Methane emission from different rice cultivation systems in Bangladesh: A model based approach

Globally, a large amount of methane (CH 4 ) emission is generated from agricultural systems including rice farming in many Asian countries including Bangladesh. However, a reliable estimate of CH 4 emission from rice cultivation is not available, particularly considering the different rice types (i.e., local land races, high yielding varieties (HYV), and hybrid varieties) grown under diverse conditions. Here, we estimated current and future CH 4 emission (both rate and amount) from different rice farming systems in Bangladesh using the IPCC Tier1 method. Model based estimates were validated with a pilot survey and with other studies. Across all rice types grown in different seasons, (i.e., Aus from March to August, Aman from July to December and Boro from December to June), the estimated CH 4 emission in 2020 was at 2348 Gg CH 4 yr − 1 (95% CIs of 799–5628 Gg CH 4 yr − 1 ) while a slightly higher CH 4 emission was estimated at 2376 Gg CH 4 yr − 1 for the year 2060 after considering a 0.5 % cultivable rice land migration to non-agricultural activities. We also found signi�cant differences in CH 4 emission rates among types of rice cultivation and growing season. Average across all seasons, the highest CH 4 emission was from hybrid varieties (225 kg CH 4 ha − 1 yr − 1 ), while the lowest (128 kg CH 4 ha − 1 yr − 1 ) from local land races. In contrast, the same local land races showed the highest emission rates when normalized against yield. Across all rice types, the largest CH 4 emission was during Aman season accounting 61% of the total annual emission. Our model-based estimates reasonably compare with survey-based estimates (r 2 = 0.94, p < 0.01) and fall within one standard deviation of a log-normal distribution of measurement based CH 4 emissions. Our �ndings, therefore, provide a deeper insight into CH 4 emissions from rice cultivation systems.


Introduction
Global warming is one of the most pressing challenges, which is affecting the global energy balance and many other related processes.This unwanted increase in the global surface temperature is affecting lives, living conditions and welfare on the planet (Hoegh-Guldberg et al. 2018).The underlying driving force of global warming is the increase in concentrations of different greenhouse gases (GHGs) such as carbon dioxide (CO 2 ), methane (CH 4 ), and nitrous oxide (N 2 O).Anthropogenic contributions such as fossil fuel burning and land use change are the most responsible agents for the increased atmospheric concentrations in CO 2 , CH 4 , and N 2 O, accounting for 64%, 17%, and 6% of the total radiative forces, respectively (IPCC 2013) Methane is considered as the second dominant greenhouse gas after CO 2 since it is 28 times more powerful than CO 2 , signifying its contribution (~ 16%) to the global warming potential (GWP) (Aydin et al. 2010).Although CH 4 emission from fossil fuel burning is considered to be low, the CH 4 concentration in the atmosphere reached to 1803 ppb in 2011, a 2.5 times increment compared to the pre-industrial era (722 ppb in 1750) (IPCC 2013).This clearly indicates that anthropogenic contributions from sources other than fossil burning are equally important for changing the radiative balance.About 60% of the global CH 4 is emitted from anthropogenic sources, whereas 40% of the emissions come from natural sources (Karakurt et al. 2012).The major natural sources of CH 4 are wetlands and oceans, termites, wild res, grassland (Jones et al. 2005;Agarwal et al. 2009;Cai et al. 2011) while the major anthropogenic sources include enteric fermentation in ruminants, anaerobic decay of organic matter in rice paddies, municipal solid waste land lls, cattle ranching, and manure management (Guisasola et al. 2009;Yang et al. 2009;Park et al. 2011;Chang et al. 2012;Yusuf et al. 2012).While the agricultural sector contributes the highest amount of CH 4 , accounting for 53% of global emissions (Yusuf et al. 2012), rice paddies alone contribute over 11% of anthropogenic CH 4 emissions (IPCC 2013), placing paddy rice cultivation at the top of all other cereals (Linquist et al. 2012).The carbon foot print of global rice production is large (49 ± 4.5 Mt CO 2 -e yr − 1 ) (IPCC 2014).The carbon foot print of rice cultivation may escalate further in the future due to additional rice production for meeting the food demand of the increasing global population.
The CH 4 emission from rice paddies depends on many factors including water management (Wang et al. 2012), organic fertilizer input (Feng et al. 2013), nitrogen (N) fertilizer use (Banger et al. 2012), soil temperature (Khalil et al. 1998), and rice cultivars (Qin et al. 2015).It is also in uenced by elevated CO 2 , temperature (Dijkstra et al. 2012;Bhattacharyya et al. 2013) and soil properties (i.e., SOC and soil pH) (Yan et al. 2005).Among these factors, water management and organic matter application are usually positively associated with CH 4 production.
In Bangladesh, rice is traditionally cultivated in three seasons namely, Aus (from March to August), Aman (July -December), and Boro (December -June) with diverse growing conditions (details of rice growing conditions can be found in SI).Among the three seasons, rice cultivation in Aus and Aman seasons are mostly rainfed while in Boro season, except for deep water rice, it is cultivated entirely under irrigated conditions (Shelley et al. 2016).The formation of CH 4 occurs under anoxic conditions, and CH 4 emission varies widely when rice is grown in irrigated or rainfed conditions (Haque et al. 2019).Additionally, various types of rice, i.e., high yielding varieties (HYV), hybrid varieties, and local land races (cultivars) are grown in each of the three rice growing seasons in Bangladesh, differing in growth period, cultivation conditions and productivity.It is expected that CH 4 emission from modern varieties are higher than local varieties due to longer growth periods, chemical fertilizer input, and water management practices.In contrast, this scenario could be completely opposite when emission rates are normalized against yield, since the yield of local rice is lower than of HYV.However, it is not known how CH 4 emission changes for rice types grown under diverse conditions.
While eld-based measurements could provide reliable estimates of CH 4 emission, it is often di cult to do so for the whole country considering all rice growing seasons and types.As an alternative, modelling of CH 4 emissions using measurement based data was proposed and advocated (Yan et al. 2005;IPCC 2006IPCC , 2019)).Among the various model-based approaches, IPCC 2019 guidelines for National Greenhouse Gas Inventories are considered to be acceptable and most used for reporting of GHG emissions at the national scale.These guidelines are either inventory based (i.e., Tier 1 and Tier 2) or model-based (Tier 3).
The Tier 1 approach provides for simple estimations, based on a generalized default emission factor and scaling factors while the Tier 2 approach uses nation speci c default and scaling factors.In contrast, Tier 3 approach typically estimates CH 4 emissions by combining detailed geospatial information into a model.Since, the data required for Tier 2 and 3 to estimate CH 4 emission are not available and Tier 1 allows for using agronomical data (i.e., crop cultivation area, yield, etc.) that are typically collected by the Department of Agricultural Extension, Ministry of Agriculture, we made a national CH 4 emission inventory using the IPCC 2019 Tier 1 method.We also estimated the contributions of different rice types and seasons.We believe that the Tier 1 method could prove a reliable estimate of CH 4 emission for Bangladesh since most of the input data is available.
Following a similar approach, a number of studies estimated CH 4 emission for Bangladesh (Bachelet and Neue 1993;ALGAS 1995;Rahman et al. 1998;MoEF 2012MoEF , 2002;;Yan et al. 2003Yan et al. , 2009; U.S. EPA 2012; Khan and Saleh 2015;Peters et al. 2017).However, none of these studies considered the large diversity in rice farming systems such as rice growing season, soil topography, soil organic matter content and organic amendments, rice type (hybrid varieties, HYV, and traditional land races) and growth period.Thus, the projected CH 4 emission estimates by these studies are possibly quite far from the real emission while the projections among these studies are also diverse requiring a reliable estimate.For example, using the IPCC 2006 methodology, Yan et al. (2009) estimated CH 4 emission at 1660 Gg CH 4 yr − 1 from all types of rice elds in Bangladesh for the year 2000.Recently, Khan and Saleh (2015) estimated the national CH 4 emisison from the rice elds in Bangladesh for 2009 at 1146 Gg CH 4 yr − 1 using the same methodology.Moreover, use of more complex models such as CH 4 MOD 2.5 , AIRS, SCIAMACHY, and GOSAT created further uncertainty with large variability in estimates ranging between 438 and 6680 Gg CH 4 yr − 1 (Khan and Saleh 2015;Peters et al. 2017).
The acreage of different rice types is under continuous changes due to the adoption of modern varieties (hybrid and HYV) replacing local land races.Population pressure may accelerate the adoption process further necessitating a reliable future projection to estimate its contributions to global radiative forcing.Considering the diversity in rice types and growing conditions, the current study was conducted to estimate CH 4 emission from different rice types grown under different growing seasons for 2020 to 2060.

Rice cultivation systems in Bangladesh
In Bangladesh, rice is grown in three possible rice growing seasons namely Aus,Aman, and Boro (Siddique et al. 2016a).Among the three rice growing seasons, rice in Aman (broadcast and transplanted) is generally planted between the month of July and August and harvested from November to December whereas rice in Boro is planted from December to early February and harvested between April and June.The Aus rice is either broadcast seeded or transplanted during March and April after the pre-monsoon shower and harvested between July and August (Shelley et al. 2016).The rice types grown in different seasons are categorised as local land races, inbred modern varieties known as HYV and, heterosis exploring rice, the hybrid varieties.The three different rice types are grown during all three seasons, except for hybrid varieties that are usually not grown during Aus.More detail about rice cultivation in Bangladesh is provided in the SI.

Data collection
The annual harvested area and yield of different rice types (local, HYV and hybrid) in three possible growing seasons (Aus,Aman, and Boro) were collected from the Department of Agricultural Extension (DAE) and Bangladesh Bureau of Statistics (BBS) (BBS 2020a, b, c;DAE 2021).The data were documented for each of the administrative districts.Moreover, the growth duration of different rice cultivars were collected from the Seed Certi cate Agency (SCA 2014).

Methods for estimating methane emission
The IPCC 2019 Tier 1 guideline was used for estimating CH 4 emissions from rice cultivation systems in Bangladesh.The calculation procedure is outlined in Fig. 1.Having data of the administrative districts, the national emission was calculated from the CH 4 emissions in different districts while the CH 4 emission from each district was calculated separately for different rice types (local, HYV and hybrid) grown under three different growing seasons (Aus, Aman, and Boro).

Calculation of CH 4 emission from rice cultivation systems
In the Tier 1 method, CH 4 emissions from different rice farming systems were estimated using Eq. ( 1) See equation 1 in the supplementary les section.
Where, CH 4 rice is the annual CH 4 emission from rice cultivation in Gg CH 4 yr -1 , EF i,j,k is the daily emission factor for i, j, and k conditions in kg CH 4 ha -1 day -1 , t i,j,k is the cultivation period of rice for i, j, and k conditions in days, A i,j,k is the annual harvested area of rice for i, j, and k conditions in ha yr -1 , and i, j, and k represent the different ecosystems, such as, water regimes, type and amount of organic amendments, and other conditions under which CH 4 emissions from rice may vary while 10 -6 is the conversion factor of methane emission to Gigagram.
As shown in equation ( 1), the daily speci c emission factor was calculated by multiplying the baseline emission factor (EF) with various scaling factors (SF) to account for the contributing factors using Eq. ( 2) See equation 2 in the supplementary les.
Where, EF i,j,k is the adjusted daily emission factor for a particular harvested area, EF c is the baseline emission factor for continuously ooded elds without organic amendments, SF w is the scaling factor to account for the differences in water regime during the cultivation period, SF p is the scaling factor to account for the differences in water regime during season prior to rice cultivation, SF o is the scaling factor accounting for both type and amount of organic amendments applied, and SF s,r is the scaling factor for soil type, rice cultivar etc.The latter scaling factor was not considered in this study.

Baseline emission factor for continuously ooded elds without organic amendments
The IPCC provides a default CH 4 emission of 0.85 kg ha -1 day -1 with con dence intervals (CI) of 0.58-1.26for a continuously ooded rice cultivation during the entire growth period without any organic amendments for the South Asia region (IPCC 2006(IPCC , 2019)).Since there is no country speci c data for Bangladesh, we used this value as the base line emission factor.
2.6 Scaling factor for water regime pre and during rice cultivation IPCC provided scaling factors for pre and in-season water regime (SF w ) highlighting two different scenarios, i.e., an aggregated case and a disaggregated case (Table S1).The SF w for different rice types grown in different seasons in Bangladesh were based on Yan et al. (2005), customized from the IPCC scaling factors by considering the ecology of rice cultivation, soil physiography, climatic data and management practices (Table 1).

Scaling factor for organic amendment
IPCC 2019 provided the scaling factor for both type and amount of organic amendment incorporated during the rice cultivation (Table S3).According to IPCC guidelines, the scaling factor for organic amendments was calculated using Eq.(3) See equation 3 in the supplementary les section.
Where, SF o is the scaling factor for both type and amount of organic amendment applied, ROA i is the application rate of organic amendment i, in dry weight for straw and fresh weight for others in t ha -1 , CFOA i is the conversion factor for inclusion of the relative effect of organic amendment i.
The organic matter amendments in rice elds of Bangladesh include rice straw and compost or cow dung amendments.The straw incorporation rate was calculated from the projection of straw yields using the grain yield data, collected from the Department of Agricultural Extension, Ministry of Agriculture while the straw retention rate was determined by synthesizing physiography, rice types, and growing seasons.For detailed calculations see the SI (Tables S5 and Figs.S1-3).
In Bangladesh, the application of compost/farm yard manure (FYM) is diverse while the rate of application is unknown.It is also not possible to have a valid rate of compost/FYM application since there is no documented information in the Department of Agricultural extension (DAE) in Bangladesh.
Generally, agriculture lands receive a variable amount of Compost/FYM.Based on expert's opinion, we considered that the average compost/FYM application rate is 2.5 t ha -1 as a fresh weight, irrespective of land types.However, we did not consider the compost/FYM application for local rice cultivation because it is principally cultivated in rainfed low land and deep-water conditions.On the other hand, green manure is also being applied in rice elds, but the incorporation rate is very low.So, in this study, we did not consider any green manure amendments.For timing of organic matter amendments, we considered that the organic amendment was made within 30 days prior to seedling transplanting (BRRI 2018).

Growth duration of rice
In order to calculate the duration of rice cultivation, we followed a systematic scheme where the weighted average value was computed using the lifespan of individual rice cultivars and the harvested area by applying Eq. ( 4) See equation 4 in the supplementary les section.
The lifespan of modern cultivars was taken from the Seed Certi cation Agency (SCA 2014).On the other hand, we used the literature data and expert judgment for estimating growth duration of local land races (Rahman et al. 2008, 2009, Siddique et al. 2013, 2016b, Akter et al. 2016, 2017, 2018;Halder et al. 2016;Islam et al. 2018a, b;Mia et al., in preparation) 2.9 Validation of the estimate with a survey A survey was conducted among the farmers in six different districts (Bagerhat, Barishal, Patuakhali, Tangail, Rangpur and Cox's bazar) to validate the scaling factors including 20 farmers from each district.The questionnaire and procedure of survey conduction can be found in SI (S 3.3).Emission factors for each of the districts were calculated for different types of rice (local, HYV and hybrid) grown in different seasons (Aus, Aman and Boro) using the farmers' data.Next, survey based CH 4 emission rate was plotted against the model-based estimate to determine the t while this plot was also compared with one-to-one line.

Spatial distribution of CH 4 emission
The annual CH 4 emission from different districts was visualized using ArcGIS 10.8.

Prediction of future emission
For future projections of methane emission up to 2060, the rice acreage was analysed using a suitable curve tting on the historical acreage (Fig. S4).Using the parameters of the best t equations, we calculated the future acreage of rice.Similarly, we analysed the hybrid coverage increment between 2007 and 2017.An average of this increment was ~1%, which we used for projecting the hybrid rice acreage in the future.Additionally, we considered that 0.5% of the land will be migrated due to the non-agricultural activities although this gure was suggested to be ~1% (Hasan et al. 2013).However, all other possible changes in soil-atmospheric conditions including climate change were not considered for.Finally, the emission of each of the rice types was projected by multiplying acreage with respective current emission rate of that rice types.

Statistical analysis
Calculated emission rates were analysed using JMP 8.0 (SAS Institute) to examine effects of rice type (i.e., Local, HYV, and Hybrid) and cultivation season (Aus, Aman, and Boro).One-way analysis of variance (ANOVA) was performed on mean emission rates from different districts while the districts were used as random factor.The means were separated using Tukey's HSD (Honestly Signi cant Differences) considering a probability of 5%.

Spatial CH 4 emission
Spatial distribution of methane emission can help to make policy makers aware about their roles in reducing emissions.In Bangladesh, the annual CH 4 emission rates varied widely among different regions (divisions) (p<0.001,Fig. 2).Across all rice growing seasons, the largest emission rate was estimated at 283 kg CH 4 ha -1 yr -1 from Aman rice in Chittagong while the lowest rate (258 kg CH 4 ha -1 yr -1 ) was estimated in Shylet.Similarly, the largest annual emission rate was estimated at 173 kg CH 4 ha -1 yr -1 from Boro rice in Shylet while the lowest estimation (99 kg CH 4 ha -1 yr -1 ) was in Barisal.The highest annual CH 4 emission rate in Aus season was from Dhaka (59 kg CH 4 ha -1 yr -1 ), while the lowest rate was estimated at 56 kg CH 4 ha -1 yr -1 for Mymensingh.When accounting for land area for each rice type and season, the variability was evident for different districts (Fig. 3, see details in Figs.S6-8 and Table S7).As shown in Fig. 3, methane emission was relatively large in the district Dinajpur, Mymensingh, Naogaon, and Sunamganj, while it was relatively small in Bandarban, Chuadanga, Dhaka, Jhalakati, Khagrachhari, Madaripur, Manikganj, Meherpur, Munshiganj, Narayanganj, Rajbari, Rangamati, and Shariatpur districts.
Averaged across all rice types and growing seasons, our estimated national CH 4 emission rate from rice cultivation was 2348 Gg CH 4 yr -1 with 95% CIs of 799-5628 Gg CH 4 yr -1 in the year 2020.

Annual CH 4 emissions from different rice types
Across all rice growing seasons, the CH 4 emission rate varied signi cantly for different rice types (p<0.001,Fig. 4A).The CH 4 emission rate from hybrid varieties was 76%, and 37% higher than for local varieties and high yielding varieties, respectively.When emission was scaled against yield, the largest emission rate was estimated for local rice varieties (88 g CH 4 kg -1 ), or 69% and 34% larger than hybrid and HYV varieties, respectively (Fig. 4B).

Annual CH 4 emission rates from different rice season
Average across all rice types, the estimated CH 4 emission rate varied signi cantly among rice growing seasons (p<0.001,Fig. 5A).Among the rice growing seasons, the CH 4 emission rate was signi cantly larger in Aman (242 kg CH 4 ha -1 yr -1 ), followed by Boro (160 kg CH 4 ha -1 yr -1 ) and Aus season (58 kg CH 4 ha -1 yr -1 ).Similarly, the yield normalized CH 4 emission also varied among rice growing season with the largest emission rate (109 g CH 4 kg -1 ) from Aman cultivation while it was 75% lower from Aus rice cultivation when compared with Aman cultivation (Fig. 5B).

Validation of the methane emission rate
The survey based CH 4 emission rate was linearly related to our model-based estimates (r 2 =0.94, p<0.01).
However, the survey-based estimate is lower by ~10% than our model-based estimate and thus, the plot fell below the one-to-one line with more inclination for the HYV and hybrid rice (Fig. 6).

Future projection of methane emission
The projection of total CH 4 emission from different rice types is presented in Fig. 7 (see details in Fig. S5).
Provided that all management remains similar and changes in rice cultivation follow the past trends (Fig. S4), the total CH 4 emission was projected to increase from 2348 Gg CH 4 yr -1 in the year 2020 to 2392 Gg CH 4 yr -1 in 2060, a 1.87% increase in CH 4 emission when no land would be migrated from agricultural to non-agricultural purposes.However, the increment will be only 1.18% if 0.5% of the agricultural lands would migrate to non-agricultural activities.Moreover, the projection without land migration showed that the CH 4 emission from modern rice varieties (HYV and hybrid) would increase by 3.5% in 2060 whilst the emission from local varieties would decrease by 9% (Fig. S5).

CH 4 emission from different rice cultivation systems in Bangladesh
Diversity in rice growing conditions and management practices can affect CH 4 emission rates.The CH 4 emission rates varied across different regions of the country because of the speci city of rice adoptions under diverse agro-climatic conditions.Among different districts, the cumulative CH 4 emissions from three rice growing seasons was higher in Dinajpur, Mymensingh, Naogaon, and Sunamganj districts.This was possibly due to larger acreage of rice cultivation with modern varieties (Table S7).In addition, water management could also have contributed to this larger emission rates, since irrigation (i.e., intermittent drainage) is also practised.
Rice types (local vs HYV), when grown in different conditions can generate different amounts of CH 4 .
Across growing seasons, the CH 4 emission rates from modern rice varieties i.e., hybrid and HYV were 76 and 37% higher than for local varieties (Fig. 4).This was probably associated with multiple factors, speci cally water regime in and before growing season, and organic matter amendment (Wang et al. 2012;Feng et al. 2013).First, local Aus rice cultivars are usually grown without irrigation and often are affected by drought.Therefore, the in-season water management emission factor (EF) is considered as 0.16 while its pre-season water regime is quite dry, i.e., long drainage, (EF ~0.89,Table 1).Moreover, in the Boro season, it is usually grown in the lowlands as deep-water rice.Therefore, the in-season water management EF is low, 0.06.In contrast, the assigned in-season water management emission factors for HYV and hybrids are much higher since they are grown with multiple drainage in the Boro season (EF= 0.55) and regular rainfed conditions (EF=0.54) in the Aman season.Similarly, the pre-season water management EF was also high, i.e., 1.0 and 2.41 during Aus, Aman and Boro season, respectively (Table 1).Secondly, the HYV and hybrids are cultivated with high inputs including organic matter and nitrogenous fertilizers (Table S3 and S4) while local rice is cultivated with minimal or without any organic or inorganic fertilizers.
The yield normalized CH 4 emission from local rice varieties was higher than modern rice varieties (HYV and Hybrid) (Fig. 4) suggesting that local rice has a much higher carbon foot print than modern cultivars.Therefore, reducing carbon foot print of rice production without compromising yield, HYV and hybrid cultivation should be considered.
The rice growing seasons in Bangladesh with large variability in growing conditions including water regimes can affect CH 4 emission rates.These changes in growing conditions also change the adaptability of rice varieties (local, HYV and hybrids).The estimated CH 4 emissions from different rice growing seasons were different with the largest rates from different rice grown during Aman season (Fig. 5).The higher EF associated with in-and pre-season water regime likely contributed to the larger CH 4 emission.For example, the EF for pre-season was 2.41 in Aman season vs 1.00 in Boro (Table 1).A relatively larger emission from the rice grown during Boro season than Aus season, was due to a larger coverage of HYV and hybrid (99% in Boro vs 88% in Aus season).Moreover, the duration of the Boro rice was longer than rice grown in other seasons.For instance, the average duration of HYV and hybrid rice was 149 d during Boro season while it was 117 d for the same rice grown in Aus season (data not presented).

Comparison between earlier estimates of CH 4 emission from rice elds using IPCC methodology
Methane emission estimates for Bangladesh reported in different studies were quite diverse ranging between 375 and 3110 Gg CH 4 yr -1 with a median value of 1032 Gg CH 4 yr -1 (Fig. 7).Across all rice types grown in different seasons, the estimated annual CH 4 emission rate in our study was 2348 Gg CH 4 yr -1 (Fig. 7), which was higher than most of the previous estimates made using the same method (ALGAS 1995;MoEF 2002MoEF , 2012;;Sass 2003;Yan et al. 2009;U.S. EPA 2012;Khan and Saleh 2015;FAO 2020).This discrepancy was possibly due to not accounting for several contributing factors, such as duration of rice cultivation, water management, and organic matter amendments in the other studies, but which, were included in the current study (MoEF 2002(MoEF , 2012;;Yan et al. 2009;Khan and Saleh 2015).Besides, we used topography customized scaling factors for water management practices (during and before rice cultivation), which, were not considered in the previous estimation studies (MoEF 2002(MoEF , 2012;;Khan and Saleh 2015).Although, it was not possible to access data for organic matter and straw incorporation, we used projected values using literature data and expert judgement (see Figs. S1-3 and Table S5).Therefore, it is not unlikely that our estimate surpassed most of the previous estimates.However, given the importance of organic matter incorporation for CH 4 emission in rice (e.g., Yan et al. 2005;Das and Adhya 2014), we believe that this type of management needs to be included for more accurate estimates of CH 4 emission.Our estimate is also somewhat inconsistent with estimates based on GIS and processbased simulation models.Different values are also evident between model (log mean=3.25,dispersion=0.43)and GIS based estimates (log mean of 3.01 and dispersion of 0.30) (Fig. S10).Multiple factors could be responsible for these observed differences.For instance, GIS and processes-based estimates applied different methodological approaches while the models also require different input variables (see supplementary Fig. S9) (Yao et al. 2006).
We compared our estimated CH 4 emission with the eld-based emission rates of 1081 eld experiments in different countries across the world (Fig. 9).The log normal mean emission rate of these experiments was 0.10 kg CH 4 ha -1 d -1 with a standard deviation of 0.53 kg CH 4 ha -1 day -1 .Our mean estimated daily CH 4 emission rate was 0.07 kg CH 4 ha -1 day -1 which was within one standard deviation of the normal probability distribution.Similarly, the estimated daily emission rate from local, HYV and hybrid rice types (averaged across season) and during Aus, Aman, Boro (average across rice types) were also close, within one standard deviation of the normal distribution.However, our mean estimate was lower than the average value across all observations.Moreover, we observed a signi cant relationship between the survey-based CH 4 emission rates with model-based estimates (r 2 =0.94, p<0.01) although survey-based estimate underestimated the emission by ~10%.Considering all these together, we believe that our estimate is reliable although estimates based on actual measurements could provide a more accurate result.

Limitations of the study
Methane emissions from rice elds do not only depend on the duration of rice growing season, water regime during and before the cultivation season, and types and amount of organic matter amendments, but also on various agro-climatic factors, such as soil pH, rice cultivars, root exudates, etc., which were not considered in this study.It is also possible that individual rice cultivars may be different under each of the rice type categories (i.e., local, HYV and hybrids), which however, were not considered in this study.In future work, inclusion of these contributing factors could improve the CH 4 emission estimates from rice elds in Bangladesh.We acknowledge that assignment of scaling factors and default emission factors carry some uncertainties since these were based on literature and modelled data.Moreover, our prediction of CH 4 emission carries some uncertainties since the calculation was based on literature data and on expert opinion.

Conclusions
This study estimated CH 4 emission rates from different rice cultivation systems in Bangladesh using the IPCC Tier 1 method.The CH 4 emission from all rice types grown under three different seasons was estimated at 2348 Gg CH 4 yr -1 with 95% CIs of 799-5628 Gg CH 4 yr -1 .Our estimate exceeds most of the previous estimates, possibly due to the inclusion of additional contributing factors to CH 4 emission, while it falls within one standard deviation of the normal probability distribution of measurement-based estimates.Additionally, our results indicated that production of modern rice varieties (HYV and hybrid) contributed 87% of the annual methane emission while rice grown in the Aman season contributed the highest CH 4 emission (61%).In contrast, the yield normalized CH 4 emission showed a completely opposite scenario with the largest emission rate from local rice varieties, 67% higher than for modern rice varieties (HYV and hybrid).Moreover, our results provide compelling evidence that the emission will not signi cantly change by 2060, if 0.5% agricultural land migration to non-agricultural activities are included.Therefore, our ndings provide important insights into CH 4 emissions from different rice types grown under diverse growing conditions.

Figures Figure 1
Figures

Figure 7
Figure 7 Methane emission by sectors: A comprehensive review of emission sources and mitigation methods.Renew Sustain Energy Rev 16:5059-5070.https://doi.org/10.1016/j.rser.2012.04.008Table * Values within the parentheses are the customized scaling factors used in this study (adopted from IPCC 2019 guidelines)