Upper-bound theoretical carbon mitigation potential of substituting mineral fertilizers with livestock manure nutrients in China: evidence from provincial panel data

Abstract

Introduction:

Manure nutrient recycling is an important pathway for reducing mineral fertilizer dependence and improving agricultural nutrient circularity, but its mitigation effect should be distinguished from actual realized emission reduction. This study estimates the upper-bound theoretical carbon-mitigation potential of substituting mineral fertilizer nutrients with livestock and poultry manure-derived nutrients in China.

Methods:

Using provincial panel data for 31 provinces from 2005 to 2024, manure resources and manure-derived N, P₂O₅, and K₂O supplies were quantified, and upper-bound ER was calculated using nutrient-specific fertilizer emission factors. Dagum Gini decomposition, global Moran’s I, and the Spatial Durbin Model were applied to examine regional inequality, spatial clustering, and conditional spatial associations.

Results:

Results show that national upper-bound ER declined from 2.183 × 10⁷ t CO₂-eq in 2005 to 1.677 × 10⁷ t CO₂-eq in 2019, rebounded to 2.021 × 10⁷ t CO₂-eq in 2023, and slightly decreased to 1.962 × 10⁷ t CO₂-eq in 2024. The western region contributed 42.43% of the national total in 2024. Spatial dependence was significantly positive. Manure resource intensity had positive direct and total effects, whereas fertilizer intensity had negative direct and total effects. Agricultural GDP was positively associated with mitigation potential, while mechanization, urbanization, and cropping structure showed heterogeneous local and spillover effects. Robustness and regional analyses confirmed a stable local resource-endowment effect but region-specific spillover patterns. These estimates represent theoretical upper bounds rather than realized mitigation.

Discussion:

China possesses substantial but spatially uneven theoretical carbon-mitigation potential from manure-based mineral fertilizer substitution. Converting this technical upper bound into realized mitigation requires region-specific improvements in manure collection, treatment, transport, nutrient matching, and field application. Policies should therefore shift from maximizing manure return alone toward spatially coordinated, agronomically appropriate, and environmentally controlled nutrient recycling.

1 Introduction

Agricultural decarbonization has become an important component of China’s pathway toward carbon peaking and carbon neutrality (Chen et al., 2022; Huang et al., 2025). Although energy and industry dominate China’s total greenhouse gas emissions, agriculture remains a major source of non-CO₂ emissions, especially methane from livestock production and rice cultivation and nitrous oxide from cropland nutrient inputs (Du et al., 2022; Song et al., 2023). Recent inventory-based studies show that China’s agricultural non-CO₂ emissions have entered a stabilization stage since the mid-2010s, but livestock- and cropland-related emissions still account for a large share of the agricultural emission structure (Gao et al., 2025). Scenario assessments further indicate that agricultural mitigation technologies could help China’s agricultural non-CO₂ emissions peak earlier and reduce emissions substantially by 2060, highlighting the importance of nutrient management, fertilizer reduction, and manure recycling in agricultural climate policy (Chen et al., 2022).

For a long time, China’s agricultural growth has relied heavily on mineral fertilizer inputs (Zhang et al., 2025). While mineral fertilizers have contributed greatly to crop production and food security, excessive or inefficient fertilizer application has also increased nutrient losses, soil and water pollution, and fertilizer-related greenhouse gas emissions (Zhang et al., 2024). Reducing mineral fertilizer dependence without compromising crop productivity is therefore a key challenge for green agricultural transformation. Organic fertilizer substitution, especially the recycling of livestock and poultry manure nutrients to replace part of mineral fertilizer inputs, provides a potential pathway for coordinating crop production, nutrient circularity, and carbon mitigation (Ning et al., 2022; Niu et al., 2024). Livestock and poultry manure contains considerable nitrogen (N), phosphorus (P), and potassium (K) resources and can serve as an important nutrient source for cropland. From the perspective of circular agriculture, manure recycling transforms livestock waste into crop nutrients, thereby reducing both waste-disposal pressure and mineral fertilizer demand. A national farmer-survey-based assessment showed that manure-derived N, P, and K inputs accounted for 19.0%, 25.5%, and 31.1% of total N, P, and K inputs in China’s croplands, respectively, with manure nutrient utilization showing clear spatial differences between eastern and western China (Zhang Q. et al., 2023). Meanwhile, manure management research has increasingly shifted from resource recovery alone toward integrated goals of pollution reduction, carbon accounting, and carbon-neutral livestock production (Zhang L.,E.,R. et al., 2023).

Empirical evidence from field experiments also supports the agronomic and environmental value of partial manure substitution, although the effects vary by crop system, substitution ratio, soil condition, and nutrient balance. For example, manure replacing synthetic fertilizer in a winter wheat–summer maize system was found to maintain or improve yield sustainability and reduce the crop carbon footprint under appropriate substitution levels (Niu et al., 2024). Chicken manure substitution for mineral nitrogen under reduced nitrogen input has also been shown to improve crop yield and soil fertility in North-Central China (Ning et al., 2022). However, long-term substitution of chemical phosphorus fertilizer with organic manure may involve trade-offs in phosphorus use efficiency and potential phosphorus losses if nutrient inputs are not properly balanced (Lu et al., 2021). These findings indicate that manure substitution has mitigation potential, but this potential cannot be equated with automatically realized emission reduction.

Despite these advances, three gaps remain in the current literature. First, many studies focus on field-scale substitution effects, manure nutrient utilization, or agricultural emission inventories, while fewer studies quantitatively connect provincial manure-derived NPK resources with fertilizer-related carbon-mitigation potential over a long time span. Second, existing studies have not sufficiently distinguished between theoretical substitution potential and actual realized mitigation. In practice, manure recycling is constrained by collection efficiency, treatment losses, transportation radius, cropland absorption capacity, seasonal nutrient demand, and farmers’ adoption behavior. Therefore, estimates based on complete nutrient recycling and equivalent fertilizer substitution should be interpreted as upper-bound theoretical potential rather than actual emission reductions. Third, manure resources, crop demand, livestock production, and agricultural policies are unevenly distributed across regions. China’s crop–livestock spatial mismatch and regional differences in manure treatment conditions require region-specific recycling strategies (Bai et al., 2021), yet the spatial inequality and spillover patterns of manure-based carbon-mitigation potential remain insufficiently examined.

To address these gaps, this study constructs a provincial panel accounting and spatial econometric framework to assess the upper-bound theoretical carbon-mitigation potential of mineral fertilizer substitution by livestock manure nutrients in China. Conceptually, the study is grounded in three linked perspectives. Resource endowment theory suggests that livestock manure resources provide the material basis for manure-derived NPK nutrient supply. Circular economy theory explains how manure nutrients can be returned to cropland to substitute mineral fertilizers and avoid fertilizer-related emissions. Spatial spillover theory further implies that manure-based substitution potential may be spatially correlated because livestock production, fertilizer dependence, cropping structure, agricultural services, and environmental policies are linked across neighboring or economically connected provinces.

Specifically, this study uses provincial panel data for 31 Chinese provinces from 2005 to 2024 to estimate livestock and poultry manure resources, manure-derived N, P₂O₅, and K₂O nutrient supplies, and the associated upper-bound theoretical ER. The Dagum Gini decomposition is applied to identify regional inequality and its sources, global Moran’s I is used to test spatial clustering, and the Spatial Durbin Model is employed to examine local associations and spatial spillover patterns. To avoid overinterpreting the accounting results, this study further conducts sensitivity analyses under alternative manure substitution rates and fertilizer emission-factor perturbations. The main contributions are threefold: first, it links manure nutrient accounting with fertilizer-related carbon-mitigation assessment at the provincial scale; second, it reveals the temporal evolution, regional inequality, and spatial dependence of manure-based upper-bound ER; and third, it explicitly distinguishes theoretical mitigation potential from actual emission reduction, thereby providing a more cautious empirical basis for region-specific manure recycling and agricultural carbon-mitigation policies.

2 Conceptual framework

The upper-bound theoretical carbon-mitigation potential of manure-based mineral fertilizer substitution is formed through the interaction of manure resource endowment, manure-derived nutrient supply, cropland nutrient demand, technological and institutional conditions, agricultural organization, and spatial linkages among provinces. This study therefore constructs its conceptual framework based on resource endowment theory, circular economy theory, and spatial spillover theory, while incorporating the technological, institutional, and organizational conditions that may enable or constrain manure nutrient recycling.

First, from the perspective of resource endowment theory, livestock and poultry manure resources constitute the material basis for manure-derived nutrient supply. Provinces with larger livestock sectors and greater manure resources per unit of cropland generally possess a larger potential supply of manure-derived N, P, and K nutrients. These nutrient resources provide the physical foundation for substituting mineral fertilizers. Manure resource intensity is therefore expected to be positively associated with upper-bound theoretical carbon-mitigation potential, particularly through local direct effects. However, resource abundance alone does not guarantee actual carbon mitigation because the conversion of manure resources into usable fertilizer nutrients depends on collection, storage, treatment, transportation, field application, and cropland absorption conditions.

Second, circular economy theory explains how livestock manure can be transformed from an agricultural waste stream into a recyclable nutrient resource. In a linear agricultural production system, untreated manure may generate environmental pollution while crop production remains dependent on mineral fertilizers. In a circular agricultural system, manure-derived nutrients are returned to cropland, reducing mineral fertilizer demand and avoiding fertilizer-related emissions. Within this pathway, manure resource endowment determines the potential nutrient supply, while agricultural economic scale and cropping structure influence crop nutrient demand and cropland absorption capacity. The resulting estimate is defined in this study as the upper-bound theoretical carbon-mitigation potential under complete manure nutrient recovery and equivalent mineral fertilizer substitution.

The conversion of resource endowment into feasible nutrient substitution is further conditioned by technological capacity, institutional support, and agricultural organization. Agricultural mechanization may reduce the labor and operational costs of manure transport and field application, although general mechanization does not necessarily represent manure-specific collection, composting, spreading, and nutrient-management capacity. Agricultural science and technology input may improve manure treatment, nutrient testing, application precision, and recycling efficiency. Environmental regulation can encourage pollution control and manure treatment investment, but it may also induce the closure or relocation of livestock farms. Government agricultural support may promote infrastructure construction, technical services, and organic fertilizer use, although its effectiveness depends on whether public expenditure is directed toward manure-specific recycling activities. The number of farms reflects the development of specialized agricultural operating entities and the organizational basis for manure collection and utilization, but it does not directly measure the average operating scale or actual manure-recycling capacity of individual farms.

Third, spatial spillover theory suggests that manure-based fertilizer substitution is not only a local process but may also be associated with interprovincial linkages. Livestock production, fertilizer-use behavior, agricultural mechanization, urbanization, cropping structure, environmental regulation, public agricultural support, technological investment, and agricultural service systems may display spatial correlation across geographically adjacent or economically connected provinces. Neighboring provinces may share similar production structures, policy environments, input markets, transport networks, and technical service systems. These common conditions may generate spatial clustering in upper-bound theoretical mitigation potential.

At the same time, the physical transferability of manure nutrients is constrained. Manure is bulky, heterogeneous in nutrient concentration, costly to transport, and subject to nutrient losses during storage and movement. Consequently, provinces may exhibit similar levels of theoretical mitigation potential because they share comparable agricultural structures, even when manure itself is not transported across provincial boundaries. The spatial dependence identified in this study therefore represents broader interregional associations in resource endowment, production conditions, institutional environments, and agricultural services rather than unrestricted cross-provincial movement of untreated manure.

Based on this framework, manure resource intensity is expected to have a positive local association with upper-bound theoretical carbon-mitigation potential. Agricultural economic scale may increase potential nutrient demand and organizational capacity, while cropping structure may affect the compatibility between manure nutrient supply and cropland absorption. High chemical fertilizer intensity may indicate stronger dependence on mineral fertilizer-based production systems and may therefore be negatively associated with manure substitution potential. Mechanization and agricultural science and technology input may improve the technical conditions for manure utilization, but their effects depend on whether they support manure-specific treatment and field application.

Urbanization may operate through opposing channels. It can improve infrastructure, public services, environmental governance, and agricultural service markets, but it may also reduce agricultural land, reallocate rural labor, restrict livestock production, and intensify crop–livestock spatial separation. Environmental regulation may promote manure recycling through stricter pollution-control requirements, while simultaneously reducing local manure supply through livestock-sector adjustment. Government agricultural support may facilitate manure treatment and utilization, but broad agricultural expenditure may not necessarily target manure-specific activities. The number of farms may reflect stronger agricultural organization, although a larger number of operating entities does not automatically imply larger average farm size or better manure-recycling performance. The signs and magnitudes of these variables are therefore determined empirically rather than imposed in advance.

The spatially lagged explanatory variables may also exhibit heterogeneous effects. Technological services, policy practices, agricultural machinery, organic fertilizer markets, and production conventions may diffuse across provincial boundaries and generate positive spillovers. Conversely, neighboring provinces may compete for agricultural services, organic fertilizer products, transport capacity, cropland nutrient demand, or production factors, producing negative spatial associations. The empirical model therefore examines both local effects and spatial spillover effects rather than assuming a uniform direction of interprovincial influence.

It should be emphasized that this conceptual framework distinguishes upper-bound theoretical mitigation potential from actual realized emission reduction. The upper-bound potential estimated in this study represents the maximum technical mitigation space under complete manure nutrient recovery and equivalent mineral fertilizer substitution. Environmental regulation, government support, technological input, mechanization, urbanization, and agricultural organization do not mechanically determine this accounting value. Instead, they are introduced as conditioning factors associated with its spatial distribution and with the institutional and technical environment in which manure recycling may be implemented. Actual mitigation remains dependent on manure collection efficiency, treatment and storage losses, transportation feasibility, nutrient-demand matching, field application efficiency, farmer adoption, and manure-specific policy support. This distinction is essential for interpreting the spatial econometric results and designing feasible manure-recycling strategies.

3 Data sources and methods

3.1 Study area and data sources

This study covers 31 provincial-level administrative units in mainland China (provinces, autonomous regions, and municipalities) over the period 2005–2024. All data were obtained from official and publicly available statistical sources in China. Livestock data, including the stocks and slaughter numbers of major livestock and poultry categories, were collected from the China Rural Statistical Yearbook (2006–2025 editions). Crop production data, mainly crop planting areas, were obtained from the China Statistical Yearbook (2006–2025 editions). Other coefficients required for indicator accounting were taken from authoritative domestic and international studies and are reported with citations in the subsequent calculation steps (see Table 1).

3.2 Accounting for livestock manure, nutrient resources, and upper-bound theoretical carbon-mitigation potential

This study constructs a stepwise accounting framework to quantify livestock and poultry manure resources, manure-derived nutrient resources, and the upper-bound theoretical carbon-mitigation potential generated by mineral fertilizer substitution. The accounting process includes three steps. First, annual manure resources are estimated by combining livestock and poultry quantities with species-specific daily manure excretion coefficients and feeding-cycle parameters. Second, manure-derived N, P, and K resources are calculated using nutrient-content coefficients and are further converted into N, P₂O₅, and K₂O equivalents to ensure consistency with mineral fertilizer nutrient accounting. Third, the manure-derived N, P₂O₅, and K₂O resources are converted into upper-bound theoretical carbon-mitigation potential using nutrient-specific fertilizer emission factors.

The accounting boundary covers the major livestock and poultry categories available in provincial statistical data, including cattle, horses, donkeys, mules, sheep, pigs, and poultry. The total manure resource refers to the annual fresh-weight manure excreted by these livestock and poultry categories. The nutrient resource refers to the N, P, and K contained in the estimated manure resources. The carbon-mitigation indicator used in this study refers to the upper-bound theoretical mitigation potential under the assumption that manure-derived nutrients can be fully recycled and used to substitute equivalent mineral fertilizer nutrients. Therefore, this indicator measures the maximum technical mitigation space embodied in manure nutrient resources rather than actual realized emission reductions.

3.2.1 Total manure resource

The total manure resource was estimated according to livestock and poultry quantities, daily manure excretion coefficients, and feeding-cycle parameters. Let , , and denote province, year, and livestock type, respectively. The livestock and poultry types considered in this study include cattle, horses, donkeys, mules, sheep, pigs, and poultry. The annual manure resource of livestock type in province and year is estimated as shown in Equation 1:where represents the annual manure resource of livestock type in province and year , measured in tonnes; represents the statistical quantity of livestock type , measured in heads or birds; is the daily manure excretion coefficient of livestock type , measured in kg per head or bird per day; is the feeding cycle of livestock type , measured in days; and 1,000 is the conversion coefficient from kilograms to tonnes.

The construction of depends on the statistical characteristics of different livestock and poultry categories. For long-cycle livestock, including cattle, horses, donkeys, mules, and sheep, year-end stock numbers are used because these animals are generally raised over a relatively long period and their annual manure generation is closely related to standing populations. For short-cycle livestock and poultry, especially pigs and poultry, annual slaughter or output numbers are used because their production cycles are shorter and several batches may be raised within 1 year. The feeding-cycle parameter is therefore introduced to align different livestock quantity bases with annual manure generation. This treatment ensures that the manure resource accounting reflects both livestock scale and species-specific production cycles.

The total manure resource of province in year is obtained by summing across all livestock and poultry types as shown in Equation 2:where represents the total livestock and poultry manure resource in province and year , measured in tonnes, and represents the number of livestock and poultry categories included in the accounting framework.

The daily manure excretion coefficients and feeding-cycle parameters are taken from Xu et al. (2022). Specifically, the feeding cycles are set as 300 days for cattle, 365 days for horses, donkeys, mules, and sheep, 180 days for pigs, and 55 days for poultry. The corresponding daily manure excretion coefficients are 18.00 kg/day for cattle, 10.20 kg/day for horses, 8.56 kg/day for donkeys and mules, 2.00 kg/day for sheep, 2.44 kg/day for pigs, and 0.12 kg/day for poultry. These parameters are consistent with the coefficients reported in Table 1 and are applied uniformly across provinces and years to ensure comparability.

3.2.2 Nutrient resources

Based on the estimated manure resources, the nutrient resources contained in livestock and poultry manure were calculated using species-specific nutrient-content coefficients. Let , , and denote the contents of N, P, and K in manure from livestock type , respectively. These coefficients are expressed as percentages of fresh manure weight. The N, P, and K resources from livestock type in province and year are calculated as shown in Equations 3–5:where , , and represent the N, P, and K resources contained in manure from livestock type in province and year , respectively.

The total manure-derived N, P, and K resources in province and year are obtained by summing across all livestock and poultry categories as shown in Equations 6–8:where , , and denote the total manure-derived N, P, and K resources in province and year , respectively.

To maintain consistency with mineral fertilizer nutrient accounting, elemental phosphorus and potassium are further converted into P₂O₅ and K₂O equivalents. The conversion coefficient from elemental P to P₂O₅ is 2.29, and the conversion coefficient from elemental K to K₂O is 1.20 (Li et al., 2026). Therefore, manure-derived P₂O₅ and K₂O resources are calculated as Equations 9, 10:where and represent manure-derived phosphorus and potassium resources expressed as P₂O₅ and K₂O equivalents, respectively. Together with , these converted nutrient indicators provide the basis for estimating the upper-bound theoretical carbon-mitigation potential from mineral fertilizer substitution.

3.2.3 Upper-bound theoretical carbon-mitigation potential from mineral fertilizer substitution

This study estimates the carbon-mitigation potential generated by substituting mineral fertilizer nutrients with manure-derived nutrients. To avoid overinterpretation, the estimated indicator is defined as the upper-bound theoretical carbon-mitigation potential rather than actual realized emission reduction. The baseline accounting assumes that manure-derived N, P₂O₅, and K₂O nutrients are fully collected, effectively treated, transported to cropland, and equivalently substituted for the same amounts of mineral fertilizer nutrients. Under this assumption, the avoided fertilizer-related emissions represent the maximum technical mitigation space embodied in manure nutrient resources.

Let denote the upper-bound theoretical carbon-mitigation potential in province and year . Let , , and represent the manure-derived nutrient resources expressed as N, P₂O₅, and K₂O, respectively. The nutrient-specific emission factors of nitrogen, phosphate, and potash fertilizers are denoted by , , and , respectively. The baseline values are set as 2.116, 0.636, and 0.180 t CO₂-eq per t nutrient for N, P₂O₅, and K₂O fertilizers, respectively (Xu et al., 2022). The upper-bound theoretical carbon-mitigation potential is calculated as Equation 11:where represents the maximum avoided fertilizer-related emissions that could be achieved if all manure-derived nutrients were effectively recycled and used to replace equivalent mineral fertilizer nutrients.

This accounting framework should be interpreted as a technical-potential assessment. It does not imply that the estimated ER has been fully realized in actual agricultural production. In practice, the realized mitigation effect is constrained by manure collection efficiency, storage and composting losses, treatment capacity, transportation radius, timing mismatches between manure supply and crop nutrient demand, field application efficiency, and farmers’ willingness to substitute mineral fertilizer with organic manure. Therefore, the baseline ER provides an upper-bound benchmark for evaluating the spatial distribution and temporal evolution of manure-based fertilizer-substitution potential across provinces.

3.2.4 Sensitivity analysis of substitution rates and emission factors

Because the baseline ER is calculated under the assumption of complete nutrient substitution, sensitivity analysis is conducted to examine how the estimated mitigation scale changes under alternative manure recycling and substitution conditions. First, we introduce different effective substitution-rate scenarios. Let denote the proportion of manure-derived nutrients that can be effectively recycled and substituted for equivalent mineral fertilizer nutrients. The scenario-specific ER is calculated as Equation 12:where . The 30% scenario represents a conservative recycling condition with substantial constraints in collection, treatment, and transport; the 50% scenario represents a moderate substitution condition; the 70% scenario represents a relatively high recycling condition; and the 90% scenario represents a near-complete but still practically constrained substitution condition. These scenarios allow us to distinguish between the upper-bound technical potential and more conservative estimates under partial manure recycling and utilization.

Second, considering that fertilizer emission intensities may change with energy structure, production technology, and supply-chain efficiency, this study further tests the sensitivity of ER to the choice of emission factors. The baseline emission factors for N, P₂O₅, and K₂O fertilizers are jointly perturbed by ±20%. Let denote the perturbation coefficient, where . The adjusted emission factors are expressed as Equation 13:

Accordingly, the emission-factor-adjusted ER is calculated as Equation 14:where represents the upper-bound theoretical mitigation potential under the low-emission-factor scenario or high-emission-factor scenario .

It should be noted that, when the substitution rate or emission factors are adjusted by a common positive scalar across all provinces and years, the absolute magnitude of ER changes proportionally, whereas relative spatial rankings, Dagum Gini coefficients, and Moran’s I statistics remain mathematically unchanged. Therefore, the sensitivity analysis is mainly used to assess the uncertainty in the numerical scale of ER, while the spatial disparity and clustering results are interpreted on the basis of the baseline upper-bound ER.

3.3 Dagum gini decomposition and identification of the sources of regional disparities

This study measures inter-provincial inequality and decomposes its sources in provincial carbon mitigation potential to reveal the extent of spatial imbalance and its structural determinants in China’s livestock-manure fertilization pathway. Although the conventional Gini coefficient can quantify the magnitude of overall inequality, it cannot clearly distinguish whether such inequality mainly stems from within-region differences or between-region differences. Moreover, when regional distributions overlap (i.e., cross-region rank reversals occur), standard decomposition approaches may lead to interpretational bias. Therefore, this study adopts the Dagum Gini coefficient and its decomposition framework, which allows for distributional overlap and rigorously decomposes overall inequality into three components: within-region inequality, net between-region inequality, and transvariation intensity (Qin et al., 2022). This structured decomposition helps identify the dominant contributors to overall inequality and provides a basis for subsequent heterogeneity analysis and policy discussion.

The inequality object is provincial carbon mitigation potential. Let denote the carbon mitigation potential in province in year , the number of provinces, and the national mean of in year . The overall Dagum Gini coefficient in year is defined as Equation 15:

This index captures the relative dispersion of provincial carbon mitigation potential normalized by the national mean; a larger value indicates greater inter-provincial inequality. To examine dynamic evolution, this study computes annually and further decomposes it to identify how the sources of inequality change over time.

To enhance the structural and policy relevance of inequality interpretation, the 31 provinces are grouped into four macro-regions—eastern, central, western, and northeastern China. Suppose the sample is partitioned into regional groups. Group contains provinces and has group mean . Then the overall Gini coefficient can be decomposed as Equation 16:where denotes the contribution of within-region inequality, the contribution of net between-region inequality, and the contribution of transvariation intensity. Within-region inequality reflects the average disparity among provinces within the same region. It is computed by first obtaining the within-group Gini coefficient for each region and then aggregating them using weights that account for regional size and mean level. The within-group Gini coefficient for region is calculated as Equation 17:

Accordingly, the within-region contribution is calculated as Equation 18:where the weight jointly reflects the region’s population share (number of provinces) and its mean mitigation level relative to the national total, ensuring comparability and additivity of regional contributions to overall inequality.

In the Dagum framework, between-region inequality is further decomposed into net between-region inequality and transvariation intensity. Net between-region inequality captures the separable differences in mean levels and relative positions across regions, whereas transvariation intensity captures the additional inequality generated by distributional overlap between regions. To implement this split, the cross-group inequality between any two regions and is first calculated. The between-group Gini coefficient is calculated as Equation 19:

Dagum’s method further introduces a relative impact coefficient to measure the degree of overlap and net dominance between the two regional distributions. Intuitively, when the distributions of two regions overlap less and the disparity is largely attributable to differences in overall levels, tends to be larger and the between-region inequality is primarily explained by the net difference. Conversely, when the distributions overlap substantially and mean gaps alone cannot explain the disparity, tends to be smaller and the contribution of transvariation intensity increases. Accordingly, net between-region inequality and transvariation intensity can be expressed as Equations 20, 21:

This decomposition enables not only the quantification of overall inequality, but also a structured interpretation of whether inequality is dominated by within-region disparities or net between-region gaps, and how much is attributable to distributional overlap (transvariation). In turn, it provides more informative evidence for understanding the mechanisms behind the spatial disparities in carbon mitigation potential from livestock manure fertilization.

3.4 Spatial autocorrelation analysis

To examine whether the upper-bound theoretical carbon-mitigation potential generated by manure-based mineral fertilizer substitution exhibits spatial dependence across provinces, this study applies global Moran’s I to test the spatial clustering pattern of provincial ER (Yang et al., 2023). Spatial autocorrelation analysis provides a preliminary statistical basis for introducing spatial econometric models, because significant spatial dependence implies that the ER of one province may not be independent of the ER of neighboring or economically connected provinces.

Let denote the upper-bound theoretical carbon-mitigation potential of province in year , and let denote the national mean in year . The global Moran’s I statistic is calculated as Equation 22:where n is the number of provinces; is the element of the spatial weight matrix ; and . A positive and statistically significant Moran’s I indicates spatial clustering, meaning that provinces with high ER tend to be adjacent or closely connected to provinces with similarly high ER, and provinces with low ER tend to be connected to provinces with similarly low ER. A negative value indicates spatial dispersion, while a value close to zero indicates weak or no global spatial association.

The construction of the spatial weight matrix is crucial because it defines the assumed spatial linkage structure among provinces. This study uses two types of spatial weight matrices. The baseline matrix is a first-order geographic contiguity matrix. If province and province share a common boundary, ; otherwise, . The diagonal elements are set to zero, namely, . To ensure comparability across provinces with different numbers of neighbors, the matrix is row-standardized so that each row sums to one. As shown in Equation 23:where denotes the row-standardized contiguity weight.

Considering that manure recycling, fertilizer substitution, agricultural production factors, and policy diffusion may be shaped not only by geographic adjacency but also by economic similarity, this study further constructs an economic–geographic composite matrix for robustness analysis. The geographic component is measured by inverse geographic distance between provincial capital cities. As shown in Equation 24:where is the great-circle distance between province and province . The economic component is measured by the inverse difference in the average agricultural economic scale between provinces during the sample period. As shown in Equation 25:where and represent the average agricultural GDP of provinces and over 2005–2024. The constant term one is added to avoid division by zero. The economic–geographic composite weight is then defined as Equation 26:and . The composite matrix is also row-standardized. As shown in Equation 27:

The contiguity matrix captures direct geographic adjacency, while the economic–geographic matrix captures a broader linkage structure combining geographic proximity and agricultural economic similarity. Comparing results under these two matrices helps assess whether the observed spatial dependence is sensitive to the assumed spatial interaction structure.

3.5 Spatial durbin model

Based on the global spatial autocorrelation results, this study further employs a Spatial Durbin Model (SDM) to examine the spatial association and spillover effects of upper-bound theoretical ER across provinces (Liu and Chen, 2025). Compared with non-spatial panel models, the SDM explicitly incorporates both the spatial lag of the dependent variable and the spatial lags of explanatory variables. It therefore allows ER in one province to be associated not only with local explanatory variables, but also with neighboring provinces’ ER and neighboring provinces’ agricultural production conditions.

The dependent variable is the upper-bound theoretical carbon-mitigation potential from manure-based fertilizer substitution. To reduce heteroscedasticity and scale differences, the dependent variable is transformed into logarithmic form. The core explanatory variable is manure resource intensity, which reflects manure resource supply per unit of cropland and captures the local resource endowment for manure-based nutrient substitution. The control variables are selected according to four interconnected dimensions: agricultural production and nutrient demand, technological capacity, institutional support, and agricultural organization. Chemical fertilizer application intensity reflects the dependence of agricultural production on mineral fertilizer inputs. Agricultural GDP represents the overall scale of the agricultural economy, while cropping structure captures differences in crop nutrient demand and cropland absorption conditions. Agricultural mechanization and agricultural science and technology input intensity reflect the technical conditions supporting manure collection, transportation, treatment, and field application. Urbanization is included to account for changes in land use, rural labor allocation, and crop–livestock spatial separation. Environmental regulation intensity and government agricultural support intensity represent the institutional environment and public-sector support associated with agricultural pollution control and resource recycling. The logarithm of the number of farms is included to capture the development of specialized agricultural operating entities and the organizational basis for manure collection and utilization.

The baseline two-way fixed-effects SDM is specified as Equation 28:where denotes the logarithm of the upper-bound theoretical carbon-mitigation potential in province and year ; is a vector of explanatory variables, including manure resource intensity and control variables; is the element of the spatial weight matrix; is the spatially lagged dependent variable; and denotes the spatially lagged explanatory variables. The parameter captures the spatial dependence of ER across provinces; measures the association between local explanatory variables and local ER; and captures the association between neighboring provinces’ explanatory variables and local ER. Province fixed effects control for time-invariant provincial characteristics such as geographic endowments, long-term agricultural structure, and institutional background, while year fixed effects absorb common shocks such as national agricultural policies, macroeconomic fluctuations, and technological progress. is the random error term.

The SDM is preferred because it nests both the Spatial Autoregressive Model (SAR) and the Spatial Error Model (SEM) under specific restrictions. If the coefficients of the spatially lagged explanatory variables are jointly zero, namely, , the SDM can be simplified to a SAR model. If the restriction holds, the SDM can be simplified to a SEM model. Therefore, the validity of using SDM rather than SAR or SEM should be supported by formal model comparison tests.

In spatial models, the estimated coefficients of SDM cannot be directly interpreted as marginal effects because changes in one province may affect other provinces through spatial feedback loops. Therefore, following the spatial econometric decomposition framework, this study reports direct effects, indirect effects, and total effects. The direct effect measures the average association between a local explanatory variable and local ER after accounting for spatial feedback. The indirect effect measures the average spillover association between neighboring provinces’ explanatory variables and local ER. The total effect is the sum of the direct and indirect effects. This decomposition provides a more accurate basis for interpreting the local and spatial spillover pathways of manure resource intensity and other agricultural production factors.

It should be emphasized that the SDM results are interpreted as conditional spatial associations rather than strict causal effects. Although two-way fixed effects can control for time-invariant provincial heterogeneity and nationwide annual shocks, they cannot fully eliminate potential endogeneity caused by omitted time-varying factors, reverse causality, or simultaneous changes in livestock production, environmental governance, manure treatment infrastructure, and agricultural modernization. Therefore, the spatial econometric analysis is used to identify robust spatial dependence and possible spillover patterns, while causal mechanisms are discussed cautiously and supported by robustness tests where feasible.

3.6 Model diagnostics and robustness strategy

To justify the spatial econometric specification, this study conducts a series of model diagnostic and robustness tests. First, global Moran’s I is used to examine whether provincial ER exhibits significant spatial dependence. If ER shows significant positive spatial autocorrelation, non-spatial panel models may suffer from omitted spatial dependence, and spatial econometric models become necessary.

Second, Lagrange Multiplier tests are conducted based on the residuals of the non-spatial panel model. Both LM-lag and LM-error tests are used to identify whether spatial dependence is more likely to exist in the dependent variable or in the error term. Robust LM-lag and Robust LM-error tests are further applied to distinguish the two forms of spatial dependence when both standard LM tests are significant. Significant results from these tests indicate that spatial effects should be explicitly modeled.

Third, LR and Wald tests are used to examine whether the SDM can be simplified to SAR or SEM. The restriction is tested to determine whether the SDM can be reduced to SAR, while the restriction is tested to determine whether the SDM can be reduced to SEM. If these restrictions are rejected, the SDM is statistically preferred because it can simultaneously capture spatial dependence in the dependent variable and spatial spillovers from explanatory variables.

Fourth, fixed-effects specification tests are conducted to determine the appropriate panel structure. Province fixed effects control for unobserved time-invariant regional heterogeneity, while year fixed effects control for common temporal shocks. Given that manure resources, agricultural production systems, and policy environments differ substantially across provinces and change over time, the baseline model adopts a two-way fixed-effects specification. This setting reduces bias from unobserved provincial characteristics and national annual shocks.

3.7 Regional classification

For the Dagum Gini decomposition and regional heterogeneity analysis, this study adopts the widely used four-region classification of China—eastern, central, western, and northeastern China—to reflect structural differences in resource endowments, livestock production layout, agricultural production conditions, and socio-economic development stages. Following the commonly applied “four economic zones” convention used in official statistics and related studies, provinces are categorized according to geographic location and development gradient as follows.

The eastern region includes Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and Hainan. The central region includes Shanxi, Anhui, Jiangxi, Henan, Hubei, and Hunan. The western region includes Inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang. The northeastern region includes Liaoning, Jilin, and Heilongjiang.

This classification facilitates the identification of within-group and between-group disparity sources in inequality decomposition, and it also supports subsequent regional heterogeneity testing in the dynamic panel framework, thereby enabling a discussion of how the effects of manure resource intensity on manure-based mitigation outcomes may differ across distinct regional contexts and the corresponding policy implications (Figure 1).

FIGURE 1

4 Results

4.1 Temporal evolution of livestock and poultry manure resources and manure-derived nutrient supply

From 2005 to 2024, livestock and poultry manure resources in China showed a clear decline–rebound–slight decline trajectory. National manure resources decreased from 1.53 × 10⁹ t in 2005 to 1.10 × 10⁹ t in 2019, indicating a contraction in livestock manure supply during the earlier period. After 2019, manure resources recovered markedly and reached 1.32 × 10⁹ t in 2023, before slightly declining to 1.28 × 10⁹ t in 2024. Overall, the 2024 level remained 16.48% lower than that in 2005, suggesting that the recent recovery had not fully reversed the long-term decline.

The regional structure of manure resources also changed over time. The western region consistently contributed the largest share and became increasingly important, with its national share rising from 38.43% in 2005 to 46.31% in 2024. In contrast, the eastern region’s share declined from 32.91% to 25.60%, reflecting a relative contraction of manure resource supply in more urbanized and land-constrained areas. The central region remained broadly stable, while the northeastern region accounted for a relatively small but slightly increasing share. This pattern indicates that China’s manure resource endowment has become more concentrated in resource-oriented and agriculture-dominant regions. Manure-derived nutrient supply followed a similar temporal pattern. Total manure-derived N, P, and K supply decreased from 1.98 × 10⁷ t in 2005 to 1.52 × 10⁷ t in 2019, then recovered to 1.84 × 10⁷ t in 2023 and slightly declined to 1.79 × 10⁷ t in 2024. Compared with 2005, the 2024 level was 9.75% lower, a smaller decline than that of total manure resources. This suggests that manure-derived nutrient supply was relatively more resilient than manure quantity, possibly because changes in livestock structure affected manure quantity and nutrient composition differently. In terms of nutrient composition, manure-derived nitrogen remained the largest component throughout the study period, followed by phosphorus and potassium. In 2024, manure-derived N, P, and K supplies were 7.14 × 10⁶ t, 5.66 × 10⁶ t, and 5.05 × 10⁶ t, respectively. Over the whole period, nitrogen accounted for approximately 40.07% of total manure-derived nutrient supply, phosphorus for 31.57%, and potassium for 28.36%. This nitrogen-dominant structure provides the main nutrient basis for subsequent mineral fertilizer substitution and upper-bound theoretical ER accounting.

Overall, China’s manure resources and manure-derived nutrient supply experienced a long-term decline before 2019, followed by a recovery after 2019 and a slight decrease in 2024. Spatially, manure resource endowment and nutrient supply became increasingly concentrated in the western region, while the eastern region’s contribution declined. These results indicate that manure-based nutrient recycling potential is shaped not only by national livestock production changes but also by regional shifts in livestock resource distribution (Figure 2).

FIGURE 2

4.2 Upper-bound theoretical ER from manure-based mineral fertilizer substitution in China

The upper-bound theoretical carbon-mitigation potential from manure-based mineral fertilizer substitution showed a clear stagewise trajectory during 2005–2024. Under the complete nutrient-recycling and equivalent-substitution assumption, national ER reached 2.183 × 10⁷ t CO₂-eq in 2005, then declined continuously to 1.677 × 10⁷ t CO₂-eq in 2019, representing a decrease of 23.17%. After 2019, ER rebounded to 2.021 × 10⁷ t CO₂-eq in 2023, indicating a recovery of manure-derived nutrient supply and substitution space. In 2024, ER slightly decreased to 1.962 × 10⁷ t CO₂-eq, down by 2.89% compared with 2023, but still 16.99% higher than the 2019 trough. Overall, the 2024 level remained 10.11% lower than that in 2005, suggesting that although the recent recovery was evident, the long-term upper-bound mitigation space had not fully returned to its early-period level.

This temporal pattern is closely related to changes in livestock and poultry production, manure resource supply, and agricultural structural adjustment. The decline before 2019 may reflect the combined effects of livestock-sector restructuring, strengthened environmental regulation of livestock breeding, and the spatial reorganization of animal production. The rebound after 2019 is likely associated with the recovery of livestock production capacity, especially the restoration of pig production, together with policy efforts to promote manure resource utilization and fertilizer reduction. However, because the ER indicator used here is an upper-bound theoretical estimate, the rebound should be interpreted as an increase in potential substitution space rather than evidence that equivalent actual emission reductions were fully realized.

Regionally, ER displayed a stable but increasingly resource-oriented spatial pattern. The western region consistently contributed the largest share of national upper-bound ER and became more prominent over time. In 2005, the eastern, central, western, and northeastern regions contributed 7.22 × 10⁶, 5.31 × 10⁶, 8.15 × 10⁶, and 1.14 × 10⁶ t CO₂-eq, respectively, accounting for 33.07%, 24.35%, 37.36%, and 5.23% of the national total. By 2024, the corresponding ER values were 5.41 × 10⁶, 4.69 × 10⁶, 8.33 × 10⁶, and 1.19 × 10⁶ t CO₂-eq, with shares of 27.59%, 23.89%, 42.43%, and 6.09%, respectively. These results indicate a gradual shift in the spatial center of manure-based substitution potential toward the western region, where livestock resources and cropland carrying capacity provide a stronger resource basis for manure nutrient recycling.

The eastern region showed a marked decline in both absolute ER and national share. This pattern may be related to the contraction and relocation of livestock production under land constraints, environmental regulation, and urban expansion. In contrast, the western region maintained a high and rising contribution, indicating that manure-derived nutrient resources are increasingly concentrated in resource-endowed and agriculture-oriented regions. The central region remained relatively stable, while the northeastern region contributed a smaller absolute amount but showed a slight increase in national share, reflecting its continued role as an important agricultural production base.

At the provincial level, high ER values were mainly concentrated in provinces with large livestock production, abundant manure resources, and relatively strong cropland nutrient absorption capacity. In 2024, Guangxi, Sichuan, Shandong, Yunnan, and Hubei ranked among the leading provinces, with upper-bound ER values of 1.62 × 10⁶, 1.39 × 10⁶, 1.19 × 10⁶, 1.12 × 10⁶, and 1.05 × 10⁶ t CO₂-eq, respectively. These provinces represent the main contributors to national manure-based fertilizer-substitution potential. Nevertheless, their estimated ER values should be understood as technical upper bounds. The actual realization of such potential depends on manure collection rates, treatment and storage losses, transportation feasibility, cropland demand matching, and farmers’ adoption of organic fertilizer substitution.

Overall, the results show that China’s upper-bound theoretical ER from manure-based fertilizer substitution experienced a decline–rebound–slight decline trajectory during 2005–2024. The spatial pattern became increasingly concentrated in the western region and several major agricultural provinces. This indicates that manure nutrient recycling has considerable theoretical mitigation potential, but the conversion of this potential into actual emission reductions requires improvements in manure collection, treatment, transport, and field application systems (Figure 3).

FIGURE 3

4.3 Sensitivity analysis of upper-bound theoretical ER

Given that the baseline ER is estimated under the assumption of complete recovery and equivalent substitution of manure-derived nutrients, sensitivity analysis was conducted to evaluate how the estimated mitigation scale changes under alternative substitution-rate and emission-factor scenarios. The results show that the absolute magnitude of ER is sensitive to the assumed substitution rate and fertilizer emission factors, whereas the temporal trajectory and spatial structure remain stable under uniform scenario settings.

First, under different effective substitution-rate scenarios, national ER decreases proportionally as the substitution rate declines from the upper-bound benchmark. In the baseline upper-bound scenario, national ER was 2.183 × 10⁷ t CO₂-eq in 2005, declined to 1.677 × 10⁷ t CO₂-eq in 2019, rebounded to 2.021 × 10⁷ t CO₂-eq in 2023, and then slightly decreased to 1.962 × 10⁷ t CO₂-eq in 2024. Under the conservative 30% substitution scenario, the corresponding values were 0.655 × 10⁷, 0.503 × 10⁷, 0.606 × 10⁷, 0.589 × 10⁷t CO₂-eq, respectively. Under the 50%, 70%, and 90% substitution scenarios, the estimated ER increased proportionally, but all scenarios retained the same stagewise pattern of decline before 2019, recovery during 2019–2023, and a slight decrease in 2024 (Table 2).

These results indicate that the baseline upper-bound ER should not be interpreted as realized emission reduction. Instead, it represents the maximum technical mitigation space under complete manure nutrient recycling and equivalent mineral fertilizer substitution. When practical constraints such as collection efficiency, treatment losses, transport distance, and field application efficiency are considered, the realized ER would be lower than the upper-bound estimate. However, the temporal pattern remains robust across substitution-rate scenarios. Specifically, ER decreased by 23.17% from 2005 to 2019, increased by 20.47% from 2019 to 2023, decreased by 2.89% from 2023 to 2024, and remained 10.11% lower in 2024 than in 2005.

Second, the sensitivity of ER to fertilizer emission factors was examined by jointly perturbing the N, P₂O₅, and K₂O fertilizer emission factors by ±20%. Under the low-emission-factor scenario, national ER in 2024 was 1.570 × 10⁷ t CO₂-eq, while under the high-emission-factor scenario it increased to 2.355 × 10⁷ t CO₂-eq. In 2005, the corresponding values were 1.746 × 10⁷ and 2.620 × 10⁷ t CO₂-eq, respectively. Although the numerical scale changed substantially, the overall temporal trajectory remained unchanged (Table 3).

The emission-factor sensitivity results suggest that the estimated magnitude of upper-bound ER depends on the selected fertilizer emission coefficients. This is important because fertilizer production technologies, energy structures, and supply-chain efficiencies may change over time. Nevertheless, when emission factors are perturbed uniformly across provinces and years, the main trend remains stable, indicating that the observed decline–rebound–slight decline pattern is driven primarily by changes in manure-derived nutrient resources rather than by the specific numerical values of the emission factors.

It should also be noted that the substitution-rate and emission-factor scenarios in this study are applied as uniform proportional adjustments. Therefore, they change the absolute ER magnitude but do not alter provincial rankings, regional shares, Dagum Gini coefficients, or Moran’s I statistics. In 2024, for example, the western region remained the largest contributor under all sensitivity scenarios, accounting for 42.43% of national ER, followed by the eastern region at 27.59%, the central region at 23.89%, and the northeastern region at 6.09%. This indicates that the main spatial conclusion is not dependent on the complete-substitution assumption, although the practical mitigation scale is strongly affected by the actual manure recycling and substitution rate.

4.4 Regional inequality in upper-bound theoretical ER: Dagum Gini decomposition

The Dagum Gini decomposition was used to examine regional inequality in upper-bound theoretical ER across the four macro-regions. Overall, interprovincial inequality remained relatively stable during 2005–2024. The national Dagum Gini coefficient fluctuated within a narrow range of 0.338–0.353, declining from 0.349 in 2005 to 0.338 in 2013, rising to 0.353 in 2018, and then returning to 0.347 in 2024. This indicates that the spatial disparity in manure-based upper-bound ER was persistent but did not show a clear monotonic trend of either convergence or divergence.

The decomposition results show that the source structure of inequality changed over time. Transvariation intensity was the largest contributor in most years, indicating that regional distributions of upper-bound ER overlapped substantially and that inequality could not be explained only by simple mean differences among regions. However, its contribution gradually weakened compared with the early period. In contrast, net between-region differences became more important in the later period, suggesting that the regional gradient in manure-based substitution potential became clearer. The within-region contribution remained relatively stable, indicating that changes in overall inequality were mainly driven by between-region gaps and distributional overlap rather than by sharp changes within individual regions.

Within-region inequality also showed clear regional differences. The eastern region experienced a gradual increase in internal inequality, with its within-region Gini rising from 0.343 in 2005 to 0.403 in 2024. This suggests growing differentiation among eastern provinces in manure resource endowment, cropland absorption capacity, and manure recycling conditions. The central region showed a clear convergence trend, with its within-region Gini decreasing from 0.241 to 0.123 over the same period. The western region maintained relatively high internal inequality, reflecting the coexistence of high-potential provinces and provinces with more limited manure-based substitution capacity. The northeastern region had the lowest internal inequality overall, although a slight increase appeared in the later years (Table 4).

4.5 Spatial clustering of upper-bound theoretical ER: global Moran’s I

Global Moran’s I was used to examine whether upper-bound theoretical ER exhibited spatial clustering across provinces. The results show that provincial upper-bound ER was characterized by positive spatial autocorrelation during the study period, and the clustering pattern became more stable after 2009.

Under the geographic contiguity matrix, Moran’s I was positive in all selected years. The value was 0.045 in 2005 and was not statistically significant, indicating weak spatial dependence in the early period. From 2009 onward, Moran’s I became statistically significant and remained positive. It increased from 0.169 in 2009 to 0.247 in 2019, suggesting that the spatial clustering of upper-bound ER strengthened during this period. Although Moran’s I declined slightly after 2019, it remained significant in 2021, 2023, and 2024, with values of 0.238, 0.234, and 0.226, respectively. This indicates that provinces with high upper-bound ER tended to be adjacent to provinces with similarly high ER, while low-ER provinces also tended to cluster spatially.

The results based on the economic–geographic matrix are broadly consistent with those from the contiguity matrix. Moran’s I under this matrix was generally smaller, but it was significant from 2007 onward and followed a similar temporal pattern, strengthening before 2019 and weakening slightly afterward. This suggests that the spatial dependence of upper-bound ER is not only associated with geographic adjacency, but also related to broader economic–geographic linkages among provinces (Table 5).

4.6 Model diagnostics and specification tests for spatial econometric modeling

Before estimating the Spatial Durbin Model, this study conducted a series of diagnostic and specification tests to examine whether spatial econometric modeling was necessary and whether the SDM was preferred over alternative spatial models. The tests were based on the balanced provincial panel dataset for 31 provinces from 2005 to 2024.

First, the LM tests indicate the presence of spatial dependence. The LM-lag statistic is 5.723 and is significant at the 5% level, while the Robust LM-lag statistic is 12.259 and is significant at the 1% level. These results suggest that the upper-bound theoretical ER of one province is spatially associated with the ER levels of neighboring provinces. The standard LM-error test is not significant, but the Robust LM-error test is significant at the 1% level, indicating that spatial error dependence cannot be fully excluded. Overall, the LM diagnostics suggest that spatial effects should be explicitly incorporated into the econometric specification.

Second, model comparison results show that the SDM performs better than SAR and SEM. The log-likelihood of the SDM is 437.490, which is higher than that of SAR and SEM. In addition, the SDM has the lowest AIC and BIC values among the three spatial models, indicating better model fit. This suggests that a more flexible spatial specification is needed to capture both spatial dependence in the dependent variable and spatial spillovers from explanatory variables.

Third, the LR and Wald tests further support the use of the SDM. The LR test for SDM versus SAR is 127.629, and the LR test for SDM versus SEM is 133.498; both are significant at the 1% level. Similarly, the Wald tests also reject the restrictions that the SDM can be simplified to SAR or SEM. These results indicate that the spatially lagged explanatory variables should not be omitted and that the SDM cannot be reduced to either SAR or SEM. Therefore, the SDM is statistically preferred for the baseline spatial econometric analysis.

Finally, the Hausman test is significant, indicating that the fixed-effects specification is preferred over the random-effects specification. Accordingly, the baseline model adopts two-way fixed effects to control for time-invariant provincial heterogeneity and common year-specific shocks. This setting helps reduce bias from unobserved regional characteristics and nationwide policy or market changes.

Taken together, the diagnostic results provide statistical support for using the SDM with two-way fixed effects. However, the estimated spatial effects should still be interpreted as conditional spatial associations rather than definitive causal effects, because potential endogeneity may remain due to omitted time-varying factors, reverse causality, and simultaneous changes in livestock production, manure treatment capacity, and agricultural modernization (Table 6).

4.7 Spatial Durbin model estimation and decomposition of spatial spillover effects

Before estimating the Spatial Durbin Model, multicollinearity among the explanatory variables was examined using the variance inflation factor (VIF) and tolerance statistics. The VIF values ranged from 1.07 to 2.12, with the highest value observed for manure resource intensity, while all tolerance values were between 0.47 and 0.94. These values were well within the commonly accepted thresholds of VIF <10 and tolerance >0.10, indicating that no serious multicollinearity existed among the explanatory variables. The variables were therefore retained in the subsequent spatial econometric analysis.

The Spatial Durbin Model showed significant spatial dependence in the upper-bound theoretical carbon-mitigation potential. The coefficient of the spatially lagged dependent variable was 0.24 and significant at the 1% level, indicating that provinces with higher manure-based mitigation potential tended to be spatially associated with neighboring provinces exhibiting similarly high values. This result confirms that the spatial distribution of upper-bound mitigation potential is not independent across provinces and supports the continued use of the SDM rather than a conventional non-spatial panel model.

The coefficient of manure resource intensity remained significantly positive at 0.08. The effect decomposition further showed a significantly positive direct effect of 0.08 and a significantly positive total effect of 0.11, whereas the indirect effect of 0.03 was not statistically significant. These results indicate that manure resource intensity was primarily associated with the upper-bound theoretical mitigation potential within the province itself. Provinces with more abundant manure resources per unit of cropland have a larger potential supply of manure-derived nutrients and therefore possess greater technical space for replacing mineral fertilizers. However, this resource advantage did not generate a statistically robust cross-provincial spillover effect, which is consistent with the bulky nature of manure, its relatively high transport costs, and the constraints imposed by storage losses, transport distance, and local cropland absorption capacity.

Chemical fertilizer intensity showed a significantly negative direct effect of −4.04 and a significantly negative total effect of −8.75, while its indirect effect was not statistically significant. This suggests that greater dependence on mineral fertilizer was negatively associated with manure-based substitution potential. Provinces characterized by intensive chemical fertilizer use may have established production practices, input markets, and agricultural service systems that are more strongly oriented toward mineral fertilizers, thereby weakening the relative role of manure-derived nutrients in fertilizer substitution. Nevertheless, because the dependent variable represents an upper-bound theoretical potential rather than realized substitution, this result should be interpreted as a conditional association rather than evidence that fertilizer use directly reduces actual manure recycling.

Agricultural economic scale, measured by the logarithm of agricultural GDP, had a significantly positive direct effect of 0.52 and a significantly positive total effect of 0.45. Its indirect effect was statistically insignificant. This indicates that provinces with larger agricultural economies generally had greater local mitigation potential, possibly because a larger agricultural production system creates greater crop nutrient demand, stronger cropland absorption capacity, and more favorable conditions for organizing manure collection and field application. The absence of a significant indirect effect suggests that these advantages remained predominantly local rather than being transmitted systematically across provincial boundaries.

Agricultural mechanization exhibited a more complex spatial pattern. Its direct effect was significantly positive at 0.21, indicating that improved mechanization was positively associated with local upper-bound mitigation potential. Mechanized transport and field application may reduce the labor and operational constraints involved in manure utilization. However, its indirect effect was significantly negative at −0.47, resulting in a negative total effect of −0.26 that was only marginally significant at the 5% level. This finding suggests that general agricultural mechanization in neighboring provinces does not necessarily promote local manure recycling. Highly mechanized regions may reinforce specialized crop-production systems, attract agricultural services and production factors, or intensify interregional competition, while manure recycling requires specialized equipment for collection, composting, transport, and field return that is not fully captured by general mechanization indicators. Therefore, the local facilitating effect of mechanization was offset by its negative spatial association.

Urbanization did not have a significant direct effect, but its indirect effect was significantly negative at −2.92, resulting in a significantly negative total effect of −2.68. This indicates that the association between urbanization and manure-based mitigation potential operated mainly through spatial spillovers. Higher urbanization in neighboring provinces may be accompanied by agricultural land conversion, labor reallocation, stricter constraints on livestock production, and greater spatial separation between livestock farming and cropland. These changes may weaken regional crop–livestock linkages and reduce the conditions necessary for manure nutrient recycling in surrounding provinces. The insignificant local effect also suggests that urbanization may simultaneously improve infrastructure and environmental governance while constraining agricultural space, producing offsetting effects within the same province.

Cropping structure showed a significantly positive direct effect of 1.02 but a significantly negative indirect effect of −0.88. Consequently, its total effect was positive but statistically insignificant. The positive direct effect indicates that a cropping structure with greater nutrient demand or stronger capacity to absorb organic nutrients was associated with higher local substitution potential. In contrast, changes in cropping structure in neighboring provinces may create competition for organic fertilizer resources, agricultural service capacity, and manure transportation networks, generating a negative spatial association. The insignificant total effect reflects the coexistence of strong local promotion and offsetting interprovincial effects (Tables 7, 8).

4.8 Robustness checks for the spatial effects of manure-related endowment

Four robustness checks were conducted by replacing the core explanatory variable, changing the spatial weight matrix, replacing the dependent variable, and excluding observations from the COVID-19 period. The results consistently support the positive local association between manure-related resource endowment and upper-bound theoretical carbon-mitigation potential.

When manure-derived nutrient resource intensity was used as the core explanatory variable, its coefficient was 7.99 and remained significantly positive at the 1% level. This result indicates that the baseline finding does not depend on measuring manure-related endowment solely by the physical quantity of manure resources. Provinces with greater manure-derived nutrient supply per unit of cropland also exhibited greater theoretical potential for replacing mineral fertilizer nutrients.

Under the economic–geographic composite spatial weight matrix, the coefficient of the core explanatory variable remained positive and significant at the 1% level. Its spatially lagged term was also significantly positive, suggesting that the relationship between manure resource endowment and mitigation potential was not restricted to direct geographic adjacency. Economic similarity and broader interprovincial agricultural linkages may also contribute to spatial associations in manure-based fertilizer-substitution potential. Nevertheless, the spatial autoregressive coefficient became significantly negative under this specification, indicating that the direction and magnitude of spatial dependence are sensitive to the definition of spatial connectivity.

When the dependent variable was replaced by carbon emission intensity, the core explanatory variable retained a significantly positive coefficient of 0.09. Its spatially lagged term was also significantly positive, indicating that manure resource endowment was positively associated not only with the absolute scale of upper-bound mitigation potential but also with the alternative intensity-based measure. However, the spatial autoregressive coefficient was statistically insignificant under this specification, suggesting that the spatial dependence of the alternative dependent variable was weaker than that of the baseline measure.

After excluding observations from the COVID-19 period, the coefficient of manure resource intensity remained significantly positive at 0.08. The positive local association was therefore not driven by abnormal changes in livestock production, agricultural input markets, or interregional economic activity during the pandemic. The spatial autoregressive coefficient also remained significantly positive at 0.25, which was close to the baseline estimate and confirmed the persistence of spatial clustering after the abnormal period was removed.

Several control variables also showed broadly similar patterns under the variable-replacement, dependent-variable-replacement, and sample-exclusion specifications. Chemical fertilizer application intensity remained significantly negative, whereas agricultural GDP, agricultural mechanization, and cropping structure generally retained significantly positive local coefficients. The negative spatial associations of agricultural mechanization and urbanization also remained evident in several specifications. In contrast, the coefficients and spatial effects of some control variables changed under the economic–geographic composite matrix, demonstrating that the estimated spillover mechanisms are more sensitive to the assumed spatial linkage structure than the local effect of manure resource endowment (Table 9).

4.9 Regional heterogeneity in total effects across eastern, central, western, and northeastern China

To identify regional differences in the factors associated with upper-bound theoretical carbon-mitigation potential, the full sample was divided into eastern, central, western, and northeastern regions. The results reveal substantial heterogeneity in both local associations and spatial spillover patterns across the four regions.

Manure resource intensity remained significantly positive in all four regions, confirming that local manure resource endowment constitutes the fundamental material basis for manure-based mineral fertilizer substitution. However, the magnitude of the association differed considerably. The coefficient was largest in northeastern China at 1.77, followed by central China at 0.77, whereas the coefficients in eastern and western China were both 0.08. The stronger association in northeastern China may reflect the combination of relatively concentrated livestock production, large-scale grain cultivation, and substantial cropland nutrient demand. In central China, the close coexistence of crop and livestock production may facilitate the matching of manure nutrient supply with cropland demand. Although the coefficients were smaller in eastern and western China, their statistical significance indicates that manure resource intensity remained an important local correlate of theoretical mitigation potential.

The spatially lagged manure resource intensity was significantly positive in the eastern, central, and northeastern regions but insignificant in the western region. This indicates that manure resource endowment was associated with positive cross-provincial spatial linkages in eastern, central, and northeastern China. Provinces in these regions may share livestock-production systems, agricultural service networks, organic fertilizer markets, and policy environments with neighboring provinces. In contrast, the insignificant spatial lag in western China suggests that its manure resource advantage remains predominantly localized, potentially because long transport distances, dispersed agricultural production, complex terrain, and relatively weak interprovincial service linkages constrain the spatial transmission of manure-related benefits.

The spatial autoregressive coefficient was significantly positive in eastern and central China, with values of 0.30 and 0.24, respectively. This demonstrates significant positive spatial dependence in upper-bound mitigation potential within these two regions. The corresponding coefficients were insignificant in western and northeastern China, indicating that the overall spatial clustering mechanism was weaker in these regions after other explanatory variables and their spatial lags were controlled for. Thus, the spatial dependence observed at the national level appears to be driven primarily by eastern and central China.

Chemical fertilizer application intensity had a significantly negative local coefficient only in eastern China. This result suggests that strong dependence on mineral fertilizers may more clearly constrain manure-based substitution potential in the eastern region. Intensive agricultural production, established fertilizer supply chains, and relatively strong crop–livestock separation may reduce the relative role of manure-derived nutrients. Chemical fertilizer intensity was not statistically significant in the other three regions, indicating that its association with theoretical substitution potential depends on regional production systems and input-use structures. Its spatially lagged coefficients were insignificant in all four regions, providing no evidence of a stable cross-regional spillover effect.

Agricultural economic scale also exhibited marked regional variation. Agricultural GDP was significantly positively associated with mitigation potential in eastern China but significantly negatively associated with it in central China. The positive eastern coefficient may reflect better-developed agricultural service systems, stronger organizational capacity, and greater investment capacity for manure collection and field application. In central China, agricultural economic expansion may not necessarily improve the spatial matching of livestock manure and cropland nutrient demand and may instead reinforce fertilizer-intensive or specialized production systems. No significant local association was found in western or northeastern China. The spatial lag of agricultural GDP was significantly negative in eastern China but significantly positive in northeastern China, further demonstrating that the spatial implications of agricultural economic development vary across regional production structures.

Agricultural mechanization had a significantly positive local coefficient only in central China. This suggests that mechanization may be particularly important in reducing the labor and operational constraints associated with manure transportation and field application in the central region. Although the local coefficient was insignificant in eastern China, its spatially lagged term was significantly positive, indicating that mechanization in neighboring eastern provinces may facilitate cross-regional agricultural services, equipment sharing, and manure-return operations. Neither local nor spatial effects were significant in western and northeastern China, possibly because general mechanization does not adequately represent manure-specific collection, treatment, and application capacity in these regions.

Urbanization showed significantly negative local coefficients in eastern, central, and western China, whereas its local association was insignificant in northeastern China. Urbanization may reduce available agricultural land, accelerate rural labor outmigration, impose tighter constraints on livestock production, and intensify the spatial separation between livestock farming and cropland. These effects appear particularly evident in eastern, central, and western China. However, the spatially lagged urbanization rate was significantly positive in central, western, and northeastern China. This suggests that urbanization in neighboring provinces may generate positive spatial associations through infrastructure improvement, agricultural service development, livestock-production relocation, or increased demand for standardized waste treatment and organic fertilizers. The coexistence of negative local and positive neighboring associations indicates that urbanization operates through different within-province and cross-provincial pathways.

Cropping structure exhibited opposite regional patterns. In eastern China, its local coefficient was significantly positive, while its spatial lag was significantly negative. A locally favorable cropping structure may strengthen demand for organic nutrients and improve manure–cropland matching, whereas similar adjustments in neighboring provinces may intensify competition for manure resources and agricultural services. In western China, cropping structure had a significantly negative local coefficient but a significantly positive spatial lag. This result may indicate that local crop structures are not always compatible with manure nutrient supply, while cropping adjustments in neighboring provinces may create broader markets, service networks, or interregional nutrient-demand linkages. Neither local nor spatial effects were statistically significant in central and northeastern China.

Agricultural science and technology input intensity had a significantly positive local coefficient only in western China. Technological and financial support may be particularly important in overcoming the western region’s constraints related to transport distance, dispersed production, treatment capacity, and field application. In regions with more developed agricultural infrastructure and service systems, the marginal association of general science and technology expenditure may be weaker. Environmental regulation intensity, government agricultural support intensity, and the number of farms did not exhibit statistically significant local or spatial effects in any region. These macro-level indicators may not adequately capture manure-specific regulations, subsidies, treatment facilities, or the scale and operational characteristics of livestock farms (Table 10).

5 Discussion

5.1 Interpreting ER as upper-bound theoretical mitigation potential

A central contribution of this study is to quantify the upper-bound theoretical carbon-mitigation potential generated by substituting mineral fertilizer nutrients with manure-derived N, P₂O₅, and K₂O. This distinction is important because the ER estimated here does not represent actual realized emission reduction, but rather the maximum technical mitigation space under complete manure nutrient recovery and equivalent mineral fertilizer substitution. In practice, realized substitution depends on manure collection efficiency, storage and composting losses, transport distance, nutrient availability, crop nutrient demand, application timing, and farmers’ willingness to use organic fertilizer. Recent national survey evidence shows that manure nutrients already contribute to cropland nutrient inputs in China, but the shares of manure-derived N, P, and K remain far below complete substitution, confirming that a substantial gap exists between theoretical nutrient supply and actual utilization (Zhang Q. et al., 2023).

This interpretation is also consistent with recent research on manure management and carbon neutrality, which emphasizes that manure recycling can contribute to emission mitigation only when the whole chain of collection, treatment, transport, and agronomic application is effectively organized (Zhang L.,E.,R. et al., 2023). Therefore, the estimated ER should be understood as a benchmark for identifying where manure-based fertilizer substitution potential is concentrated, rather than as evidence that the corresponding amount of CO₂-eq emissions has already been avoided. This clarification responds directly to the concern that complete nutrient substitution may overstate the practically achievable mitigation effect.

The sensitivity analysis further reinforces this point. When alternative substitution rates are applied, the absolute magnitude of ER changes proportionally, but the decline–rebound–slight decline trajectory remains unchanged. This suggests that the temporal pattern observed in this study is mainly driven by changes in manure-derived nutrient resources, whereas the absolute scale of mitigation depends strongly on the effective substitution rate. Similar caution is needed for fertilizer emission factors, because fertilizer-related emissions may change over time with energy structure, production technology, and supply-chain efficiency. Studies on fertilizer-related climate impacts have shown that emissions arise from both fertilizer production and use, and that reducing fertilizer-related emissions requires improvements across the fertilizer value chain rather than only at the field-application stage.

5.2 Temporal changes and the post-2019 rebound

The results show that manure resources, manure-derived nutrients, and upper-bound ER all followed a stagewise trajectory of decline before 2019, rebound during 2019–2023, and slight decline in 2024. This pattern suggests that manure-based substitution potential is closely linked to changes in livestock production structure, especially the contraction and recovery of livestock capacity. The pre-2019 decline may be associated with livestock-sector restructuring, stricter environmental regulation, spatial relocation of animal production, and changes in farm-scale organization. Policy reviews indicate that China has strengthened manure pollution control and recycling regulations since the late 2010s, including action plans and technical guidance for manure utilization, which reshaped livestock waste management and increased compliance pressure on livestock farms (Wei et al., 2021).

The rebound after 2019 is one of the most policy-relevant findings of this study. It likely reflects the recovery of livestock and poultry production capacity, the restoration of manure nutrient supply, and the continued implementation of manure resource-utilization policies. Recent policy reviews show that China’s manure resource-utilization policies have evolved from pollution control toward integrated resource recovery and green livestock development, which helps explain why manure recycling capacity and policy attention increased in the later period (Lin et al., 2025). However, the rebound should still be interpreted as a recovery in potential substitution space rather than proof of realized mitigation. Without adequate collection, processing, and field-return systems, an increase in manure resources may also increase environmental pressure rather than automatically reduce fertilizer-related emissions.

This finding is consistent with field-scale and meta-analytic evidence showing that manure substitution can produce agronomic and environmental benefits under appropriate substitution ratios, but excessive or poorly managed substitution may create new nutrient-loss risks. For example, long-term field evidence from a winter wheat–summer maize system shows that manure replacing synthetic fertilizer can improve yield sustainability and reduce the crop carbon footprint under suitable management (Niu et al., 2024). A meta-analysis of China’s major grain crops also found that replacing synthetic nitrogen fertilizer with manure increased crop yield and nitrogen-use efficiency, but the effect varied by nitrogen application rate, substitution rate, climate condition, and soil property (Ren et al., 2023). These studies support the interpretation that manure substitution has meaningful mitigation potential, but that the realized outcome depends on management conditions rather than resource quantity alone.

5.3 Spatial inequality and the regional concentration of manure-based substitution potential

The spatial results show that upper-bound ER is unevenly distributed across provinces and is increasingly concentrated in the western region and several major agricultural provinces. This pattern is consistent with the regional mismatch between livestock production, cropland nutrient demand, and manure treatment capacity. Previous research has emphasized that China requires region-specific manure treatment and recycling technologies because crops, feed, livestock, and manure resources are unevenly distributed across space (Bai et al., 2021). The westward concentration of upper-bound ER observed in this study therefore reflects not only resource endowment, but also the broader spatial reorganization of livestock and crop production.

The Dagum Gini results suggest that overall interprovincial inequality remains relatively stable, while the contribution of net between-region differences becomes more prominent. This means that the spatial pattern is not simply a matter of individual high- or low-value provinces, but reflects an increasingly clear regional stratification. Regions with abundant manure resources and cropland absorption capacity have greater theoretical substitution space, whereas regions with limited livestock resources, higher urbanization pressure, or stronger crop–livestock separation may face a narrower local substitution base. Evidence from national manure-utilization data also indicates strong spatial heterogeneity in manure nutrient input and utilization across China’s croplands, supporting the need to interpret manure recycling through a regional lens (Zhang Q. et al., 2023).

At the same time, high upper-bound ER does not necessarily mean high realized environmental benefit. If livestock-dense regions lack adequate cropland absorption capacity or if manure is not transported and applied efficiently, nutrient surpluses may increase ammonia volatilization, nitrate leaching, or nitrous oxide emissions. Recent multi-objective optimization research suggests that manure solutions can generate co-benefits for crop yield, nitrogen emissions, and climate impact only when substitution rates and spatial allocation are optimized (Yin et al., 2025). Therefore, the spatial concentration of upper-bound ER should be treated as a planning signal: it identifies where manure-derived nutrient resources are large, but further policy design is needed to convert this potential into actual mitigation.

5.4 Spatial dependence and spillover interpretation

The positive and statistically significant spatial autoregressive coefficient confirms that the upper-bound theoretical carbon-mitigation potential is spatially interdependent across Chinese provinces. This result is broadly consistent with recent evidence that livestock-related carbon emissions and agricultural low-carbon performance exhibit spatial clustering and cross-regional associations in China (Zhou et al., 2025). Such dependence does not necessarily indicate direct transmission of mitigation outcomes between provinces. Rather, neighboring provinces may share similar livestock-production structures, crop systems, fertilizer-use practices, agricultural service markets, infrastructure conditions, and policy environments, causing their upper-bound mitigation potentials to evolve in related spatial patterns.

Manure resource intensity retained a significantly positive direct effect and total effect after the other agricultural, technological, and institutional conditions were controlled for. This result confirms that manure resource endowment is the primary material basis of the theoretical space available for mineral fertilizer substitution. National survey evidence also shows pronounced spatial differences in manure nutrient application, with the manure shares of total N, P, and K inputs generally lower in eastern China and higher in western China (Zhang Q. et al., 2023). Provinces with more manure resources per unit of cropland can theoretically supply more manure-derived nutrients and therefore avoid a greater amount of fertilizer-related emissions under the complete-substitution assumption.

However, the indirect effect of manure resource intensity was not statistically significant in the national model, indicating that the resource-endowment association remained predominantly localized. This finding is consistent with the physical and economic characteristics of manure: it is bulky, heterogeneous in nutrient concentration, costly to store and transport, and subject to nutrient losses during collection, treatment, and movement. Farm-survey research across China has identified transport costs, uncertain manure quality, insufficient application equipment, and weak manure exchange markets as major barriers to replacing mineral fertilizer with manure (Zhang et al., 2021). City-level evidence further demonstrates that the spatial decoupling of livestock manure supply from crop nutrient demand contributes to fertilizer dependence, while long-distance manure transport may be economically infeasible in many cases (Kang et al., 2023). Consequently, a province’s manure abundance does not automatically translate into a strong benefit for neighboring provinces.

The robustness analysis provides additional support for this interpretation. The positive local association remained significant when manure-derived nutrient intensity replaced manure resource intensity, when the dependent variable was changed to an intensity-based measure, and when the COVID-19 period was excluded. By contrast, the spatially lagged core variable and spatial autoregressive coefficient were more sensitive to the spatial weight matrix and dependent-variable definition. This pattern suggests that the local resource-endowment relationship is the most stable empirical finding, whereas cross-provincial spillovers depend more strongly on how interregional connectivity is defined.

The results therefore distinguish between the spatial concentration of potential and the physical transferability of manure nutrients. Spatial clustering may arise because adjacent provinces have similar agricultural structures, but actual interprovincial nutrient redistribution requires processing, standardization, transport, quality certification, and recipient-cropland capacity. Recent national optimization studies similarly conclude that spatially reallocating manure or livestock production can reduce regional nutrient surpluses and greenhouse gas emissions, but only when crop demand, transport costs, environmental thresholds, and livestock-product supply are considered jointly (Kang et al., 2024). Thus, the significant spatial dependence found in this study should be interpreted as evidence for coordinated nutrient planning rather than as evidence that untreated manure can circulate freely across provincial boundaries.

5.5 Agricultural production structure and enabling conditions for manure substitution

Chemical fertilizer application intensity exhibited significantly negative direct and total effects. This association indicates that provinces characterized by stronger mineral fertilizer dependence tend to have less favorable conditions for manure-derived nutrient substitution after other factors are held constant. Long-term reliance on mineral fertilizer may become embedded in input-distribution channels, agronomic recommendations, machinery configurations, farmers’ production routines, and expectations regarding nutrient availability. Survey evidence suggests that manure users and non-users differ substantially in their perceptions of manure quality, costs, technical risks, and labor requirements, and that fertilizer reduction requires coordinated technical services and economic incentives rather than manure availability alone (Zhang et al., 2021). The negative effect should nevertheless be understood as a conditional association, because the estimated upper-bound potential is not a measure of the actual quantity of fertilizer replaced in agricultural production.

Agricultural GDP showed positive direct and total effects, suggesting that a larger agricultural economy provides a stronger demand and organizational foundation for manure nutrient utilization. Provinces with larger agricultural sectors generally have greater crop nutrient demand, larger production and service markets, and stronger capacity to organize manure processing, transport, testing, and field application. However, agricultural economic expansion does not necessarily guarantee environmentally efficient manure use. Recent studies of crop–livestock decoupling show that agricultural specialization can simultaneously expand production capacity and weaken the circular linkage between livestock farms and crop farms (Han et al., 2023; Kang et al., 2023). The positive agricultural GDP effect therefore represents a larger potential absorption and organization capacity rather than proof that agricultural growth automatically produces actual carbon mitigation.

Agricultural mechanization produced a positive direct effect but a negative indirect effect. The positive local effect is plausible because machinery can reduce labor requirements and improve the efficiency of transporting and applying organic nutrient products. Nevertheless, general mechanization indicators mainly describe crop-production equipment and may not capture manure-specific machinery for collection, solid–liquid separation, composting, transport, injection, spreading, and precision nutrient management. A national choice experiment found that crop farmers preferred policy packages combining field guidance, financial support, and machinery services, demonstrating that manure application requires specialized service provision rather than generic mechanization alone (Zhang et al., 2022). The negative indirect effect may therefore reflect regional competition for agricultural machinery services and production factors or the expansion of highly specialized cropping systems in neighboring provinces.

Urbanization had no significant direct effect but generated a significantly negative indirect and total effect. Urbanization may reduce agricultural land, reallocate rural labor, tighten spatial restrictions on livestock production, and intensify the separation between livestock farms and cropland. These effects are likely to extend beyond administrative boundaries because livestock relocation, agricultural labor migration, land conversion, and input-service restructuring often occur at a regional scale. The continuing decoupling of crop and livestock production has been identified as a major structural cause of fertilizer overuse and manure nutrient imbalance in China (Kang et al., 2023). At the same time, urbanization may improve roads, environmental services, financing, and waste-treatment capacity, which can offset part of the local negative influence and help explain the insignificant direct effect.

Cropping structure showed a positive direct effect and a negative indirect effect. The positive direct effect indicates that local crop composition and nutrient demand determine the capacity of cropland to absorb manure-derived nutrients. A national meta-analysis found that the yield and nitrogen-use-efficiency responses to manure substitution differed across wheat, maize, and rice systems and were further conditioned by substitution rate, climate, soil organic matter, and available phosphorus (Ren et al., 2023). Field evidence also shows that moderate manure substitution can maintain crop productivity, increase soil organic carbon storage, and reduce the carbon footprint, whereas excessively high substitution ratios may reduce yield sustainability (Niu et al., 2024). Therefore, crop-side demand and agronomic suitability are as important as livestock-side nutrient supply.

The negative spatial effect of cropping structure may indicate competition or mismatch in regional nutrient allocation. When neighboring provinces expand crops with high organic nutrient demand, they may compete for commercial organic fertilizers, processing facilities, transport services, and specialized application machinery. Conversely, a manure-rich province may still face nutrient surpluses when its local crop composition, seasonal demand, or cropland distribution does not match manure nutrient availability. Numerical experiments at catchment and regional scales likewise show that the environmental and agronomic outcomes of crop–livestock integration depend on identifying an appropriate manure-substitution ratio and achieving spatial nutrient balance (Chen et al., 2023). The coexistence of positive local and negative neighboring effects consequently explains why the total effect of cropping structure was statistically insignificant.

Environmental regulation intensity, government agricultural support intensity, agricultural science and technology input intensity, and the number of farms did not show statistically significant national direct, indirect, or total effects. These findings should not be interpreted as evidence that regulation, public support, technology, and agricultural organization are unimportant. Micro-level evidence shows that policy cognition can increase livestock farmers’ investment in manure recycling facilities and that cooperative membership and land leasing can mediate this relationship (Wang et al., 2021). Crop farmers also express a clear preference for combined subsidy, technical-guidance, and machinery-service packages rather than isolated policy instruments (Zhang et al., 2022). Agricultural socialized services have similarly been found to promote farmers’ organic fertilizer use by reducing information, technology, and market barriers (Ren, 2023).

The lack of significance in the provincial model is more likely to reflect differences between the conceptual mechanisms and the macro-level proxies used in the analysis. General environmental expenditure does not distinguish enforcement against livestock pollution from investment in manure recycling; broad agricultural fiscal expenditure does not measure manure-specific subsidies; and general science and technology expenditure does not directly represent the adoption of manure-treatment or field-application technologies. Similarly, the number of farms captures the prevalence of specialized operating entities but not their average scale, livestock density, cropland holdings, or actual capacity to recycle manure. These variables may exert stronger effects on the realization rate of the theoretical potential than on the upper-bound accounting value itself.

5.6 Regional heterogeneity and policy implications

The regional analysis confirms that manure resource intensity has a positive local association with upper-bound theoretical carbon-mitigation potential in eastern, central, western, and northeastern China. However, the magnitude of the coefficient, the significance of the spatially lagged variable, and the roles of the control variables differ considerably. These differences reinforce the view that manure recycling cannot be promoted through a uniform national implementation model. China’s manure utilization already displays a pronounced regional pattern, with lower manure nutrient shares in eastern cropland and higher shares in western China (Zhang Q. et al., 2023). Regional policies therefore need to distinguish manure resource availability from crop absorption capacity, processing conditions, transport feasibility, and institutional readiness.

In eastern China, manure resource intensity had significant local and spatially lagged effects, and the spatial autoregressive coefficient was also significantly positive. The eastern region is therefore characterized by both strong local resource dependence and relatively close interprovincial linkages. Nevertheless, chemical fertilizer intensity and urbanization had negative local effects, while agricultural GDP had a positive local effect. This combination suggests that eastern China possesses substantial economic and organizational capacity but faces severe land constraints, fertilizer lock-in, intensive urbanization, and crop–livestock separation. Spatial optimization research for eastern China shows that food and feed self-sufficiency, livestock distribution, manure nitrogen balance, and transport flows must be considered jointly to avoid regional manure surpluses (Li et al., 2024). Eastern policy should therefore prioritize the production of concentrated and standardized organic fertilizers, cross-provincial service networks, and cooperation between specialized livestock and crop farms rather than the long-distance movement of untreated manure.

The eastern result for cropping structure—a positive local coefficient and negative spatial lag—further suggests that provinces may improve their own manure absorption capacity while competing with neighboring provinces for organic nutrient products and application services. Policy coordination should consequently extend beyond individual provincial fertilizer-reduction targets. Regional nutrient-balance platforms could identify manure-surplus and nutrient-deficit areas, while transport support should be limited to treated products whose nutrient value justifies movement costs. Such arrangements would help prevent one province’s livestock relocation or fertilizer-reduction policy from transferring nutrient pressure to another province.

In central China, manure resource intensity also showed significant local and neighboring associations, and the positive spatial autoregressive coefficient indicates a relatively strong regional interaction system. Central China contains major grain-producing and livestock-producing areas, giving it considerable potential to reconnect crop nutrient demand with manure supply. However, agricultural GDP had a negative local coefficient, whereas mechanization had a positive local effect. This pattern implies that agricultural expansion alone may reinforce specialization and fertilizer-intensive production unless it is accompanied by manure-specific machinery, nutrient budgeting, and cross-farm coordination. Evidence at the city level shows that stronger crop–livestock integration can reduce chemical fertilizer use, but the degree and effect of integration vary spatially (Kang et al., 2023). Central China should therefore develop county-level nutrient budgets, third-party manure collection and application services, and contracts linking specialized livestock farms with grain producers.

Urbanization had a negative local but positive spatially lagged effect in central China. Locally, urban expansion and rural labor loss may weaken crop–livestock circulation; regionally, urbanization in neighboring provinces may improve transport, service markets, investment capacity, and demand for standardized environmental management. This coexistence suggests that urbanization does not have a uniform effect but changes the spatial organization of agricultural production. Policies should make use of regional infrastructure gains while preventing the displacement of livestock pollution toward less-regulated counties.

Western China exhibited a significant positive local manure resource effect but no significant spatially lagged manure effect or overall spatial autoregressive dependence. This finding indicates that western manure-resource advantages are largely localized. National manure-use data show relatively high manure nutrient shares in western cropland, but recent spatial assessments also identify western livestock-intensive areas as potential nutrient-surplus regions where cropland availability and manure management efficiency do not always match resource supply (Zhang Q. et al., 2023; Xu et al., 2026). The region’s long transport distances, dispersed settlements, heterogeneous terrain, and uneven service capacity limit spontaneous cross-provincial recycling.

Agricultural science and technology input intensity was significantly positive only in western China, indicating that technological support may have a comparatively high marginal value where basic collection, treatment, transport, and application conditions remain constrained. However, technological investment should target operational bottlenecks rather than only the construction of facilities. Appropriate priorities include decentralized composting, solid–liquid separation, nutrient concentration, pelletization, quality testing, cold- and arid-region treatment processes, and machinery suited to dispersed cropland. Recent national optimization evidence indicates that interprovincial redistribution can narrow nutrient imbalances, but its feasibility depends on controlling transport distance and associated greenhouse gas emissions (Xu et al., 2026). Western policy should therefore combine local processing with selective transport of high-value organic fertilizer products, rather than subsidizing unrestricted movement of raw manure.

The opposite local and neighboring effects of cropping structure and urbanization in western China also demonstrate the importance of regional complementarity. Local cropping systems may not provide sufficient or seasonally matched nutrient demand, whereas crop demand and infrastructure development in neighboring provinces may create outlets for processed manure products. A regional market can support this complementarity only when nutrient contents, pathogen control, heavy-metal risks, product quality, and transport costs are clearly regulated.

Northeastern China exhibited the largest local coefficient for manure resource intensity and a significantly positive spatially lagged manure effect, although the spatial autoregressive coefficient itself was not significant. The combination of concentrated grain production, relatively large operational units, and substantial cropland nutrient demand provides a strong material basis for manure substitution. Moderate manure substitution has been shown to improve yield sustainability, increase soil organic carbon storage, and reduce cropping-system carbon footprints under wheat–maize rotations (Niu et al., 2024). The northeastern strategy should therefore connect manure nutrient recycling with black-soil conservation, soil organic matter restoration, and low-carbon grain production.

Nevertheless, the northeastern estimates must be interpreted cautiously because the regional subsample contains only three provinces. The strong coefficient of manure resource intensity may partly reflect the concentrated livestock and cropland structure of these provinces, while the lack of a significant spatial autoregressive coefficient suggests that the overall regional feedback mechanism is less stable than the individual spatially lagged effects. Policy recommendations should therefore emphasize coordinated demonstration projects and province-specific nutrient budgets rather than inferring a uniform northeastern spillover mechanism.

Overall, the heterogeneity results support a differentiated policy framework. Eastern China requires high-value manure processing, reduction of crop–livestock separation, and cross-provincial market coordination; central China requires county-level nutrient matching and specialized field-return services; western China requires targeted technological investment and economically feasible processing and transport systems; and northeastern China should integrate manure substitution with black-soil protection and large-scale grain production. National optimization studies likewise show that crop–livestock spatial reallocation and structural adjustment can reduce manure nitrogen surplus and livestock-related greenhouse gas emissions when regional crop demand and environmental carrying capacity are simultaneously considered (Kang et al., 2024). The national policy objective should therefore shift from maximizing manure return rates toward achieving spatially balanced, agronomically appropriate, and environmentally controlled nutrient circulation.

5.7 Endogeneity, uncertainty, and limitations

The SDM estimates should be interpreted as conditional spatial associations rather than causal effects. Province and year fixed effects control for persistent provincial heterogeneity and nationwide annual shocks, but they cannot fully eliminate reverse causality, simultaneity, measurement error, or omitted time-varying factors. For example, livestock-sector adjustment, fertilizer policy, manure infrastructure, environmental enforcement, and agricultural service development may change simultaneously and jointly affect both manure resource intensity and the estimated upper-bound mitigation potential.

A more fundamental limitation arises from the construction of the variables. The dependent variable is calculated from manure-derived N, P₂O₅, and K₂O resources, while manure resource intensity is also derived from livestock manure availability. Their positive relationship is therefore partly embedded in the accounting structure. The SDM is useful for examining whether this resource-based potential remains spatially associated with agricultural production, institutional, and technological conditions, but it cannot establish that increasing manure production causes actual emission reductions. The core coefficient should primarily be interpreted as evidence that resource endowment structures the geographic distribution of theoretical substitution potential.

Although the expanded model controls for environmental regulation, agricultural support, science and technology input, and farm organization, these variables remain broad provincial proxies. They do not directly measure manure-specific treatment-facility capacity, facility utilization, actual recycling subsidies, organic fertilizer quality, field-return machinery, manure collection networks, farmers’ adoption, or the share of manure nutrients actually replacing mineral fertilizer. Micro-level studies demonstrate that manure recycling decisions depend on policy cognition, cooperative participation, land access, field guidance, machinery services, perceived risks, and financial incentives (Wang et al., 2021; Zhang et al., 2022). The absence of significant effects for the macro proxies may therefore reflect measurement aggregation rather than the absence of underlying policy or technological mechanisms.

The upper-bound ER also assumes complete nutrient recovery and equivalent substitution of mineral fertilizers. In reality, nutrient availability is reduced by losses during collection, storage, treatment, and field application, while the fertilizer value of manure depends on mineralization timing, nutrient ratios, soil properties, climate, crop type, and application method. Meta-analytic evidence shows that manure substitution effects vary substantially with substitution ratios, total nitrogen inputs, climate, soil organic matter, and crop systems (Ren et al., 2023). Field experiments likewise indicate that moderate substitution may maintain yields and reduce carbon footprints, whereas very high substitution rates may reduce yield or create nutrient imbalances (Hu et al., 2023; Niu et al., 2024). Thus, the accounting results indicate maximum technical space, not a forecast of realized mitigation.

The use of constant nutrient coefficients and fertilizer emission factors introduces additional uncertainty. Livestock feeding practices, manure composition, treatment technology, fertilizer-production energy sources, and supply-chain efficiency can vary across provinces and over time. The sensitivity analysis addresses uncertainty in the numerical magnitude of the estimates, but uniform proportional perturbations do not alter spatial rankings or inequality measures. Future uncertainty analysis should therefore use nutrient-specific and region-specific distributions rather than only common scaling factors.

Provincial data also conceal substantial within-province heterogeneity. Livestock farms and cropland may be spatially separated within the same province, and provincial nutrient balance does not guarantee that manure can be economically transported from surplus counties to nutrient-demanding counties. Recent national research shows that manure nutrient surpluses and deficits coexist across provinces and that optimized redistribution must satisfy ecological thresholds while minimizing transport distance and application-related emissions (Xu et al., 2026). County-, watershed-, and farm-level analyses are therefore needed to identify feasible manure collection radii, processing locations, recipient cropland, and seasonal application pathways.

Finally, the estimated spatial effects depend on the selected spatial weight matrix. The robustness results show that the local manure-resource association remains stable, whereas the spatially lagged core variable and the spatial autoregressive coefficient change under alternative connectivity definitions. Future research could compare contiguity, geographic-distance, economic-distance, transport-network, livestock-product-trade, and organic-fertilizer-flow matrices. Quasi-experimental policy designs, instrumental-variable spatial models, farm-level surveys, and facility-level datasets could further distinguish theoretical resource potential from feasible substitution and actual carbon mitigation.

6 Conclusion

This study estimated the upper-bound theoretical carbon-mitigation potential of substituting mineral fertilizer nutrients with livestock and poultry manure-derived nutrients in China. Based on provincial panel data for 31 provinces from 2005 to 2024, manure resources, manure-derived N, P₂O₅, and K₂O supplies, and the corresponding upper-bound ER were quantified. Dagum Gini decomposition, global Moran’s I, and the Spatial Durbin Model were then used to examine regional inequality, spatial clustering, and conditional spatial associations. It should be emphasized that the ER calculated in this study represents a theoretical upper bound under complete manure nutrient recovery and equivalent fertilizer substitution, rather than actual realized emission reduction.

The results show that national upper-bound ER followed a decline–rebound–slight decline trajectory. It decreased from 2.183 × 10⁷ t CO₂-eq in 2005 to 1.677 × 10⁷ t CO₂-eq in 2019, rebounded to 2.021 × 10⁷ t CO₂-eq in 2023, and slightly declined to 1.962 × 10⁷ t CO₂-eq in 2024. Spatially, the western region remained the largest contributor, accounting for 42.43% of national upper-bound ER in 2024. The Dagum Gini results indicate that overall interprovincial inequality remained relatively stable, while net between-region differences became more prominent, suggesting clearer regional stratification in manure-based substitution potential.

The spatial econometric results confirm significant positive spatial dependence in upper-bound theoretical mitigation potential. Manure resource intensity had significantly positive direct and total effects, whereas its indirect effect was statistically insignificant, indicating that manure resource endowment primarily operates through localized rather than stable cross-provincial channels. Chemical fertilizer application intensity showed significantly negative direct and total effects, while agricultural economic scale had significantly positive direct and total effects. Agricultural mechanization produced a positive direct effect but a negative indirect effect, suggesting that general mechanization supports local manure utilization but does not necessarily provide manure-specific collection, treatment, and field-application services across regions. Urbanization generated significant negative indirect and total effects, whereas cropping structure exhibited a positive direct effect and a negative indirect effect, resulting in an insignificant total effect.

Environmental regulation intensity, government agricultural support intensity, agricultural science and technology input intensity, and the number of farms did not exhibit statistically significant national direct, indirect, or total effects. This does not imply that these institutional and technological conditions are unimportant. Rather, the provincial indicators used in this study are broad proxies and may not capture manure-specific regulation, subsidies, treatment infrastructure, application technologies, or the organizational capacity of individual farms. These factors may influence the practical realization of the estimated potential more strongly than its upper-bound accounting value.

The robustness tests show that the positive local association between manure-related resource endowment and upper-bound theoretical mitigation potential remained significant after replacing the core explanatory variable, changing the spatial weight matrix, using an alternative dependent variable, and excluding the COVID-19 period. The spatially lagged core variable and the spatial autoregressive coefficient were more sensitive to model specification, indicating that the local resource-endowment effect is more robust than the estimated cross-provincial spillover mechanism.

Regional heterogeneity was pronounced. Manure resource intensity remained significantly positive in eastern, central, western, and northeastern China, but its magnitude and spatial transmission differed across regions. Significant positive spatial dependence was mainly observed in eastern and central China, while the spatially lagged manure resource variable was positive in eastern, central, and northeastern China but insignificant in western China. Agricultural GDP, mechanization, urbanization, cropping structure, and agricultural science and technology input also showed different signs and levels of significance across regions. In particular, agricultural science and technology input was significantly positive only in western China, suggesting that technological support may have a higher marginal value where manure collection, processing, transport, and field-application conditions remain relatively constrained.

These findings suggest that manure management should shift from a narrow waste-treatment approach toward an integrated framework of nutrient circulation, fertilizer substitution, and carbon mitigation. Eastern China should reduce crop–livestock separation and develop standardized, high-value manure-derived fertilizer products and coordinated regional service markets. Central China should strengthen county-level nutrient budgeting and specialized manure-return services connecting livestock and crop producers. Western China should prioritize appropriate processing technologies, local treatment capacity, and economically feasible transport systems. Northeastern China should integrate manure nutrient recycling with black-soil conservation, soil organic matter improvement, and low-carbon grain production.

Future research should incorporate actual manure collection and recycling rates, treatment and storage losses, manure-specific policy support, field-application efficiency, and finer-scale crop–livestock matching data. County-, farm-, and watershed-level analyses, together with stronger causal identification strategies, would help distinguish upper-bound theoretical potential, technically and economically feasible substitution, and actual realized carbon mitigation more clearly.

References

• 1

  BaiZ.WangX.WuX.WangW.LiuL.ZhangX.et al (2021). China requires region-specific manure treatment and recycling technologies. Circ. Agric. Syst.1, 1–8. 10.48130/CAS-2021-0001

  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  ChenM.CuiY.JiangS.ForsellN. (2022). Toward carbon neutrality before 2060: trajectory and technical mitigation potential of non-CO2 greenhouse gas emissions from Chinese agriculture. J. Clean. Prod.368, 133186. 10.1016/j.jclepro.2022.133186

  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  ChenL.WangY.YangN.ZhuK.YanX.BaiZ.et al (2023). Improving crop-livestock integration in China using numerical experiments at catchment and regional scales. Agric. Ecosyst. and Environ.341, 108192. 10.1016/j.agee.2022.108192

  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  DuM.AntunesJ.WankeP.ChenZ. (2022). Ecological efficiency assessment under the construction of low-carbon city: a perspective of green technology innovation. J. Environ. Plan. Manag.65 (9), 1727–1752. 10.1080/09640568.2021.1945552

  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  GaoY.LiZ.HongS.YuL.LiS.WeiJ.et al (2025). Recent stabilization of agricultural non-CO2 greenhouse gas emissions in China. Natl. Sci. Rev.12 (4), nwaf040. 10.1093/nsr/nwaf040

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  HanZ.HanC.ShiZ.LiJ.LuoE. (2023). Rebuilding the crop-livestock integration system in China ——based on the perspective of circular economy. J. Clean. Prod.393, 136347. 10.1016/j.jclepro.2023.136347

  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  HuY.LiD.WuY.LiuS.LiL.ChenW.et al (2023). Mitigating greenhouse gas emissions by replacing inorganic fertilizer with organic fertilizer in wheat–maize rotation systems in China. J. Environ. Manag.344, 118494. 10.1016/j.jenvman.2023.118494

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  HuangJ.LuH.DuM. (2025). Regional differences in agricultural carbon emissions in China: measurement, decomposition, and influencing factors. Land14 (4), 682. 10.3390/land14040682

  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  KangX.DuM.LiuQ.DuH.ZouW.ZhaoL.et al (2023). City-level decoupling between livestock and crop production and its effect on fertilizer usage: evidence from China. Sci. Total Environ.905, 167115. 10.1016/j.scitotenv.2023.167115

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  KangJ.DingW.ChangN.YiX.ZhangJ.LiH. (2024). Optimized crop-livestock coupling to reduce agricultural manure-N surplus and greenhouse gas emissions in China. J. Clean. Prod.467, 142835. 10.1016/j.jclepro.2024.142835

  • CrossRef
  • Google Scholar
  • View reference in article

• 11

  LiY.ShiY.DengX.SunZ.AccatinoF. (2024). Increasing food and feed self-sufficiency and avoiding manure N surplus in eastern regions of China through a spatial crop-livestock optimisation model. Agric. Syst.217, 103911. 10.1016/j.agsy.2024.103911

  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  LiM.ZhangX.WangY.HouJ.TianX.ZhangW.et al (2026). Microbial mechanisms driving carbon emissions and humification in inoculant-enhanced swine manure composting: insights into inoculation timing. J. Environ. Chem. Eng.14 (2), 121298. 10.1016/j.Jece.2026.121298

  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  LinH.JiaoH.LinH.XuX. (2025). The evolution of policies for the resource utilization of livestock manure in China. Agriculture-Basel15 (2), 153. 10.3390/agriculture15020153

  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  LiuZ.ChenC. (2025). Can the integration of agriculture and tourism improve the rural human settlement Environment? Based on panel data of 30 provinces over 2008–2022 in China. Front. Sustain. Food Syst.9, 9–2025. 10.3389/fsufs.2025.1673999

  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  LuY.GaoY.NieJ.LiaoY.ZhuQ. (2021). Substituting chemical P fertilizer with organic manure: effects on double-rice yield, phosphorus use efficiency and balance in subtropical China. Sci. Rep.11 (1), 8629. 10.1038/s41598-021-87851-2

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  NingL.XuX.ZhangY.ZhaoS.QiuS.DingW.et al (2022). Effects of chicken manure substitution for mineral nitrogen fertilizer on crop yield and soil fertility in a reduced nitrogen input regime of north-central China. Front. Plant Sci.Volume, 13–2022. 10.3389/fpls.2022.1050179

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  NiuJ.SaeedQ.WangW.ZhangR.LiuL.LvF.et al (2024). Manure replacing synthetic fertilizer improves crop yield sustainability and reduces carbon footprint under winter wheat–summer maize cropping system. J. Environ. Manag.358, 120936. 10.1016/j.jenvman.2024.120936

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 18

  QinQ.YanH.LiuJ.ChenX.YeB. (2022). China’s agricultural GHG emission efficiency: regional disparity and spatial dynamic evolution. Environ. Geochem. Health44 (9), 2863–2879. 10.1007/s10653-020-00744-7

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  RenZ. (2023). Effects of risk perception and agricultural socialized services on farmers' organic fertilizer application behavior: evidence from Shandong Province, China. Front. Public Health11, 11–2023. 10.3389/fpubh.2023.1056678

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  RenK.SunY.ZouH.LiD.LuC.DuanY.et al (2023). Effect of replacing synthetic nitrogen fertilizer with animal manure on grain yield and nitrogen use efficiency in China: a meta‐analysis. Front. Plant Sci.14, 14–2023. 10.3389/fpls.2023.1153235

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  SongS.ZhaoS.ZhangY.MaY. (2023). Carbon emissions from agricultural inputs in China over the past three decades. Agriculture13 (5), 919. 10.3390/agriculture13050919

  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  WangY.WangJ.WangX.LiQ. (2021). Does policy cognition affect livestock farmers' investment in manure recycling facilities? Evidence from China. Sci. Total Environ.795, 148836. 10.1016/j.scitotenv.2021.148836

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  WeiS.ZhuZ.ZhaoJ.ChadwickD. R.DongH. (2021). Policies and regulations for promoting manure management for sustainable livestock production in China: a review. Eng. Agric.8 (1), 45–57. 10.15302/j-fase-2020369

  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  XuS.-Q.ChenW.-J.XieL.-Q.ChaiS.-Y.JiaK.-X.WeiY.-Q. (2022). Organic waste resources and nutrient utilization potential in China. J. Plant Nutr. Fertilizers28 (8), 1341–1352. 10.11674/zwyf.2021663

  • CrossRef
  • Google Scholar
  • View reference in article

• 25

  XuK.ZhangQ.-Q.CaiY.-Y.LuoX.-J.ZhaiY.-Q.LiuW.-R.et al (2026). Diagnosing spatial decoupling of manure and crop nutrients in China: drivers and multi-objective optimization for sustainable redistribution. Npj Sustain. Agric.4 (1), 11. 10.1038/s44264-025-00120-x

  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  YangM.ChuJ.LiZ.LiuX.YuF.SunF. (2023). An examination of regional variations in pesticide usage and grain yield in China before and after the double reduction policy’s adoption. Pol. J. Environ. Stud.32 (2), 1887–1898. 10.15244/pjoes/157550

  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  YinY.GuoS.XuQ.LiuJ.WangH.ZhuangY.et al (2025). Co-benefits for cropland yield, nitrogen emissions, and climate impact through multi-objective optimization agricultural manure solutions. Nat. Commun.16 (1), 6415. 10.1038/s41467-025-61885-w

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 28

  ZhangT.HouY.MengT.MaY.TanM.ZhangF.et al (2021). Replacing synthetic fertilizer by manure requires adjusted technology and incentives: a farm survey across China. Resour. Conservation Recycl.168, 105301. 10.1016/j.resconrec.2020.105301

  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  ZhangT.MengT.HouY.HuangX.OenemaO. (2022). Which policy is preferred by crop farmers when replacing synthetic fertilizers by manure? A choice experiment in China. Resour. Conservation Recycl.180, 106176. 10.1016/j.resconrec.2022.106176

  • CrossRef
  • Google Scholar
  • View reference in article

• 30

  ZhangL.ReaihanE.AliM. M.LinH.ZhangS.JinS.et al (2023a). Livestock and poultry manure management from the perspective of carbon neutrality in China. Front. Agric. Sci. and Eng.10 (3), 341–362. 10.15302/J-FASE-2023509

  • CrossRef
  • Google Scholar
  • View reference in article

• 31

  ZhangQ.ChuY.YinY.YingH.ZhangF.CuiZ. (2023b). Comprehensive assessment of the utilization of manure in China’s croplands based on national farmer survey data. Sci. Data10 (1), 223. 10.1038/s41597-023-02154-7

  • CrossRef
  • Google Scholar
  • View reference in article

• 32

  ZhangL.JiangG.XiaoR.HouK.LiuX.LiuX.et al (2024). An appropriate amount of straw replaced chemical fertilizers returning reduced net greenhouse gas emissions and improved net ecological economic benefits. J. Clean. Prod.434, 140236. 10.1016/j.jclepro.2023.140236

  • CrossRef
  • Google Scholar
  • View reference in article

• 33

  ZhangZ.-M.WuL.GaoT.-Y.LiuT.-H.WangC.XuM.-G.et al (2025). Effects of chemical fertilizer and organic material application on soil organic carbon fractions in black soils of northeast China: a meta-analysis. Huan Jing Ke Xue= Huanjing Kexue46 (8), 4947–4960. 10.13227/j.hjkx.202407132

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 34

  ZhouJ.ChenC.WuL.WangH. (2025). Spatiotemporal dynamics and spatial spillover effects of carbon emissions in china’s livestock economic system. Sustainability17 (10), 4611. 10.3390/su17104611

  • CrossRef
  • Google Scholar
  • View reference in article