residential, tertiary and administrative), in the transport sector (road transport) and in the waste management sector (Fig. 4). Moreover CoM baseline energy consumption was compared with IEA energy data on final energy consumption in the building sector for different energy sources (Fig. 3). This ensures consistency between the two comparisons because IEA energy data feeds the underlying activity data of EDGAR for the energy related sectors. EUROSTAT data on energy consumption was also used to further check the robustness of results. Evidence from the comparison is discussed in Sect. 4.2.

This section is focused on the description of similarities and differences between the CoM approach to emission (and energy) accounting and EDGAR v4.2, IEA, EUROSTAT inventories (see also Olivier and Berdowski, 2001; Janssens-Maenhout et al., 2012).

EDGAR is a joint project of the European Commission DG JRC and the Netherlands Environmental Assessment Agency (PBL). It provides past and present global anthropogenic emissions of greenhouse gases and air pollutants by country on a spatial grid (EDGAR, 2011).

The International Energy Agency (IEA) is an autonomous intergovernmental organization established in the framework of the Organisation for Economic Co-operation and Development (OECD) in 1974. It provides worldwide detailed energy balances and CO2 emission estimations at national level.

EUROSTAT provides data from the annual greenhouse gas inventory compiled by the European Environment Agency (EEA) on behalf of the EU. Estimates of greenhouse gas emissions are produced for a number of sources which are delineated in sectors primarily according to the technological source of emissions, as devised by the IPCC.

Common and divergent characteristics of the above mentioned databases are succinctly presented in Table 7. The approach of EUROSTAT is very similar to IEA and thus only minor differences can be found between the two datasets. For this reason, the comparison between approaches for emission accounting is developed in more details for CoM, EDGAR and IEA inventories. i.

The EDGAR definition of energy consumption in the tertiary, residential and transport sectors covers local, in-situ emission sources, whereas the CoM considers, not only the in-situ emissions sources but also, the emissions which occur due to the consumption of energy from carriers which could emit outside the city's territory, such as electricity and heat/cold (scope 2 emissions). Even if the consumption of electricity and heat/cold (delivered as final commodity to the user) does not imply emissions at the place of consumption, but at the production site, the CoM considers the users to be co-responsible for the production of the electricity and heat/cold during the supply chain and therefore for the corresponding emissions at the place of production. This is particularly important for those countries where electricity from fossil fuel and heat from district heating plants are widely used. For this reason, the “CoM sample 2013” was compared to international databases (see Sect. 4.2) only for those sectors and energy carriers whose emissions are related to energy produced at the place of final consumption.

Furthermore, cities that adopted the LCA approach for emission accounting, implicitly included, in the above mentioned sectors, the emissions related to the supply chain of the fuels (see Sect. 3.1) in their inventories. In EDGAR, these emissions are allocated to other sectors (e.g. railway and marine transport, industrial processes etc.) and, sometimes, to other countries, in case the fuels used for energy production are imported. Therefore, in order to render the inventories more comparable, the emissions included in the LCA inventories of the “CoM sample 2013” were converted to direct emissions (see Sect. 3.5) for the comparison of inventories developed in Sect. 4.2.

The emission reduction target is specified as a ratio of the expected reduction of emission by 2020 and the level of emissions in the baseline year. The emission reduction target has to be at least 20 %, in absolute or per capita terms. All signatories have to submit an official action plan with the detailed set of actions to reduce emissions, including project management information (implementation time frame, responsible bodies, costs) and impact estimations for 2020 per action and per sector (such as expected energy savings, green energy production [MWhy-1] and CO2 reduction [ty-1]). At the time of the assessment, only the CO2 reduction estimation per macro-sector was compulsory. As a consequence, the type and quality of data varies significantly across cities and requires a preliminary screening to build the reference CoM sample.

The reported inventories of greenhouse gas emissions and energy consumption at municipal level are based on data provided by cities participating in the CoM. The “CoM sample 2013” is built starting from the full sample of SEAPs that were accepted before 14 March 2013. However, inconsistencies in the data led to the exclusion of some cities from the sample.

First, towns with extreme and non-reasonable declared total emissions per capita, energy consumption per capita and estimated emission reduction per capita were excluded from the CoM sample to drop possible coding errors.

Subsequently, detailed checks on the internal consistency of inventories were conducted.

As mentioned in Sect. 3.2., cities in the CoM are asked to report energy consumption and emissions per type of energy carrier, disaggregated by macro-sector (buildings, transport and others), and further subdivided by sector (i.e. consumption of diesel for municipal fleet and related emissions; consumption of gasoline for private transport in the municipality and related emissions). They should report as well subtotals by macro-sectors and sectors (across energy carriers). The declared total by sector was required not to be greater (in absolute value) than the computed corresponding figure (summing disaggregated figures per energy carrier). Detected inconsistencies greater than 5 % of the total have been manually corrected when expert judgment allowed to infer the committed mistake. Remaining cities have been excluded by the CoM sample. No consistency test has been carried out for the sectors in the group “Other” (e.g. waste and waste water treatment). Most of the emissions from these sectors derive from sources other than the energy carriers included in the CoM.

The same approach was applied to totals by energy carrier (within macro-sectors). However, reported total consumption (emissions) by energy carrier can be substantially greater than the sum of its reported components per sectors. It happens when data on consumption (emissions) is available only aggregated at the macro sectorial level and it is not possible to disaggregate it by sector. In this case, total consumption (emissions) is reported for the macro-sector, while figures for sectors are reported only if available. As a consequence, the city is not included in the sample only if the computed total is greater than the declared total, by more than 5 % (unless expert judgment allowed to infer the committed mistake and correct it).

Finally, implicit emission factors were computed dividing emissions and energy consumption as declared by subsector and energy carrier. Some of the implicit emission factors were found to be incompatible with internationally accepted reference values. Cities with implicit emission factors above 1 tonnes CO2 eq/MWh for fuels and above 2 tonnes CO2 eq/MWh for electricity were excluded from the sample if no reasonable justification for the choice was found and no clear mistake to be corrected was identified. Figure 5 reports the distribution of implicit emission factors for subtotals of sectors, by energy carrier, for the cities included the “CoM Sample 2013”. The reference value provided in the CoM guidelines is also highlighted with a vertical line, when relevant. The figure shows that the adopted procedure for data validation excludes values that are not acceptable according to the relevant literature and international guidelines. However, dubious cases remain (outliers within the acceptable range). Few cities may have misinterpreted the type of fuel to be associated with reported data. Some cities have zero emissions for positive energy consumption (implicit emission factor is zero) or zero energy consumption for positive emissions (implicit emission factor is missing). This may be related to lack of information for specific categories, low incidence of the energy carrier in the sector considered, exclusion from the reduction target of emissions for the selected energy carrier (but energy consumption was reported for completeness) or mistakes. Since these cases (as highlighted by Fig. 5) do not prove to be clearly erroneous, they are not excluded from the sample. Nonetheless, they should be used with caution.

As a result, the relative difference between declared and computed totals by subsector for all cities in the sample is not greater than 5 % in absolute value. Computed totals by energy carrier do not exceed declared totals by more than 5 %. Implicit emission factors lie within an acceptable range, as derived from theory.

The distribution of signatories in the resulting “CoM sample 2013”, by city size is reported in Table 2, while the distribution of signatories by country is reported in Table 1. Spain and Italy account for 80 % of the sample.

As signatories have a choice between two calculation methodologies (standard IPCC or LCA approach), baseline emissions have been transformed from supply chain emissions (LCA approach) to direct emissions (IPCC approach). In order to render the emission inventories more comparable within the sample and also with other emission databases (see Sect. 3.3), when referring to the emissions reported in the LCA inventories, a conversion factor of 0.885 was applied. That is equal to the average ratio between IPCC emission factors and the recommended LCA emission factors for energy products (Covenant of Mayors Guidelines, Bertoldi et al., 2010). This is assumed to be representative of direct emissions embedded in LCA inventories.

Finally, it is not possible to perform a conversion to reconcile different reporting units (CO2 and CO2eq) therefore all the data reported are considered as CO2eq. This leads to underestimate the quantity of non-CO2 emissions in the sample. However, the data still provides a reliable idea of direct emissions of towns, as CO2 is responsible for an average of 85 % of GWP in Europe (EDGAR, 2011).

The most common reference year for the BEI is either 2005 or 2007. Only 24 cities adopted 1990 for the BEI, as suggested in the CoM guidebook (Bertoldi et al., 2010). Moreover, there are strong differences from country to country. 1990 is the most common year in Germany and Sweden, while French cities chose more recent years as reference for the BEI (after 2000). 53 % of Italian SEAPs and almost all British SEAPs selected 2005. Furthermore, 80 % of Spanish signatories took 2007 as reference year and almost 50 % of Portuguese SEAPs opted for 2008.

Clearly, the choice of the reference year is crucial in determining the efforts required to meet emission reduction targets. For example, in countries such as Germany and Sweden, which experienced a negative trend in emissions from the beginning of 1990 (EDGAR, 2011), the choice of 1990 as reference year could be seen as a way to ease the work needed to meet the selected target. Conversely, a positive trend in emissions can be observed in most countries, until the drop related to the recent economic crisis. Therefore, the selection of a recent pre-crisis baseline year for the BEI may be related to strategic behaviour, i.e. to lower the effort required to meet emission reduction targets. Yet, this could also be related to real scarcity of appropriate data at municipal level: while in Nordic countries, where there is an older tradition of local GHG mitigation policies, signatories joining the CoM are already developing an action plan which generally refers to 1990 as a base year, signatories from the South often rely on more recent data, generally extracted from studies at regional level, not yearly available.

With regard to the methodology to calculate emission inventories, most SEAPs applied the IPCC approach and focused on CO2eq (see Table 3). This can be related to the higher complexity of the LCA approach which requires computing emissions related to the supply chain of each energy product (Cerutti et al., 2013a, b).

Disaggregated emissions by macro-sector and sector are reported in Table 4. The category “unassigned emissions” report emissions that were not assigned to a particular sector, as it was not mandatory. Buildings accounts for 65 % of total emissions in the inventory, followed by the transport (31 %) and the other sectors (4 %). Overall, one third of emissions are not properly attributed to sectors within the macro-sectors. This might be due to unavailability of detailed data at the city level to meet the desired breakdown in the CoM.

Emissions reported in the BEI vary considerably from country to country. Nonetheless, a common pattern can be identified for the distribution of emissions between macro-sectors in the selected countries (Table 5). First, cities focus on the sectors that are identified as key for the CoM (see Sect. 3.2). The share of emissions reported in other sectors is always less than 1 % of total emission, with the exception of Spain (7 %). Moreover, the building sector accounts for more than half of total emissions in the inventories, with the only exception of France (32 %). Its significance grows to more than 75 % in Germany, Italy and the United Kingdom.

Average emissions and energy consumption per capita from the “CoM sample 2013” are compared to country level emissions from EDGAR v4.2 and energy consumption from IEA. The comparison with EUROSTAT yields similar result as with IEA data and it is not reported (see the similarities in emission accounting reported in Sect. 3.3).

The comparison of emission per capita is relevant for the following sectors: residential buildings; tertiary buildings; private and commercial transport; waste management. For these sectors, emissions per capita are reported and compared to those reported at the country level from EDGAR v4.2. The per capita values in EDGAR were obtained using data gathered for CO2, CH4 and N2O emissions from 1990–2008. Because not all the years are equally represented as baseline years in the CoM, the EDGAR “per capita” average, used as comparison, was calculated as a weighted average, considering yearly data from 1990 to 2008, with the weight of each year given by the sum of the inhabitants of those signatories choosing that year as baseline.

Country and sector-specific emissions per capita from the “Covenant sample 2013” were calculated based only on the cities that reported disaggregated emission by sector. The inclusion of cities that reported zero missions in some sectors, as a result of their decision not to disaggregate emissions, would artificially decrease the computed level of emissions per capita in some sectors.

Keeping in mind the limitations mentioned in Sect. 3.3 regarding the inter comparability of data between the EDGAR database and the CoM sample, emissions in the commercial sector are higher than the national average for those countries where very large cities have a higher share from the total population of the sample (France, Great Britain) and lower for those where the sample contains small and medium cities (Italy, Spain, Sweden). This trend was expected given the fact that large cities are usually services providers also for the population of the surrounding areas. The CoM per capita emissions for Germany are much lower than expected because many very large cities were excluded from the calculation, given the fact that they did not report data disaggregated by subsector.

The CoM average for per capita emissions in the transport sector are around 20 % higher or lower than the national average with the very notable exception of France where the emissions in CoM are almost the double of the national average and Italy where the emissions per capita are about half the national average. This could highlight some inconsistencies in the methodology for building the inventories in the transport sector. While the recommendation of the Covenant Guidelines is that the basis for calculating the activity data for the energy inventory in the transport sector should be the mileage in the territory and the average consumption per kilometre per type of vehicle and type of fuel, the practice shows that not all the cities follow this exact methodology. Especially the small municipalities tend to exclude the energy consumption due to the transiting traffic and to account only for those vehicles registered in their territory and only for the mileage related to those vehicles on their territory. In the case of France, the significant difference between the national and the CoM per capita emissions are due to the high value declared by big cities, which represent a high share from the sample population for France. We can speculate that these cities have more complete data regarding the traffic, including data regarding the supper emitting vehicles on their territory, and that they also act as a pole for the daily commute of a significant population living outside their territories.

Per capita emissions in the residential sector have a similar variation of ±25 % between the two datasets, with the notable exception of Sweden where the CoM average is 47 % higher than the national average.

Besides the data on emissions, a very important part of the CoM inventory is the data on final energy consumption. As a prerequisite for the emission inventory, the final energy consumption inventory follows the same structure.

CoM data on final energy consumption in the buildings sector (including all key sectors for the CoM: administrative, commercial and residential) has been compared with IEA data for 2005. This is one of the best represented base year in the CoM sample.

Final energy consumption was grouped by the four main categories of energy carriers: electricity, heat/cold, fossil fuels and renewable energy sources.

As shown in Fig. 3 regarding energy consumption per capita, CoM electricity and fossil fuel consumption per capita are comparable to the national data with a variation of maximum 43 %, the consumption of heat/cold and renewable energy sources is subject to higher variation. As expected, for heat and cold the CoM values are generally higher than the national averages. Even more, as the national per capita value for heat consumption is already very low in the representative countries, in comparison with the consumption of other category of energy carriers (e.g. fossil fuel), the significant share of big cities with high per capita value (in Germany and Italy), raised considerably the per capita value of the Covenant sample, reaching values of almost six times bigger than the national average (Italy).

The opposite trend is observed for the final consumption of renewable energy sources transformed in heat at the place of consumption such as: biofuels, biomass, geothermal pumps etc. For this category of energy carriers, the value reported in the CoM is much lower than the national averages. Without a more in depth analysis we cannot explain the cause of this variation. We could only assume that this parameter is very much related to the local availability of the renewable sources and that sometimes the characteristics of the supply chain of these energy carriers make it difficult for the municipalities to collect reliable data (e.g. for biomass).

Overall, even if there are some inconsistencies related to the data input or the methodologies used to gather the activity data, the Covenant values are similar to the national averages. Nevertheless, given the voluntary character of the movement and the absence of a more strict control of the data reported, the outlier values have to be considered with caution.

We can conclude that, the “Covenant sample 2013” provides valuable data to support the analysis of heterogeneity in final energy consumption and greenhouse gases emissions at city level.

The voluntary nature of the project can be seen as a limiting factor with regard to the accuracy and completeness of data collected by the CoM. Furthermore, signatories voluntarily commit to an emission reduction target which is not legally binding, so expected emission reductions might not be achieved.

In the SEAP sample, only one third of signatories (35 %) decided to adopt the minimum target of 20 % emission reduction. Most of them committed to higher targets: 43 % of the signatories adopted a reduction target between 20 and 25 %, 10 % of the signatories targeted a reduction between 25 and 30 %, while 12 % of the signatories committed to a target of over 30 %. Figure 2 shows the total emission reduction potential by targeted emission reduction.

Big cities tend to adopt higher reduction targets than smaller cities. As a result, total emission reduction potential of the 394 towns that committed to exactly 20 % (20 Gt of CO2eq. by 2020) is lower than total emissions that could be reduced by the 128 town that committed to more than 30 % (34 Gt of CO2eq.).

Overall, the GHG reduction potential of small cities (less than 50 000 inhabitants) is about 17 % (69 Mt CO2eq.) of the projected reduction potential of the entire project. Signatories with reduced resources often choose low-cost solutions, such as awareness raising and behavioural change or they launch joint projects with neighbouring municipalities (joint SEAPs).

Expected emission reduction by city is estimated in the paper according to two methods. The first method estimates emission reduction according to the committed target, as a share of baseline emissions reported in the BEI. The second method computes the sum of expected emission reduction associated to actions planned in the SEAP. Total emission reduction according to the two procedures is reported in Table 6. Overall, cities tend to set an emission reduction target that is lower than the expected total reduction from planned actions. This can be related to a cautious approach to the flagship target, or a proactive approach to the planning of actions for emission reduction. Nonetheless, there are also cities that have not planned all the actions to be undertaken by 2020, as needed to reach the pledged emission reduction target. They planned in greater detail medium and short term actions, while they set the general strategy for subsequent years. These cities have a total expected level of emission reduction that is lower than the target, by now.

The data of the CoM sample does not allow to perform a detailed assessment of emission reduction potential per specific action. In fact, most SEAPs report only the estimated total emission reductions per field of action, as the estimation of emission reduction for a single action is not mandatory. Furthermore, SEAPs are elaborative instruments that plan simultaneous actions (Dall'O' et al., 2013), thus the expected effect of combined actions may not be further disentangled. Therefore, the analysis is restricted to emission reduction estimates per field of action, as reported in Table 6.

The largest share of potential CO2 and CO2/capita reduction is expected from the buildings sector, followed by the transport sector. A relevant number of actions are planned in the field “Working with citizens and stakeholders”. Low expected emission reduction is associated to them even if the importance of raising awareness is widely acknowledged (Kousky and Schneider, 2003). It epitomizes the difficulty to quantify the reduction potential of “soft” actions that do not directly include technical measures to improve resource efficiency (e.g. Heidrich et al., 2013; Rybski et al., 2013).