Bangladesh has one of the highest rates of pollution in the world and is experiencing dramatic environmental degradation in the face of explosive population growth. Urban population rose from less than 40 million in 2006 to more than 55 million by 2015 [10]. The present low cost of electricity plays a large role in the economic growth and development of Bangladesh. At the country’s independence from Pakistan in 1971 only 3% of the population had access to the electricity and by 2009, 2012 and 2017 this figure increased to 47%, 59.6% and 80% respectively [11]. However, there is a continuous struggle to overcome energy crisis. Population growth, system losses, and the country’s prior dependence on a now dwindling supply of natural gas are major factors in the struggle. In 2017 the grid capacity was 15,821 MW, including 16,070 MW imported and 2200 MW captive power capacity [11] [12], and 16,070 MW including off-grid RE [13]. By June 2018, the installed generation capacity including captive power, was approximately 18,753 MW. Although this shows a considerable increase in installed capacity in a small range of years, there is still increasing demand for electricity. The government has committed itself at looking for solutions from all angles including importing fuel sources, installing new power plants, and buying electricity from independent power producers to meet the demands. An intergovernmental agreement has been signed between Bangladesh and Russia to install a 2-generation nuclear power plant at Rooppur and Pabna (60 years lifetime). There will be 2 units of 1200 MW each nuclear plant. Rooppur is expected to be commissioned 2024 [14]. This future electricity source gives hope to the government that electricity cost will be competitive though initial investment is very high. Bangladesh is also attracting private investment and independent power producers, and these are producing 54% of the total installed electricity [12] [15] [16]. In recent years, there have also been improvements in technical and distribution losses which decreased from 16.9% in 2009 to 12.2% in 2017 [12].

Due to shortages in natural gas, Bangladesh is increasingly switching to oil and coal-based power generation [17]. Oil and coal are highlighted as the first track, short term option to meet the rising electricity demand in this country [12]. These fuel sources are preferable as they can generate a significant amount of power within the shortest time. However, oil-based power plants are costly compared to gas-based plants, and oil contributes to extensive environmental pollution. These factors impede the process of oil being a long-term solution to alleviate the shortage of natural gas in Bangladesh. The use of indigenous coal is also being encouraged. The plan is to reach 23,692 MW generation from coal by 2022 [18]. Coal, however, is also environmentally costly in terms of mining and pollution [16]. In fact, Bangladesh is predicted to be one of the most carbon emitting countries by 2025 financial year with 83 * 10⁶ tons of CO₂ from 132,520 GWh electricity generation [19].

The PSMP-2016 of Bangladesh [20] is a resource that lists all available energy producing projects in the country and is updated approximately every 5 - 6 years. It documents that 10% of the nation’s grid should be achieved from RE by 2020 and 20% by 2030. These targets are set according to the availability of RE sources in the country. Nevertheless, all of these RE sources together fail to meet proposed grid demands for electricity. RE projects need to be at least developed in accordance as planned to account for a total of 2651MW of electricity generation by the year 2022 [12]. If so, ongoing and forecasted projects RE could feed the national energy grid at a share of around 10% by 2022 [12] [21]. Due to heightened energy demand, the reality for Bangladesh’s power grid in the short term is a mixed energy system consisting of both FF and RE projects.

Research and development are paving the way toward decreasing the cost of RE, although more research needs to be conducted on not just installing different kinds or renewable energy, but increasing the energy capacity of RE and other sustainable technologies. Furthermore, there needs to be more direction on how best to implement RE practices for energy generation capacity and carbon emission reduction in a mixed energy system.

Several methods have been used to investigate the carbon emission in different sectors and regions. Utilizing the zero-sum gains data envelopment analysis (ZSG-DEA) model, Wu et al. [22] examined the carbon reduction potential of China’s key industries and found the electric power, iron and steel industries have the greatest potential for emissions reduction. Lin and Ahmed [23] applied a Logarithmic Mean Divisia Index (LMDI) method to model a variety of factors with carbon emission reduction potential in Pakistan and found population and GDP contributed to increased carbon emissions. The effect of FF was mixed. The LMDI model and input output structural decomposition analysis has also used to investigate the long-term relationship among carbon emissions and the main factors affecting energy intensive industries in China. These authors illustrated that industrial scale and labor production are the main contributors to carbon emission whereas energy intensity has a negative contribution [24]. Yu et al. [25], constructed a panel quantile model to examine how the development of renewable energy and a decline in energy intensity relate to carbon emission reductions in China. Recently, Li et al. [3], used a dynamic Computable General Equilibrium (CGE) model to investigate how the energy supply-side and demand-side policies for energy saving and carbon emission reduction could be synergistic. Xuan et al. [26] used the difference-in-difference (DID) model to promote the reduction of carbon emissions in China’s trading policies. The results show that carbon emission trading policies, economic development, technological research level and opening to the outside world can significantly reduce carbon dioxide emission intensity and promote carbon emission reduction. Using a constrained performance index measure (CPIM) model Ding et al. [27] analyze the strong relationship between energy saving performance and carbon ER in China. Using a panel quantile regression model method, Khan et al. [9] show that financial development has increasing influence on RE consumption. These studies have used many different variables like foreign direct investment, GDP, economic growth, urbanization, and democracy to account for population use of resources [27] [28]. Not only are there discrepancies between the methods and outcome many of these studies are in more developed countries. Quantifying environmental pollution using per-head indicators can also be somewhat insufficient when comparing economically developed countries with underdeveloped countries. This is due to large differences in social life, socio-economic status, large gaps in energy-demand and industry in poorer countries [4]. Research objectives would be improved if energy generating activities were assessed individually, by source, instead of per head.

Indirect effects of energy direction also exist, alongside direct effects. These have important consequences on the environment. Some findings confirm carbon emissions have a direct effect on energy production output [9]. In addition, for energy sources to reach their electricity form, they are required to pass through processes that result in different power capacities and release different amounts of carbon per unit of power generated. Furthermore, different fuel sources can emit different quantities of carbon per unit of power generated. For example, CEI intensity from coal-based power generation is nearly twice the CEI of gas-based power generation [19]. RE engages with negligible direct carbon emissions than fossil fuels. In recent years, much attention has been given to extended input-output models (E-IO) to account for energy use and environmental pollution. Notable works conducted by Cumberland 1966 [29], Leontief and Ford 1972 [30], Bullard-Herendeen 1975 [31], and Griffin and Gregory 1976 [32], have written extensively about the model. E-IO models can provide an appropriate framework for analyzing carbon emission by accounting for each activity that happens in the energy cycle. This method can account for the indirect influence of carbon emissions on RE as well as direct effects. For instance, Peng et al. [33], employed the E-IO model to investigate the potential pathways toward energy-related carbon emission reduction in heavy industrial regions of China. They investigate the potential ways for these zones to achieve energy-related carbon emission reductions, with three northeastern provinces accounting for about 1/10 of China’s energy consumption and 1/6 of China’s CO₂ emissions. Jiang et al. [34], use input-output analysis to focus on structural carbon emission during energy supply and energy demand in Chinese industries and find that in the Chinese energy system, the carbon emissions of traditional high-carbon energy sources (such as raw coal, diesel, and fuel oil) slowly decline, whereas use of low-carbon clean energy sources (such as natural gas) increase. Ren et al. [35] consider an Epsilon-Based Measure (EBM) Data envelopment analysis model from an input-output perspective in the solar photovoltaic industry to analyze carbon emission reduction and efficiency. Luo et al. [36] use the E-IO model to compare and analyze the specific driving forces of change in carbon emissions based on the disaggregation of China's power industry. They illustrate that the embodied carbon emissions of each clean energy sector are relatively lower than the carbon emissions produced by thermal power generation. Recently, Chard et al. [37] use the input-output analysis of environmental expansion to quantify the carbon footprint of selected Australian equity investments. The analysis also calculates the representative employment footprint of these investments to observe the broader impact of the transition to a low-carbon economy.

The present study wishes to address if replacement of a single FF by RE can mitigate carbon emissions. It is still unclear as to where and how in the energy nexus RE should be prioritized, especially in developing countries. Therefore, the current study quantifies energy production and direct carbon emission output of FF and RE resources in Bangladesh and engages an ER model. The study utilizes the 5-year trajectory proposed in the PSMP-2016 which is close to a new review. The ER model is useful as it considers both baseline and projected scenarios and can comprehensively capture the details of both scenarios and compute the emission reduction for the investigated sources [38]. We not only assess the impact of listed energy sources from the PMSP-2016, by using the ER model with sensitivity analysis, we can analyze the degree of carbon emission reduction when any FF is replaced by RE technology built to equal energy generation capacity. This model can then assess where, in the mixed energy nexus, CE reduction can be best implemented.

On this premise, this study employs 3 main steps; 1) we quantify the relationship between energy generation capacity from each source and the relative carbon emission with the support of E-IO analysis; 2) we examine RE effects on carbon reduction by adopting an ER model; and finally, 3) the baseline carbon emission and ER results are subjected to sensitivity analysis. Using this modeling system, we can assess each FF one at a time, against RE, to find which type of FF is best replaced by RE to enhance carbon emission reduction in a mixed energy system. By adopting this approach, we yield useful insight into relationships that exist in the carbon-energy nexus in Bangladesh.

Table 1 presents the direct effects on energy generation capacity activity from each type of energy source and their respective carbon emissions in Bangladesh over 2017-2021. Within-year total energy generation capacity and carbon emission out are shown in the bottom row, and the total energy generation capacity and carbon emission output over the total five year-period for each energy source and their percentage share are shown in the last 4 rows.

Total energy generation capacity increases in Bangladesh by 95,928.1 MW from the baseline year 2017 to the projected year, 2021. This energy generation capacity is influenced to different extents by each FF and RE. Total coal and gas energy generation capacity is 7021 MW and 52,587 MW, respectively. Indigenous oil accounts for 37,916 MW of total energy generation capacity, while imported oil accounts for 5800 MW. In Scenario 1, total carbon emission output from FF energy generation capacity over the 2017-2021 period is 27,676,782 tCO₂eq from coal and 131,061,522.3 tCO₂eq from gas. Carbon emission output from total oil sources is 145,521,822 tCO₂eq. Of the oil sources, 126,214,780.8 tCO₂eq is from indigenous oil, and 19,307,040 tCO₂eq from imported oil.

In Scenario 2 we analyzed the PMSP-16 values using coal as the import model. Imported coal is compared along with indigenous coal, indigenous oil, and indigenous gas. The sum of energy generation capacity from indigenous coal and imported oil, create total energy generation capacity from coal. Likewise, the sum of carbon emission output from indigenous coal and imported coal, make up the total carbon emission output for coal. Within-year total energy generation capacity and carbon emission output are shown in the bottom row, and the total energy generation capacity and carbon emission output over the total five year-period for each energy source and their percentage share are shown in the last 4 rows (Table 3).

Figure 5 then illustrates each FF source and total combined RE sources in Scenario 2.

Energy generation capacity from coal imports in the PSMP-16 remain the same from year to year, so there is no difference in the imported total energy generation capacity output from imports across the 5-year study trajectory in Scenario 1 compared to Scenario 2. Total carbon emission output by total coal sources does change across the trajectory. In Scenario 2 total coal contributes to 50,540,382 tCO₂eq with imported and indigenous coal providing 22,863,600 tCO₂eq and 27,676,782 tCO₂eq, respectively. In Scenario 1 imported oil provides more carbon output (145,521,822 tCO₂eq) and this is due to the high percentage share of energy capacity and carbon emission output from indigenous oil. Therefore, holding imports stable, we can show that energy generation capacity by indigenous coal is lower and releases lower levels of carbon than that of oil.

Total energy generation capacity for RE combined over the 5 years is not different from Scenario 1 and amounts to 10,856.5 MW. Carbon emission output from RE is 110,376 tCO₂eq over the same time-period. For illustrative purposes, the combined RE sources are again disaggregated into individual renewable energy sources as the percentage share of total energy generation capacity and percentage share of total carbon emissions from each source over 5 years and shown in Figure 6. Of the RE sources, solar power (4.59%), then wind power (3.77%), contribute to the greatest percentage share of total energy generation capacity.

We looked at energy generation capacity and carbon emission output from each FF source and RE in the baseline year (2017) and the projected year (2021). In Scenario 1, coal sources combined provide an extra 2254 MW in 2021 compared to 2017. There is no change in energy capacity from imported coal between the years, so it is indigenous coal that contributes for the difference in energy capacity. In 2017, indigenous coal provides 2.83% of the share energy generation capacity and 4.10% of the share of carbon output, whereas in 2021 energy generation capacity rises to 10.57% and carbon output 15.61%. In 2021 compared to 2017, an extra 1225 MW is provided by indigenous gas and an extra 2319 MW from indigenous oil. Carbon emission output from total coal provides an extra 8,885,268 tCO₂eq in 2021 compared with 2017. As there is no difference in carbon tCO₂eq from imported coal between baseline years 2017 to project year 2021 from imported coal, it is indigenous coal that contributes to the excess carbon output in 2021. Indigenous gas accounts for an extra 3,036,873 tCO₂eq of carbon output in 2021.

Figure 7 illustrates the percentage share of energy generation capacity by each FF energy source and all the RE sources combined for Scenario 2 in 2017 (Panel (A)), the first year of the trajectory and 2021 (Panel (B)). The percentage share of carbon emissions by each energy source within this year are also illustrated. The data within each year of the trajectory are mentioned in the text.

In Scenario 2, gas shows the greatest percentage share of energy generation capacity within the baseline year (2017) which is the same as Scenario 1. However, in 2021, at the end of the trajectory, this drops more than 10% to 42.68%

(Figure 7(a) compared with Figure 7(b)). A decrease in the percentage share of energy generation capacity by gas occurred in most years in the trajectory at (53.97% in 2017, 46.05% in 2018, 44.61% in 2019, 45.70% in 2020 and 42.68% in 2021). The percentage share of carbon emission output within each year also decreases over time (except for 2020), dropping approximately 10% by 2021. This measure of carbon output is 49.09% in 2017, 41.80% in 2018, 41.01% in 2019, then 42.84% and 39.65% in 2021, respectively. Therefore, as in Scenario 1, the percentage share of carbon emissions by gas in Scenario 2 closely follows energy generation capacity output, most likely declining as percentage share of energy generation capacity decreases.

The percentage share of energy generation capacity by indigenous oil in 2021 (30.60%) was almost the same as 2017 (30.91%), with yearly fluctuations throughout the rest of the trajectory (37.05% in 2018, 34.5% in 2019, 32.76% in 2020, and 30.60% in 2021). This was probably due to other sources changing in the energy mix. Carbon emission output by indigenous oil contributes 37.75% of the share in 2017, 45.16% in 2018, 42.55% in 2019, 41.23% in 2020 and 38.17% in 2021.

The share of energy generation capacity by all sources of coal in 2017 is 9.09%. Despite some small fluctuations, the percentage share of energy generation capacity by coal rises to 14.98% by 2021 (9.02% in 2018, to 11.22% in 2019, 10.66% in 2020). Of the coal sources, indigenous coal provides 2.83% of the share of energy generation capacity in 2017, and this source rises steadily over time (3.68% in 2018, 6.25% in 2019 and 5.94% in 2020). Notably, by 2021, the projected values rise almost 4-fold to 10.57%. Coal imports initially contributed the greater percentage share of energy generation capacity by coal sources at 6.26%. However, this declines over time (5.34%, 4.97%, 4.72% and 4.41%; 2018-2021, respectively). Therefore, the use of indigenous coal appears to take precedence in of Bangladesh within the projected study period. This is in line with the Bangladeshi government propositions introduced in the literature review.

Carbon emission output by total coal rises over the study period. In 2017, the percentage share of carbon emission output by all sources of coal amounts to 13.15%. At the end of the study trajectory, the percentage share of carbon emission output from all coal sources rises to 22.12% (13.02% in 2018, 16.40% in 2019, and 15.89% in 2020). This rise is mainly due to an increase in indigenous coal. The percentage share of carbon emission output by indigenous coal is observed to rise almost 4-fold from 4.10% in 2017 to 15.61% in 2021. The values for indigenous coal in other years are 5.31% in 2018, 9.14% in 2019 and 8.85% in 2020. That indigenous coal is a factor contributing to the projected rise in carbon emission output by coal is supported by a drop in the percentage share of energy generation capacity by imported coal. The latter coal source contributes to 9.05% of the baseline year (2017) total carbon emission output, whereas by 2021 this measure was lower (6.52%).

As in Scenario 1, the percentage chare of energy generation capacity by RE sources combined increases substantially over time. RE sources contribute 6.03% of the share of energy generation capacity in 2017, 7.88% in 2018, 9.70% in 2019 and 10.88% in 2020. The share of energy generation capacity by RE in 2021 is nearly double that of the baseline year (11.74%). There is also an increase in the percentage share of carbon emission by combined RE sources over the years, and this is sourced from biomass (0.01% in 2017 to 0.05% in 2021).

We next disaggregated the RE sources and quantitated energy generation capacity and carbon emissions. Results are given in Table 4 and Figure 8(a) and Figure 8(b).

The results from Scenario 2, were the same as Scenario 1 Table 4 vs Table 2. Increases in the percentage share of RE over the trajectory was attributed to an increase in the percentage share of energy generation capacity by solar power and wind power over the trajectory (Figure 8(a) and Figure 8(b)).

It is of particular interest to see how carbon emissions change from each energy

source when they each have an equal increase in energy generation capacity. We analyzed this model for Scenario 1 and Scenario 2. For Scenario 1, in which oil is accounted as an import, calculations reveal that a 1% increase in total oil, indigenous oil and imported oil energy generation capacity increases carbon emission output by 1.25%. For each of the oil sources, respectively. A 1% increase in indigenous coal-based energy generation capacity increases carbon emission output to a greater degree (1.48%). Interestingly, a 1% in indigenous gas-based energy generation capacity results in the lowest (0.93%) carbon emission output. A 1% increase in RE based energy generation capacity results in very little carbon emission output (0.0042%).

In Scenario 2 which accounts for coal as an import, calculations reveal that a 1% increase in total coal, indigenous coal, and imported coal-based energy generation capacity increases carbon emission output by 1.46% for all coal sources, respectively. A 1% increase indigenous oil-based energy generation capacity causes slightly lower (1.23%) carbon emission output. When indigenous gas-based energy generation capacity increases by 1%, carbon emission output is 0.92%. Therefore, in both scenarios, gas shows the lowest carbon emission output of the FF sources relative to equal increases in energy generation capacity.

In Scenario 2 a 1% increase in RE based energy generation capacity increases carbon emission output by 0.0042%, so is not different than Scenario 1. Distinctly, energy generation capacity from RE can potentially reduce a significant amount of carbon emissions if it replaces any of the FF. This is analysed further in Sections 4.3 and 4.4.

Baseline carbon emissions are calculated by multiplying the combined RE sources and each individual FF and with its carbon emission intensity (CEI). Each energy source is converted to MWh. The RE project energy generation capacity is then established. The project emission was the subtracted from the BE to obtain the emission reduction. Afterwards, the carbon emissions reduced by RE from FF based energy generation is presented as the ER. Results for each FF type in Scenario 1 are presented in Table 5. These include oil-based imports, indigenous oil, the other indigenous imports, coal, and gas.

In Scenario 1, indigenous gas has the highest BE of carbon in 2017 (1,487,596.74 tCO₂eq), followed by indigenous oil (1,144,041.98 tCO₂eq). Imported oil attributes to lower BE levels of carbon although indigenous coal has the lowest BE carbon emission (124,173 tCO₂eq). The findings are primarily governed by each energy sources CEI. These are 0.9 kWh 0.76 KWh and 0.57 kWh for coal, oil, and gas, respectively. In 2021, indigenous gas again shows the highest BE (3,339,618.25 tCO₂eq) and the increase in BE of indigenous oil increased almost 3-fold. Indigenous coal shows the lowest BE carbon emissions in 2017 and this increases over 10-fold over the 5-year trajectory so that in 2021, the BE value of carbon output from indigenous coal is 1314578.16 tCO₂eq. This latter finding highlights the massive growth in indigenous coal use. Except for biomass, the RE sources in the present study do not show carbon output. The CEI from biomass is only 0.28 kWh. In this regard, the sum of RE CEI values is low. Therefore, there are only minor changes in ER by RE.

We next looked at the ER for Scenario 2. These values included imported oil, and the indigenous FF’s oil, coal, and gas. The results are presented in Table 6.

In Scenario 2, the values for BE in 2017 are similar to those of Scenario 1, the exception being imported coal. The BE of imported coal in 2017 is 274,363.2 tCO₂eq whereas imported oil for the same time-period in Scenario 1 is less (231,684.48 tCO₂eq). The higher values of imported coal across the study period are primarily attributed to the higher CEI of coal. Again, the ER values by RE in Scenario 2 are not extraordinarily different from BE values.

Sensitivity analysis was conducted on emission reduction results. The average baseline emission and average emission reduction stand for the mixed energy baseline emissions and emission reduced by RE, respectively, when RE is modelled into the system at equal energy generation capacity. Sensitivity analysis was conducted for both scenarios in the first year (2017) and projected year (2021) of the trajectory.

Results for Scenario 1 in 2017 are illustrated in Figure 9(a) and Figure 9(b). Results in Figure 9(a) shows the baseline carbon emission effect and Figure 9(b) shows the carbon ER effect on each FF after RE is modelled into the equation and used at equal generation capacity as the FF it replaces. Values given are at −0.25 and +0.25 which are random numbers generated in sensitivity modelling using one factor at a time. Is it observed that at −0.25% and +0.25%, the BE of carbon is significantly higher from indigenous gas-based power and then from indigenous oil. The lowest baseline carbon emissions came from imported

oil and indigenous coal, respectively. Figure 9(b) clarifies Section 4.4 that RE has no effect on carbon ER for any of the FF and carbon output remains very similar to BE levels.

Scenario 1: 2017

In 2021, (Figure 10(a) and Figure 10(b)), sensitivity analysis shows that BE carbon levels (Figure 10(a)) from gas power generation and indigenous oil are very similar and carbon output from both sources is higher than 2017. On the other-hand imported oil is significantly lower than average emissions, and then indigenous coal.

Scenario 1: 2021

By observing Figure 10(b), one can see that there is clearly no effect of RE on reducing carbon emission output from any of the FF.

The procedure is repeated for Scenario 2. Figure 11(a) illustrates the sensitivity results for the baseline carbon emission effect and Figure 11(b) shows carbon ER by RE on each FF when RE is modelled to replace a FF one factor at a time. In 2017, gas power generation shows the highest baseline levels of carbon emission output and indigenous oil is also higher than average emissions (Figure 11(a)). Indigenous and imported coal, respectively, show significantly less baseline emissions of carbon output. RE has no effect of reducing carbon output from any of the fossil fuels (Figure 11(b)).

Scenario 2: 2017

In 2021, random number generation in the sensitivity analysis shows that gas power generation and indigenous oil show much higher BE carbon levels than 2017 (Figure 12(a)), and they are both higher than average emissions. BE from indigenous oil being almost the same as gas. Imported coal has the lowest BE and indigenous coal is also lower than average emissions. Again, RE has no effect on carbon ER (Figure 12(b)).

Scenario 2: 2021