Introduction

Methane (CH4) and nitrous oxide (N2O) are two major anthropogenic greenhouse gases (GHGs), with global warming potentials exceeding 27 and 273 times that of carbon dioxide (CO2), respectively, over a 100-year time horizon1. Inland freshwater ecosystems play key parts in exchanging carbon (C) and nitrogen (N) between land, ocean and atmosphere, and are responsible for a significant proportion of CH4 and N2O emissions to the atmosphere globally2,3,4. However, these assessments have not yet fully accounted for the characteristics of freshwater aquaculture systems, despite their recognized potential to be a much more intensive GHG emitting source than natural waters due to the extensive application of aquafeeds and fertilizers5,6. Previous assessments indicate that only 25% of the input N and 50% of the feed C consumed by cultured species was converted into biomass, with the remainder excreted into water, released to the atmosphere or preserved in sediments5,6,7,8, rendering a significant anthropogenic source of CH4 and N2O.

Despite growing acknowledgment of the importance of freshwater aquaculture as a major source of CH4 and N2O, the magnitude and dynamics of these emissions remain conspicuously absent or implicit from the Intergovernmental Panel on Climate Control (IPCC) Assessment Report and from the most global and regional GHG budgets1,3,9,10,11,12. A few studies on CH4 and N2O emissions from aquaculture show considerable uncertainties, varying by a factor of three with strong spatial and temporal variability13,14,15. This high uncertainty is particularly concerning as global freshwater aquaculture is projected to increase rapidly by 200% by 2050 to meet global animal protein demand16,17.

Some efforts have been made to quantify CH4 and N2O emissions from inland aquaculture. However, most of the efforts have been limited to small ponds, and emissions from other aquaculture waters, including rivers, lakes, reservoirs and rice fields, have been explicitly excluded to avoid the risk of “double accounting” of these ecosystem types in other assessments, which undoubtedly leads to an underestimation of CH4 and N2O emissions from freshwater ecosystems13,18,19,20. Furthermore, management decisions on the choice of cultured species, drainage and aeration, along with climate feedbacks such as microbial responses to climate warming and changes in hydrology, which are not well represented in state-of-the-art estimates, can alter CH4 and N2O fluxes13,14,21,22,23. In addition, many previous analyses at regional and global scale remain poorly constrained due to the limited geographical distribution of observations. Most commonly, these studies use simple averages of measured gas emissions from a few well-characterized stations to represent a country or a large region, resulting in large uncertainties3,13,24, unknown spatial patterns21,24, and missing emission sources14,21,24. Therefore, a comprehensive quantification of CH4 and N2O emissions from freshwater aquaculture is needed to better understand aquatic systems emissions and the future GHG contribution of the aquaculture sector.

China is the world’s largest freshwater aquaculture producer, accounting for half of the global production and production growth during past two decades17,25. Since the Reform and Opening Up (1978) to the 12th Five-Year Plan period (2014), aquaculture production has increased by 40-fold to 29.4 million tons (Mt) and is expected to reach 50 Mt by 203026,27. Both statistical data and remote sensing monitoring revealed a doubling of aquaculture areas over the last four decades28,29,30. Driven by the growing demand for animal protein globally, the volume and area of freshwater aquaculture in China is projected to further increase in the coming decades9,26,31. These characteristics have made China a hotspot region for aquacultural GHG emissions, which, nevertheless, remains poorly studied in a spatially and temporally explicit manner.

Here, we address this gap in information by using a nationwide measurement of gas fluxes and four decades of aquaculture data. Specifically, we presented a meta-analysis of CH4 and N2O fluxes from five freshwater aquaculture systems in China, including ponds, lakes, reservoirs, ditches and paddy fields (Fig. 1). We estimated CH4 and N2O emissions from each aquaculture system at the provincial and national levels for the period 1980–2022, based on annual aquaculture data from the China Fishery Statistical Yearbook (CFSY). We then explored the socioeconomic and environmental drivers of the dynamic changes in China’s freshwater aquaculture emissions using the Logarithmic Mean Divisia Index (LMDI) method. We also examined emission differences with IPCC Tier 1 methods and previously available estimates.

Fig. 1: Spatial distribution of flux sites with available CH4 and N2O data for freshwater aquaculture in China.
figure 1

Stars, triangles and circles represent flux measurements for CH4, N2O and both GHGs, respectively. The symbols are further marked with different colors for pond, lake, reservoir, paddy field and ditch aquaculture systems. The size of the symbols indicates the number of observations for each GHG-system-site. The map shows provincial mean production of freshwater aquaculture in China from 1980 to 2022.

Full size image

Results

Variations in aquaculture CH4 and N2O emissions

We compiled 358 measurements of CH4 and N2O from 158 peer-reviewed publications, covering 151 sites across 24 of the 30 freshwater aquaculture provinces in mainland China (Supplementary Fig. 1). We examined differences in gas fluxes between and within aquaculture systems, cultured species and management practices (Fig. 2). All the freshwater aquaculture systems considered for the present meta-analysis are sources of CH4 and N2O to the atmosphere, with an average CH4 flux of 4.6 ± 7.2 mg m−2 h−1 and N2O flux of 36.9 ± 76.6 ug m−2 h−1. The mean CH4 flux from ditches is highest at 12.1 mg m−2 h−1, followed by paddy fields (9.0 mg m−2 h−1) and ponds (5.7 mg m−2 h−1) (Fig. 2a). In contrast, the CH4 fluxes from the other two aquaculture water bodies, are 15 times lower than those from ponds and paddy fields, with 0.5 mg m−2 h−1 for lakes and 0.7 mg m−2 h−1 for reservoirs. These fluxes are broadly consistent with previous synthesis from 310 lakes and 153 reservoirs in China (0.3–1.9 mg m−2 h−1)32. The much higher CH4 fluxes from anthropogenic aquaculture waters are mainly due to much greater feed inputs, higher perimeter to surface ratios and shallower waters, which means more substrate for methanogenic respiration, higher loads of terrestrial carbon and less time for CH4 removal by oxidation13,20,33,34,35.

Fig. 2: CH4 and N2O fluxes from freshwater aquaculture in China.
figure 2

CH4 fluxes among different aquaculture systems (a), cultured species (b) and under different management practices (c). Corresponds to d–f, but for N2O fluxes. The top and bottom of the box indicate the 25th and the 75th quartile, respectively. The line within each box and the empty square represent the median and mean, respectively. Whiskers mark the 10th and 90th percentiles. Different letters above the bars indicate that groups are statistically different (P < 0.05) in ANOVA or independent samples t-test. Numbers in parentheses indicate sample size. PO ponds, LA lakes, RS reservoirs, PR paddy fields, DT ditches.

Full size image

Similar results are found for N2O fluxes with two times higher emissions from ponds (47.7 ug m−2 h−1), paddy fields (47.3 ug m−2 h−1) and ditches (22.7 ug m−2 h−1) than from lakes (15.3 ug m−2 h−1) and reservoirs (23.2 ug m−2 h−1) (Fig. 2d). This can be explained by a higher total organic nitrogen content in sediments as a result of extensive nitrogen application. In addition, a much higher redox potential due to more readily degradable organic matter in anthropogenic waters can favor denitrification, which produces more N2O than nitrification6,21,36.

Among different cultured species, CH4 fluxes of 11.8 ± 7.3 mg m−2 h−1 from shrimp farming and N2O fluxes of 37.7 ± 33.9 ug m−2 h−1 from crab farming are significantly higher than those from fish and mixed species (Fig. 2b, e). This may be due to the increased availability of organic substrates and the highly anaerobic environment in shrimp and crab farming systems as a result of high-protein diets and benthic characteristics of shrimps and crabs6,13,37,38. More organic compounds in their feed remnants and feces, such as starch and protein, can be more easily decomposed into reaction substrates39. Benthic forage can disturbe the bottom sediment and reduces the oxygen exchange between water and sediment, thereby promoting anaerobic oxidative processes. In addition, benthic bioturbation by shrimp or crabs can trigger bubbling activity and thus increase emissions to the atmosphere, whereas less disturbance by fish can increase oxygen availability and thus inhibit emissions6,23,40.

In terms of management practices, drainage management does not result in significant differences in gas fluxes, but the use of aerators reduces CH4 fluxes by 80% and increases N2O fluxes by 47% (Fig. 2c, f). On the one hand, aeration increases the oxygen concentration in the surface layer of the substrate, reducing anaerobic environments ideal for methanogens. On the other hand, aeration promotes nitrification of NH4+-N and physical agitation by the aerators enhances N2O gas exchange at the water-air interface41,42. The results show that CH4 and N2O emissions vary significantly between aquaculture systems, with higher emissions from human-made systems. It also highlights the potential for more delicate management to reduce aquaculture CH4 emissions, albeit with a tradeoff for more N2O emissions.

Spatio-temporal patterns of aquaculture CH4 and N2O budgets

We estimated the annual CH4 and N2O emissions from each aquaculture system by multiplying the system mean fluxes by the corresponding provincial aquaculture area (Methods). Total CH4 and N2O emissions were 2.5 (0.6-4.2) Tg CH4 yr−1 and 18.3 (3.8–32.2) Gg N2O yr−1, respectively (parenthetical values represent the 10th-90th percentiles from Monte Carlo simulations), accounting for 15% and 2% of China’s terrestrial CH4 and N2O emissions on average (Fig. 3a, b). Pond and paddy field aquaculture systems contribute >75% of total emissions and 97% of the increase in their annual emissions. On a decadal scale, the fastest growth in total emissions occurred in the 1990s, representing half of the increase over the last four decades (Fig. 3c). This could be closely related to a series of marine protection laws enacted by the government in the mid-1990s, in particular the seasonal moratorium on marine fishing, which stimulated the development of inland aquaculture25,26. On a provincial basis, the top five aquaculture provinces, along the Yangtze river basin, are Sichuan, Jiangsu, Hubei, Hunan and Anhui, emitting 75% of the combined emissions (reported as global warming potential CO2-eq) of CH4 and N2O (Fig. 3d and Supplementary Fig. 2).

Fig. 3: Aquaculture CH4 and N2O budgets in China.
figure 3

National total emissions of CH4 (a) and N2O (b) from five freshwater aquaculture systems between 1980 and 2022, calculated by multiplying the system mean fluxes by the corresponding provincial aquaculture area. The histogram shows the uncertainty in aquaculture emissions, with the red lines representing the 10, 50 and 90% quantiles of the Monte Carlo simulation. The pie chart shows the percentage of total emissions from each aquaculture system. c Mean CH4 (left axis) and N2O (right axis) emissions for the 1980s (1980–1989), 1990s (1990–1999), 2000s (2000–2009) and 2010s (2010–2022). d Provincial mean emissions (CO2-eq) of CH4 and N2O. Because several provinces are not clear enough to distinguish the emissions from each aquaculture system, we have mapped them (emissions below 2 Tg CO2-eq yr−1) in Supplementary Fig. 2. The full names of all administrative units are listed in Supplementary Table 1. 1 Tg = 103 Gg.

Full size image

We found that pond and paddy field systems are not only the largest freshwater aquaculture emitters of CH4 (1.9 (0.04–3.7) Tg CH4 yr−1) and N2O (13.3 (0.6–26.4) Gg N2O yr−1), but also have the fastest growing rate of emissions (0.02 Tg CH4 yr−1 and 0.2 Gg N2O yr−1) (Fig. 4 and Supplementary Fig. 3). Unexpectedly, both CH4 and N2O emissions from paddy fields are comparable to those from ponds and are even growing faster, suggesting that emissions from the rice-fish system should not be overlooked in the national GHG budget along with those from aquaculture ponds. In contrast, CH4 and N2O emissions from lakes and reservoirs are much lower due to the lower flux intensity combined with a relatively small aquaculture area in these water bodies (Fig. 2 and Supplementary Figs. 4 and 5). Ditches are the third most important source of CH4 and N2O after ponds and paddy fields, with a slight but insignificant decreasing trend.

Fig. 4: CH4 and N2O emissions from different freshwater aquaculture systems.
figure 4

a Mean of national total CH4 (bottom axis) and N2O (above axis) emissions from pond, lake, reservoir, paddy field and ditch aquaculture systems. Total emissions were estimated by multiplying the mean fluxes (Fig. 2a, d) by the aquaculture area of the aquaculture system. Bars and error bars represent mean±s.d. b Trends in CH4 and N2O emissions from each aquaculture system during the period 1980-2022. *** and ns indicate P < 0.01 and non-significant, respectively.

Full size image

On spatial variations, provincial mean CH4 and N2O emissions are 0.1 Tg CH4 yr−1 and 0.7 Gg N2O yr−1, but there is a 200- and 90-fold difference between the highest and lowest emitting provinces, respectively (Fig. 5a, b). The corresponding values are 0.4 Tg CH4 yr−1 in Jiangsu province (2.7 Gg N2O yr−1 in Hubei province) and 0.002 Tg CH4 yr−1 in Qinghai province (0.03 Gg N2O yr−1 in Beijing, a municipality in China). However, the emission hotspots are highly concentrated in central China in Sichuan, Hunan, Hubei, Anhui, Jiangsu provinces, which emit 45 Tg CO2-eq yr−1. Over the last 43 years, total emissions increased by 1.8 Tg CO2-eq yr−1 (Supplementary Fig. 6), but their spatial pattern shows a clear difference before and after 2000 (Fig. 5c, d). During the period 1980–1999, annual emissions increase significantly in 80% of the provinces, but decrease in 40% of the provinces after 2000, of which 30% are in coastal areas. In contrast, the emission intensity per unit of protein decreases exponentially due to the much faster increase in production (Fig. 5e, f), although the emissions intensity has stalled over the past ten years. Mean emission intensities are higher in northeast, northwest and southwest China, where aquaculture production per unit area is generally less than 2500 kg ha−1 (Supplementary Fig. 4c, d).

Fig. 5: Spatial patterns of CH4 and N2O emissions from freshwater aquaculture in China.
figure 5

Provincial mean emissions of CH4 (a) and N2O (b) in 1980–2022. Provincial trends in combined emissions (CO2-eq) of CH4 and N2O in 1980–1999 (c) and 2000-2022 (d). Spatial distribution (e) and temporal changes (f) of emission intensities for protein production over the last 43 years.

Full size image

Drivers of aquaculture emissions

We used an LMDI analysis (Methods) to quantify the contribution of seven socioeconomic and environmental factors to changes in China’s freshwater aquaculture emissions, including population changes, economic growth, dietary change, food security, aquaculture product quality, emission intensity, and freshwater constraints (Fig. 6). Between 1980 and 2022, the 329.7% increase in aquaculture emissions was dominated by rapid economic growth (red bars), which alone would have driven aquaculture emissions to increase by 139.4% and 244.1% during the periods 1980–2000 and 2000–2022, respectively. The next most important driver of increasing emissions was the improved quality of aquaculture products (protein content, orange bars) and dietary changes (green bars) toward more animal-based foods43. Higher quality and greater demand for animal protein-stimulated aquaculture resulted in a 121.9% and 81.0% increase in emissions over the two 20-year periods, respectively. Population growth also pushed aquaculture emissions upward steadily by 12.3% in 1980–2000 and 11.0% in 2000–2022 (blue bars).

Fig. 6: Contribution of seven socioeconomic and environmental drivers to the changes in China’s aquaculture emissions in the periods 1980–2000 and 2000–2022.
figure 6

The length of each bar reflects the contribution of each factor to changes in aquaculture emissions during the corresponding period. The value below the bar indicates its contribution proportion.

Full size image

Although total emissions were increasing, two factors effectively reduced emissions, restraining the growth rate. The strongest constraint is freshwater resources (light blue bars), which, in the absence of other factors, could have reduced emissions by 148.9% and 257.4% in the period 1980–2000 and 2000–2022, respectively, completely offsetting the emissions increases caused by economic growth. This is mainly because freshwater scarcity can not only directly constrain the potential expansion of freshwater aquaculture but also indirectly limit aquaculture production by impacting feed production17. The next most important factor was the decrease in emission intensity per unit of protein (purple bars), which accounted for a 52.6% and 60.8% decrease in total emissions, respectively. Increased food security due to improved food production capacity (water productivity) caused aquaculture emissions to decrease by 3.0% in the period 1980-2000 and to increase by 4.1% in the period 2000-2022 (deep blue bars).

Discussion

Using an extensive observation dataset of direct measurements of CH4 and N2O fluxes covering five freshwater aquaculture systems in China, we estimated total CH4 and N2O emissions from freshwater aquaculture to be 2.5 (0.6–4.2) Tg CH4 yr−1 and 18.3 (3.8–32.2) Gg N2O yr−1. These emissions are equivalent to the total CH4 emissions (2.6 Tg CH4 yr−1) and 60% of the N2O emissions (29.6 Gg N2O yr−1) from natural rivers, lakes and reservoirs in China18,32,44. They account for more than half of the national CH4 and N2O budgets of terrestrial aquatic ecosystems (including inland waters and natural wetlands), as the result of five times higher CH4 and twice higher N2O effluxes from freshwater aquaculture than the average for other inland waters (Supplementary Fig. 7), though they occupy less than 10% of inland water surface area45,46,47, and exceeds the magnitude of soil methane uptake by the methanotrophs48. In addition, the 1.9 (0.04–3.7) Tg CH4 and 13.3 (0.6–26.4) Gg N2O emitted annually by the two largest emitters (ponds and paddy fields) result in the same radiative forcing as 55 Tg CO2 yr−1 on century time scales, offsetting about two years of China’s terrestrial GHG sink (29 Tg CO2-eq yr−1)48. These emissions of CH4 and N2O from freshwater aquaculture in China, even if they do not increase over the next 80 years, can introduce a global mean radiative forcing of at least 0.01 W m−2 by the end of the 21st century49.

We compare our estimates in our study with IPCC methodological guidelines by re-estimating aquaculture emissions using the IPCC Tier 1 method50,51. The IPCC Tier 1 estimates grossly underestimate CH4 emissions and overestimate N2O emissions, with biases increasing by 0.6% and 8% per year, respectively, as aquaculture area expands and production increases (Supplementary Fig. 8a, b). This is because the IPCC default CH4 fluxes are two times lower than those derived from our meta-analysis, especially for pond and paddy field systems. For N2O, the default yield emission factor (1.69 gN2O kg−1yield) is higher than that derived from field experiments for all aquaculture species in China (Supplementary Fig. 9). The proportional differences suggest that the IPCC default coefficient is not suitable for applications in China.

Although the comparisons with published freshwater CH4 and N2O emissions13,14,21,23,42,52,53,54 were difficult due to the differences in accounting approaches, system classification, spatial and sector coverage, and statistical treatments, we made this tentative comparison (Supplementary Fig. 8c, d) to illustrate the differences. Our estimates are close in magnitude to one previous estimate based on the classification of aquaculture intensification levels21. However, our estimates are higher than recent species-based estimates for N2O55, and lower than similar estimates for CH4 based on different aquaculture system types14. The relatively lower CH4 emissions derived here are largely the result of the additional term, ‘coastal wetland reclamation’, included in ref. 14, combined with lower CH4 flux intensities from ponds, lakes and reservoirs in our estimates. These low flux intensities are probably due to the fact that our synthesized field measurements provide additional data coverage over aquaculture regions in China, not just the highly productive central and eastern regions as before (Fig. 1 and Fig. 1 of ref. 14). The lower species-based estimates of N2O emissions are mainly due to the lower yield emission factors resulting from the higher feed conversion rates used in the previous calculation52. Nevertheless, the fairly close emissions to ref. 21 give us confidence in the robustness of our estimates in covering different systems and levels of management intensity, as well as a comprehensive spatial and temporal coverage over the past four decades in China.

Our estimates are subject to some limitations related to the practical constraints of performing gas flux measurements at higher spatial and temporal resolution. First, measurements from at least one culture period are likely to capture major variability in emission fluxes, but nevertheless, the collection of multi-year data with overlapping periods will reduce the uncertainties in the calculations, considering intensified human impacts and climate warming. Second, despite the relatively homogeneous management practices and environmental conditions, and the significant differences in gas emissions (Fig. 2), the mean value used for the same aquaculture system may not be representative enough of broader aquaculture systems at the country scale. More spatially resolved data are needed, particularly in the northern, western and southwestern farming areas (Supplementary Fig. 10), to improve the robustness of upscaling. In addition, although the statistical aquaculture area is accurate, there is a risk of double counting with inland waters and rice fields due to challenges in clearly identifying aquaculture and non-aquaculture water bodies. In fact, the issue arises from the fact that inland waters and rice fields do not exclude aquaculture areas in their accounts, which, in turn, could lead to an overestimation of the GHG emissions from freshwater ecosystems19,20,32,54. Finally, there are some inherent correlations between the socioeconomic drivers, such as dietary changes and economic development. Future work using finer resolution datasets and improved decomposition model is recommended to reduce the uncertainty of driver analysis.

Despite the uncertainties, this study provides a systematic and spatio-temporally explicit estimate of CH4 and N2O emissions from freshwater aquaculture in the world’s largest aquaculture country. The results show that inland aquaculture emissions are higher than previously thought4,56,57, highlighting that their omission from global GHG budget estimates may largely overestimate the magnitude of the terrestrial GHG sink. As a result of climate warming, CH4 and N2O emissions are expected to increase across freshwater aquaculture systems due to higher activities of methanogens, nitrifying and denitrifying bacteria58,59. Although CH4 and N2O fluxes from these microorganisms are currently poorly constrained, it is intuitive to assume that they increase with increasing eutrophication and temperature3,6,13. This is supported by a generally positive relationship between CH4 and N2O emissions and temperature across biomes6,60. Of the world’s top 21 aquaculture producers, 90% are located in low and mid-latitudes, where temperatures are projected to rise significantly61. The results of this study highlight the need to closely track CH4 and N2O emissions from major aquaculture producers such as India, Indonesia and Bangladesh to better understand the global GHG budget and explore climate mitigation opportunities.

With the ever-increasing demand for animal protein and the stagnation of capture fisheries and mariculture, freshwater aquaculture is becoming a key -indeed the only- option for an affordable, accessible and stable supply of aquatic food17,62,63. By 2050, global aquatic food production will need to double to meet rising population- and income-driven demand64. The resulting expansion of aquaculture industry is expected to substantially increase aquatic emissions, thereby making inland aquaculture play a more important role in the global CH4 and N2O cycles65. Nevertheless, the emission intensity per unit of protein from aquaculture products is still much lower than that of the world’s major protein-producing livestock13, which is responsible for one-third of global anthropogenic CH4 and N2O emissions66,67. In addition to lower GHG emissions, aquaculture requires less freshwater and land than agriculture and livestock, providing a sustainable solution for reducing food production pressures on the environment while ensuring global economies, livelihoods, and nutritional security22,68.

Although the lower environmental impact of aquaculture than that of many terrestrial meat production systems, there is still the need to rapidly reduce the GHG footprint of the aquatic farming sector. Achieving the Paris Agreement’s goal of limiting global temperature rise to 1.5 or well below 2 °C above pre-industrial levels requires large and rapid reductions in GHG emissions69. A new and recent estimate of the remaining carbon budget (the net amount of CO2 humans can still emit) for a 50% chance of keeping warming to 1.5 and 2 °C is equivalent to a few years and less than ten years of CO2 emissions, respectively70. Despite efforts that have reduced the intensity of aquaculture emissions by 50% over the past two decades (Fig. 5f), a reduction in emissions have to be achieved in a way that does not endanger food supply.

One effective option for reducing aquaculture non-CO2 emissions is the extensive adoption of aerated systems given much lower CH4 fluxes from intensified systems with aeration (Fig. 2). Despite the aeration would increase N2O emissions, the estimated reduction of CH4 emissions (175 Tg CO2-eq) would outweigh the global warming potential (GWP) of increased N2O emissions (5 Tg CO2-eq) by 35 times either at 20-year horizon or at 100-year horizon, assuming the aeration system can be sufficiently equipped for all freshwater aquaculture systems. In addition, studies also reveal significant yield improvement through aeration in pond and paddy field systems21,71, rendering the aeration measurements a possible solution to mitigate climate change in a manner that does not endanger food security. If half of China’s aquaculture area (4 Mha based on area in 2022) is equipped with aerated systems, the combined emissions of CH4 and N2O will be reduced by 38% from 169 to 105 Tg CO2-eq. We suggest that sustainable management of the growing freshwater aquaculture should be considered as a viable climate change mitigation pathway to benefit planetary health, food security and human well-being.

Methods

CH4 and N2O fluxes dataset

China’s freshwater aquaculture systems can be classified into ponds, lakes, reservoirs, paddy fields and ditches according to the aquaculture waters28. We first compiled a database of CH4 and N2O fluxes measured in five aquaculture systems from peer-reviewed journal articles and dissertation in the Web of Science Core Collection (http://isiknowledge.com), Google Scholar (http://scholar.google.com) and China Knowledge Infrastructure (CNKI, https://www.cnki.net/). The Boolean search string (GHG|Greenhouse gas OR CH4|methane OR N2O|nitrogen oxide) AND (aquaculture|fish|shrimp|crab|freshwater) OR (pond|lake|reservoir|rice field|paddy field|ditch|canal|inland waters) were used as keywords until August 2023. More than 5000 publications matched the search, combined with the relevant references of these articles.

We then filtered and processed the initial data based on seven criteria: 1) Fluxes were measured in freshwater aquaculture systems or in aquaculture systems with salinity less than 0.5 ppt in China51, excluding coastal wetland reclamation aquaculture and marine aquaculture; 2) Measurements covered at least one culture period to capture the main variability in fluxes; 3) Gas fluxes were either derived from laboratory measurements of water samples or measured directly in situ; 4) Data reported CH4 and/or N2O fluxes and detailed information on the experiments, including measurement methods, management practices, cultured species, surface area, water depth and observation time; 5) When experiments in the same aquaculture system were reported in multiple studies, we used the study that reported the largest number of sampling dates. If experiments in the same aquaculture system used different management practices, we considered them as different measurements3,13,14; 6) If the publication did not clearly describe whether lakes and reservoirs were farmed or not, we made a distinction based on news reports and Wikipedia. If no information in news reports or Wikipedia was found, the data was discarded to avoid bias; 7) Only CH4 and N2O fluxes with exact geographic locations were included. Where coordinates were not available but site description was sufficient, we obtained approximate latitude and longitude from Google Maps. This resulted in a collection of 358 CH4 and N2O fluxes from 158 published literature, covering 151 sites and 24 provincial administrative units in China. Note that the abnormally high/low flux measurements were removed from our analysis. In addition, due to changes in the administrative regions of some provinces in China, we combined the provincial data to ensure temporal consistency. For example, Chongqing was not a provincial administrative region and belonged to Sichuan Province before 1997; thus, we combined the data of Chongqing and Sichuan for the period from 1998 to 2022 and treated them as one provincial region. These provinces contributed 90% of the national freshwater aquaculture volume28. For CH4 fluxes, we included studies that provided both diffusive and ebullitive fluxes from the open-water surface either separately (for example, via bubble traps or acoustic surveys for ebullition and via thin-boundary-layer modelling or floating chambers for diffusion) or together (for example, via floating chamber or eddy covariance methods).3,21. We finally extracted data directly from the tables and texts of the filtered publications or by digitizing graphs using the Engauge Digitizer software (version 11.0).

Estimation of CH4 and N2O budgets

We estimated CH4 and N2O emissions from each aquaculture system by multiplying the mean emission fluxes by the system-specific aquaculture area in each province from 1980 to 2022. Aquaculture areas were obtained from the CSFY28. The CSFY is a comprehensive statistical resource that reflects the overall development of China’s fisheries industry. Jointly compiled by the National Bureau of Statistics and the Ministry of Agriculture and Rural Affairs, it draws on data from various sources, including agricultural censuses, fisheries enterprise surveys, government records, and supplementary information collected through sampling and questionnaires. The content encompasses aspects related to fisheries production, economy, society, disasters, and resource environment. It reports the freshwater aquaculture area of each system at provincial level. We calculated the 100-year GWP as the combined emissions using the following equation:

$${{{\rm{GWP}}}}=27* {{{{\rm{CH}}}}}_{4}+273* {{{{\rm{N}}}}}_{2}{{{\rm{O}}}}$$
(1)

where GWP is CO2 equivalence (Tg CO2-eq yr-1), CH4 and N2O denote annual emission of CH4 (Tg CH4 yr-1) and N2O (Tg N2O yr-1), respectively.

We then calculated the emission intensities, defined as the combined emissions per unit of protein in aquatic products.

$${{{\rm{EF}}}}=\frac{{{{{\rm{E}}}}}_{{{{\rm{GHG}}}}}}{0.5* {\sum }_{{{{\rm{i}}}}=1}^{{{{\rm{n}}}}}{{{{\rm{P}}}}}_{{{{\rm{i}}}}}* {{{{\rm{C}}}}}_{{{{\rm{i}}}}}}\,$$
(2)

where EF is the emission intensity per kg of protein for all species (fish, shrimp, crab and others), EGHG is the combined emissions of CH4 and N2O (Tg CO2-eq yr−1). i is the number of species. Pi and Ci are the production and protein content of specific species, respectively. The constant 0.5 is the proportion of edible meat in aquatic products13,72. According to the United States Department of Agriculture (USDA, https://www.usda.gov/). Ci is 22 g, 24 g, 19 g and 15 g protein per 100 g meat for fish, shrimps, crabs and other species (e.g., mussels, snails, clams, etc.) respectively.

Comparison with IPCC Tier1 methods and previous estimates

We calculated annual CH4 and N2O emissions for the period 1980–2022 using the IPCC-guided methods and default emission coefficients used to estimate aquaculture emissions (Supplementary Fig. 9). CH4 emissions were calculated by multiplying mean emission rates (kg CH4 ha−1 yr−1) from each aquaculture system by the corresponding area. N2O emissions were estimated by multiplying yield emission factors (kg N2O per kg production) by provincial production obtained from the CSFY. We compared with IPCC-based estimates at a national level to examine and attribute emissions differences. We also compared with currently available estimates of freshwater aquaculture emissions in China, which differ in accounting boundary, data source, estimation approach, time period and GHG, to check the robustness of our results.

Decomposition analysis

Decomposition analysis (DA) methods have been used extensively to quantify the contribution of socioeconomic drivers to changes in environmental pressures73,74. The LMDI is one of the most popular DA methods, originally developed for time series analysis using data with sufficient temporal and sectoral detail. The advantage of the LMDI approach is that it can be easily applied to any data at any level of aggregation and allows for optimal decomposition, i.e., no unexplained residual terms, consistent aggregation, and satisfactory additivity75,76. As a result, many studies have used LMDI to provide policy-relevant insights, for instance by identifying the driving forces of energy77, emission78, crop production79 and water use80. We used it to quantify the relative influence of seven socioeconomic drivers on freshwater aquaculture emissions in China, including population, economy, diet, aquaculture product quality, food production capacity, emission intensity and freshwater. Details of the specific data used to calculate the indicator can be found in Supplementary Table 2. All data are taken from the China Statistical Yearbook compiled by the National Bureau of Statistics.

In this study, we decompose the combined emissions of CH4 and N2O at national scale as follows:

$${{{\rm{C}}}}={{{\rm{P}}}}\times \frac{{{{\rm{G}}}}}{{{{\rm{P}}}}}\times \frac{{{{\rm{W}}}}}{{{{\rm{G}}}}}\times \frac{{{{\rm{F}}}}}{{{{\rm{W}}}}}\times \frac{{{{\rm{A}}}}}{{{{\rm{F}}}}}\times \frac{{{{\rm{PRO}}}}}{{{{\rm{A}}}}}\times \frac{{{{\rm{C}}}}}{{{{\rm{PRO}}}}}$$
(3)
$$={{{\rm{P}}}}\times {{{\rm{Y}}}}\times {{{\rm{R}}}}\times {{{\rm{T}}}}\times {{{\rm{D}}}}\times {{{\rm{Q}}}}\times {{{\rm{E}}}}$$

where C is total CH4 and N2O emissions converted to CO2 equivalents; G is gross domestic product (GDP); W is national water resources; F is per capita consumption of staple foods; A is per capita consumption of aquatic products; PRO is freshwater aquaculture protein production. PRO was calculated based on the denominator of Eq. (2). Thus, according to Eq. (3), C can be represented by the following factors:

P is population;

\({{{\rm{Y}}}}=\frac{{{{\rm{G}}}}}{{{{\rm{P}}}}}\) is GDP per capita and measures economic growth;

\({{{\rm{R}}}}=\frac{{{{\rm{W}}}}}{{{{\rm{G}}}}}\) the freshwater demand to make per unit of GDP, representing the freshwater resource constraint;

\({{{\rm{T}}}}=\frac{{{{\rm{F}}}}}{{{{\rm{W}}}}}\) the amount of freshwater used to produce each unit of food, which indicates national food production capacity and reflects food security;

\({{{\rm{D}}}}=\frac{{{{\rm{A}}}}}{{{{\rm{F}}}}}\) is the share of aquatic product in per capita food consumption and represents dietary changes;

\({{{\rm{Q}}}}=\frac{{{{\rm{PRO}}}}}{{{{\rm{A}}}}}\) is the proportion of freshwater aquaculture protein in the aquatic product consumed, indicating the quality of the aquaculture product; and

\({{{\rm{E}}}}=\frac{{{{\rm{C}}}}}{{{{\rm{PRO}}}}}\) is the CO2-eq emissions per unit of protein from freshwater aquaculture and represents the emission intensity.

The change of aquaculture emissions in the period from the base year t-1 to the final year t is calculated as

$$\Delta C={C}^{t}-{C}^{t-1}$$
(4)

and further decomposed to seven factors as

$$\Delta C= \, L\left({C}^{t},{C}^{t-1}\right){{{\mathrm{ln}}}}\left(\frac{{P}^{t}}{{P}^{t-1}}\right)+L\left({C}^{t},{C}^{t-1}\right){{{\mathrm{ln}}}}\left(\frac{{Y}^{t}}{{Y}^{t-1}}\right)+L\left({C}^{t},{C}^{t-1}\right){{{\mathrm{ln}}}}\left(\frac{{R}^{t}}{{R}^{t-1}}\right)\\ +L\left({C}^{t},{C}^{t-1}\right){{{\mathrm{ln}}}}\left(\frac{{T}^{t}}{{T}^{t-1}}\right)+L\left({C}^{t},{C}^{t-1}\right){{{\mathrm{ln}}}}\left(\frac{{D}^{t}}{{D}^{t-1}}\right)\\ +L\left({C}^{t},{C}^{t-1}\right){{{\mathrm{ln}}}}\left(\frac{{Q}^{t}}{{Q}^{t-1}}\right)+L\left({C}^{t},{C}^{t-1}\right){{{\mathrm{ln}}}}\left(\frac{{E}^{t}}{{E}^{t-1}}\right)$$
$$\,=\Delta {C}_{P}+\Delta {C}_{Y}+\Delta {C}_{R}+\Delta {C}_{T}+\Delta {C}_{D}+\Delta {C}_{Q}+\Delta {C}_{E}$$
(5)

Here, \(\Delta {C}_{P}\), \(\Delta {C}_{Y}\), \(\Delta {C}_{R}\), \(\Delta {C}_{T}\), \(\Delta {C}_{D}\), \(\Delta {C}_{Q}\) and \(\Delta {C}_{E}\) are the change in aquaculture emissions attributed to population variation, economic growth, freshwater limitation, food security, dietary adjustment, aquaculture product quality improvement and emission intensity change, respectively. In Eq. (5), \(L\left({C}^{t},{C}^{t-1}\right)\) is the logarithmic mean of the CO2-eq in the base year t-1 and the final t, calculated as

$$L\left({C}^{t},{C}^{t-1}\right)=\left\{\begin{array}{c}\frac{{C}^{t}-{C}^{t-1}}{{{\mathrm{ln}}}{C}^{t}-{{\mathrm{ln}}}{C}^{t-1}},\quad{for}{C}^{t} \, \ne \, {C}^{t-1}\\ {C}^{t},\hfill \quad {for}{C}^{t} \, = \, {C}^{t-1}\end{array}\right.$$
(6)

Uncertainty analysis

We quantified the uncertainty in the current emission estimates using Monte Carlo simulations, including uncertainty from the CH4 and N2O fluxes. We performed simulations for each aquaculture system at the provincial level. For each simulation, we generated a total of 5000 random values from a normal distribution around the mean and standard deviation for that aquaculture system-gas flux. National CH4 and N2O emissions were the sum of the emissions from the five aquaculture systems across all aquaculture provinces for each interaction. The uncertainty range was reported as the 10th–90th percentiles of the resulting emission distribution3,19.

Statistical analysis

Differences in CH4 and N2O fluxes between aquaculture systems and cultured species were examined using a Least Significant Difference test in a one-way analysis of variance (ANOVA). The effects of management practices on fluxes were examined using an independent samples t-test. Linear regression and curve estimation were used to analyze trends in estimated emissions and emission intensities over the last four decades. All statistical analyses were performed using SPSS (version 26.0).