Towards Sustainable Development Goal 7 “Universal Access to Clean Modern Energy”: National Strategy in Rwanda to Scale Clean Cooking with Bottled Gas

: More than 90% of Rwandans rely on polluting solid fuels to meet their cooking needs. The negative impacts on health, climate, and the environment have led the Rwandan government to set a target of halving that number to 42% by 2024. A National Master Plan to promote scale up of liqueﬁed petroleum gas (LPG) has been developed to deﬁne (i) the necessary market conditions, (ii) public and private sector interventions, and (iii) the expected societal impacts. Findings are reported from modelling scenarios of scaling LPG use towards the 2024 policy target and the 2030 target for “universal access to clean modern energy” (SDG7). Household LPG use is projected to increase from 5.6% in 2020 to 13.2% by 2024 and 38.5% by 2030. This level of adoption could result in a reduction of 7656 premature deaths and 403,664 disability-adjusted-life-years (DALYs), as well as 243 million trees saved. Reductions in carbon dioxide and black carbon emissions equivalents (CO 2 e and BCe, respectively) are estimated to reach 25.6 million MT and 14.9 MT, respectively, by 2030. While aggressive policy intervention is required, the health, environmental, and developmental beneﬁts are clear. Implementation of the Rwanda National LPG Master Plan will provide a model for other sub-Saharan African countries to address the priorities for cessation of reliance on solid fuels as an energy source.


Introduction
Rwanda, a small landlocked country situated in East Africa, is the most densely populated country in sub-Saharan Africa with a population density of 1 per 525 square meters and a total population of 12.952 million (2020) [1]. In 2019, almost one-fifth (17%) of the population lived in cities and urban environments [2], projected to increase to one-third (35%) by 2024 [3]. Rwanda's gross domestic product (GDP) per capita was US$ 801.7 in 2019 [2], representing an overall growth of 9.4% annual change [4], aligned with the government's intention to transform Rwanda's current agricultural-based economy to one that is industry-and service-based.
Similar to most countries in sub-Saharan Africa (SSA), the household energy market in Rwanda is dominated by cooking energy, and cooking energy is dominated by reliance on solid fuels. The primary fuels used for cooking are firewood (used by as many as 83% of rural households of whom more than a quarter can gather firewood for free) and charcoal (which is the most common fuel in urban settings, with more than 40% of urban households relying on it) [5]. While use of liquefied petroleum gas (LPG) or bottled gas for cooking has doubled in Rwanda since 2016 [6], still only 5.6% of households report using it as a primary cooking fuel in 2020 [5]. Other clean fuels are even less prevalent in household cooking, with 0.23% of households reporting using biogas and 0.19% using electricity for cooking [5].

Materials and Methods
In line with the main study aims, there were two main foci to the analyses: firstly, to model the potential proportion of the population using LPG under varying conditions affecting market growth. This included (i) a scenario of no policy or market intervention with current demographic, market and LPG industry trends projected into the future and (ii) a scenario with implementation of specific policies and interventions, in addition to increased private sector investment supportive of LPG market expansion. Secondly, it included an analysis of the potential impacts of increasing LPG adoption to policy target levels (by 2024 and 2030) on (i) the environment (reduced deforestation and emissions detrimental to climate) and (ii) health (averted premature mortality and morbidity (DALYs) from reductions in exposure to household air pollution (PM 2.5 )).

Enhancing Adoption of LPG through Market Enhancement: Modelling Impacts towards Government Adoption Targets
Rwanda has seen a small, but significant increase in adoption of LPG over the last 5 years owing to increased availability of and accessibility to LPG, specifically in urban environments. The aim of the National LPG Master Plan was to consider how to significantly build on this growth through the setting of conditions that positively affect and stimulate further LPG adoption and corresponding availability and accessibility, above current growth trends. Specific consideration was given to the potential effects of policies and interventional measures that (i) improve access to LPG (with an emphasis on home delivery), (ii) implement and fully enforce the branded cylinder recirculation model (BCRM) of LPG distribution in accordance with world standards and international best practices [35], (iii) mandate public institutions to transition to clean cooking/LPG use from traditional use of wood fuel, and (iv) promote LPG transition (especially among urban and middleincome households) through consumer campaigns. Alongside these market enhancements and increased supply, it is anticipated that LPG adoption would also increase with the population size and rate of urbanisation over time.
To model the potential of stimulated LPG adoption to achieve target levels by 2024 and 2030, analyses contrasted three scenarios including (i) a baseline (business-as-usual scenario) where no intervention is employed; (ii) an interventional scenario where market enhancements (informed by the Master Plan) in the form of increased retail density, home delivery, and institutional LPG use are promoted; and (iii) a policy target scenario incorporating the interventional scenario enhancements with additional regulatory/legislative measures governing fuel use. The three scenarios are described below: • Business-as-usual (BAU) scenario. In this scenario, existing demographic, market, and LPG industry trends were projected to continue according to current trends to 2024 and to 2030, assuming no additional government/private sector-led intervention. Under this scenario, the majority of the LPG adoption growth relates to external factors such as increased population size and rate of urbanisation with an assumption that the private sector will continue to invest and expand LPG availability (in bulk and through additional cylinders) according to business as usual. • Interventional scenario. This scenario models a combination of interventions and market measures designed to simulate population adoption of LPG and to build consumption of LPG in the country. The modelled market developments include increased consumer access to LPG primarily by enhancing retail point density (including home delivery), implementing best practices in LPG distribution (fully enforcing the BCRM), and mandating LPG use in public institutions that currently rely on polluting fuels. • Policy target scenario (NST-1). This scenario represents the government's aspirational target of 40% of the population, including public institutions, adopting LPG for cooking by 2024. In addition to conditions set out under the interventional scenario, pro-LPG mandates are added that create significant additional urban and institutional LPG demand and anti-biomass fuel mandates (e.g., a charcoal ban) that make competing biomass fuels less available or unavailable, wherever practical.

Data Sources for LPG Adoption Projections in the Residential Sector
The LPG adoption modelling scenarios were informed by nationally representative data sources used to project trends of population uptake of LPG into the future. These included the latest available nationally representative population-based surveys with information on self-reported use of household fuels, including (i) the 2016 World Bank's multi-tier framework (MTF) survey on household access to electricity and cooking practices (sample n = 3300 households) [6] and (ii) the National Survey on Cooking Fuel Energy and Technologies in Households, Commercial, and Public Institutions in Rwanda (CFET) [5], commissioned by MININFRA (sample of n = 5020 households and 469 institutions). Fuel and population data of public institutions were obtained through direct correspondence with the relevant authorities in the country, desk review, and information contained in the CEFT dataset [5]. Population data for the modelling were based on the Rwandan National Institute of Statistics (NISR)'s official projections for population size and urbanization rates [36]. LPG fuel price data over time were modelled using RURA published data [37]. To set the modelling scenarios for LPG adoption, a number of informed assumptions were made based on available data, and are summarised in Table 1. Awareness raising and regulatory conditions The modelled market enhancements underpin consumer-education campaigns and full enforcement of the best practice LPG distribution model (i.e., the branded cylinder recirculation model).

Policy intervention
This model assumed that the policy target level of 40% LPG adoption by 2024 is achieved through (i) accelerated investment by the private sector and GoR to ramp up LPG supply and access even faster than the interventional case and (ii) anti-biomass fuel mandates (e.g., charcoal ban enforced in Kigali and other main cities) and pro-LPG fuel mandates/regulation.
Note. Urbanisation, population, and average household size estimates are taken from the Rwandan National Institute of Statistics (NISR) projections [36]. LPG use per capita is a median value for mixed users from the CFET, 2020 survey [5]. Consumer price is based on RURA 2019 LPG price data [37] and IMF-based projections of future RWF-USD exchange rates [4].
In addition, for all scenarios, a sub-analysis of LPG household uptake by socioeconomic status (SES) was performed to consider differential trends in adoption patterns over time by SES according to the official Rwanda Ubudehe categorisation, which, as of 2020, classified Rwandan households according to four categories [38,39] (reorganised under five categories since February 2021). It is worth noting that Ubudehe categories are not based on income ranges alone, but include a combination of income, assets, education, and household expenditure, formed by the government to inform social welfare programs [38,39].

Statistical Analysis
To project the increased proportion of LPG adoption according to the different modelling scenarios, a probit cumulative link model with scale effects and flexible thresholds was fitted to the MTF [6] and CFET datasets [5]. The dependent variable (LPG adoption) was a three-level categorical variable based on self-reported cooking fuel in the last month (prior to completion of the household survey) and was coded as 'no LPG', 'LPG + other fuels' (mixed use), and 'LPG exclusive use'. The choice of independent variable was theorydriven and pragmatic from the limited set of common variables between the two datasets. We used (i) survey year, (ii) socioeconomic status of the household (Ubudehe), (iii) geo- graphical context (Kigali urban city, other urban areas, or rural), (iv) distance from LPG source/retail point, and (v) LPG price per Kg as independent variables. The model allows for the independent variables to have additive and multiplicative effects with the latter interpreted as effects on the scale (or dispersion) of a latent distribution. To identify which of the independent variables had multiplicative effects, we fitted models with different sets of them as multiplicative variables and we used the Bayesian information criterion for model selection. The final model used year, socioeconomic status of the household, regional context, distance from LPG source/retail point, and LPG price per Kg as additive variables and year, regional context, retail distance, and LPG price as multiplicative variables. We used the model to predict the dependent variable in the years 2024 and 2030 under assumptions concerning improvements in socioeconomic circumstance (projecting existing favourable trends), rates of urbanisation (based on NISR projections), distance from LPG source, and LPG price. Specifically, assumptions for the modelling scenarios are presented in Table 1. All analyses were conducted using R version 4.0.1. [40].

Evaluating Impacts on Environment, Climate, and Health of Scaled LPG Adoption for Cooking
Estimation of the health, climate, and environmental impacts from the increased adoption of LPG for cooking in the residential and public institutions sectors, under the three scenarios, was assessed according to the following outcomes:

•
Environment and Climate Impacts. The impacts on environment and climate of displacement of solid fuels for cooking through increased adoption of LPG were assessed through (i) averted deforestation (number of trees saved that would otherwise be used as source of firewood or for charcoal production), (ii) averted CO 2 emissions (measured as CO 2 equivalents or CO 2 e (including CO 2 , methane (CH 4 ), and nitrous oxide (N 2 O)), (iii) averted BC emissions equivalents (including BC, organic carbon (OC), carbon monoxide (CO), and total non-methane organic compounds (TNMOC)), and (vi) the economic value of averted CO 2 e emissions in terms of carbon financing.

•
Health. Health impacts from reductions in household air pollution (PM 2.5 ) from combustion of solid fuels through increased adoption of LPG was assessed through (i) averted number of premature HAP related-deaths, (ii) averted HAP-related morbidity (disability-adjusted life years (DALYs)), and (iii) the potential economic value of the projected reductions in mortality and morbidity according to the main global burden of disease (GBD) outcomes.
Data Sources and Methodology Used for the Environmental, Climate, and Health Impact Assessments To estimate the impacts on health, climate, and environment of scaled LPG adoption (household and institutional), estimates from the projected proportions of the population using LPG for each time horizon (2024 and 2030) for each of the three modelling scenarios were used. Environmental impacts (averted deforestation/trees saved) were calculated based on existing household and institutional use of firewood and charcoal that was converted to equivalent wood consumption, using a ratio of 7 based on data from Tanzania [41]. Deforestation was defined according to the fraction of non-renewable biomass (fNRB) (this is the percentage of wood removed unsustainably from forests, applicable to both charcoal and firewood). The fNRB was estimated to be 59% (52-66%) for Rwanda based on Bailis et al. [42] with tree mass set at 100 kg [43] (see Section S2.1 in the Supplementary Materials for detailed procedures and equations used to estimate the deforestation impacts).
To calculate reductions in CO 2 e emissions resulting from projected household and institutional transition from solid fuels to LPG (under each scenario), the Gold Standard Technologies and Practices to Displace Decentralized Thermal Energy Consumption (TPDDTEC) Methodology was adopted. This methodology estimates total CO 2 e emissions by calculating the CO 2 equivalent (CO 2 e) emissions of three greenhouse gases-CO 2 , CH 4 , and N 2 O. The emission rates were multiplied by the global warming potential (GWP 100 ) conversion factors for each gas (32 for CH 4 and 298 for N 2 O) (see Table S2). As a second step, the CO 2 e emissions for different fuel use were calculated by multiplying household level fuel consumption by the net calorific value of the fuel and average stove efficiencies using global averages obtained from the literature [14,15] (see Table S3). This results in the energy use per fuel (megajoules delivered or MJd), which was multiplied by the CO 2 e emissions factor (in g/MJd) to obtain the total CO 2 e emissions (in grams, which was then converted to metric tonnes). The CO 2 e tonnage differential was calculated by subtracting the CO 2 e emissions under the interventional scenario from CO 2 e emissions in the BAU scenario.
Similarly, to calculate the BCe emissions, reductions per annum from scaled adoption of LPG, the same four-step approach was used following the gold standard methodology: (i) the BCe emissions per unit of fuel use were calculated using the TPDDTEC Guidelines on Black Carbon Quantification Methodology [44]; (ii) the BCe emission per fuel was multiplied by the GWP 100 conversion factor for BC (equal to 660) based on the IPCC 2013 report [20], but increased by 14% based on Etminan et al. [45]; and (iii) the global warming potential of BCe emissions per fuel type was then multiplied by the total consumption per fuel in kg. This calculation estimated the BCe emissions from fuel use, and they were calculated for LPG, charcoal, and firewood, reflecting the fuel projections trends under the various adoptions scenarios up 2030.
The economic value of reduced CO 2 equivalent emissions was estimated by multiplying the total emissions averted by 2030 by the prevailing price of carbon for the latest available year (2019). The price of carbon was derived from existing carbon values and was set at a mean of US$ 3.8 of CO 2 equivalent (ranging from US$ 1.4 to 4.3) and the transacted volume in 2019 corresponded to 6.4 MT CO 2 equivalent [46]. The economic value of reduced carbon was estimated using the observed prevailing carbon price in Africa and multiplying it by the carbon emissions averted.
To estimate averted HAP-related deaths and disability (ADALYs) achieved through scaled adoption of LPG under each of the three modelling scenarios, 'pre'-and 'post'intervention (i.e., LPG adoption) PM 2.5 exposure values were used as inputs into the Household Air Pollution Intervention Tool (HAPIT version 3.1.1) [47]. This is an analytical tool based on established global burden of disease methods incorporating five disease outcomes that are causally associated with PM 2.5 exposure from household fuel combustion-COPD, lung cancer, ischemic heart disease, and stroke in adults as well as pneumonia in children. HAPIT is widely used for modeling health impacts of interventions to reduce HAP exposure. Full details of the values used for health impact modelling are provided in the Supplementary Materials (Section S2.3).
The economic value of averted HAP-related deaths was estimated by multiplying the annual GDP per capita for 2019 in Rwanda (US$ 801.7) [2] by the total number of averted deaths for Rwandan working age adults (aged 15-64 years). Similarly, the economic value of HAP-related averted DALYs (ADALY) was calculated by multiplying the annual GDP per capita for 2019 by the number of ADALYs for working age adults (aged 15-64 years).
The full list of assumptions, equations, and values used to calculate climate and health impacts is provided in the Supplementary Materials (Sections S2.2 and S2.3).

Enhancing Adoption of LPG through Market Enhancement: Modelling Impacts towards Government Adoption Targets
The results from the three modelling scenarios were considered according to both domestic (e.g., household) and institutional use of LPG following the two time horizons of 2024 (Rwanda government policy target) and 2030 (SDG 7). Household LPG adoption was projected to increase from 5.6% in 2020 to 10.4% in 2024 and 24% in 2030 under the business-as-usual scenario (i.e., without market enhancement) (Figure 1). Under this scenario, the greatest growth was projected for urban settings, with 65.9% of the households in Kigali city projected to use LPG in 2024 (from 45.1% in 2020), increasing to 86.4% in 2030. Similarly, 25.2% households in 'other urban' areas were projected to be using LPG in 2024, increasing to 56.7% in 2030. By contrast, for rural contexts, only very limited growth was projected (0.6% of the households estimated to adopt LPG by 2024 rising to 3.4% in 2030).

Enhancing Adoption of LPG through Market Enhancement: Modelling Impacts towards Government Adoption Targets
The results from the three modelling scenarios were considered according to both domestic (e.g., household) and institutional use of LPG following the two time horizons of 2024 (Rwanda government policy target) and 2030 (SDG 7).

Projections of Household LPG Adoption
Household LPG adoption was projected to increase from 5.6% in 2020 to 10.4% in 2024 and 24% in 2030 under the business-as-usual scenario (i.e., without market enhancement) (Figure 1). Under this scenario, the greatest growth was projected for urban settings, with 65.9% of the households in Kigali city projected to use LPG in 2024 (from 45.1% in 2020), increasing to 86.4% in 2030. Similarly, 25.2% households in 'other urban' areas were projected to be using LPG in 2024, increasing to 56.7% in 2030. By contrast, for rural contexts, only very limited growth was projected (0.6% of the households estimated to adopt LPG by 2024 rising to 3.4% in 2030). With the intervention scenario (implementation of market enhancement strategies as described above), household LPG adoption was projected to increase to 13.2% in 2024 and to 38.5% by 2030 (Figure 1). While increases in LPG adoption under this scenario are primarily driven by urban contexts (expected to reach 89.2% by 2030 in urban areas other than Kigali and 94.4% in urban Kigali), LPG adoption in rural areas was also estimated to increase modestly (projected to reach 12.4% in 2030).
In contrast to the business-as-usual and interventional scenarios, the policy target scenario for household adoption of LPG was 33.6% in 2024 and 47.2% by 2030 (Figure 1). With the intervention scenario (implementation of market enhancement strategies as described above), household LPG adoption was projected to increase to 13.2% in 2024 and to 38.5% by 2030 (Figure 1). While increases in LPG adoption under this scenario are primarily driven by urban contexts (expected to reach 89.2% by 2030 in urban areas other than Kigali and 94.4% in urban Kigali), LPG adoption in rural areas was also estimated to increase modestly (projected to reach 12.4% in 2030).
In contrast to the business-as-usual and interventional scenarios, the policy target scenario for household adoption of LPG was 33.6% in 2024 and 47.2% by 2030 (Figure 1).
To understand the projected LPG use under the interventional scenario according to household income, projections over time were stratified by SES (according to the four Ubudehe categories) ( Figure 2). While the proportion of homes using LPG with the largest income (highest SES category) had the greatest predicted growth in LPG adoption under the intervention scenario (33.2% in 2020; 51.6% in 2024; 75% in 2030), adoption of LPG is also projected to significantly increase in households with lower incomes. Even in the poorest quartile of income (lowest category of SES), 19% of households are projected to be using LPG by 2030 ( Figure 2). Ubudehe categories) (Figure 2). While the proportion of homes using LPG with the largest income (highest SES category) had the greatest predicted growth in LPG adoption under the intervention scenario (33.2% in 2020; 51.6% in 2024; 75% in 2030), adoption of LPG is also projected to significantly increase in households with lower incomes. Even in the poorest quartile of income (lowest category of SES), 19% of households are projected to be using LPG by 2030 (Figure 2).  (Table S1).

Overall LPG Adoption Projections (Residential, Institutional, and Commercial Sectors)
One of the priorities for the government was to include public institutions in projections, given the potential to mandate fuel use in this sector. The use of LPG by the following institutions was modelled: schools, prisons, police stations (with temporary detention centres), healthcare facilities, military barracks, and refugee camps. Other non-public institutions were also included in the analyses including private catering institutions such as hotels and restaurants and other industrial users of LPG.
In 2020, approximately 4500 MT of LPG was estimated to be consumed by both public and commercial institutions (based on [5] and commercial data). With effective national mandate of public institutions to use LPG for clean cooking, it is expected that almost full conversion can be achieved by 2024, leading to predicted growth of 29,670 MT of consumed LPG in 2024 and 34,760 MT in 2030 (based on NISR population projections calculated for each category of institution between 2024 and 2030). While it is assumed that all institutions will adopt LPG within this time frame, it should be noted that is unlikely that all institutions will use LPG exclusively within this time horizon. In addition, LPG target projections in boarding schools and prisons took into account the existing level of adoption of biogas, which has been promoted in institutions since 2000 [48] as well as for rural household uptake since 2007 [49,50]. It is estimated that only about 40% of the total institutional cooking needs are met through biogas [5], and the remaining 60% was projected to be LPG from 2024 onwards.  (Table S1).

Overall LPG Adoption Projections (Residential, Institutional, and Commercial Sectors)
One of the priorities for the government was to include public institutions in projections, given the potential to mandate fuel use in this sector. The use of LPG by the following institutions was modelled: schools, prisons, police stations (with temporary detention centres), healthcare facilities, military barracks, and refugee camps. Other non-public institutions were also included in the analyses including private catering institutions such as hotels and restaurants and other industrial users of LPG.
In 2020, approximately 4500 MT of LPG was estimated to be consumed by both public and commercial institutions (based on [5] and commercial data). With effective national mandate of public institutions to use LPG for clean cooking, it is expected that almost full conversion can be achieved by 2024, leading to predicted growth of 29,670 MT of consumed LPG in 2024 and 34,760 MT in 2030 (based on NISR population projections calculated for each category of institution between 2024 and 2030). While it is assumed that all institutions will adopt LPG within this time frame, it should be noted that is unlikely that all institutions will use LPG exclusively within this time horizon. In addition, LPG target projections in boarding schools and prisons took into account the existing level of adoption of biogas, which has been promoted in institutions since 2000 [48] as well as for rural household uptake since 2007 [49,50]. It is estimated that only about 40% of the total institutional cooking needs are met through biogas [5], and the remaining 60% was projected to be LPG from 2024 onwards.
Modelling of the three scenarios according to projected adoption of LPG uptake from all sectors (residential, institutional, and commercial) is summarised in Figure 3. Modelling of the three scenarios according to projected adoption of LPG uptake from all sectors (residential, institutional, and commercial) is summarised in Figure 3. Across all sectors, LPG adoption was projected to increase from 5.6% in 2020 to 16.8% in 2024 and 30.4% in 2030 under the business-as-usual scenario (e.g., without market enhancement). By contrast, the interventional scenario (enhanced market conditions) projected 19.6% of the population to be using LPG in 2024, reaching 44.9% in 2030. According to this scenario, the NST-1 policy goal of 40% LPG adoption across all sectors will be reached by 2030.

Impacts on the Environment and Climate
Population transition to LPG for cooking is estimated to have a dramatic impact on prevention of forest degradation and corresponding loss of biodiversity in Rwanda. If the intervention scenario is effectively implemented to the predicted 44.9% adoption by 2030 (for all sectors), 83 million trees could be saved compared with lower levels of increased adoption following the business-as-usual scenario ( Table 2).  Across all sectors, LPG adoption was projected to increase from 5.6% in 2020 to 16.8% in 2024 and 30.4% in 2030 under the business-as-usual scenario (e.g., without market enhancement). By contrast, the interventional scenario (enhanced market conditions) projected 19.6% of the population to be using LPG in 2024, reaching 44.9% in 2030. According to this scenario, the NST-1 policy goal of 40% LPG adoption across all sectors will be reached by 2030.

Impacts on the Environment and Climate
Population transition to LPG for cooking is estimated to have a dramatic impact on prevention of forest degradation and corresponding loss of biodiversity in Rwanda. If the intervention scenario is effectively implemented to the predicted 44.9% adoption by 2030 (for all sectors), 83 million trees could be saved compared with lower levels of increased adoption following the business-as-usual scenario ( Table 2). Achieving levels of LPG adoption in line with the NST-1 policy target (projected to reach 53.6% in 2030) could save an additional 86 millions trees over the intervention scenario, for a total of 354 million trees in 2030 (Table 2). Figure 4 summarises the projected reduction in harvested trees by 2030 (SDG 7 target) (see Figure 4 compared with the 2020 adoption levels).
Note: *This is an approximation including effects on household and institutional cooking.   In terms of climate impacts, the transition from charcoal and firewood to LPG for cooking will decrease total and per capita carbon emissions through two mechanisms: (i) decreased emissions from combustion of solid fuels and (ii) decreased fuel production (in the case of charcoal). The total cumulative CO 2 e emissions reductions between 2021 and 2030 were estimated to be between 28.5 million (under the intervention scenario) and 40.1 million emissions averted (under the NST-1 policy scenario) (Table 2). Similarly, the overall cumulative (2021-2030) reductions in BCe emissions resulting from scaled transition to LPG for cooking in Rwanda are projected to be 17.1 million emissions averted if the interventional scenario is fully implemented or 25.8 million emissions averted if the NST-1 policy target is achieved, assuming no further increase in adoption of LPG from levels in 2020.
Finally, the cumulative economic value of averted CO 2 e emissions following national transition to LPG for clean cooking over 2021-2030 was estimated to be US$ 108.2 million (intervention scenario) and US$ 152.4 million (NST-1 policy scenario), assuming no further increase in adoption of LPG from levels in 2020.

Health Impacts
The cumulative number of deaths saved (adults and children) from the adoption of LPG among residential cooks between 2021 and 2030 was estimated to be 3832 (under the interventional scenario) and 7656 (under the NST-1 policy scenario) (Table 3). Similarly, the cumulative number of ADALYs was estimated to range from 201,333 to 403,664 from 2021 to 2030. The averted deaths and DALYs generated by transitioning to LPG by public institutions (impacting cooks, catering staff, and to a less extent residents of public institutions such as school pupils and so on) could not be modelled owing to the paucity of data in this sector. Figure 5A,B present the results for year 2030 alone in comparison to no further increase in adoption of LPG from levels in 2020. Note. The impacts on health are modelled only for the household sector. They are based on the five health conditions (COPD, lung cancer, ischemic heart disease, and stroke in adults as well as pneumonia in children) known to be causally associated with HAP for Global Burden of Disease calculations in 2013. Since this time other health outcomes (e.g., diabetes, asthma and adverse pregnancy outcomes (2019)) are also used in GBD calculations and therefore our estimates are likely to be conservative.
(2021-2030). Economic value of averted deaths and DALYs N/A US$ 9.8 million US$ 19.5 million Note. The impacts on health are modelled only for the household sector. They are based on the five health conditions (COPD, lung cancer, ischemic heart disease, and stroke in adults as well as pneumonia in children) known to be causally associated with HAP for Global Burden of Disease calculations in 2013. Since this time other health outcomes (e.g. diabetes, asthma and adverse pregnancy outcomes (2019)) are also used in GBD calculations and therefore our estimates are likely to be conservative.
These values are estimated to lead to a total economic value (based on the prevailing average wage rate times the labour time and productivity gained from the averted deaths and DALYs) of US$ 9.8 million (under the intervention scenario) and US$ 19.

Discussion
Similar to a growing number of countries in sub-Saharan Africa, transition to clean  These values are estimated to lead to a total economic value (based on the prevailing average wage rate times the labour time and productivity gained from the averted deaths and DALYs) of US$ 9.8 million (under the intervention scenario) and US$ 19.5 million (under the NST-1 policy scenario) cumulatively for the years 2021 to 2030.

Discussion
Similar to a growing number of countries in sub-Saharan Africa, transition to clean modern energy for cooking, including LPG, is being targeted as a public health and environmental priority in Rwanda [51,52]. Accelerated expansion of LPG in residential and institutional sectors requires effective policies and investment necessary to stimulate this change and the Rwandan government is implementing this through its Master Plan. The current analysis provides some context in moving towards the government biomass fuel reduction target investigating the impact of factors that will stimulate LPG adoption over time (both those that can and cannot be controlled). An 'intervention scenario', including strategies to increase market expansion and accessibility of LPG and mandating institutional use of LPG, estimates that, with effective implementation, adoption of LPG can be expected to increase from 5.6% in 2020 to one-fifth of the population (19.6%) by 2024. While this falls short of the government target, extrapolating to the SDG 7 "universal access to clean modern energy" time horizon of 2030, it is estimated that the target can be achieved over this extended time horizon, with almost half of the population (44.9%) using LPG in 2030.
Achieving this level of adoption is estimated to have significant positive impacts on both the environment and climate. Expansion to 45% population adoption of LPG by 2030 (as conservatively modelled in the intervention scenario) will (i) save an estimated 268 million trees cumulatively (compared with 185 million saved under BAU conditions), (ii) reduce CO 2 e emissions to 28.5 million MT (compared with a reduction of 18.6 MT under BAU conditions), (iii) reduce BCe emissions by 17.1 MT (compared with a reduction of 11.1 MT under BAU conditions), and (iv) yield estimated economic savings of US$ 108.2 million from these reduced environmental and climate detriments (compared with a reduction of US$ 70.6 million under the BAU scenario). Achieving this expansion of LPG will also significantly improve public health with an estimated 2,590 premature deaths from exposure to HAP being averted and an additional 135,716 DALYs being averted from associated reduction in HAP-related morbidity (with values estimated as a difference between the totals achieved under the intervention scenario compared to BAU conditions, see Table 2). These positive gains in health are estimated to result in economic savings of US$ 9.8 million between 2021 and 2030.
The tangible benefits from scaling adoption of clean cooking with LPG, including environment and health, make a compelling case for supporting the conditions necessary for expansion of the LPG market through the National LPG Master Plan. It will also be necessary to accompany policy and regulation with investment and interventions to positively stimulate adoption, especially by households.

Scaling Adoption of LPG to Government and SDG7 Time Horizons
It is clear that LPG uptake is positively correlated with increasing urbanization at societal level and increasing socio-economic status and level of education (particularly for the head of the household) at an individual/household level. Unsurprisingly, our results find that increasing trends of LPG adoption are substantially greater in higher income/SES strata than lower strata, particularly in urban contexts, consistent with findings reported in other studies [53][54][55][56]. However, of note is the projected increase in adoption by 2030 in the lowest SES stratum with 19.1% of the poorest households adopting LPG by 2030 if the intervention scenario conditions are achieved (see Figure 2).
The potential for national policies, private sector investment, and the donor community to address inequality in income/SES to facilitate access to LPG for clean cooking by resource poor households warrants consideration. International research in sub-Saharan Africa and other lower-and-middle-income countries (LMIC) highlights that mechanisms to reduce the up-front cost of switching to LPG can accelerate its adoption, especially when combined with education and sensitization efforts. These include consumer innovations such as microfinance for initial purchase of LPG equipment [57][58][59] and the potential for paying for LPG in small amounts through pay-as-you-go smart meter technology (PAYG LPG) [35,60]. This latter commercial innovation is receiving increasing attention for its potential in opening up clean cooking with LPG to low-income households given (i) no (or greatly reduced) upfront LPG equipment costs and (ii) the ability to pay for LPG in small increments (although at higher costs per kg) [35,60,61]. Another attraction of this model for consumers is that consumption can be tracked through the LPG smart meters in real time, facilitating timely home delivery to ensure a household never runs out of LPG for cooking [60]. While no PAYG LPG company currently operates commercially in Rwanda, BBOXX Ltd. (an electricity off-grid company offering supply through PAYG technology) conducted pilot studies of a PAYG LPG system in 2019-2020 in urban Kigali. While their preliminary results were encouraging, an expansion of the pilot work has not been initiated [61]. A successfully implementation of PAYG LPG expansion may facilitate consumer adoption of LPG in line with the government's national LPG road map with the benefit of opening up access to clean cooking for the poorest segments of society that may benefit from paying in small fuel amounts.
Proximity to outlets selling LPG, as a proxy for access, was identified to be an important predictor of its uptake, reinforcing the requirement for private sector-led investment stimulated by government intervention to scale adoption according to aspirational targets. The modelling found that a minimum distance of 2 km is required to ensure household uptake of LPG. This finding is consistent with previous research in sub-Saharan Africa (Cameroon, Ghana, and Kenya) that quantifies distance from an LPG retail point as a key driver for more use of the fuel [32][33][34]60]. Home delivery will address the issue of proximity providing convenient access to LPG without the need for travel to secure the fuel-national data suggested that, currently, one-quarter (27%) of urban households have access to home delivery of the LPG cylinder refills [5], which is a high rate for sub-Saharan Africa. However, outside of urban areas, home delivery might become more problematic without a greater density of retail outlets and poor road quality may add home delivery costs.
Interestingly, the cost of LPG refills was found to be less important than access, although it is considered to be an important driver of uptake and, specifically, of exclusive use. As of 2020, the LPG retail price in Kigali city was estimated to be an average of 1060 Rwf/kg (1.09 US$/kg) [37]. The cost of an LPG cylinder deposit (for first time users) plus the initial gas ranged from 38,000 RwF (US$ 38.9) for a 6 kg filled cylinder to 80,000 RwF (US$ 82) for a 12 kg filled cylinder. Modelling of the costs of LPG found only a very minimal correlation between LPG price and usage. While it might be expected that increases in LPG price might correspond to a decrease in usage, such an analysis was not possible given the paucity of available data from households using LPG from national data sources [5]. We hypothesise that, in the early stages of LPG market development where the majority of the population is rural with limited or no access to LPG fuel, expanding access to cylinders and LPG equipment will be the main driver of LPG uptake. Fluctuations in price will impact levels of adoption of LPG only once an accessible and reliable supply is established.

Impacts of Scaling LPG Adoption to Government and SDG7 Time Horizons on Environment, Climate, and Public Health
An increasing number of countries in sub-Saharan Africa are setting aspirational targets for transition to LPG to address the significant detrimental effects on environment, climate, and public health of reliance on solid fuels and kerosene for energy [51,52]. Despite this, there is a paucity of evidence on the potential positive impacts that can be achieved by scaling LPG for clean cooking. Indeed, concern has been expressed over the potential additional negative contributions to the climate that might result from the promotion of a fossil fuel such as petroleum-based LPG (although entirely a by-product of petroleum sector refining and production of other fuels). Our findings are fully consistent with recent modelled evidence of the combined climate and health impacts from increased population adoption of LPG in Cameroon [13], across 40 diverse LMIC settings [62] and other countries [63][64][65]. Rosenthal et al. [62] conducted modelling based on a theoretical intervention scenario (25,000 homes) whereby solid fuel use was replaced with cleaner cooking options including (i) improved solid fuel stoves, (ii) an advanced combustion (e.g., with fan assistance) solid fuel stove, and (iii) use of LPG for cooking. The modelling was conducted over a 3-year period with an assumption of 60% adoption. Compared with the improved stove intervention, use of LPG was found to lead to an estimated fourfold reduction in morbidity (averted DALYs) with a corresponding reduction of more than 100,000 tons of CO 2 e emissions. The authors concluded that programs focusing on scaling adoption of LPG were more effective in terms of both health and climate gains than those involving improved and advanced biomass burning cookstoves [62]. Similar to the current analysis, Kypridemos et al. [13] modelled the climate and health impacts of scaling LPG adoption in Cameroon to an aspirational target by the SDG7 time frame of 2030. The Cameroon government's target of 58% of the population using LPG by 2030 (from 20% in 2014) through successful implementation of its national LPG Master Plan was found to result in an estimated 28,000 saved lives and a reduction of 770,000 DALYs [13]. In addition to reductions in health-damaging household air pollution (PM 2.5 ) that led to this health gain, emissions of BC and other climate forcing pollutants were found to be reduced by one-third. These reductions, together with reductions in non-renewable harvesting of wood for fuel, were found to lead to a global cooling effect of −0.1 milli • C by 2030. When forecast for a longer theoretical time frame of 2100 (more relevant for climate impacts), the global cooling effect was found to be −0.7 milli • C using a conservative fraction of non-renewable harvested wood. Consistent with these results is our finding that the health gain that can be achieved from scaling adoption of LPG in Rwanda in line with the NST-1 policy target (a theoretical 7,656 saved lives and 403,664 ADALYs) is not at the expense of climate through scaled adoption of a fossil fuel (with an estimated reduction in 28.5 million MT of CO 2 e emissions and the preservation of 354 million trees). It is also important to note that the estimated health gain from our modelling is likely to underestimate the true impacts of sustained and exclusive adoption of LPG for all domestic cooking. Firstly, modelled reductions in exposure to PM 2.5 from LPG adoption were based on values taken from the HAPIN trial for Rwanda [66] that found households using LPG for cooking (intervention) in the field had an average exposure of 45 µg/m 3 (Johnson et al. personal communication). This level of personal exposure is higher than the WHO Indoor Air Quality Guideline Interim Target 1 level for Health of 35 µg/m 3 , and is likely indicative of community contamination from ambient air pollution (mixed use of LPG and solid fuels was unlikely in the HAPIN trial given that LPG fuel was given for free). Second, PM 2.5 from solid fuel use is now known to be causally associated with a variety of other health issues (including adverse pregnancy outcomes and diabetes) considered in later iterations of global burden of disease estimation than that used for the HAPIT tool. In addition, other conditions including cataracts in adult women, tuberculosis, and laryngeal cancer [67], which are not considered in burden of disease estimation PM 2.5 , have not been included in estimates of health impact for the Rwandan modelling. Fourthly, the modelling was restricted to household personal exposures (main cook and children) and did not include potential detrimental health impacts on catering staff operating in public institutions. Finally, as noted in the study by Kypridemos et al. [13], global warming is likely to increase the global burden of disease [68], thus any reductions in global warming potential from population transition away from polluting solid fuels can potentially reduce this associated disease burden. This has not been considered in the current health impact modelling.

Supply Considerations to Facilitate Population Adoption of LPG for Clean Cooking and Role of the Donor Community
As highlighted in the analysis, supply of LPG (including density of retail outlets) will be a critical issue in facilitating population transition (domestic and institutional) away from reliance on polluting solid fuels to clean cooking with LPG. This supply needs to be sufficient to fulfil scaled adoption across Rwanda, requiring necessary investment in provision of LPG cylinders, stoves, infrastructure, and distribution networks. Rwanda currently relies entirely on importing its supply of LPG for both residential and commercial markets. While value added tax on the fuel is zero-rated, there are currently LPG import taxes including a 1.5% infrastructure development tax and a 1.2% African Union tax.
One exciting new initiative that might benefit countries that currently solely rely on imported LPG, like Rwanda, is the potential development of domestic bioLPG production that can lead to greater diversification of LPG supply, reduce the average LPG wholesale price in the country, and hedge against importation price increases. BioLPG (chemically identical to fossil-derived LPG and derived from fully renewable sources) can be produced from feedstocks such as municipal solid waste (MSW) from appropriately engineered landfills sites [69,70]. Advantages of bioLPG include (i) its renewable credentials that can secure international investment, (ii) its resistance to currency devaluation against the U.S. dollar (currently an issue for imported LPG), and (iii) the fact that it can be blended with imported LPG. Of course, it also retains all the benefits of the current credentials of petroleum-based LPG as a scalable fuel (e.g., highly efficient, not detrimental to climate, and well-liked by consumers). In Kigali, and perhaps other major urban conurbations in Rwanda, there is potential to produce bioLPG from processing of MSW that would secure and enhance the Rwandan LPG market-a potentially important future step in its clean energy road map.
The donor community/Official Development Assistance (ODA) funding also has a key role to play in an effort to support scaling up of clean cooking. Rwanda has been the first country to benefit from the World Bank (WB) Clean Cooking Fund initiative [71]. Under the WB-promoted Rwanda Energy Access and Quality Improvement project, Rwanda will be receiving over US$ 150 million of funding to support access to modern energy for households and public institutions [2,72].

Conclusions
Effective implementation of the National LPG Master Plan for transition away from inefficient polluting solid fuels in Rwanda, replaced by LPG for clean cooking in homes and institutions, will have significant positive gains in terms of health, environment, and climate. The findings from the modelling of an 'intervention scenario' of scaling LPG use towards the 2024 policy target suggest that household LPG use is projected to increase from 5.6% in 2020 to 13.2% by 2024 and 38.5% by 2030. If achieved, this level of adoption would result in the reduction of 7,656 premature deaths and 403,664 disability-adjusted-life-years, as well as 243 million trees saved. Reductions in CO 2 and BC emissions equivalents are estimated to reach 25.6 million MT and 14.9 MT, respectively, by 2030.
To achieve aspirational targets of transition to clean cooking within the SDG 7 time horizon of 2030, significant public and private sector mobilisation is required to effect the required market changes to scale population adoption of LPG. This will include creating favourable conditions for new consumers to adopt LPG as a main source of cooking fuel (e.g., through positive regulation and supporting private sector investment and innovation that can facilitate both adoption and sustained use of LPG, including the likes of pay-asyou-cook smart meter technology and microfinance). The Rwandan National LPG Master Plan, commissioned and supported by the Ministry of Infrastructure, is an important initial step in mapping out the conditions necessary to achieve this scale and the positive health, environmental, and climate gain that can be achieved in this relatively short time frame.
Funding: This study was sponsored by the European Union's Infrastructure Trust Fund through its Clean Cooking for Africa Programme (CCAP), administered by the German Development Bank KfW.

Data Availability Statement:
Most data is contained within the article or Supplementary Material. For any additional data, please contact the corresponding author.