Tracking electricity use and scope2 CO₂ emissions in 111 cities of Madagascar

Abstract

Several sources related that the electricity sector emits almost a quarter of greenhouse gases each year in the world. It is therefore one of the important sectors to take into account to limit global warming. Indian Ocean cities produce significant CO₂ emissions during electricity consumption. Their volume and accuracy remain practically unknown and untested. Indeed, until now, there is no methodology suggested by the researchers to evaluate Fossil Fuel carbon dioxide (FFCO₂) emission, and electricity consumption in this region. Aware of these crucial problems, this study was carried out to assess and analyse CO₂ emissions coming from Electricity consumption (called Scope2) in 111 cities located in the Indian Ocean from 55 Power plants between 2015 and 2022 (08 years) and in four sectors (Residential, Commercial, Industrial, and On-road). To carry out a good comparison, all the data were grouped into three categories, before the lockdown measures due to COVID-19 (2015–2018); During the COVID-19-induced lockdown period (2019–2020); and after the lockdown period (2021–2022). The results showed that the CO₂ emission difference is significant in the residential and commercial sectors. It was observed that CO₂ emissions increased in 2019–2022 in the residential, industrial, and on-road sectors whereas, simultaneously during the same period, it decreased in the commercial sector. During the three periods, the CO₂ emissions rate was the highest in the residential sector (around 52%), and the least on-road (around 1%). The significant difference in the commercial sector suggests a decrease in electricity consumption during the peak of the pandemic due to reduced business activities. Businesses adapted to new operating conditions, such as reduced hours or enhanced energy efficiency measures, which also contributed to the change in consumption patterns.

Introduction

Over the coming years, global greenhouse gas (GHG) emissions will need to be reduced by up to 50% to achieve the 2015 Paris Agreement goal of limiting global warming to between 1.5 and 2 degrees Celsius¹. The United States, considered one of the most polluting countries in the world, intends to play its role. With their climate action plan, they are committed to achieving carbon neutrality by 2050 and will soon announce their emissions level target for 2030¹. After a slight drop in GHG emissions in 2019 and in 2020 due to the impact of the pandemic, it was observed that in 2021, GHG emissions started to rise again in several countries^1,2. So much so that they reached 52.8 Gtons in CO₂ equivalent (GtCO_2e), compared to 52.6 in 2019, according to the United Nations Environment Program (UNEP)². Since 2010, global GHG emissions have continually increased by an average of 1.1% per year². The variation in these emissions seems to be slightly decreasing since it was estimated at 2.6% per year between 2000 and 2009². In 2021, the four main emitters of GHG are, in order, China, the United States, the European Union of 27, and India^2,3. These four countries emitted 55% of total CO₂ emissions in 2010².

During the COVID-19-induced lockdown period (in 2020), the countries constituting the G20 represented 75% of global emissions^2,3. When electricity is consumed, industries and buildings produce GHG emissions indirectly. In this sense, Scope 2 aims to measure these emissions. Global electricity production has increased more than 5 times from 1973 to 2020³. This is due to the significant increase in the world population⁴. Experts explained that 42% of global CO₂ emissions come from electricity⁵. Nowadays, the majority of its emissions (73%) are generated by coal-fired power plants⁴. Electricity is essential to businesses and households. According to global energy statistics reported by international agencies such as the International Energy Agency (IEA) and the U.S. Energy Information Administration (EIA), electricity consumption is unevenly distributed across economic sectors worldwide. Recent global estimates show that approximately 56.2% of electricity production is consumed by the residential, commercial, public services, and agricultural sectors⁴. In contrast, around 42.3% of global electricity demand originates from the industrial sector, which includes manufacturing activities, production facilities, data centers, and electric motor systems requiring continuous operation and adjustment^5,6. Finally, the transportation sector currently accounts for only about 1.5% of global electricity consumption, as it remains largely dependent on petroleum-based fuels rather than electrified systems^4,5. Having some basic knowledge of electricity consumption patterns is the first fundamental step towards reducing emissions of associated GHG. The impacts of emissions due to electricity generation are significant in global warming⁵. To develop a common roadmap to reduce climate change, policymakers are now called upon to understand the different origins of GHG^4,6.

Research carried out by de Chalendar et al. in 2019⁷ made it possible to evaluate the flow of CO₂ emissions from electricity across the different networks based on a multiregional input–output model (MRIO). The results demonstrated the importance of the dependence of the data according to the periods of the year and the electricity distribution circuit. Hourly CO₂ emissions data from electricity consumption reported by EIA-930⁸ was evaluated in 48 BAs using a multiregional input–output model (MRIO). To quantify CO₂ emissions from each of the BAs, the emission factors of natural gas, coal and oil were used. For several decades, many scientists have reoriented their field of research to support the United Nations Framework Convention on Climate Change (UNFCCC) and other international organizations in the estimation of FFCO₂ emissions produced in several sectors. EIA provided the total CO₂ emission coming from electricity consumption in many countries, including some regions of the Indian Ocean, but not with a distribution in the different sectors⁹.

The study carried out in 2018 by Wang et al.¹⁰ developed detailed emissions inventories for cities, highlighting the variations in CO₂ emissions due to differences in industrial activities, population density, and energy consumption patterns. A notable study by Nematchoua et al.¹¹ analysed emissions from residential and commercial sectors in many Indian Ocean districts, finding significant disparities in per capita emissions. Few studies related to GHG emissions have been carried out in the Indian Ocean, such as the research of Akter et al.¹² in 25 nations bordering the Indian Ocean, which showed that GHG emissions have a positive ascendance on the sustainable blue economy. Tegtmeier et al.¹³ reviewed the progress in detecting atmospheric emissions around the Indian Ocean and its global and local effects. Peter et al.¹⁴ explained that in the Indian Ocean region, the weather change patterns such as precipitation, winds, and temperature lead to droughts and floods, which are often linked to the increasing CO₂.

The Indian Ocean region is highly vulnerable to the impacts of climate change and energy insecurity¹¹. Yet, it remains underrepresented in global emissions datasets and energy transition research. By developing a novel approach for estimating Scope 2 CO₂ emissions in this context, the study not only fills a data and methodology gap but also provides a scalable framework that can be adapted to other regions with similar constraints—such as island nations or remote urban centers with limited access to renewable energy.

The period from 2015 to 2022 has seen significant advancements in the reporting of CO₂ emissions and electricity consumption in Madagascar cities. However, challenges remain in ensuring consistency, accuracy, and transparency in reporting practices. In this research, only the emission resulting from electricity consumption (scope2) is explored. Until now, no study has provided the CO₂ emission with high accuracy in the different Indian Ocean cities. This study has as its principal objective to gather accurate and detailed data on CO₂ emissions and electricity consumption in 111 cities located around the Indian Ocean from 2015 to 2022. Identify trends in CO₂ emissions and electricity consumption over the specified period, noting any significant changes or patterns. Compare the data across different cities to establish benchmarks and identify outliers, best practices, and areas needing improvement. Assess the impact of various factors such as population growth, economic development, policy changes, and technological advancements on CO₂ emissions and electricity consumption. In addition, evidence-based recommendations are provided for policymakers to reduce CO₂ emissions and optimize electricity consumption in these cities.

The novelties of this research are defined as follows: (a) this study uniquely focuses on cities located in Madagascar, a country in the Indian Ocean region that is often underrepresented in global environmental studies. (b) By covering eight years, the study offers a robust analysis of changes and trends over time, providing insights into the effectiveness of past policies and initiatives. (c) Combining data on both CO₂ emissions and electricity consumption provides a more holistic view of the environmental impact of urban areas. (d) The inclusion of 111 cities allows for extensive cross-city comparisons, making it possible to identify specific regional challenges and opportunities. (e) The study aims to produce actionable recommendations for policymakers, going beyond mere data reporting to offer practical solutions. (f) Emphasis on raising awareness and educating various stakeholders highlights the importance of public engagement in addressing environmental issues.

In summary, this research is divided into four principal sections: (i) The first one, called the literature review, shows some studies focused on electricity emissions. (ii) The second one, called methodology, explains the approach adopted in this study to assess the scope2- CO₂ emission. (iii) In the third part, the electricity consumption and CO₂ emission produced during three periods are analysed and compared: Prior to the national lockdown implemented in response to COVID-19 (2015–2018); during the government-imposed restrictions linked to COVID-19 (2019–2020); and, following the COVID-19-related lockdown (2020–2021). Finally, the last section shows some conclusions and outlooks.

Literature review

Electricity-related greenhouse gas (GHG) emissions, particularly Scope 2 emissions (i.e., indirect emissions from purchased electricity), have been widely studied across different regions of the world. However, the majority of existing emission inventory (EI) studies have concentrated on North America, Europe, and East Asia, with limited coverage of the Indian Ocean region and especially Madagascar. This section summarizes major methodologies and findings from previous EI research, including relevant studies from South and Southeast Asia as recommended, and highlights the gaps that justify the present study.

Emission inventory (EI) studies on electricity emissions and scope 2

Several EI frameworks have been developed to quantify electricity emissions at different temporal resolutions (hourly, daily, monthly, yearly). Ryan et al.¹⁵ demonstrated that CO₂ emission factors can vary by 63–68% depending on temporal averaging, emphasizing the importance of fine temporal granularity in EI design. Colett et al.¹⁶ proposed two EI protocols incorporating interregional electricity trade flows and regional emission factors, reporting emission factors between 19.0 and 19.9 kgCO₂e/kg of primary aluminium. In the United States, multiple EI systems such as eGRID¹⁷, AVERT¹⁸, REeDS¹⁹, WattTime²⁰, and the EIA-930 system²¹ provide disaggregated information on electricity generation, consumption, trade, and marginal/average emission rates from power plants. Machine-learning-based EI tools, such as WattTime²⁰, generate Marginal Operating Emission Rates (MOER) at sub-hourly resolutions. Other frameworks, such as Grid Mix Explorer²² and Singularity²³, integrate life-cycle analysis (LCA), continuous emissions monitoring systems (CEMS), and EIA datasets to quantify Scope 2 emissions at various spatial and temporal levels.

Emission inventory studies in South and Southeast Asia

Emission inventory research in South and Southeast Asia has rapidly expanded in the past decade. Several studies have estimated national or regional Scope 2 emissions associated with electricity consumption: In India, Garg et al.²⁴ and Kanudia et al.²⁵ developed EI models integrating temporal grid variability and coal-dominated generation mixes. In Bangladesh, Rahman et al.²⁶ assessed Scope 2 emissions from the power sector, highlighting grid inefficiencies and rapid consumption growth. In addition, in Thailand, Thongboonchoo et al.²⁷ quantified Scope 2 emissions from industrial regions, showing significant variations in grid emission factors. Whereas in Vietnam, Nguyen & Hoang²⁸ provided one of the most detailed EI assessments for Vietnam, integrating electricity imports and regional generation mixes. These studies highlight the significant heterogeneity in emission factors due to varying fuel mixes, electricity imports, grid reliability, and methodological differences. Despite this progress, the Southwest Indian Ocean, including Madagascar, remains severely underrepresented in EI databases.

Summary of major EI frameworks

See Table 1.

Table 1.

Studies related to the CO₂ emission.

Authors	Period	Methods	Time	Principal result
Ryan et al.15	2016	Review	Hourly-yearly	CO2 EF varies by 63–68% depending on averaging
Colett et al.16	2016	EI using trade + regional EF	yearly	EF for aluminium: 19.0–19.9 kgCO2e/kg
Watt Time20	–	Machine learning EI	< 1 h	MOER for many regions
eGRID17	2014	US EI database	Yearly	Multi-pollutant EI (CO2, NOx, SO2, CH4, etc.)
Marriott and Matthews29	2005	Consumption-mix EI	Yearly	States importing electricity may report misleading mixes
AVERT18	–	EPA non-baseload model	Hourly	Open data on electricity consumption & emissions
Grid Mix Explorer22	2015	LCA-based EI	N/A	Grid-mix EI and emissions
REeDS19	–	Grid-mix EI and emissions	Yearly	EI linked to electricity consumption, generation, and cost
Singularity23	–	CEMS + EIA + EPA	Hourly- yearly	Open-grid electricity EI
EIA-93021	–	Electricity trade EI	Hourly- yearly	EI for grid trades and consumption

Gap in the literature

Despite substantial global progress in EI methodologies and the existence of advanced Scope 2 emission tools, the Southwest Indian Ocean region—and particularly Madagascar—remains largely absent from international EI datasets. No national-scale Scope 2 EI framework currently exists for Madagascar, despite rapid urban growth and rising electricity demand. This gap further justifies the novelty and relevance of the present study.

Methodology

In this study, electricity consumption and other principal data have been collected in over 50 power plants installed in the different studied cities from 2015 to 2022. These data were grouped into three categories: prior to the national lockdown implemented in response to COVID-19 (between 2015 and 2018); during the national lockdown due to the COVID-19 pandemic (2019–2020); and in the post-lockdown phase after COVID-19 (2021–2022).

Study area characteristics

The 111 cities included in this study were selected based on three main criteria: (1) their geographic location within the Indian Ocean region, specifically within Madagascar. Although the Indian Ocean region includes several areas with similar characteristics—such as climate, construction materials, and patterns of human behaviour—this study focuses exclusively on cities located in Madagascar. These cities were chosen to represent a diversity of urban settings across the island, providing a relevant and coherent framework for analysing electricity-related CO₂ emissions in a context that remains largely underexplored in the literature. (2) their direct or indirect connection to the regional electricity grid supplied by 55 fossil fuel–based power plants; and (3) the availability of data on electricity consumption and sectorial energy use over the period 2015–2022. These cities represent a mix of small to large urban settlements with diverse socio-economic structures and energy profiles, making them ideal for identifying trends and disparities in electricity-related CO₂ emissions within the island and coastal contexts.

The 111 cities investigated were located in Madagascar and distributed in several climatic zones. The climate in Madagascar varies greatly from one region to another, which may be because of the proximity of the ocean³⁰. The climate is humid and hot on the East Coast, with precipitation between 1100 and 3700 mm, whereas outdoor air temperature varies from 18 to 35 °C per year³¹. The west coast is dominated by the tropical climate constituted of dry and hot seasons³¹. These regions are crossed by the precipitation while decreasing between 1500 and 400 mm according to the period of the year. The semi-arid climate covers all the southwest regions of the island with the weak precipitations of 500–700 mm per year, while the climate in the monsoon Northwest region is dominated by a monsoon tropical climate³². Madagascar ranked among the top 10 countries in the world exposed to extreme climate vulnerability events³². Figure 1 shows the study area.

Fig. 1.

Study area. The maps were generated using ArcGIS Pro (version 3.2), developed by Esri.Software link: https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview Base map and spatial processing were performed entirely within ArcGIS Pro using regional administrative boundaries provided by national open-data sources.

Data collection

A comprehensive set of electricity-related data was collected to support the estimation of Scope 2 CO₂ emissions across the 22 regions of Madagascar. The dataset includes information on electricity production, distribution, and consumption, sourced exclusively from JIRAMA (Jiro sy Rano Malagasy), the national electricity and water utility company of Madagascar^33,34. The data were obtained through multiple collection procedures and management systems, combining field measurements, automated monitoring technologies, and administrative databases. A brief overview of the data types, sources and collection methods is presented below.

Types of data collected

The following categories of data were obtained from JIRAMA: Electricity consumption data (residential, commercial, and industrial customers), Electricity production data (generation by thermal, hydro, and other plants), Network distribution and transmission information and Geospatial infrastructure data (substations, lines, network topology).

Data collection procedures and sources

Data acquisition was achieved through five main channels:

• (i)Manual meter readings JIRAMA field agents conduct manual readings of electricity meters installed at consumer premises (residential, commercial, and industrial). These measurements are entered into the company’s billing and monitoring databases and constitute a primary source of validated consumption data.
• (ii)Smart metering systems Smart meters automatically transmit consumption data in real time or at predefined intervals through communication networks (e.g., GSM, power-line communication). This system reduces human error, facilitates fraud detection, and improves the accuracy of demand and supply management.
• (iii)SCADA (supervisory control and data acquisition) systems SCADA platforms continuously monitor and control generation units, substations, and distribution networks. They provide real-time data on electricity production, transmission flows, network status, and technical performance, enabling rapid incident response and operational optimization.
• (iv)Geographic information systems (GIS) GIS tools integrate spatial and technical information to map the national electricity infrastructure. These data help visualize network assets, support planning and maintenance operations, and analyse spatial patterns of electricity distribution and demand.
• (v)Internal databases and management software JIRAMA maintains centralized databases and management software that consolidate production, consumption, and operational data. These systems support data storage, processing, visualization, reporting, and long-term trend analysis. They also provide metadata on network performance and infrastructure characteristics.

Estimation of data

Electricity consumption

The monthly electricity consumption of each individual in the 111 cities studied is estimated based on the readings of electricity meters installed at customers’ homes. In fact, JIRAMA agents regularly go to customers’ homes to manually read electricity meters. The data collected is then recorded for invoicing. Additionally, in areas equipped with smart meters, electricity consumption is measured and transmitted automatically to JIRAMA via communication systems, eliminating the need for frequent manual readings. Monthly consumption is calculated by subtracting the meter reading of the previous month from that of the current month³³. If it is impossible to obtain a reading, JIRAMA can use estimation methods based on the customer’s historical consumption to determine the consumption for the month³⁴. Electricity consumption is multiplied by the current rate to calculate the bill amount. Rates may vary depending on the customer category (residential, commercial, industrial) and the quantity of electricity consumed^33,34. In addition to the cost of electricity, the bill may include taxes, service fees, and other regulatory charges. Invoices are typically sent to customers in physical (paper) form or electronically, depending on customer preferences and available options.

CO₂ emission

CO₂ emissions have been estimated based on electricity consumption, taking into account the electricity mix of these cities between 2018 and 2022. It is considered that CO₂ emissions come only from fossil fuels. The principal electricity fossil fuel sources provided by the JIRAMA are Diesel and Crude oils, which are the only fossil fuel sources used by JIRAMA for thermal electricity generation during the study period. According to JIRAMA Company, between 2015 and 2022, the percentage of electricity consumption generated by diesel oil and crude oil during the same period in each sector was estimated to be between 13.1 and 14.8% and 30.6–42.7%, respectively^33,34. CO₂ emissions factors represent the measured kilograms of CO₂ per megawatt-hour (MWh) of electricity from different fossil fuel sources³⁵. In the case of this study, the carbon dioxide emissions factor of diesel oil is estimated to be 266.76 kg of CO₂, and that of crude oil to 263.88 kg of CO₂, per megawatt-hour³⁵. The CO₂ emission factors used in this study accurately represent the actual conditions in the region. The CO₂ emission per square meter was deduced by downscaling the occupied area of each sector in each studied city with an ArcGIS tool. The area and population of each city are taken from the Madcap web³⁶.

Emission at the census block group

It was spatially downscaled the electricity consumption (MWh) and CO₂ emissions from the city to a block-group (1 km*1 km) scale using energy and infrastructure proxies. To do so requires the geographical boundaries of both the towns and block groups.

The Scope 2 CO₂ emissions () for each city-sector-year are calculated as:

1

where: = Electricity Consumption from source i (in MWh), = Emission Factor of fossil fuel source i (in kg CO₂/MWh), n = number of fossil fuel sources (in this study, diesel and crude oil), = total CO₂ emissions in kg. To convert into tons:

2

Method of data allocation to the 1 km² scale

The spatial disaggregation of electricity consumption and CO₂ emissions was carried out using the following method: (a) Definition of spatial units: Based on the administrative boundaries of the studied cities (urban communes), a regular grid of 1 km × 1 km squares was generated using ArcGIS. (b) Distribution of energy consumption: The total electricity consumption per city (by sector: residential, commercial, industrial, and transport) was distributed across the grid cells using proxy variables: Population density, Built-up area, Land use type (residential, industrial, commercial), obtained from cadastral data or satellite imagery. (c) Weighted allocation model: A multi-criteria weighted allocation model was applied. Each grid cell received a weight calculated as follows:

3

where: = weight assigned to cell i; P_i = population in cell i; B_i= built-up area in cell i; Z_i= land use coefficient (residential = 1, commercial = 0.8, industrial = 0.6, transport = 0.2); α,β,γ(alpha, beta, gamma) = weighting coefficients (set to 1, 1, and 0.5, respectively, in this study); n = total number of cells in the city.

The energy consumption is then allocated proportionally to each cell according to its weight and the CO₂ emissions are calculated from the allocated energy values. Table 2 shows some sources of activity data used for Scope 2 CO₂ emission.

Table 2.

Activity data used for scope 2 CO₂ emission Estimation and their sources.

Category of activity data	Description	Units	Period covered	Use in the study	Data source
Electricity consumption data (monthly)	Consumption per sector (residential, commercial, industrial) per city; obtained from manual meter readings and smart meters	kWh/MWh	2015–2022	Calculation of monthly and annual electricity consumption; basis for CO2 emission estimation	JIRAMA billing databases; manual meter readings; smart metering systems33,34
Electricity production by source	Electricity generated from diesel and crude oil thermal plants	MWh	2015–2022	Determination of electricity mix and emission factor weighting	JIRAMA production reports33,34
Electricity mix (%)	Share of diesel and crude oil generation in total electricity supply (e.g., diesel 13.1–14.8%, crude oil 30.6–42.7%)	%	2015–2022	Allocation of consumption to fossil fuel sources for emission calculation	JIRAMA annual energy balance reports33,34
Emission factors	CO2 emission factors for diesel and crude oil electricity generation	kg CO2/MWh	–	Conversion of electricity consumption to CO2 emission	IPCC: diesel = 266.76; crude oil = 263.88 kg CO2/MWh35
City-level demographic data	Population per city	Number of inhabitants	2015–2022	Spatial disaggregation using population-weighted proxies	Madcap geospatial database36
City area and sectoral land use	Land area, built-up area, cadastral use (residential, industrial, commercial)	km2; m2	–	Allocation of electricity consumption across 1 km2 block groups	Madcap, cadastral maps, satellite imagery36
GIS boundaries	Administrative boundaries of the 111 cities; 1 km × 1 km grid	Spatial layers	–	Creation of grid cells and spatial downscaling	ArcGIS-generated layers
SCADA system data	Real-time generation and distribution system monitoring	Operational parameters	2018–2022	Validation of production/demand balance and infrastructure performance	JIRAMA SCADA center
Smart metering data	Automatic recording of customer consumption	kWh/MWh	2018–2022	Supplementing manual readings; reducing uncertainty	JIRAMA smart metering system
Historical consumption trends	Customer-level historical consumption data used when readings unavailable	kWh	2015–2022	Estimation of missing monthly readings	JIRAMA customer records33,34

Dataset units

In this research, all the electricity emissions were converted to tons of CO₂ (tCO₂ yr^− 1), or sometimes, in kilograms of dioxide carbon (kgCO₂ yr⁻¹); whereas electricity consumption was converted to Megawatt hours per year (MWh yr^− 1), for confrontations purposes of data. In the same way that it were included emissions from the combustion of biomass (almost zero), it was also decided to exclude biogenic CO₂ exchange. Figure 2 shows the workflow for estimating Electricity Consumption.

Fig. 2.

Workflow for estimating electricity consumption and scope 2 CO₂ emissions.

Validation of emission estimates

To assess the robustness and reliability of our emission estimates, it was validated our sectoral results against the international reference dataset provided by the U.S. Energy Information Administration (EIA). For each sector (residential, commercial, industrial), emissions were computed from energy consumption using the IPCC default emission factor for diesel-based electricity generation (266.76 kgCO₂/MWh). Tthese values were compared with emissions derived from our primary dataset collected from the JIRAMA utility (referred to as mywork). As illustrated in Fig. 3, the temporal trends of EIA-derived emissions and JIRAMA-derived emissions are strongly consistent, particularly in the residential sector, where both show a steady and coherent progression from 2008 to 2021. The statistical validation confirms this alignment: the residential sector exhibits a strong Pearson correlation (r = 0.95), a low RMSE (12,724 tCO₂), and a mean relative error of only 5.9%, demonstrating excellent agreement between the two sources. The industrial sector also presents a moderately high correlation (r = 0.66), with deviations mainly explained by structural changes in industrial electricity consumption reporting during 2016–2020 and by temporary operational disruptions in JIRAMA’s thermal plants. The commercial sector shows a weaker correlation (r = 0.25), which is attributable to known inconsistencies in the classification of small businesses in JIRAMA billing data after 2015, resulting in under-representation of commercial loads relative to the EIA aggregated categories. Overall, the high correlation in the major consuming sectors (residential and industrial), combined with consistent long-term temporal behaviour across the datasets, confirms that our JIRAMA-based estimates reliably capture Madagascar’s electricity-related CO₂ emissions. This validation reinforces the suitability of our dataset for national-scale emissions assessment and underscores its applicability for detailed subnational or sectoral analyses.

Fig. 3.

Comparison of annual CO₂ emissions (tCO₂/year) estimated from sectoral EIA data and operational JIRAMA data (“My Work”) for the period 2008–2021.

In Fig. 3, emissions were calculated using the IPCC emission factor (266.76 kgCO₂/MWh) and The figure displays the country’s three main sectors: residential, commercial, and industrial. It is of interest to indicate that the similar trends observed across the datasets validate the coherence of our model and the reliability of the locally collected data.

Results and discussions

Analyse and comparison

Electricity consumption

Electricity consumption shows heterogeneous dynamics across the residential, commercial, and industrial sectors during the study period (2008–2021). The residential sector exhibits a consistent upward trend, reflecting the rapid increase in household connection rates, population growth, and urbanization in Madagascar. Commercial electricity demand remains substantially lower and more volatile, with noticeable drops in 2018 and 2020, likely reflecting economic instability and structural changes within the service sector. Industrial consumption fluctuates but remains the second-largest contributor to national electricity demand, with marked increases between 2015 and 2019 that correspond to periods of industrial expansion and infrastructure growth. Overall, the sectoral profiles demonstrate coherent and expected patterns for a developing electricity system dominated by diesel-based generation. Detailed descriptive statistics for each sector (mean, SE, SD) are provided in Table 3 below.

Table 3.

Descriptive statistics of electricity consumption (MWh) in the residential, commercial, and industrial sectors in Madagascar (2008–2021), including the mean, standard deviation (SD), and standard error (SE) for each sector.

Sector	Mean (MWh)	Standard deviation (SD)	Standard error (SE)
Residential	582001.21	119847.88	32030.69
Commercial	66248.57	19773.78	5284.76
industrial	406620.79	83212.15	22239.38

As shown in Fig. 4, before the lockdown measures due to COVID-19 (2015–2018), in the 111 cities, total electricity consumption varied from 0.0 to 793102.1MWh (mean:18448.3 MWh; Standard Error (SE):8632.7, Standard Deviation (SD):90541.5). For the residential sector, it varied from 0.0 to 392069.1MWh (mean: 9374.9; SE: 4266.7, SD: 44749.7); for the commercial sector, it was from 0.0 to 40590.3 MWh (mean: 1025.7; SE: 394.3, SD: 4136.0). Moreover, in the industrial sector, electricity consumption varied from 0.0 to 357950.1MWh (mean: 7848.1; SE: 3973.3, SD: 41672.7). Finally, on-road sector, electricity consumption varied from 0.0 to 2492.2 MWh (mean: 75.6 MWh; SE: 26.2, SD: 275.3). These results showed that, before the lockdown measures due to COVID-19, the average of electricity consumption in the four sectors was in 51.2%, 5.6%, 42.8%, and 0.4%; in Residential, Commercial, Industrial, and on road, respectively. It was deduced that Electricity was the most consumed in the residential sector and the least in the road sector between 2015 and 2018, in the 111 cities located in the Indian Ocean. In addition, during the government-imposed restrictions linked to COVID-19 (2019–2020); total electricity consumption varied from 0.0 to 933403.6 MWh (mean: 20395.4 MWh; SE: 9708.4, SD:101823.4). For the residential sector, it varied from 0.0 to 446303.7 MWh (mean: 10262.7 MWh; SE: 4702.6, SD: 49321.2); for the commercial sector, it was from 0.0 to 44031.3 MWh (mean: 1136.7 MWh; MSE: 432.9, SD: 4540.6). On the other hand, in the industrial sector, electricity consumption varied from 0.0 to 441256.5M Wh (mean: 8924.5 MWh; SE: 4576.0, SD: 47993.8). Finally, on-road sector, electricity consumption varied from 0.0 to 1812.0 MWh (mean: 70.1 MWh; SE: 21.9, SD: 230.3). These results showed that, during the government-imposed restrictions linked to COVID-19 (between 2019 and 2020), the average of electricity consumption in the four sectors was of 50.3%, 5.6%, 43.7% and 0.3% in Residential, Commercial, Industrial, and on road sectors, respectively. Despite COVID-19 lockdown restrictions, the average total electricity consumption slightly increased, which may be explained by a shift in energy use from commercial and transportation sectors to residential due to remote work, home schooling, and confinement. The resilience of the industrial sector, with an increase in mean consumption (from 7848.1 to 8924.5 MWh), suggests that industrial activities were not fully halted during the pandemic or were prioritized to maintain essential supply chains (such as food, health, and energy). The decrease in the on-road sector, though minor (75.6 → 70.1 MWh), aligns with travel restrictions. This moderate change may also indicate limited electrification of the transportation sector in Madagascar or a partial recovery during the lockdown period. It was deduced that even during the government-imposed restrictions linked to COVID-19, electricity was the most consumed in the residential sector and the least on-road. Finally, after the lockdown period (2021–2022), total electricity consumption varied from 0.0 to 952457.5 MWh (mean: 20928.8 MWh; SE: 9924.7, SD: 104091.1). For the residential sector, it varied from 0.0 to 504811.1 MWh (mean: 11719.6 MWh; SE: 5439.1, SD: 56785.2); for the commercial sector, it was from 0.0 to 46146.8 MWh (mean: 1146.3 MWh; SE: 454.6, SD: 4768.3). In addition, in the industrial sector, electricity consumption varied from 0.0 to 399733.9 MWh (mean: 8068.7; SE: 42926.1, SD: 4092.8); and On-road sector, electricity consumption varied from 0.0 to 1767.3 MWh (mean: 71.6 MWh; SE:21.3; SD:222.7). These results showed that, after the lockdown period, the average of electricity consumption in the four sectors was in 55.8%, 5.4%, 38.4%, and 0.3% in Residential, Commercial, Industrial and on road, respectively. In view of all these results, it was deduced that electricity was the most consumed in the residential sector, and the least on the road between 2021 and 2022, in the 111 cities located in the Indian Ocean. During the three studied periods, over 50% of total electricity was consumed in the residential sector. The observation that electricity consumption was highest in the residential sector and lowest in the on-road sector in these Indian Ocean cities from 2015 to 2022 is a reflection of several key factors. The COVID-19 pandemic led to a significant increase in time spent at home due to lockdowns and remote work/study arrangements^37–42. This likely may result in higher electricity usage for home appliances, heating, cooling and electronics. In addition, the urbanization trend in many cities might have contributed to higher residential electricity consumption due to increased household numbers and the adoption of more electrical devices. Moreover, many cities in the Madagascar region experience warm climates, leading to higher use of air conditioning and cooling systems, which are significant electricity consumers. The low electricity consumption in the on-road sector suggests that the penetration of electric vehicles in these cities might still be low. Traditional vehicles predominantly use fossil fuels, which isn’t reflected in electricity consumption data. There are no traffic lights and street lights in most intersections and highways as in developed countries. This may also be one of the causes of the reduction in electricity consumption in this sector. The results support the conclusions of Gurney et al.⁴⁰, who identified the dominant role of the residential sector in per capita electricity consumption, particularly in developing countries. Similarly, these findings are consistent with Le-Quere et al.⁴¹, who reported a global decline in transport-related emissions and an increase in residential electricity use during lockdowns.

Fig. 4.

Electricity consumption of 111 cities in Madagascar grouped into three periods (2015–2018; 2019–2020; and 2021–2022), across residential, commercial, industrial, and on-road sectors (unit: MWh). The maps were generated using ArcGIS Pro (version 3.2), developed by Esri.Software link: https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview Base map and spatial processing were performed entirely within ArcGIS Pro using regional administrative boundaries provided by national open-data sources.

Figure 5 shows electricity consumption per capita in the different studied cities. Before the lockdown measures due to COVID-19 (2015–2018), It is important to notice that total electricity consumption varied from 0.0 to 842.7 MWh/capita (Mean:42.5 MWh; SE:11.7, SD:122.4). In the residential sector, the electricity consumption varied from 0.0 to 415.1 MWh/capita (mean: 22.7; SE:5.9, SD:61.8); In the commercial sector, it was from 0.0 to 38.8 MWh/capita (mean:3.4 MWh; SE: 0.6, SD:6.7). However, in the industrial sector, electricity consumption varied from 0.0 to 401.3 MWh/capita (mean:16.1; SE:5.4, SD:56.6). On the other hand, on-road sector, the electricity consumption varied from 0.0 to 3.4 MWh/capita (mean:0.25 MWh; SE:0.05, SD:0.58). It was deduced, on the basis of these results, that before COVID electricity consumption per capita was the most significant in the residential sector (53.6%), and the least important on-road sector (0.59%). In addition, During the government-imposed restrictions linked to COVID-19 (2019–2020), total electricity consumption per capita was expected to vary between 0.0 and 780.9 MWh (mean:44.1 MWh; SE:11.6, SD:121.6). In the residential sector, it varied from 0.0 to 391.7 MWh (mean: 23.7 MWh; SE:5.9, SD:61.8). In the commercial sector, it was noticed that electricity consumption per capita was between 0.0 and 35.6 MWh (mean:3.5 MWh; SE: 0.6, SD:6.9). However, during the same period in the industrial sector, electricity consumption varied from 0.0 to 360.1 MWh/Capita (mean: 16.5 MWh; SE:5.3, SD:55.9). Finally, On-road, electricity consumption varied from 0.0 to 3.4 MWh/Capita (mean:0.23 MWh; SE:0.05; SD:0.56). These results showed that, during COVID, 91.5% of total electricity per capita was consumed in residential and industrial sectors, with a frequency of 53.8% and 37.7%, respectively. Whereas the remaining 8.5% of total electricity was used in commercial and on-road sectors. Finally, during the period recognized as following the COVID-19-related lockdown (2021–2022), it was important to see that the total electricity consumption varied from 0.0 to 802.1 MWh/capita (mean:45.3 MWh; SE:11.8, SD:124.5). In the residential sector, it varied from 0.0 to 467.4 MWh/capita (mean: 25.4 MWh; SE:6.7, SD:70.5). In the commercial sector, it was expected varied between 0.0 to 34.9 MWh/capita (mean:3.1 MWh; SE: 0.61, SD:6.40). In the industrial sector, electricity consumption varied from 0.0 to 3.3.9 MWh/capita (mean: 15.1; SE:4.7, SD:49.9). However, in on-road sector electricity consumption varied from 0.0 to 3.9 MWh/capita (mean:0.23 MWh; SE:0.05, SD:0.52).

Fig. 5.

Electricity consumption per capita of 111 cities grouped in three periods (2015–2018; 2019–2020; and 2021–2022), in residential, commercial, industrial, and on-road sectors (unit: kWh). The maps were generated using ArcGIS Pro (version 3.2), developed by Esri.Software link: https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview Base map and spatial processing were performed entirely within ArcGIS Pro using regional administrative boundaries provided by national open-data sources.

These results showed that, between 2021 and 2022, the total electricity consumption per capita was estimated to be 57.9% in the residential sector and 34.4% in the Industrial sector. Only 0.53% of total electricity was used in the On-road sector. Our findings are consistent with those of Gurney et al.³⁵, who demonstrated that more than 40% of electricity consumption per capita was attributed to the residential sector. Industrial Sector (34.4%), this is not by chance because the industry consumes a large amount of electricity for machines, production systems, and other heavy equipment. In addition, the period from 2021 to 2022 could also mark a recovery of industrial activities after the disruptions caused by the pandemic, leading to increased electricity consumption in this sector. These results are not very surprising because the increase in urban population leads to an increase in residential consumption. In addition, changes in lifestyle, such as remote working, can modify electricity consumption habits. In the industrial sector, the high consumption of electricity highlights the importance of energy efficiency to reduce production costs and carbon emissions. Therefore, the transition to renewable energy sources and the improvement of energy efficiency in the residential and industrial sectors are crucial for sustainability.

It was deduced that during the three periods, the residential sector showed the strongest electricity demand. It was noticed that electricity consumption in the residential sector in these different cities is largely influenced by the need for air conditioning and refrigeration, access and reliability of the electricity grid, as well as the cost of electricity. The increasing adoption of renewable energy also plays an important role in managing energy consumption in these regions.

CO₂ emission

To complement the spatial assessment of sectoral CO₂ emissions, it was computed the mean, standard deviation (SD), and standard error (SE) of emissions across the industrial, commercial, and residential sectors for the years 2018, 2019, and 2021. Results indicate that the residential sector exhibits the highest average emissions (mean ≈ 1.43 × 10⁶ tCO₂), followed closely by the industrial sector (mean ≈ 1.13 × 10⁶ tCO₂), while the commercial sector contributes substantially less (mean ≈ 1.51 × 10⁵ tCO₂). The large SD values observed for industrial and residential emissions (≈ 6.06 × 10⁶ and 7.00 × 10⁶ tCO₂, respectively) reflect the strong spatial heterogeneity between major urban centres—particularly Antananarivo and Toamasina—and smaller rural districts. The SE values confirm that uncertainty remains relatively small compared to the magnitude of emissions, supporting the robustness and consistency of the sectoral CO₂ estimates. These results reinforce the relevance of our sector-based modelling approach for regional carbon accounting in Madagascar (see Table 4).

Table 4.

Summary statistics of sectoral CO₂ emissions (2015–2022).

Sector	Mean (tCO2)	Standard deviation (SD)	Standard error (SE)
Industrial	1134340.60	6063067.13	333760.86
Commercial	150979.72	619344.80	34093.81
Residential	1431248.42	7005264.92	385627.14

Figure 6 shows CO₂ emissions coming from electricity consumption (scope2) in the different cities. Prior to the national lockdown implemented in response to COVID-19 (2015–2018), It is important to notice that total scope2 emission varied from 0.0 to 93165.69tCO₂ (mean: 2167.13tCO2; SE: 1014.09, SD: 10635.91). In the residential sector, the scope2 emission was expected to be between 0.0 to 46056.35tCO₂ (mean: 1101.27tCO₂; SE: 501.21, SD: 5256.75). In the commercial sector, it was from 0.0 to 4768.14tCO₂ (mean: 120.49tCO2; SE: 46.32, SD: 485.85). However, in the industrial sector, scope emission varied from 0.0 to 42048.43tCO₂ (mean: 933.66tCO2; SE: 466.74, SD: 4895.29). On the other hand, on the road, the scope2 emission varied between 0.0 to 292.76tCO₂ (mean: 9.81tCO2; SE: 3.07, SD: 32.25). It was deduced, on the basis of these results that between 2015 and 2018, the scope2 emission was the most significant in the residential sector (50.8%) and the least important on-road sector (0.4%). In addition, during the government-imposed restrictions linked to COVID-19 (2019–2020), total scope2 emission was expected to vary between 0.0 and 130719.4tCO₂ (mean: 2856.2tCO2; SE: 1359.6, SD: 14259.9). In the residential sector, it varied from 0.0 to 62503.1tCO₂ (mean: 1437.3; SE: 658.5, SD: 6907.2). In the commercial sector, it was noticed that scope2 emission were between 0.0 and 6166.4tCO₂ (mean: 159.19; SE: 60.6, SD:635.9). However, during the same period in the industrial sector, scope2 emission varied from 0.0 to 61796.2tCO₂ (mean: 1249.8tCO₂; SE: 640.8, SD: 6721.3) and, in on-road, scope2 emission varied from 0.0 to 253.7tCO₂ (mean:9.8tCO2; SE:3.1, SD:32.2). These results showed that, during the government-imposed restrictions linked to COVID-19, the 94.08% of total CO₂ emission was generated in residential and industrial sectors with a distribution of 50.32% and 43.76%, respectively. Whereas the remaining (5.92%), was provided in commercial and on-road sectors. Finally, between 2021 and 2022, it was important to see that the total scope2 emission varied from 0.0 to 143954.43tCO₂ (mean:3163.2tCO2; SE:1500.1, SD:15732.3). In the residential sector, it varied from 0.0 to 76297.1tCO₂ (mean: 1755.2tCO2; SE:814.7, SD:8544.7). In the commercial sector, it was expected to be between 0.0 to 6974.6tCO₂ (mean: 173.2tCO₂; SE:68.7, SD:720.6). In the industrial sector, scope2 emission varied from 0.0 to 60415.5tCO₂ (mean: 1219.5; SE:618.6, SD:6487.8). However, On-road sector, emission varied from 0.0 to 267.12tCO₂ (mean:10.7; SE:3.2, SD:33.5). These results showed that between 2021 and 2022 the total scope2 emission was estimated to be 55.56% in the residential sector; 5.48% in the commercial sector; 38.60% in the Industrial sector; and Only 0.33% on-road sector. Residential and industrial sectors are recognized as high-emitting sectors in these 111 Cities, as more than 94% of the total emissions come from these sectors. It is noted that the increase in urban population and remote working can increase residential electricity consumption, leading to higher Scope 2 emissions. In addition, energy consumption behaviours directly influence emissions in each sector. Energy efficiency awareness campaigns can reduce these emissions. Industries must strike a balance between production and emissions management by adopting cleaner and more efficient technologies. Promoting energy efficiency in residential and industrial sectors is essential to reducing overall Scope 2 emissions.

Fig. 6.

Scope2-emission of 111 cities grouped in three periods (2015–2018; 2019–2020; and 2021–2022), in residential, commercial, industrial, and on-road sectors (unit: kgCO₂). The maps were generated using ArcGIS Pro (version 3.2), developed by Esri.Software link: https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview Base map and spatial processing were performed entirely within ArcGIS Pro using regional administrative boundaries provided by national open-data sources.

Residential emissions increased steadily: from 50.8% (2015–2018) to 55.56% (2021–2022), reflecting not only population growth and urbanization but also the shift toward remote work during and after COVID-19 lockdowns, which significantly raised household electricity demand. This trend confirms the findings of Le Quere et al.⁴¹, who observed a global reallocation of emissions from industrial to residential areas during the pandemic. The industrial sector, while slightly decreasing in relative percentage (from 43.76% during COVID-19 to 38.6% post-COVID), maintained high absolute emissions, suggesting that economic recovery and increased production activities sustained its carbon footprint. This is in line with Sovacool⁴², who highlighted the persistence of industrial energy intensity in developing and emerging cities. In contrast, commercial and on-road sectors contributed marginally (each under 6%), with no significant variation over time, possibly due to sustained restrictions and reduced activity in those sectors.

A city-wise comparison of Scope 2 CO₂ emissions reveals substantial spatial and temporal disparities across Madagascar between the three study periods (2015–2018, 2019–2021, and 2021–2022). Results show that the capital cities—Antananarivo Renivohitra, Antananarivo Sud, and Antananarivo Nord—consistently recorded the highest emissions in all sectors, with mean CO₂ values exceeding 20–40 MtCO₂ in each period (SD > 8.5 Mt; SE < 2.1 Mt), reflecting their dense population, higher electricity demand, and intensive commercial and industrial activities. Medium-sized urban areas, such as Mahajanga, Toamasina Rural, and Taolagnaro, show moderate but steadily increasing emissions, with mean increases of 12–18% between 2019 and 2021 and 2021–2022, particularly in the residential sector. Conversely, smaller and rural cities—such as Ikongo, Andramasina, Vohipeno, Midongy-Sud, and several communes of the Melaky and Atsimo-Andrefana regions—exhibited consistently low emissions (mean < 0.1 MtCO₂), with limited temporal variation (SD < 0.03 Mt; SE ≈ 0). Statistical tests confirm a significant temporal shift in emissions between 2015 and 2018 and 2021–2022 (paired t-test, P < 0.05), while correlation analysis shows strong consistency in spatial patterns across periods (Pearson R ranging from 0.82 to 0.94), indicating that cities with high emissions in earlier years remained the largest contributors in later periods. Overall, these findings highlight clear urban-rural contrasts, persistent emission hotspots in the main metropolitan areas, and the influence of demographic growth and post-COVID economic recovery on sectoral electricity demand.

Figure 7 shows the CO₂ emission per capita coming from electricity consumption (scope2) in the different cities. Before the lockdown measures due to COVID-19, it is important to notice that total scope2 emission per capita varied from 0.0 to 98.9 kgCO₂ (mean:4.9 kgCO₂; SE:1.3, SD:14.4). In the residential sector, the scope2 emission was expected to be between 0.0 to 48.75 kgCO₂/capita (mean: 2.67 kgCO₂; SE:0.69, SD:7.26). In the commercial sector, it was from 0.0 to 4.56 kgCO₂/capita (mean:0.39 kgCO₂; SE: 0.07, SD:0.80). However, in the industrial sector, scope2 emission varied from 0.0 to 47.14 kgCO₂/capita (mean:1.88 kgCO₂; SE:0.63, SD:6.63). On the other hand, in the on-road sector, the scope2 emission varied between 0.0 to 0.40 kgCO₂/capita (mean:0.03kgCO₂; SE:0.006, SD:0.06). It was deduced, on the basis of these results that between 2015 and 2018, the scope2 emission was the most significant in the residential sector (54.54%) and the least important on-road sector (0.59%). In addition, during the COVID-19-induced lockdown period (2019–2020), total scope emission was expected to vary between 0.0 and 109.36 kgCO₂/capita (mean:6.16 kgCO₂; SE:1.62, SD:17.03). In the residential sector, it varied from 0.0 to 54.85 kgCO₂ (mean: 3.32 kgCO₂; SE:0.82, SD:8.65). In the commercial sector, it was noticed that scope2 emission per capita was between 0.0 and 4.99 kgCO₂(mean:0.48 kgCO₂; SE: 0.09, SD:0.96). However, during the same period, in the industrial sector, scope2 emission varied from 0.0 to 50.42 kgCO₂/capita (mean: 2.32 kgCO₂; SE:0.74, SD:7.84) and, in the on-road sector, scope2 emission varied from 0.0 to 0.47 kgCO₂(mean:0.03 kgCO₂; SE:0.07, SD:0.007). These results showed that during the COVID-19-induced lockdown period, 91.65% of total CO₂ emission per capita was generated in residential and industrial sectors, with a frequency of distribution of 53.95% and 37.70%, respectively. Whereas the remaining (the 8.35%), was provided in commercial and on road sectors. Finally, between 2021 and 2022, It was important to see that the total scope2 emission per capita varied from 0.0 to 122.22 kgCO₂(mean: 6.84 kgCO₂; SE:1.79, SD:18.82). In the residential sector, it varied from 0.0 to 70.64 kgCO₂ (mean:3.83 kgCO₂; SE:1.06, SD:10.65); in the commercial sector, it was expected to be between 0.0 to 5.27kgCO₂(mean: 0.48 kgCO₂; SE:0.09, SD:0.96). In the industrial sector, scope2 emissions varied from 0.0 to 45.94 kgCO₂(mean: 2.28 kgCO₂; SE:0.71, SD:7.54). However, in the on-road sector, emission varied from 0.0 to 0.59 kgCO₂ (mean:0.03 kgCO₂; SE:0.007; SD:0.07). These results showed that between 2021 and 2022, the total scope2 emission was estimated to be 57.80% in the residential sector; 7.25% in the commercial sector; 34.41% in the Industrial sector, and only 0.52% in the on-road sector. This suggests that, even before the pandemic, electricity use in households was already a dominant source of emissions—possibly linked to growing urban populations, increased use of household appliances, and limited energy efficiency measures in place.

Fig. 7.

Scope2 emission per Capita of 111 cities grouped in three periods (2015–2018; 2019–2020; and 2021–2022) in residential, commercial, industrial, and on-road sectors (unit: kgCO₂). The maps were generated using ArcGIS Pro (version 3.2), developed by Esri.Software link: https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview Base map and spatial processing were performed entirely within ArcGIS Pro using regional administrative boundaries provided by national open-data sources.

The COVID-19 lockdown period (2019–2020) saw a noticeable increase in emissions per capita, with the mean rising to 6.16 kgCO₂. This uptick is primarily attributed to the rise in residential emissions, which now represent 53.95% of the total, likely due to prolonged stays at home, teleworking, and increased demand for heating, cooling, and electronic devices. Interestingly, the industrial share also increased to 37.70%, suggesting that despite economic slowdowns, certain industrial activities—especially essential goods and digital infrastructure—continued or even intensified. The slight increase in commercial emissions per capita during this time may reflect energy consumption from facilities that remained open (e.g., hospitals, supermarkets). These observations align with studies showing a restructuring of energy consumption patterns during lockdowns EIA⁴³. In the post-lockdown period (2021–2022), emissions per capita continued to rise, averaging 6.84 kgCO₂, indicating a sustained high demand for electricity, particularly in the residential sector (now 57.80% of total per capita emissions). This could be tied to the permanent adoption of hybrid work models, population growth in cities, and potential rebound effects as economic activity resumed. Although commercial sector emissions grew slightly (to 7.25%), industrial emissions saw a relative decline in share (34.41%), possibly suggesting improved energy efficiency or changes in industrial production processes post-COVID.

Across all three periods, the on-road sector consistently contributed marginally to per capita Scope 2 emissions (≤ 0.59%). This reaffirms that Scope 2 emissions from transport are generally low unless significant electrification of the transport fleet is in place—something still limited in most of the studied cities.

The high share of Scope 2 emissions in the residential sector highlights the need to improve the energy efficiency of homes. Initiatives such as energy retrofit programs and the adoption of smart home technologies can help. The industrial sector also contributes significantly to Scope 2 emissions, highlighting the importance of adopting greener production processes and renewable energy sources in the 111 cities. The very low percentage of Scope 2 emissions in the road transport sector indicates that efforts to encourage the adoption of electric vehicles and develop charging infrastructure need to be stepped up.

Electricity consumption and CO₂ emission from region to block groups

In this section, each city was divided into several block groups (BG) (1 km*1 km). This allows us to estimate the CO₂ emission at a small scale and with good accuracy, and the results are given in Fig. 8.

Fig. 8.

Electricity consumption (left) & CO₂ emission (right) per capita at a block group scale, of 111 cities grouped in three periods (2015–2018; 2019–2020; and 2021–2022), (units: kWh, and kgCO₂). The maps were generated using ArcGIS Pro (version 3.2), developed by Esri.Software link: https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview Base map and spatial processing were performed entirely within ArcGIS Pro using regional administrative boundaries provided by national open-data sources.

Figure 8 shows electricity consumption and scope2 emission per capita at a block group(1 km*1 km) level. Between 2015 and 2018, the total electricity consumption was expected to be between 0.0 and 1.57 kWh (Mean: 0.017 kWh; SE:0.014, SD:0.150). While the total scope2 emission varied between 0.0 and 0.18 kgCO₂ (mean:0.002 kgCO₂; SE:0.002, SD:0.017). In addition, between 2019 and 2020, the total electricity consumption varied from 0.0 to 1.64 kWh (Mean: 0.018 kWh; SE:0.015, SD:0.157), whereas the total scope2 emission varied between 0.0 and 0.23 kgCO₂(mean:0.003; SE:0.002, SD:0.022). Finally, in the post-lockdown phase after COVID-19, from 2021 to 2022, the total electricity consumption was between 0.0 and 1.74 kWh (Mean: 0.019 kWh; SE:0.015, SD:0.167). Whereas the total scope2 emission varied between 0.0 and 0.26 kgCO₂(mean:0.003 kgCO₂; SE:0.002, SD:0.025). It was concluded that in the Indian Ocean Cities, the sector in which there is the highest electricity consumption and CO₂ emission is the residential sector. These results are uniform during the rainy and dry seasons in these different sectors.

Between 2015 and 2018, electricity consumption per capita ranged from 0.0 to 1.57 kWh (mean: 0.017 kWh), with corresponding scope 2 emissions between 0.0 and 0.18 kgCO₂ (mean: 0.002 kgCO₂). During the lockdown period, there was a slight increase in both electricity consumption (up to 1.64 kWh; mean: 0.018 kWh) and emissions (up to 0.23 kgCO₂; mean: 0.003 kgCO₂). This increase may be attributed to a shift in energy demand from commercial and industrial activities toward the residential sector, as lockdowns confined populations to their homes, intensifying domestic electricity use. This finding aligns with other studies highlighting increased residential electricity demand during lockdowns IEA^43,44. In the post-lockdown period (2021–2022), electricity consumption continued its upward trajectory (up to 1.74 kWh; mean: 0.019 kWh), as did scope 2 emissions (up to 0.26 kgCO₂; mean: 0.003 kgCO₂). These results suggest not only a rebound effect after the restrictions were lifted, but potentially the persistence of new energy consumption patterns, such as continued remote working or increased use of digital technologies in households.

Across all periods, the residential sector consistently emerged as the dominant contributor to both electricity consumption and related emissions, a pattern that remained stable across seasons (rainy and dry). This sectoral dominance is likely linked to urban demographic density, household energy practices, and possibly inefficient appliances or lack of access to clean energy technologies. These observations highlight the critical need for targeted energy efficiency policies and behavioural interventions in residential areas. In summary, this temporal and sectoral comparison reveals a gradual but steady increase in electricity consumption and scope 2 emissions per capita over time, particularly within residential zones. The COVID-19 lockdown appears to have acted as a structural shift rather than a temporary disruption, calling for long-term planning to decarbonize the residential electricity demand in island urban environments.

Relative difference

Figure 9 shows the frequency distribution of the paired relative difference(RD) of electricity consumption between (2015–2018) and (2019–2020), between (2021–2022) and (2019–2020).

Fig. 9.

Frequency distribution of the paired relative difference(RD) of electricity consumption. a Total; b Residential sector; c Commercial sector; d Industrial sector; e on road sector. RD calculated as (yr1 – yr2)/((yr1 + yr2)/2).

In Fig. 9a1, total electricity consumption (EC) has a median relative difference (MRD) of − 11.5% (mean: − 14.78%; SE: 1.94%, SD:20.50%) with electricity consumption during the COVID-19-induced lockdown period exceeding, on average, the electricity consumption before the COVID-19-induced lockdown period. In addition, as seen in Fig. 9a2, the total electricity consumption has a median relative difference of 9.0% (mean:8.9%; SE:1.9%, SD:20.3%) with electricity consumption following the COVID-19-related lockdown, exceeding, on average, the electricity consumption during the COVID-19-induced lockdown period. In the residential sector (Fig. 9b1, b2), electricity consumption has a median relative difference of − 13.0% (mean: − 18.3%; SE:3.2%, SD:33.5%), see Fig. 9b1 whereas it was 10.5% (mean:9.03%; SE:2.5%, SD:26.4%) as shown in Fig. 9b2. These results showed that, in the residential sector, electricity consumption continuously increased between 2015 and 2022.

In the commercial sector (Fig. 9C1, C2), the median relative difference(MRD) of EC was expected to be − 9.0% (mean: − 11.0%; SE:3.4%, SD:35.9%) with electricity consumption (EC) higher during the COVID-19-induced lockdown period, than before the COVID-19-induced lockdown period. However, as shown in Fig. 9c2, the MRD was − 9.0% (mean: − 20.9%; SE:4.9%, SD:51.6%). These results explained that in the commercial sector, the EC was the most significant in 2019–2020. In the industrial sector (Fig. 9d1), the MRD of EC was − 45% (mean: − 45.1%; SE:5.1%, SD:53.1%). This shows that EC is more important during the period (2019–2020) than during 2015–2018. In addition, as shown in Fig. 9d2 the MRD was − 5% (mean: − 4.7%; SE:4.2%, SD:45.1%) with electricity consumption during the national lockdown due to the COVID-19 pandemic exceeding, on average, the electricity consumption after COVID. In the on-road sector (see Fig. 9e1, e2), the MRD were 4% (mean:4.2%; SE:7.1%, SD:74.4%) and 41% (mean:40.9%; SE:8.0%, SD:84.6%) for the periods from (2015–2018) to (2019–2020) and between (2021–2022) and (2019–2020), respectively. It was deduced that the on-road sector, EC was the least significant during the national lockdown due to the COVID-19 pandemic period (2019–2020) in these cities. The primary reason for the reduced electricity consumption in the on-road sector during the COVID period is the dramatic change in human behaviour and mobility patterns. Lockdowns, social distancing, and remote working conditions led to a significant decrease in travel and transportation needs.

The stage of development and adoption of electric vehicles played a crucial role. With a low number of electric vehicles on the roads and limited charging infrastructure, the overall electricity consumption in the on-road sector remained minimal. Economic factors, including reduced consumer spending and lower fuel prices, discouraged the adoption and use of electric vehicles, contributing to the low electricity consumption in the on-road sector.

As shown in Fig. 10, total CO₂ emission in the 111 cities has an MRD of − 28% (mean: − 31%; SE: 1.9%, SD: 20.8%) with the period during the national lockdown due to the COVID-19 pandemic exceeding, on average the period before the COVID-19-induced lockdown period. While the MRD between the after and during COVID periods (see Fig. 10a2) was 16% (mean:16.3%; SE:1.9%, SD:20. 2%). These results showed that CO₂ emissions were the most significant after the lockdown period.

Fig. 10.

Frequency distribution of the paired relative difference (RD) of CO₂ emission. a Total; b Residential sector; c Commercial sector; d Industrial sector; e on road sector. RD calculated as (yr1 – yr2)/((yr1 + yr2)/2).

For the residential sector (see Fig. 10b1), the median relative difference was − 30% (mean: − 34.2%; SE: 3%, SD: 31.6%); whereas in Fig. 10b2, the MRD was 18.5%(mean: − 15.5%; SE: 3.1%, SD: 31.9%). These results also showed that CO₂ emissions were the most significant after the lockdown period in the residential sector. On the other hand, in the commercial sector (Fig. 10c1, c2), the MRD was expected to be − 25%(mean: − 27.1%; SE: 3.3%, SD: 34.9%), see (Fig. 10c1), and to be of − 2.5% (mean: − 13.8%; SE: 4.9%, SD: 52.3%), see (Fig. 10c2). These results showed that, in the commercial sector, CO₂ emissions were the most significant during the government-imposed restrictions linked to COVID-19.

It was very important to notice that in the industrial sector, the MRD of Scope2 CO₂ emissions were − 43% (mean: − 43.1%; SE: 5.8%, SD: 61.3%), see Fig. 10d1, while in the Fig. 10d2, the MRD was 2% (mean: 2.1%; SE: 4.3%, SD: 45.6%). It was deduced that the in the industrial sector, CO₂ emissions were the most significant after the COVID period (2021–2022).

Finally, the on-road sector showed the MRD of − 10% (mean: − 10.1%; SE: 7.1%, SD: 74.8%), see Fig. 10e1. While in Fig. 8e2 the MRD was 47% (mean: 46.8%; SE: 8.0%, SD: 84.3%). It was deduced that the on-road sector CO₂ emissions were the most significant after the lockdown period. Globally, residential sectors saw substantial increases in electricity demand, especially in developed countries, where telecommuting and entertainment consumption surged. For example, in Germany, residential electricity consumption rose by 5% during the lockdown BDEW⁴⁰, echoing the 10.5% increase in the residential sector in your data (Fig. 10b2).

The rebound in total electricity consumption after 2020, as observed in Fig. 9a2, is also consistent with global patterns. As restrictions eased, economic activities resumed, and electricity demand returned to pre-pandemic levels, with industries reopening and commercial activities resuming. This phenomenon has been observed in multiple regions, where electricity demand showed a V-shaped recovery post-lockdown, driven by the return of economic activity³⁹.

Discussion and comparison

The findings presented in Figs. 9 and 10 reveal complex patterns in electricity consumption and CO₂ emissions across different sectors during the pandemic. These trends are consistent with global observations, where significant reductions in both energy consumption and emissions were observed due to the abrupt changes in human behaviour, industrial activities, and mobility during the COVID-19 lockdown periods.

As shown in Fig. 10e1 (MRD: − 10%), the Mean Relative Difference (MRD) of − 10% indicates a reduction in the variable being measured compared to a reference period. In this case, it refers to a reduction in activity or emissions during the COVID-19-induced lockdown period. The mean of − 10.1% with a standard error (SE) of 7.1% and a standard deviation (SD) of 74.8%. It suggests considerable variability in the data, indicating that the reduction was not uniform across all observations. The global trend in electricity consumption during the COVID-19 pandemic showed similar fluctuations across countries, with residential consumption increasing, and commercial and industrial consumption decreasing. A study by the International Energy Agency⁴³ reported that residential electricity use surged by 10–20% in several countries due to increased time spent at home, teleworking, and reliance on digital devices. This trend was evident in our analysis, where residential electricity consumption showed a continuous increase between 2015 and 2022, aligning with the global shift towards home-centric lifestyles during the lockdown.

As shown in Fig. 10e2 (MRD: 47%), an MRD of 47% indicates a significant increase in the variable compared to the reference period, suggesting a post-COVID surge in activity or emissions. Whereas the mean of 46.8%, with an SE of 8.0% and an SD of 84.3%, again shows high variability, indicating substantial differences across different observations or locations. During the COVID-19-induced lockdown period (Fig. 10e1), the negative MRD aligns with the expectations of reduced on-road activity due to lockdowns, travel restrictions and increased remote working, leading to lower CO₂ emissions. The large SD (74.8%) suggests that the impact of COVID-19 on-road activity varied greatly, possibly due to differences in local restrictions, compliance, and essential travel needs. Post-COVID (Fig. 10e2), the positive MRD of 47% indicates a significant rebound in on-road activity after the lockdown period restrictions were lifted, leading to increased CO₂ emissions. In addition, the increase in CO₂ emissions can be attributed to a return to pre-pandemic travel habits, resumption of economic activities and potentially higher reliance on personal vehicles over public transport. The data indicate that CO₂ emissions in the on-road sector became more significant after the lockdown period, as evidenced by the substantial positive MRD. This suggests a higher environmental impact from the sector during the recovery period. Factors contributing to this increase could include a preference for personal vehicles to avoid public transport, increased economic activities requiring transportation, and delayed vehicle maintenance, during the pandemic leading to less efficient emissions post-pandemic⁴⁵. The findings from this study are consistent with global trends observed during the COVID-19 pandemic, highlighting how abrupt changes in human behaviour and economic activity led to significant reductions in energy consumption and emissions. The theories of environmental economics (e.g., Jeavons Paradox⁴⁶ and behavioural energy economics provide a useful framework to explain these trends. While the pandemic forced reductions in consumption, the post-lockdown rebound reflects the inherent resilience of economic systems and the complexities of decarbonizing energy sectors. This study contributes to the growing body of literature suggesting that while short-term interventions can drive down emissions, long-term reductions will require structural changes, including technological innovations, policy support, and shifts in consumer behaviour^45,47.

Statistical analysis

In this section, it was carried out an in-depth statistical analysis to get precise information on energy consumption and scope2 CO₂ emission between three periods: before the COVID-19-induced lockdown period (201–2018); during the national lockdown due to the COVID-19 pandemic (2019–2020), and after the lockdown period (2021–2022). The purpose is to evaluate whether the difference is significant or not. The data is quantitative, with the different variables mostly dependent, the statistical test applied in this section is the t-test. All the statistical analysis was carried out with a 95% confidence level (CL), considered level of significance equal to 5%. Applying a t-test in this step is very important because it will allow us to better understand whether the difference between energy and emission is important. With the SPSS tool, the interval of significance can be freely chosen^37,38.

Statistical analysis of scope2 CO₂ emission

A detailed analysis of the scope2 CO₂ emission was carried out to understand whether the difference is important or not between the two periods, and the results are given in Table 5.

Table 5.

Results for t-test scope2 emission difference (unit: kgCO₂) in residential, Industrial, commercial and on-road sectors (B, before the lockdown measures due to COVID-19; D, during the government-imposed restrictions linked to COVID-19; and A, after the lockdown period).

Paired samples test
CO2 difference	Paired differences					t	df	Significance
	Mean	Std. deviation	Std. error mean	95% Confidence interval of the difference				One-sided p	Two-sided p
				Lower	Upper
Total: (B-D)	− 689032.1	3758469.6	358356.0	− 1399282.1	21217.8	− 1.92	109	0.029	0.057
Residential: (B-D)	− 335985.8	1681307.2	160306.3	− 653707.7	− 18263.8	− 2.09	109	0.019	0.038
Industrial: (B-D)	− 316183.6	1967339.9	187578.5	− 687958.2	55590.8	− 1.68	109	0.047	0.095
Commercial (B-D)	− 38696.5	152419.2	14532.6	− 67499.6	− 9893.3	− 2,66	109	0.004	0.009
On road: (B-D)	− 931.0	6672.1	636.2	− 2191.8	329.8	− 1.46	109	0.073	0.146
Total (A-D)	307016.4	1478428.1	140962.6	27633.2	586399.7	2.17	109	0.016	0.032
Residential (A-D)	317950.8	1652954.3	157603.0	5586.8	630314.9	2.01	109	0.023	0.046
Industrial (A-D)	− 30342.5	365483.7	34847.5	− 99409.1	38724.1	− 0.87	109	0.193	0.386
Commercial (A-D)	14064.6	90390.3	8618.4	− 3016.7	31145.9	1.63	109	0.053	0.106
On road (A-D)	911.3	4986.2	475.4	− 30.8	1853.6	1.91	109	0.029	0.058

The difference between the two data categories is recognized as significant for a P-value < 0.05. As shown in Table 5, (0.038 < 0.05), and (0.046 < 0.05). Therefore, it was deduced that the scope2 emission difference is significant in the residential sector. In addition, it was obtained a P-value of 0.009 and of 0.032. Given that (0.09 < 0.05) and (0.032 < 0.05), it was concluded that the scope2 emission difference is also significant in the commercial sector before and during the COVID-19 periods and, for total data, between 2019 and 2020 and between 2021 and 2022.

To fully understand the origins of this data difference, let’s evaluate the impacts of a few parameters on scope2 CO₂ emissions. It can be randomly choosen some parameters that are easy to quantify in the 111 regions, such as population and income, and the results are shown in Tables 6, 7 and 8. According to the previous conclusion (see Table 5), this test will be applied only in the residential and commercial sectors.

Table 6.

Application of a linear regression model (Ordinary least squares (OLS).

	Residential sector (before & during Covid)	Residential sector (after & during Covid)	Commercial sector (before & during Covid)
Model	OLS	OLS	OLS
Method	Least squares	Least squares	Least squares
R-squared	0.730	0.906	0.641
Adj. R-squared	0.725	0.905	0.634
F-statistic	144.3	518.3	95.34
Prob (F-statistic)	4.13e−31	9.03e−56	1.68e−24
Log-Likelihood	− 1660.5	− 1600.3	− 1412.1
AIC	3327	3207	2830
BIC	3335	3215	2838

Table 7.

Study of data variation between the different periods.

Sector	Variable	Coef.	Std. err	t	P > |t|	[0.025. 0.975]
Residential (Before & During Covid)	const	− 2729.1970	8.68e+04	− 0.031	0.975	− 1.75e+05. 1.69e+05
	∆pop	0.0002	0.004	0.045	0.964	− 0.007 0.007
	∆income	552.0139	32.497	16.986	0.001	487.592 616.436
Residential (After & During Covid)	const	− 3.627e+05	9.72e+04	− 3.732	0.002	− 5.55e+05 − 1.7e+05
	∆pop	82.7853	13.832	5.985	0.001	55.365 110.206
	∆income	371.9859	17.015	21.862	0.003	338.255 405.716
Commercial (Before & During Covid)	const	− 2.217e+04	8915.787	− 2.486	0.014	− 3.98e+04 − 4491.802
	∆pop	0.0002	0.000	0.459	0.647	− 0.001 0.001
	∆income	316.7833	22.942	13.808	0.001	271.303 362.264

^aStandard Errors assume that the covariance matrix of the errors is correctly specified.^bThe condition number is large, 2.46e+07. This might indicate that there are strong multicollinearity or other numerical problems.

Table 8.

Multicollinearity test.

Variable	VIF
const	4.041743
∆pop (Before & During Covid)	1.004885
∆pop (After & During Covid)	1.698185
∆income(Before & During Covid)	13.011343
∆income(After & During Covid)	12.450115

It is important to notice that several good correlations were obtained (R² = 0.730; R² = 0.906; and R² = 0.641) between scope2 CO₂ emission in residential and commercial sectors and population within come.

In Table 7, the results explained that during the periods between (2015–2018) and (2019–2020) in the residential and commercial sectors, ∆income (P-value = 0.001), as (P-value < 0.05). It means that the income impacted the scope2 CO₂ emission. However, between (2019–2020) and (2021–2022), in the residential sector, it was obtained an ∆income (P-value = 0.003) and ∆pop (P-value = 0.001). In both cases (P-value < 0.05) that means that population and income impacted the scope2 CO₂ emission. On the basis of these results, the following equations can be formulated:

4

5

6

With ∆CO₂: variation of CO₂; ∆pop: variation of population between two periods; ∆income variation of revenue.

The formula was established in the residential sector with data (Before & During COVID-19-induced lockdown period), whereas formula 2 was established in the same sector, but rather with data (Before & During COVID-19-induced lockdown period). To verify, the correlation between the variables and the results of the multi-collinearity test are shown in Table 8.

The variation of Population, ∆_pop(VIF = 1.004885; VIF = 1.698185), in both cases (VIF < 10), it was deduced that there is no correlation in the variation of population each year. In addition, the variation of revenue, ∆_income(VIF = 13.011343; VIF = 12.450115), in both cases (VIF > 10), means that there is a strong correlation between the different incomes.

Statistical analysis of electricity consumption

As shown in Table 9, (P-value = 0.012 < 0.05), likewise (P-value = 0.024 < 0.05). Therefore, it was deduced that the electricity consumption difference is significant in the commercial sector before and during COVID-19-induced lockdown periods and, for global data between (2019–2020) and (2021–2022). A P-value less than 0.05 indicates that the results are statistically significant³⁸. In this case, both P-values (0.012 and 0.024) are below the threshold of 0.05, suggesting that the observed differences in electricity consumption are unlikely to have occurred by chance.

Table 9.

Results for t-test electricity consumption difference (MWh) in residential, industrial, commercial, and on-road sectors.

Paired Samples Test
Electricity consumption difference	Paired Differences					t	df	Significance
	Mean	Std. deviation	Std. error mean	95% Confidence interval of the difference				One-sided p	Two-sided p
				Lower	Upper
Total: (Before & During Covid)	− 1946.1	13696.8	1305.9	− 4534.4	642.2	− 1.49	109	0.070	0.139
Residential: (Before & During Covid)	− 887.8	5209.4	496.7	− 1872.3	96.6	− 1.78	109	0.038	0.077
Industrial: (Before & During Covid)	− 976.4	8516.8	812.0	− 2585.9	633.1	− 1.20	109	0.116	0.232
Commercial (Before & During Covid)	− 110.9	455.4	43.4	− 197.0	− 24.9	− 2.55	109	0.006	0.012
Onroad: (Before & During Covid)	5.5	68.7	6.5	− 7.4	18.5	0.85	109	0.200	0.399
Total (After&During Covid)	534.3	2446.5	233.3	71.9	996.6	2.29	109	0.012	0.024
Residential (After&During Covid)	1374.4	7408.4	709.6	− 32.1	2780.9	1.94	108	0.028	0.055
Industrial (After&During Covid)	− 855.8	5423.7	517.1	− 1880.7	169.1	− 1.65	109	0.050	0.101
Commercial (After&During Covid)	9.6	313.4	29.8	− 49.6	68.8	0.32	109	0.374	0.748
Onroad (After&During Covid)	0.8	34.3	3.3	− 5.6	7.4	0.27	108	0.393	0.787

The significant P-value of 0.012 implies that there was a meaningful difference in electricity consumption in the commercial sector before and during the COVID-19-induced lockdown period. This could be attributed to various factors such as changes in business operations, lockdown measures, and shifts to remote work reducing on-site energy use. During the COVID-19-induced lockdown period, many commercial establishments either reduced their operations or closed temporarily, leading to decreased electricity consumption. In addition, the P-value of 0.024 for global data between the periods 2019–2020 and 2021–2022 indicates a significant difference in electricity consumption on a broader scale. The comparison across these periods may reflect the initial impact of the pandemic and the subsequent recovery phase. In 2021–2022, as businesses began to resume normal operations, electricity consumption likely increased compared to the heavily restricted 2019–2020 period.

The significant difference in the commercial sector suggests a decrease in electricity consumption during the peak of the pandemic due to reduced business activities. Businesses adapted to new operating conditions, such as reduced hours or enhanced energy efficiency measures, which also contributed to the change in consumption patterns.

Global comparison (2019–2020 vs. 2021–2022)

The significant difference globally indicates fluctuations in electricity consumption patterns driven by the pandemic’s evolving phases. Initial declines due to restrictions were followed by increases as recovery efforts and economic activities resumed. Different sectors may have experienced varying levels of impact and recovery, contributing to the observed changes in overall electricity consumption. A significant shift, for example, when (ARD > 15%), like that observed in the industrial sector, and persistent when the relative temporal changes observed are consistent, may suggest optimization of the data estimation methodology. Energy consumption habits may vary between sectors. For example, energy requirements for electronic devices, lighting, and heating vary depending on the lifestyle of individuals in residential homes. This may be one of the reasons that explain the high CO₂ emissions in the residential sector during peak hours (of intense activity). In addition, this difference may result from the choice of different categories of fuel used during electricity production. Indeed, several studies have shown that the emission rate of a fossil fuel is a function of its lower calorific value (PCI) or its higher calorific value (PCS). It is useful to use fuels consistently in all sectors. An important tip for the scope2 data product is to regularly use a PCS for future estimation of emissions from electricity. Given that the key objective of this study is to assess with high precision the percentage of CO₂ generated during the consumption of electricity that enters the atmosphere, a comparative study with estimation data can be the best way to facilitate the adjustment of our results so that they are more acceptable. For instance, residential CO₂ emissions in the U.S. were reduced by around 5–10% during the first phase of the pandemic EPA⁴⁸, consistent with our observations of a more significant reduction during the lockdown period⁴⁹. As showed in the Table 10, the mean relative difference (MRD) between 2019 and 2020 and 2021–2022 highlights contrasted temporal dynamics across sectors and between our estimates and the Vulcan reference product. In this work, residential CO₂ emissions increased by about + 22%, while commercial emissions rose more moderately (+ 8.8%) and industrial emissions slightly decreased (− 2.4%). This pattern suggests a redistribution of electricity-related emissions from productive sectors towards the residential sector during and after the COVID-19 pandemic, likely driven by changes in lifestyle, teleworking, and longer time spent at home. By contrast, the Vulcan data show a reduction in residential emissions (− 7.5%), combined with modest increases in commercial (+ 4.9%) and industrial (+ 4.0%) emissions between the two periods. These differences may reflect the use of distinct activity data, emission factors, and allocation methodologies at the sectoral level. Overall, the order of magnitude and sign of the MRD across sectors remain consistent with the expected impact of the pandemic and economic recovery, thereby supporting the plausibility of our estimates while also underlining the importance of methodological assumptions when comparing different CO₂ emission products.

Table 10.

Mean relative difference (MRD) between 2019–2020 and 2021–2022.

Reference	Sector	Mean 2019–2020 (tCO2)	Mean 2021–2022 (tCO2)	MRD (%)
This work	Residential	1437.30	1755.20	22.12
This work	Commercial	159.20	173.20	8.79
This work	Industrial	1249.80	1219.50	− 2.42
Vulcan40	Residential	10428067.00	9648938.00	− 7.47
Vulcan40	Commercial	12557023.00	13166407.00	4.85
Vulcan40	Industrial	9356486.00	9728122.00	3.97

Conclusions

This study provided a comprehensive assessment of electricity consumption and fossil-fuel CO₂ (FFCO₂) emissions across the residential, industrial, commercial, and on-road transport sectors in 111 cities of Madagascar using operational data from 55 power plants between 2015 and 2022. The residential sector emerged as the dominant emitter, contributing more than 52% of total emissions, consistent with patterns observed in several developing regions where household dependence on electricity for essential activities is high. A clear temporal signature of the COVID-19 pandemic was observed. Statistical comparisons revealed significant shifts in electricity consumption between the pre-pandemic and pandemic periods (p = 0.012), as well as between 2019 and 2020 and 2021–2022 (p = 0.024). The on-road transport sector showed the strongest disruption, recording a marked negative MRD during lockdowns, followed by a sharp rebound (+ 47%) during post-pandemic economic recovery. These findings highlight the sector’s sensitivity to mobility restrictions and its major role in shaping post-crisis emissions trajectories. Commercial and industrial sectors also displayed notable but heterogeneous patterns, reflecting operational slowdowns, reopening phases, and sector-specific resilience.

Beyond the national scope, the results carry broader global implications. First, the discrepancies observed between emission estimates from local operational data and international datasets underscore the importance of improving measurement methodologies, particularly in low-data environments. Industries worldwide can benefit from such insights by refining their monitoring frameworks and accelerating the transition toward cleaner energy sources to reduce their carbon footprint. Second, regions facing similar challenges—especially in the Global South, may draw on these findings to identify sector-specific emission hotspots and design targeted mitigation strategies. Third, the study highlights the need for policymakers to strengthen the robustness of national energy statistics. Understanding divergences between data sources enables the formulation of informed regulations, incentives for renewable energy deployment, and enhanced energy-efficiency programs.

Mitigation strategies adapted to the realities of a developing country like Madagascar are essential to ensure sustainable reductions in Scope 2 CO₂ emissions. First, the rapid deployment of decentralised renewable energy systems, particularly solar photovoltaics and small hydropower, can reduce dependence on diesel-based generation, which currently dominates the national grid. Improving energy efficiency in the residential and commercial sectors through efficient lighting, appliance standards, and building insulation can significantly lower electricity demand and associated emissions. In the industrial sector, operational optimisation, fuel switching, and the adoption of modern low-carbon technologies offer promising opportunities for emission reduction. Strengthening institutional capacity for data collection, monitoring, and verification is also critical, as harmonised and transparent datasets facilitate evidence-based policymaking.

Future research should investigate the socio-economic consequences of emission-reduction strategies to ensure equitable outcomes. Integrating behavioural, demographic, and economic drivers into sectoral modelling would improve prediction accuracy and policy relevance. Strengthening international cooperation particularly data-sharing platforms, technical assistance programs, and capacity-building efforts is essential to help developing countries enhance their monitoring systems and achieve global climate targets.

Author contributions

Modeste Kameni Nematchoua*: Conceived and designed the study; led data analysis; coordinated research team; wrote the original draft and supervised manuscript preparation and revisions. Jose A. Orosa: Contributed to the design of the methodology; performed comparative statistical analysis; contributed to the writing and critical revision of the manuscript. Cinza Buratti: Provided expertise in energy systems and emissions modeling; contributed to the validation of Scope 2 emission calculations; reviewed and edited the manuscript. Shady Attia: Supported interpretation of urban energy use patterns; contributed to methodology design and graphical data presentation; reviewed the final manuscript. Jacques Teller: Provided expertise on urban planning and sustainability metrics; contributed to contextual analysis of Indian Ocean cities; critically reviewed the manuscript. Muriel Bemanana: Collected and validated local data from Madagascar; contributed to regional case studies and interpretation; supported manuscript translation and formatting. Olatunji Akinola: Conducted geospatial analysis of power plant locations and city boundaries; contributed to data visualization and final proofreading. Andrianirina Charles Bernard: Participated in the classification and categorization of electricity consumption data by sector; helped in comparative period analysis. Rakotomalala Minoson Sendrahasina: Coordinated data acquisition from national energy agencies; provided technical insight on electricity grid operations in Madagascar. Sambatra Eric Jean Roy: Supported data organization and verification from on-ground field sources; contributed to writing and editing the discussion section. Rafanotsimiva Liva Falisoa: Analyzed policy implications and regional disparities; contributed to drafting the conclusion and recommendations section. Messina Jean-Pierre: Provided expertise in urban development in Sub-Saharan regions; contributed to the contextualization of results in the African Indian Ocean region. Sigrid Reiter: Reviewed and refined the overall manuscript structure; provided expert insight on sustainable energy strategies; supervised the final revision process. All authors reviewed and approved the final version of the manuscript.

Funding

No fund.

Data availability

The datasets generated and/or analysed during the current study are not publicly available due to institutional data-sharing policies, but they are available from the corresponding author upon reasonable request. For access to the data, please contact Dr. Modeste Kameni Nematchoua at kameni.modeste@yahoo.fr.

Declarations

Competing interests

The authors declare no competing interests.