Introduction

Greenhouse gases (GHGs) significantly contribute to global warming, with worldwide GHG emissions reaching 59 ± 6.6 Gt CO2 eq/yr, representing a 54% increase compared to 1990 levels (IPCC 2022). Methane (CH4), the second most influential GHG, reached an atmospheric concentration of 1877 ppb in 2019, marking a 260% increase compared to pre-industrial levels (WMO 2020; UNEP 2021). At the United Nations Climate Change Conference COP26 of 2021 in Glasgow (Scotland), 105 countries pledged to reduce CH4 emissions in what was termed the ‘global methane pledge’ (Arora and Mishra 2021). Rice paddies are recognized as significant contributor to CH4 emissions in the agricultural sector, accounting for 8% of global CH4 emissions between 2008 and 2017 (Saunois et al. 2020). In South Korea, CH4 emission from rice paddies amounted to 5.7 Mg CO2 eq/yr. in 2020, constituting 27% of the total GHGs emissions in agricultural sector (ME 2022).

CH4 is generated under anaerobic conditions by microbiological processes within flooded paddies and primarily released into the atmosphere through aerenchyma of rice plants (Aulakh et al. 2001; Malyan et al. 2016). CH4 emissions from rice paddies are heavily influenced by diverse agricultural practices and environments, particularly water and organic matter management (Aulakh et al. 2001). Yagi et al. (1996) demonstrated a 42–45% reduction in CH4 emission during the cultivation period. Yagi and Minami (1990) noted that application of organic matter, such as rice straw, increased CH4 emissions by 1.8–3.5 times. Sass et al. (1994) reported a correlation between CH4 emissions and soil sand content. Watanabe and Kimura (1998) conducted a study comparing differences in CH4 emissions among three Japonica rice cultivars.

IIPCC (2019) has outlined a three-tier method for estimating CH4 emissions from rice paddies based on agricultural practices: Tier 1, 2, and 3. The Tier 1 method uses the default emission factors provided by the IPCC and applies to countries, where rice cultivation contributes relatively less to CH4 emission, or where research on country-specific emission factors is insufficient. Meanwhile, the Tier 2 method utilizes country-specific emission factors derived from field data collection, reflecting the local impact of various crop management practices. The Tier 3, the most advanced approach, involves, estimating CH4 emissions through models and monitoring networks driven by high-resolution activity data (IIPCC 2019). In South Korea, the Tier 2 approach, integrating country-specific emission factors, is employed for CH4 emissions estimations due to an insufficiency of data required for the Tier 3 method (ME 2022). Several country-specific factors have been developed to implement the Tier 2 method in South Korea. These factors include baseline emission factor (Park and Yun 2002; Kim et al. 2013), water management scaling factor (NIAS 2014), and organic matter scaling factors, such as rice straw (Ju et al. 2013), and green manure (NIAS 2014).

CH4 emissions are calculated based on daily emission factor, rice cultivation area, and cultivation period (ME 2022). However, obtaining the rice cultivation area data relies on statistical surveys, introducing uncertainty. Moreover, Baek et al. (2023) indicated a decline in CH4 emissions corresponding to the reduction in rice paddy areas in Korea, emphasizing the need for immediate surveys to accurately track changes in rice cultivation areas for CH4 emissions estimation. Furthermore, the daily emission factor for CH4 requires a scaling factor for organic matters, currently only considering the incorporation of rice straw. With increasing interest in livestock forage and upland crops, there’s been an expansion of double cropping paddy area―cultivating barely or Italian ryegrass in winter post rice harvest―in South Korea. The residue from these winter crops influences paddy CH4 emission as it acts as fertilization for organic matter (Zou et al. 2005). Therefore, achieving an accurate estimation of CH4 emissions demands a method that accurately surveys both rice paddy and winter crop cultivation areas.

Advancement in remote sensing and satellite technology have facilitated the monitoring seasonal crop cultivation (Bazzi et al. 2019; Sousa and Small 2019; Mansaray et al. 2019). Various satellites, such as Landsat, MODIS, and Sentinel have been instrumental in classifying crop cultivation. Tennakoon et al. (1992) utilized Landsat thematic mapper (TM) images and maximum likelihood method to estimated rice paddy cultivation areas and production. Hong et al. (2001) employed Landsat TM, applying rule-based, supervised, and unsupervised classification to calculate paddy area. Zhang et al. (2020) classified paddy fields using MODIS enhanced vegetation index (EVI) data and algorithms such as super vector machine(SVM)-recursive feature elimination (RFE) and stack. Bazzi et al. (2019) and Lee et al. (2021) utilized Sentinel-1 data, decision trees and random forest algorithms for estimating rice cultivation areas. Ni et al. (2021) employed Sentinel-2 images with one-class support vector machine (OCSVM) for pixel-level paddy classification. Soriano-Gonzalez et al. (2022) monitored rice cultivation and predicted yields by leveraging indices like bare soil index (BSI), normalized difference vegetation index (NDVI), and normalized difference water index (NDWI) derived from Sentinel-2 data. Among these satellites, particularly the European Space Agency’s (ESA) Sentinel-1 and Sentinel-2, offer high resolutions up to 10 m, and temporal resolutions of seven, five days, respectively. Satellite image analysis commonly employs indices such as NDVI and NDWI among various others (Huete et al. 2012; Xue et al. 2017; Ni et al. 2021). Hong et al. (2022) classified rice and winter crop cultivation fields using NDWI and NDVI calculated from Sentinel-2 imagery. In estimating CH4 emissions, satellite-derived rice paddy area data has been instrumental (Sun et al. 2017; Sudarmanian et al. 2019). While, CH4 emissions in South Korea have been estimated considering the impact of winter crops using drone images (Park et al. 2020; Jang et al. 2021), nationwide CH4 emissions estimation using drone is difficult due to their limited coverage. Therefore, utilizing satellite images is necessary for estimating nationwide CH4 emissions in South Korea. To improve the assessment of CH4 emissions from rice paddies, obtaining high-resolution data on rice paddy and winter crop cultivation areas is crucial.

This study aims to extract rice paddy and winter crop cultivation areas using satellite images and to subsequently estimate CH4 emissions based on the extracted data for the year 2020 in South Korea.

Materials and methods

Procedure of CH4 emissions from rice paddy

This study estimated CH4 emissions from rice paddies across the entirety of South Korea, spanning a geographical area between longitude 125°33′ E to 130°12′ E and latitude 34°15′ N to 38°50′ N. Given in Fig. 1 are the administrative districts and digital elevation model (DEM) in South Korea. South Korea is divided into eight provinces and 159 cities, featuring flat terrain in the west and mountainous terrain in the east (Fig. 1). In 2022, the rice cultivated area in South Korea encompassed 732,070 ha, with approximately 70% of rice paddies situated in western regions, including Gyeonggi (GG), Chungcheongnam-do (CN), Jeollanam-do (JN), and Jeollabuk-do (JB) provinces (Agricultural Area Survey 2022).

Fig. 1
figure 1

Maps of a city-level administrative district and b DEM in South Korea

Full size image

The estimation of CH4 emissions from rice paddy comprised four steps: data preparation, classification of paddy and winter crop, estimation of CH4 emission in paddy field, and spatial analysis of CH4 emission (Fig. 2). Initially, three data types were gathered: satellite images (Sentinel-1 and Sentinel-2 from ESA), statistic data on water management and rice straw incorporation sourced from the microdata of Census of Agricultural, Forestry, and Fisheries Korea (2020) and spatial data obtained from smart farm map created by the Rural Development Administration of Korea (www.data.go.kr), which was classified using 2020 aerial images. Subsequently, the areas of rice paddies and winter crop cultivation were extracted leveraging Sentinel-1 and Sentinel-2 imagery. Rice fields were classified using the backscatter coefficient of Sentinel-1 images to sort flooded paddy, while winter crop cultivation classification employed NDVI values derived from Sentinel-2 to identify areas with active vegetations growth (Lee et al. 2021; Hong et al. 2022). The accuracy of rice classification was evaluated by comparison with the smart farm map. However, due to insufficient ground truth data for winter crop cultivation, an accuracy assessment of winter crop was not feasible. Following classification, CH4 emissions were estimated at the city-level using scaling d derived from classification results and statistical data. Estimation involved employing the Tier 2 equation of IPCC guidelines and country-specific emission factors of Korea (IIPCC 2019; ME 2014). Two methods were employed for CH4 emissions calculation in South Korea, differing in their approach to organic matter. The first method considered solely rice straw as organic matter (current Tier 2 method), while the second incorporated both rice straw and winter crop as organic matter (modified Tier 2 method). The resulting CH4 emissions estimates were then visualized on a city-level map and compared between the two methodologies.

Fig. 2
figure 2

Procedure of estimating CH4 emission in this study

Full size image

Collection of satellite image data

Table details the satellite image data employed to investigate rice and winter crop cultivation areas. Sentinel-1 satellite image data were accessed via Google Earth Engine (GEE), offering preprocessed backscatter coefficient value. The sensing period of Sentinel-1 images spanned from April 1 to July 31, 2020, chosen specifically to delineate flooded paddy areas before the transplanting season. This study utilized Sentinel-1 interferometric wide-swath (IW) ground range detected (GRD) products with vertical-horizontal (VH) polarization, generally used for ground observations (Nguyen et al. 2016). A total 140 images were used in this study, featuring a temporal resolution of seven days and a spatial resolutions of 10 m (Table 1). Sentinel-2 image data were retrieved from Copernicus Open Access Hub (https://scihub.copernicus.eu/dhus) by the ESA. The sensing period of Sentinel-2 images ranged from February 1 to April 30, 2020, covering the winter crop cultivating season (Kim et al. 2018; Ahn et al. 2019). This study utilized Sentinel-2 Level-2A Bottom-Of-Atmosphere (BOA) products, offering atmospherically corrected surface reflectance. Among the 13 bands available from Sentinel-2, Red (Band 4) and Near InfraRed (NIR) (Band 8) data were utilized to calculate NDVI. A total 529 images were employed in this study, characterized a temporal resolution of five days and a spatial resolution of 10 m (Table 1).

Table 1 Information of Sentinel-1 and Sentinel-2 images for classification of rice and winter crop cultivation area
Full size table

Classification of rice and winter crop cultivation

Classification of rice paddy cultivation area

Figure 3 shows the rice paddy classification process, based on decrease in backscatter coefficient values observed during transplant season, as described by Lee et al. (2021). Specifically, pixels exhibiting consistently low backscatter coefficient values throughout the cultivation season (from April 1 to July 31, 2020) were classified as rice paddies. To create a random forest model for pixel-level classification, nine cities with cultivated areas exceeding 10,000 hectares were randomly selected for trained data. Ground truth―derived from the smart far map―were used to identify rice paddy or non-paddy areas within these cities. The scikit-learn package in Python was utilized to construct the random forest model, employing the ground truth data and minimum of backscatter coefficient value during sensing period as trained data. The dataset was split into training and validation sets at an 8:2 ratio, and the scikit-learn Random Forest model’s parameters, including the maximum depth of tree (max_depth), the number of trees (n_estimators), and the number of jobs in parallel (n_jobs), were adjusted to optimize model performance. The best-performing model achieved an accuracy of 0.76 with parameters set to max_depth was 5, n_estimators was 5, and n_jobs was 2, thus serving as the pixel-level classification model. Pixels with a backscattering coefficient around-24 or lower were classified as rice fields by this generated random forest model. To extend pixel-level classification results to plot-level classification, plots with a pixel area exceeded a defined threshold percentage of the plot area were classified as rice paddies. The threshold percentage was determined by evaluating accuracy and kappa coefficients across three cities, comparing various thresholds. The accuracy and kappa coefficient calculations were conducted using Eqs. 1, 2, and 3 (Cohen 1960).

$$p_{{\text{o}}} \left( {{\text{Accuracy}}} \right) = \frac{{{\text{TP}} + {\text{TN}}}}{{{\text{TP}} + {\text{TN}} + {\text{FP}} + {\text{FN}}}}$$
(1)
$$p_{e} = \frac{{\left( {{\text{TP}} + {\text{FP}}} \right) \times \left( {{\text{TP}} + {\text{FN}}} \right) + \left( {{\text{TN}} + {\text{FP}}} \right) \times \left( {{\text{TN}} + {\text{FN}}} \right)}}{{\left( {{\text{TP}} + {\text{TN}} + {\text{FP}} + {\text{FN}}} \right)^{2} }}$$
(2)
$${\upkappa } = \frac{{p_{{\text{o}}} - p_{e} }}{{1 - p_{e} }}$$
(3)

where \(p_{{\text{o}}}\) is overall accuracy, TP, TN, FP, FN indicate True Positive, True Negative, False Positive, False Negative for each, \({\upkappa }\) represents kappa coefficient, and \(p_{e}\) denotes ratio of correct classification accidentally.

Fig. 3
figure 3

Procedure of paddy classification using the backscatter value

Full size image

Classification of winter crop cultivation area

Figure 4 illustrates the process of classification of winter crop cultivation, primarily using NDVI analysis, which identifies vegetation areas during the winter season preceding rice cultivation (Hong et al. 2022). Winter crop cultivation areas were classified among areas classified as rice fields in summer. NDVI values were computed from the red and NIR bands of Sentinel-2 images using Eq. 4. Pixels exhibiting an NDVI surpassing the threshold of 0.6―suggested by Hong et al. (2022)―from February 1 to April 30, 2020, were designated as areas cultivated with winter crop. To extend pixel-level classification results to plot-level, plots where the ratio of pixel area exceeded a predetermined threshold of 30%―as indicated by Hong et al. (2022)―were identified as plots cultivated with winter crops. Due to a lack of ground truth data, the accuracy of the classification results for winter crop cultivation was not evaluated in this study. However, the selected NDVI threshold values of 0.6 had been previously established to differentiate between field crop and noncultivated areas in Korea (Na et al. 2016, 2017; Yoo et al. 2017). Furthermore, as reported by Hong et al. (2022), the classification method employed in this study demonstrated accuracy in specific regions (Hong et al. 2022). Hence, the classification outcomes from this study were deemed applicable for calculating CH4 emissions.

$${\text{NDVI}} = \frac{{{\text{NIR}} - {\text{Red}}}}{{{\text{NIR}} + {\text{Red}}}}$$
(4)
Fig. 4
figure 4

Source: (Modified from Hong et al.(2022))

Procedure of winter crop field classification using the NDVI

Full size image

Estimation of CH4 emissions from rice paddy

The estimation of CH4 emissions from rice paddies utilized the Tier 2-level equation of IPCC guidelines, shown by Eq. 5 (IIPCC 2019). The emissions of CH4 in rice paddy were calculated by multiplying daily emission factor of CH4, the duration of rice planting, and cultivated area of rice. The daily methane emission factor, calculated through Eq. 6, incorporates factors related to water and organic matter management, which are contingent upon the cultivation method.

$${\text{CH}}_{{4{\text{Rice}}}} = \sum \left( {{\text{EF}} \times t \times A \times 10^{ - 6} } \right)$$
(5)

where \({\text{CH}}_{{4{\text{Rice}}}}\) is annual CH4 emissions (Gg CH4/yr), EF denotes daily emission factor (kg CH4/ha/day), t indicates cultivation period of rice and A represents cultivation area of rice. The cultivation period of rice was 137 days, as used by Korea Ministry of Environment (ME 2021). The cultivation area of rice used the result of rice paddy classification.

$${\text{EF}} = {\text{EF}}_{c} \times {\text{SF}}_{{\text{w}}} \times {\text{SF}}_{{\text{o}}}$$
(6)

, where EFc means baseline emission factor (kg CH4/ha/yr). SFw represents scaling factor of water management (dimensionless), and SFo is scaling factor of organic matter management (dimensionless).

Table 2 presents the emission factors used for estimating CH4 emissions in this study, sourced from Korean country-specific emission factors provided by the Korea Ministry of Environment (ME 2014). The baseline emission factor (EFc) stood at 2.32 kg CH4/ha/yr. The incorporate water management practices, the scaling factor for water management (SFw) was determined based on the ratio of various water management methods observed during the cultivation period, documented in microdata of Census of Agricultural, Forestry, and Fisheries (2020). The water management methods were classified into categories such as continuously flooded, drainage for less than one week, one to two weeks, more than two weeks, and rainfed. Regarding organic matter, the scaling factor for organic matter (SFo) was divided into two types: rice straw incorporation and winter crop cultivation. For rice straw incorporation, it was assumed that rice straw produced in paddy was subsequently reincorporated. The scaling factor assigned to rice straw was 2.5, derived from the production range of rice straw reported in Crop production survey (2022), indicating an average range of 5–7 Mg/ha. Regarding the scaling factor for winter crop cultivation, the residue of winter crop was considered to affect CH4 emissions akin to green manure (Zou et al. 2005). Thus, the scaling factor for green manure was applied to the winter crop. As specific types of winter crops classification was difficult, Italian ryegrass, the predominant winter forage crop indicating 81.3% of winter forage crops in South Korea was chosen a representative crop (Jeong et al. 2022). The scaling factor was established, assuming that the amount of the residue of Italian ryegrass was equivalent to the application of green manure. A scaling factor of 1.98 was employed for the winter crop, based on the residue quantity of 1.5 Mg/ha from Italian ryegrass (Kim et al. 2017). The estimated CH4 emissions were converted to CO2 equivalent emissions by multiplying by 21, corresponding to the global warming potential (GWP) for CH4 (ME 2022) and the results of CH4 emissions were mapped at the city-level.

Table 2 Country-specific emission factors in Korea
Full size table
Table 3 Accuracy of paddy classification result compared to smart farm map
Full size table

Results and discussion

Classification of rice and winter crop cultivated area

Accuracy assessment of plot-level paddy classification

The plot-level classification model was optimized by setting the threshold for the ratio of pixel area within the parcel at 68%, a value determined to balance overall accuracy and kappa coefficient. Table 3 displays the overall accuracy and kappa coefficient achieved when 68% threshold for plot-level classification across three cities. The accuracy and Kappa coefficient of the classification model were 78.5% and 0.57, affirming the model’s suitability for nationwide application. To assess its applicability nationwide, the results of the rice classification in South Korea were cross-referenced with statistical data by province, as shown in Fig. 5. The comparison indicated a close resemblance between the area derived from the classification model and the statistical records, bolstering confidence in using classification results for estimating CH4 emissions.

Fig. 5
figure 5

Comparing statistic data and Sentinel-1 classified paddy area

Full size image

Calculation of rice and winter crop cultivated area

Figure 6 illustrates a city-level map delineating rice and winter crop cultivation area classified through Sentinel-1 and Sentinel-2 image classification. The extracted rice cultivated area across South Korea totaled 712,237 ha, approximately 3% less than the statistical record of 726,161 ha in 2020 (Agricultural Area Survey 2022). Predominantly, substantial rice paddy regions were concentrated in the western parts of South Korea, notably in JN (20%), CN (18%), and JB (16%), owing to the extensive, flat terrain conducive to rice growth prevalent in this region (Fig. 6a, c). Conversely, the area of winter crop cultivation in South Korea measured 117,840 ha, accounting for 17% of the rice paddy area. A significant proportion, approximately 87%, of the winter crop cultivation area was situated in southern regions, particularly in JN (44%), JB (28%), and GN (16%). This distribution indicates that the southern regions, characterized by a warmer climate, are suitable for cultivating winter crops (Seo et al. 2021) (Fig. 6b, c).

Fig. 6
figure 6

a Map of cities-level paddy area and b winter crop area. The values in parentheses represent respective paddy and winter crop areas in the province level

Full size image

In examining distribution of rice and winter crop cultivated areas, the average rice paddy area for cities stood at 4475 ha, accompanied by a standard deviation of 4194 ha. In contrast, the average winter crop area for cities was 741 ha, with a considerable standard deviation of 1336 ha (Fig. 6). These figures underscore substantial disparities in areas for different cities. As a result, the anticipated variation in CH4 emissions across the cities underscore the importance of estimating CH4 emissions at the city-level.

Estimation of CH4 emission

Calculation of CH4 emission scaling factors

Figure 7 illustrates city-level scaling factors for organic matter―specifically, rice straw and winter crops―alongside water management. As shown in Fig. 7, regions like GG and CN demonstrated higher scaling factors for rice straw due to a more substantial incorporation of rice straw (Fig. 7a). Conversely, regions such as JN, JB, and GN exhibited elevated scaling factors for winter crops, correlating with increased cultivation of winter crops (Fig. 7b). Comparing the scaling factors for winter crops with ten-year average winter temperature (from December to February) in South Korea, sourced from the Korea Meteorological Administration (https://data.kma.go.kr), reveals a boundary where the average temperature was 1 °C, aligning closely with regions showing higher scaling factors for winter crops (Fig. 7b). This suggest an anticipated increase in CH4 emissions from the modified Tier 2 method compared to CH4 emissions from the current Tier 2 method in southern regions.

Fig. 7
figure 7

Map of a organic matter scaling factor for rice straw, b winter crop, and c water management scaling factor. The values in parentheses represent the respective mean scaling factors in the province level

Full size image

The scaling factor for water management appeared higher in GW, GB and GN, with these cities predominantly situated amid high-altitude mountain ranges (Fig. 7c). This pattern arises due to the challenge of maintaining stable drainage in sloped paddy fields, leading to an extended flooded state (Jeong et al. 2010). Consequently, areas with prolonged period of flooded conditions exhibited elevated scaling factor for water management, particularly in regions characterized by sloping terrains.

The distribution analysis of scaling factors revealed an average value of 1.79 with a standard deviation of 0.28 for organic matter, while the average scaling factor for water management stood at 0.74 with a standard deviation of 0.07 (Fig. 7). This disparity highlights a greater variability in the scaling factor for organic matters compared to the scaling factor for water management. Consequently, it suggests that the organic matters would exert a more pronounced influential on CH4 emissions than water management.

Estimation of CH4 emissions

Figure 8 illustrates a comparative analysis of annual CH4 emissions and CH4 emissions per unit area in RS and RSWC. The annual CH4 emissions estimated totaled 6272 Gg CO2 eq./yr with the modified Tier 2 method, reflecting 7% increase compared to the 5867 Gg CO2 eq./yr with the current Tier 2 method. Notably, JN exhibited the highest CH4 emission, due to its expansive paddy area, while CB recorded the lowest CH4 emission owing to its smaller paddy area (Figs. 6 and 8). Moreover, the CH4 emissions per unit area resulted in 8.82 tons CO2 eq./yr with the modified Tier 2 method, presenting a 10% increase from the 8.08 tons CO2 eq./yr with the current Tier 2 method. JN displayed the highest CH4 emission per unit area, attributed to its larger scaling factor for organic matters, whereas GP depicted the lowest CH4 emission per unit area, correlating with its smaller scaling factor for organic matters (Figs. 7 and 8).

Fig. 8
figure 8

Map of annual CH4 emission and CH4 emission per hectare (a, c: RS, b, d RSWC)

Full size image

The blue circles in Fig. 8a, b represent regions abundant in rice paddy areas, while the purple circles in Fig. 9c, 9d indicate regions with significant winter crop cultivation areas. Cities falling within the blue circle aligned with regions exhibiting high CH4 emissions both Fig. 8a, b. Although some cities within the purple circle correlated with higher CH4 emissions per unit area in Fig. 8c, the majority of cities within the purple circle in Fig. 8d were associated with regions demonstrating substantial CH4 emissions per unit area. This highlights the influence of considering winter crop cultivation on CH4 emissions per unit area, emphasizing, however, that the impact of cultivated rice areas had a greater effect on CH4 emissions.

Fig. 9
figure 9

a Annual CH4 emission and b CH4 emission per unit area with paddy area, SFo, and SFw and R2 value

Full size image

Table 4 presents estimated CH4 emissions from this study, previous studies and statistics. CH4 emissions in the current Tier 2 method resemble the emissions of ME (2022), indicating similar estimation method and emission factors. However, CH4 emission reported in FAOSTAT were lower, amounting to 4271 Gg CO2 eq./yr (Table 4). FAOSTAT relied on the Tier 1 method based on IPCC guidelines, which employed baseline emission factor of 1.3 kg CH4/ha/day (IPCC 2019). This factor is smaller than Korean country-specific baseline emission factor of 2.32 kg CH4/ha/day (ME 2014). Consequently, CH4 emissions per unit area stood at 5.88 in FAOSTAT, lower compared to other studies and statistics (Table 4). The CH4 emissions estimated from a decade ago were higher compared to CH4 emissions in 2020 (Choi et al. 2013, 2018). However, when comparing CH4 emissions per unit area between the current Tier 2 method and the findings by Choi et al.(2018), a similarity is evident (Table 4). This indicates a decline in CH4 emissions corresponding to the reduction in rice paddy areas in Korea (Baek et al. 2023).

Table 4 Estimates of CH4 emissions in Korea
Full size table

Figure 9 displays CH4 emissions, CH4 emissions per unit area, and the contributing factors used in estimating CH4 emissions, including rice cultivated area and scaling factors for organic matters and water management. The chart also presents the coefficients of determination (R2) between CH4 emissions, CH4 emissions per unit area, and these contributing elements. As shown in Fig. 8, among these factors, rice cultivated areas demonstrated the strongest correlation with CH4 emissions, boasting an R2 value of 0.95. Comparatively, the scaling factor for organic matters exhibited a more pronounced correlation with CH4 emissions per unit area that the scaling factor for water management, showcasing an R2 value of 0.75 (Fig. 9). These relationships were influenced by the coefficient of variation, which was highest for rice cultivated area, followed by the scaling factor for organic matter, then for scaling factor for water management. This highlights that, rice paddy area had the most considerable influence on the disparities in CH4 emissions among cities. Moreover, CH4 emissions per unit area, when accounting for the impact of rice cultivation area, were notably affected by the scaling factor for organic matters over water management. However, it is important to note that despite JB having a relatively high scaling factor for organic matters, its CH4 emission per unit area comparatively lower (Fig. 8). This discrepancy can be attributed to JB having the smallest scaling factor for water management, indicating that water management has nonnegligible effect for CH4 emissions.

Conclusion

This study utilized Sentinel-1 and Sentinel-2 images to analyze rice and winter crop cultivated across South Korea. Using these results alongside statistical data, this study estimated CH4 emissions from rice paddy in South Korea. Mapping calculated CH4 emissions at the city-level offered a detailed view of their distribution across the country. Additionally, this study compared CH4 emissions estimated through two distinct methods: one that factored in rice straw as organic matter (current Tier 2 method) and the other that considered both rice straw and winter crop as organic matter (modified Tier 2 method).

Cities in the western regions of South Korea, known for their flat terrain, housed significant rice cultivation areas, while cities in southern regions, characterized by a warm climate, demonstrated extensive winter crop cultivation. The varied distribution of rice and winter crop cultivated areas among cities underscored the need for a city-level analysis of CH4 emissions.

Scaling factor for rice straw were notably high in cities within Gyeonggi-do and Jeollabuk-do. Cities with substantial winter crop cultivation exhibited significant scaling factors for winter crop. Additionally, cities situated amid high-altitude mountain range showed elevated scaling factor for water management.

The consideration winter crops increased both CH4 emissions and CH4 emissions per unit area. In the modified Tier 2 method, CH4 emissions totaled 6272 Gg CO2 eq./yr, a 7% increase compared to 5867 Gg CO2 eq./yr in the current Tier 2 method. CH4 emissions per unit area totaled 8.82 ton CO2 eq./yr in modified Tier 2 method, up by 10% from 8.03 ton CO2 eq./yr in current Tier 2 method. Winter crops had a more pronounced effect on CH4 emissions per unit area rather than total CH4 emissions. The study revealed that rice cultivation area exhibited the strongest correlation with CH4 emissions (R2 = 0.95), while scaling factors for organic matters showed the strongest correlation with CH4 emissions per unit area (R2 = 0.75).

The significance of this study lies in estimating city-level CH4 emissions in South Korea while considering winter crops and utilizing satellite images. The substantial CH4 emissions predominantly concentrated in the western regions of South Korea with extensive rice paddy areas and greater CH4 emissions per unit area primarily situated in the southern regions of South Korea. The city-level CH4 emissions and CH4 emissions per unit area were expected to be utilized as basic data for establishment of reduction CH4 emissions by cities.

While, considering the impact of winter crops using satellite images for estimating CH4 emissions, this study was limitations by relying on statistical data for variables like rice straw incorporation and water management. Future studies should focus on surveying agricultural practices of rice cultivation, enhancing the accuracy of CH4 emission estimation.