The spatiotemporal distribution and multi-scenarios for land use carbon metabolism: simulation and carbon reduction from the production-living-ecological space perspective Authors: Tian Chao, Wang WenHui, Zhao JinTao Published: 2026 ABSTRACT AbstractA comprehensive understanding of land use carbon metabolism characteristics from the production-living-ecological space (PLES) perspective is crucial for formulating carbon reduction strategies. As the core economic zone of northern China, the Beijing-Tianjin-Hebei (BTH) region faces severe carbon emission pressures due to rapid urbanization and intensive land use transformation. However, focusing solely on carbon metabolism calculation without considering future changes and optimization effects may prevent achieving carbon emission reduction targets. This study assessed carbon emissions and sequestration based on different land use types in PLES, constructed a multi-objective carbon reduction scenario utilizing the Dinamica-EGO model, nondominated sorting genetic algorithm II, and entropy weight-TOPSIS model, and simulated 2035 carbon reduction characteristics by coupling PLES changes. Taking the BTH region as a case study, a methodological framework and corresponding models were established. The results show that from 2000 to 2020, the total carbon emissions in the BTH region increased significantly, presenting a spatial pattern of high emissions in the southeast and low emissions in the northwest. In contrast, the overall carbon sequestration capacity showed a decreasing trend, with stronger capacity in the northwest and weaker capacity in the southeast. The multi-variable 2035 carbon emission reduction prediction model achieved an accuracy of 82.24%. The 2035 carbon reduction plan developed based on this framework outperformed the original land use plan: economic benefits, emission reduction efficiency, spatial compactness, and accessibility are projected to increase by 15.8%, 7.9%, 2.5%, and 8.3%, respectively, while carbon emissions are expected to decrease by 19.04%. The proposed PLES-based framework for carbon metabolism measurement and emission reduction simulation exhibits good applicability in regional spatial emission reduction. These findings contribute to exploring regional carbon dynamics and provide references for governments to formulate carbon reduction policies.Keywords: Spatiotemporal distribution, Carbon metabolism, Multi-scenarios simulation, Carbon reduction, Production-living-ecological space, Beijing-tianjin-hebei region Introduction Carbon emissions from fossil fuel combustion and land use are the main sources of greenhouse gases in the atmosphere [56]. These emissions, alongside massive energy consumption, further threaten regional sustainable development, human well-being, and ecosystem health, and may even lead to social unrest and imbalances in spatial structure [46]. Energy management and land use optimization are thus crucial for regional emission reduction, particularly in developing countries [47]. As a core economic hub in northern China, the industrial transformation and energy structure optimization in the Beijing-Tianjin-Hebei (BTH) region have had a significant impact on the regional ecological security pattern and the promotion of the "Dual Carbon" goals. Against the backdrop of rapid urbanization, prominent issues such as ecological space compression, concentrated energy consumption, and imbalanced functions of PLES have emerged, leading to the continuous accumulation of carbon metabolism imbalance risks, which has become a key bottleneck restricting regional sustainable development [50]. In 2023, total energy consumption in the BTH region reached 5.72 billion tons of standard coal, approximately three times that of 2000,meanwhile, over 25% of ecological land has been converted into construction land from 2000 to 2023 [50]. Coordinating the relationship between regional land, energy, and the low-carbon economy enables scholars to track these intertwined land-energy-carbon processes on spatial and temporal scales, which is critical for understanding regional carbon cycling and carbon metabolism changes [45]. Regional carbon metabolism refers to the carbon flows between different components within ecosystems and regional natural-economic systems, including the processes of carbon emission and sequestration [54]. Carbon metabolism in the production-living-ecological space (PLES) is analogous to the metabolic processes of organisms in ecosystems, outlining carbon element flows, interactions in PLES, and characteristics of structural differentiation in regional systems [51, 59]. Relevant studies have demonstrated that transitions between different PLES types alter carbon source-sink structures [33, 48]. The primary objective of regional carbon metabolism research is to explore how different elements within the PLES system affect regional carbon cycle processes. Previous studies have focused on emission processes, such as carbon emissions from industrial activities, regional construction, and human activities [10, 40], while also investigating carbon sinks in ecological spaces [61]. Current research examines horizontal carbon flows (e.g., between industries or land use types) and vertical carbon flows (e.g., between different trophic levels or within the soil-plant-atmosphere continuum) [36], and also analyzes carbon metabolism within regional socio-economic systems [39], and identifies key characteristics of implicit carbon flows in regions [17]. The transformation of land use in different PLES has a profound impact on the material cycle of ecosystems and reshapes ecosystem functions, which has become a key factor driving climate system change (Horton, 2012). To explore the dynamic evolution of carbon emissions and sequestration, carbon assessment methods have been continuously explored in recent years. Wackernagel et al. (1996) [43] first proposed the carbon footprint theory in 1992 to quantify greenhouse gases generated during energy use. By 2006, the Intergovernmental Panel on Climate Change (IPCC) released the IPCC National Greenhouse Gas Inventory Guidelines, proposing the emission factor method as an estimation method for fossil energy emissions [25]. With the advancement of environmental monitoring technology, some scholars have adopted direct measurement methods to estimate carbon emissions by directly measuring relevant physical or chemical quantities. Houghton et al. [23] estimated carbon fluxes between terrestrial ecosystems and the atmosphere from 1850 to 1990, considering the impact of different land use types on the atmospheric environment. Wei et al. (2018) used eddy covariance systems to monitor the carbon exchange process between forest ecosystems and the atmosphere in real time, revealing the spatial pattern of carbon storage in typical subtropical forest ecosystems in Zhejiang Province, China. Unlike carbon emissions, carbon sequestration is widely believed to be closely related to land cover types in ecosystems [5, 62]. The carbon sequestration potential of different land use types is usually characterized by carbon storage density [13]. Scholars typically use geographic information and remote sensing technology to estimate carbon sequestration processes [14, 44]. Thus, research on carbon metabolism is generally divided into two directions: (1) how land cover change affects carbon sequestration, and (2) changes in carbon emissions caused by energy consumption within the socio-economic system. Although there are more studies on regional carbon metabolism in the context of land use, there is still a lack of integrated research on regional integration in developing countries, especially in terms of sufficient spatial information. Moreover, research on the spatiotemporal characteristics of regional carbon metabolism in PLES is still limited, making carbon metabolism estimation based on PLES in different regions a challenging task [12, 24, 55]. Scholars have paid extensive attention to future carbon simulation, aiming to reflect future regional carbon dynamics. Since the application of remote sensing (RS) technology in carbon metabolism research, simulation studies on spatial patterns have focused on the spatial distribution of land use [6, 41]. Current research has simulated the spatiotemporal dynamics of carbon emissions and carbon sequestration among different land use types by exploring specific land use types (such as forestland, grassland, or construction land), constructing different scenario goals, and analyzing the impact of future land use changes on carbon emissions/carbon sequestration through different spatial optimization models [1, 9, 20, 38]. For instance, scholars have employed models such as the Minimum Cumulative Resistance (MCR) [18] model, Cellular Automata (CA [27, 57]), Mixed Cell Cellular Automata (MCCA) [32], Patch-generating Land Use Simulation (PLUS) model [30], System Dynamics (SD) model [60], and the Future Land Use Simulation (FLUS) model [31] to conduct spatial layout optimization research. These models demonstrate certain advantages in quantifying carbon emission intensity for individual land use types and characterizing the spatio-temporal variations of carbon budgets across different time scales. However, they primarily focus on the physical spatial transformation of land use patterns, which limits their ability to reveal the intrinsic driving mechanisms and feedback effects of the carbon cycle. Additionally, constrained by data acquisition difficulties, accuracy, and availability, simulation results exhibit inherent uncertainty. The Dinamica Environment for Geoprocessing Objects (Dinamica-EGO) model effectively addresses these limitations through its built-in carbon budget accounting module and dynamic multi-scenario simulation capabilities, enabling a direct linkage between PLES structural adjustments and carbon flow mechanisms. Compared to traditional models, it supports calibration, operation, and spatio-temporal simulation, addressing the limitation of traditional models in simulating full land use processes and enabling multi-scenario simulation [21, 28]. Nevertheless, it cannot directly optimize land use spatial configurations to achieve specific carbon emission reduction targets. Therefore, there is a need to integrate it with other models to adapt to diverse scenarios. With the advancement of computer technology, research on PLES optimization based on programming techniques has gradually evolved, with investigators attempting to integrate the aforementioned methods to explore regional spatial optimization simulations. For example, Luan et al. [34] combined the Non-dominated Sorting Genetic Algorithm II (NSGA-II) with the PLUS model, demonstrating that the algorithm’s fast non-dominated sorting and elitist strategy can efficiently address issues of regional spatial structural imbalance. Aosaf et al. [2] combined differential evolution with multi-objective genetic algorithms and found that this hybrid approach demonstrates high efficiency in solving spatial allocation optimization problems for land use in complex regions. However, existing research predominantly focuses on the allocation and optimization of land use quantity, with limited attention paid to spatial structural optimization. This study aims to integrate multiple models and algorithms to conduct PLES optimization simulations, thereby providing a valuable supplement to the existing body of research. Consequently, we established a structural framework for the regional carbon metabolism system that explores two types of hierarchical features from a PLES perspective: (a) exploring the flow characteristics of carbon emissions and carbon sequestration in PLES, and (b) establishing future carbon reduction scenarios by combining regional low-carbon goals with PLES simulation. The former can identify the characteristics of regional carbon cycles and spatial structures. The latter can identify specific areas for spatial optimization and carbon reduction and analyze the energy conservation and emission reduction potential of the region. Different from previous studies, this paper primarily focuses on the following aspects: (1) While earlier carbon metabolism research usually explored land-use carbon variations and regional carbon balance based on single land categories or independent simulation models, this study further incorporates the systematic PLES classification perspective into carbon metabolism evaluation. This refined analytical framework effectively distinguishes the functional differentiation characteristics of carbon sources and sinks in territorial space and clarifies the intrinsic influence of spatial functional transformation on regional carbon metabolic changes. (2) Optimizing land use spatial allocation and improving regional carbon metabolic efficiency are widely regarded as critical objectives for low-carbon spatial governance. Building upon the historical spatiotemporal patterns of land-use carbon metabolism, this study further realizes multi-scenario carbon reduction simulations and spatial optimization through the coupled Dinamica-EGO, NSGA-II and entropy weight-TOPSIS model system. It systematically displays the evolutionary characteristics of regional carbon metabolism and the comprehensive benefits of optimized land use structure under different development scenarios, which holds practical implications for formulating territorial spatial planning and low-carbon regulation policies. The methodology not only quantifies structural changes of land-use carbon metabolism under different development frameworks but also establishes actionable optimization paradigms for regional spatial decarbonization strategies. (3) Due to differences in territorial functional attributes and land-use transformation patterns, significant disparities exist in carbon metabolic levels and carbon reduction potential across production, living, and ecological spaces. Based on multi-scenario simulation results and model coupling verification, this study investigates the spatial optimization effects under different development scenarios and compares the carbon reduction advantages and adaptability of diverse PLES allocation schemes. The findings provide targeted insights for formulating context-specific spatial carbon reduction strategies, thereby supporting the achievement of regional dual-carbon goals and sustainable territorial development. Study area and data sources Study area The BTH region (113°04’E-119°53’E, 36°01’N-42°37’N), the largest integrated region in northern China, holds an important position in China’s economic development, encompassing Beijing, Tianjin, and Hebei Province. There are significant differences in natural resources, ecological environments, and socio-economic development within the BTH region. By the end of 2021, the resident population and GDP of the BTH region were 132 million and 8.56 trillion RMB, respectively. The urbanization rate was 76.0%, with 85.1% in Beijing, 81.3% in Tianjin, and 61.5% in Hebei Province (Fig. 1) [50]. The location of the BTH region Economic development and rapid population growth in the BTH region have converted large amounts of ecological land into construction land (e.g., transportation land, public facilities land, industrial and mining land, urban-rural residential land) and led to significant energy consumption. Thus, the BTH region is facing a series of issues such as uneven energy distribution, tight housing supply, and low construction efficiency. Data sources and processing Land use and PLES data were interpreted from Landsat TM and ETM+ remote sensing images. Cloud-free Landsat images for 2000, 2005, 2010, 2015, and 2020 were acquired from the Geospatial Data Cloud (http://www.gscloud.cn/), which provides global coverage of free, continuously updated remote sensing datasets. Each image underwent atmospheric and geometric correction to eliminate distortions from sensor errors and terrain effects. Classification accuracy was assessed using the Kappa coefficient, with all images achieving an accuracy exceeding 83%, meeting the requirements for subsequent spatial analysis (Gorr and Kurland, 2020). A classification scheme was established, encompassing eight land categories: cropland, forestland, grassland, wetland, water bodies, urban land, rural residential land, and unused land. Energy data were used to construct the carbon metabolism accounting system. Municipal-level energy consumption data, including raw coal, oil, natural gas consumption, and industrial energy use were collected. These statistical data (in tabular format) were converted into 30m×30m raster data through inverse distance weighting interpolation, ensuring spatial consistency with land use data for subsequent model integration. To unify energy consumption data within a consistent statistical scope, missing data were supplemented with sector-specific surveys. For the Dinamica-EGO model, 10 driving factors were selected to characterize the spatial location of PLES: Terrain factors include digital elevation model (DEM) and slope (derived from DEM), with a resolution of 30m; socio-economic factors consisting of population density and GDP data (original 1km) resampled to 30m using bilinear interpolation to match the study’s spatial resolution; proximity factors involving distance variables (from city centers, main roads, and river systems) calculated using ArcGIS 10.8 based on relevant vector data; neighborhood effects quantified via the 3×3 Moore window method: referring to historical PLES transition rules, weights were assigned to each PLES type, and the index value was calculated by weighting the proportion of adjacent PLES types [21], with data derived from long-term land use datasets; and policy variables including ecological protection red lines and basic farmland protection areas converted from vector to 30m×30m raster data based on land use datasets to reflect policy constraints on spatial distribution. All driving factors were normalized to the range [0,1] to eliminate scale discrepancies and geospatial datasets were unified to the WGS84 projection Table 1. Table 1Data information and sourcesDataResolutionTime periodData sourceLand use and PLES data30m2000, 2005, 2010, 2015, 2020Geospatial Data Cloud (http://www.gscloud.cn/)Energy data-2000-2020Statistical Yearbook of China; Statistical Yearbooks of Beijing, Tianjin, and Hebei ProvincePopulation density1 km2000-2020Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC)GDP1 km2000-2020Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC)DEM30m2020Geospatial Data Cloud (http://www.gscloud.cn/)Slope30m2020Geospatial Data Cloud (http://www.gscloud.cn/)Distance variables30m2020National Administration of Surveying, Mapping and Geoinformation (NASG)Neighborhood effects30m2000, 2005, 2010, 2015, 2020Land use datasetsPolicy variables-2015Beijing-Tianjin-Hebei Coordinated Development Plan Data information and sources Materials and methods The study framework A regional carbon metabolism accounting model from the PLES perspective was constructed to quantify regional carbon emissions and sequestration, comprising three components: (1) carbon metabolism accounting; (2) spatial pattern assessment; and (3) future simulation of PLES-based carbon emission reduction. First, regional carbon emissions and sequestration were estimated using a carbon accounting model. Second, the spatial distribution characteristics of PLES carbon metabolism were analyzed with remote sensing support. Finally, based on PLES spatial evolution results, simulated changes in PLES carbon emission reduction were derived by integrating the nondominated sorting genetic algorithm II (NSGA-II) and entropy weight-TOPSIS. Fig. 2 illustrates the overall study framework. Framework coupling carbon accounting, spatial pattern and future optimization Classifying production-living-ecological space The carbon metabolism process involves multiple agents, subsystems, and pathways in both socio-economic and natural ecosystems [52]. Building on previously constructed models of regional carbon metabolism networks [51], this study defines a regional carbon metabolism network from the PLES perspective. In this network, regional carbon metabolism identifies interactions between natural and social systems within PLES; these systems serve as nodes to form two distinct types of spatial flow pathways [3]. Based on the multifunctionality of land use types and considering the actual conditions of the BTH region, PLES was categorized into three types [22]: (1) Production space (P): agricultural production space (APS, corresponding to cropland), industrial production space (IPS, corresponding to industrial and mining land), and transportation production space (TPS, corresponding to transportation land); (2) Living space (L): urban living space (ULS, corresponding to urban land) and rural living space (RLS, corresponding to rural residential land); (3) Ecological space (E): forest ecological space (FES, corresponding to forestland), grassland ecological space (GES, corresponding to grassland), water ecological space (WES, corresponding to water bodies and wetland), and other ecological space (OES, corresponding to unused land) (Fig. 3). Interactions within PLES, such as the transformation of ecological space into production space and the agglomeration of urban and rural living spaces around transportation corridors, directly govern the evolution of the carbon metabolism process by altering regional carbon source-sink relationships and energy consumption patterns. Specifically, APS generates carbon emissions through agricultural activities such as fertilization and irrigation, while adjacent GES can partially offset such emissions through vegetation carbon sequestration. IPS releases substantial industrial carbon emissions via fossil energy consumption, with neighboring FES forming an emission reduction buffer due to its high carbon sink capacity. ULS and RLS produce household carbon emissions from commuting, building energy consumption, and other activities, which can be mitigated by carbon absorption from surrounding WES and ecological green spaces. Additionally, carbon emissions from road network operations in TPS can be partially neutralized through the carbon sink function of ecological corridors along the routes. This dynamic flow, functional synergy, and trade-offs of carbon sources and sinks among production, living, and ecological spaces constitute the core characteristics of carbon metabolism evolution. Classification of PLES Carbon metabolism accounting Carbon metabolism in this study refers to the total process of carbon flows in PLES, including carbon emissions (from production and living spaces as carbon sources) and carbon sequestration (from ecological spaces as carbon sinks). To quantify carbon in specific PLES types, the relationship between carbon metabolism and PLES was established through PLES classification. Based on PLES functional attributes, ecological space (e.g., forests, grasslands, water bodies) was treated as carbon sinks for carbon sequestration calculations, while production space (agriculture, industry, transportation) and living space (urban and rural areas) were treated as carbon sources for carbon emission calculations. As outputs of regional carbon metabolism, carbon emissions originate from various human activities: (1) emissions from agricultural activities and industrial production; (2) emissions from road transportation; (3) emissions from human habitation and daily life; and (4) emissions from the tertiary industry. These emissions were linked to specific PLES types, including APS, IPS, TPS, ULS and RLS, and estimated using Eqs. (1)-(5). where CEaps, CEips, CEtps, CEurls denote carbon emissions from APS, IPS, TPS, and ULS and RLS, respectively; C, M, F, and G represent fertilizer applied amount, agricultural land planting area, agricultural film applied amount, and irrigation area, respectively; Kc, Km, Kf, and Kg denote the fertilizer emission coefficient (895.6 kg·t-1), agricultural planting emission coefficient (0.01647 tC·hm-2), agricultural film emission coefficient (5180kg·t−1), and irrigation emission coefficient (266.48kg·hm−2), respectively [8],CE, L, P and O represent energy consumption, low-level heat generation of fuel (TJ/Gg), carbon content per unit of energy calorific value (tC/TJ), and energy combustion oxidation rate(%), respectively; 44/12 denotes the molecular weight conversion ratio of CO2 to C; j represents energy type (see Appendix A for CO₂ emission parameters); Et and Ec represent fossil fuel consumption and electricity generation during transportation (kWh), respectively; Ef denotes the electricity carbon emission factor (1.246 kg CO2/kWh); Ult and GDP% represent the proportion of tertiary industry carbon emissions and tertiary industry GDP, respectively; Elt represents tertiary industry energy consumption (calculated using Eq. 2). Referring to relevant research [37, 42], PLES carbon sequestration was calculated using PLES area and empirical coefficients (Table 2). These coefficients are consistent with the temperate ecosystem characteristics of the BTH region, as verified by local forest and grassland carbon sequestration studies [49, 53], ensuring the rationality of regional carbon sink estimation. Table 2Carbon sequestration coefficients of PLESPLES typesSequestration coefficients (t/hm2·a)ReferencesAPS0.007[7, 16]FES0.6552[16]GES0.567[16, 49]WES0.025[7]OES0.0005[53] Carbon sequestration coefficients of PLES Spatial simulation modeling This section simulated PLES to predict future spatial changes and spatial emission reduction characteristics, comprising two parts: (1) future PLES spatial simulation (constructing a PLES evolution model integrating macro-prediction and micro-allocation); and (2) analysis of PLES carbon emission reduction potential. Spatial carbon emission distribution and emission reduction potential during PLES optimization were clarified via future PLES spatial simulation. For Part 1, macro-level PLES area was predicted using population size and construction land area, while micro-level impacts of PLES evolution factors were analyzed using a logistic regression model; spatial simulation was then performed with the Dinamica-EGO model [4]. For Part 2, carbon emission reduction scenarios under different targets were established based on the government’s 2035 carbon reduction goals. Finally, these targets were combined with PLES spatial simulation to simulate spatial carbon emission distribution (Cui and Li, 2020). Part one: simulation of PLES spatio-temporal change The Dinamica-EGO model is a geo-environmental modeling platform based on the Markov model, which uses meta-cellular automata to simulate the quantity and spatial location of different land types. It achieves higher accuracy in complex spatial simulations, particularly for multi-type land use change predictions [4]. This study employed the Dinamica-EGO model to simulate the spatiotemporal dynamics of PLES, with transition probabilities determined from historical PLES data. The specific process is as follows: A linear regression model was used to estimate PLES land demand (Eq. 6). First, new living space area (X) was derived from population, GDP, and construction land area prediction models. The areas of production space and ecological space converted to living space were set as x1 and x2 (where x1+ x2=X). The area of ecological space converted to production space (Y) and production space converted to ecological space (Z) were calculated using Eq. (7): Ten factors were selected to determine PLES spatial location through a transformation function, including (1) global variables (elevation, slope, population density, per capita GDP, distance from the city center, distance from main roads, distance from river systems, neighborhood effects); (2) policy variables (ecological protection red lines, basic farmland protection areas); and (3) random factors (Eqs. 8-11). After multiple training iterations of the Dinamica-EGO model, the conversion interval step size and total step window size were set to 2 and 1-35, respectively. The cell size was 30m × 30m, with a total of 892,389 cells. The conversion rule function is as follows: where is the probability of conversion to living space; , and RA are conversion probabilities of ecological spaces based on global, policy, and random factors, respectively; is the global factor value (from the logistic regression model); confij denotes a conditional function, where the suitable living space=1, restricted construction area=0.5, prohibited construction area=0; γ represents the random number, ranging from 0 to 1; σ represents a random variable, ranging from 1 to 10. The Dinamica-EGO model was run with 2010 PLES data as the reference to simulate 2020 PLES. Simulation accuracy was assessed using Kappa coefficients for quantity (Kappan) and location (Kappam). If the difference between these coefficients was less than a set threshold, PLES simulation was considered stable, and the simulation was terminated; otherwise, model parameters were reset for iterative optimization. Using 2020 data, the model simulated 2035 PLES spatio-temporal changes (Fig. 4). Flow chart of the PLES simulation process Part two: multi-objective spatial optimization (1) Objectives. To clarify the carbon emission reduction potential of PLES optimization, each type of PLES was further categorized into nine functional subtypes (Table 3). Aligned with the national "Dual Carbon" goals and local energy-saving and emission reduction targets in the BTH region, multi-objective spatial optimization of PLES was conducted by integrating quantitative and spatial dimensions, encompassing four core objectives under four scenarios: (1) Low-carbon quantitative objectives: Maximize PLES economic benefits (EB) and minimize PLES carbon emissions (CE), as formulated in Eq. (12). These objectives balance development efficiency and emission control by linking economic output and carbon intensity of each PLES subtype. (2) Low-carbon spatial objectives: Optimize PLES spatial compactness (COM) and accessibility (ACC), as shown in Eq. (13). Compactness reflects the aggregation degree of production/living spaces to reduce redundant land use, while accessibility emphasizes the connectivity between living spaces and key infrastructure (e.g., transportation networks, public services) to lower commuting-related emissions: Table 3Economic benefit coefficient and carbon emission density parameters of PLES typesVariablesPLES typesEk (104 yuan/hm2)Ck (t/hm2)x1APS61.31.69x2IPS1042.6610.71x3TPS101.68.72x4ULS42.723.77x5RLS30.24.23x6FES31.710.39x7GES122.560.55x8WES10.022x9OES10.0005 Economic benefit coefficient and carbon emission density parameters of PLES types where Ek and Ak represent the economic benefit coefficient and area of PLES types, respectively; Ck is the carbon emission density of PLES. The EB parameters were referenced from land use economic benefit studies in the North China Plain [31, 50] and adjusted to the BTH region’s industrial structure and land use efficiency. The CE parameters were calculated based on regionally calibrated emission factors [15, 25] combined with municipal energy consumption data. where m is the number of unit spaces; m and n are the number of grid rows and columns in the study area, respectively; dij is the relative distance between plots (i, j); T is the total PLES area; drl is the average distance from urban-rural living spaces to transportation land and city centers. TO comprehensively considers both quantity and spatial optimization objectives, the calculation formula is as follows: (2) Constraints. Constraints were set based on national spatial planning and land use planning in the BTH region. The total PLES area was 21.7 million ha, and minimum area requirements for each PLES type (Table 4), determined by land use layout regulations and development land standards. Meanwhile, to control total PLES conversion costs, maximum conversion limits were set (Eq. 15): Table 4Minimum area requirements of PLES typesVariablesPLES typeMinimal area requirement (million ha)x1APS7.80x2IPS1.98x3TPS0.75x4+x5ULS and RLS2.10 Minimum area requirements of PLES types where is the PLES conversion cost coefficient, when k ≠ k’, ckk’=1, otherwise it is 0. This study restricted the conversion of each PLES type to no more than 30% of its original area and prohibited transformation of ecological functional areas. Algorithm-based optimization process Non-dominated sorting genetic algorithm II The non-dominated sorting genetic algorithm II (NSGA-II) is an improved version of the non-dominated sorting genetic algorithm (NSGA). Incorporating an elitist strategy, it extends genetic algorithms for multi-objective optimization and is rooted in the concept of Pareto optimality [12]. Pareto optimality refers to the optimal paths and methods for multi-objective functions, it pursues global optimality rather than that of a single objective, acknowledges inherent trade-offs among objectives, and ultimately yields a set of optimal solutions. Currently, NSGA-II has been widely applied in fields such as land use allocation [12, 35]. Post-evaluation based on entropy weight-TOPSIS The NSGA-II algorithm generates numerous Pareto optimal solutions, making it difficult to determine a final result (Fig. 5) [9]. Thus, the entropy weight-TOPSIS model was used to identify the optimal PLES emission reduction solution. The calculation of entropy weight depends on the data distribution of Pareto optimal solutions, which ensures the objectivity of preference weight distribution in different scenarios. This model calculates distances between evaluation objects and ideal/worst solutions, ranks objects, and determines their superiority [26]. Weights were determined using the entropy weight method; Euclidean distances from positive and negative ideal solutions were calculated (Eqs. 16-17) to derive relative proximity. Optimal solutions were then ranked based on proximity to ideal values (Eq. 18): The modeling modules of the algorithm-based optimization process where and denote distances from scheme i to the positive and negative ideal solutions, respectively; and represent the maximum and minimum values of solution i for index j; is an element in the weighted standardized decision matrix. Results and discussions Analysis of carbon metabolism Total carbon emissions in 2020 increased by 42741.67× 104 t compared to 2000, showing a 3.27-fold increase during these 20 years. IPS and TPS were the primary sources, accounting for 38% and 31% of total emissions, respectively. Urban and rural living spaces followed, while APS contributed minimally (approximately 4%) (Fig. 6). Spatiotemporal patterns of PLES carbon metabolism (2000-2020) Carbon sequestration capacity in the BTH region was weak and insufficient to offset emissions. Total carbon sequestration decreased by 1.24%, from 439.98×104 t in 2000 to 429.92×104 t in 2020. FES dominated carbon sequestration, contributing 67.34% of total absorption, with sequestration increasing from 273.65×104 t in 2000 to 300.20×104 t in 2020. GES and APS accounted for 26.33% and 3.95% of total sequestration, respectively. WES and OES had negligible sequestration capacity (≤0.4%). Overall, carbon sequestration in the BTH region was primarily driven by FES, GES, and APS. Spatial carbon metabolism distribution To analyze the spatial distribution of carbon metabolism, the natural breaks method was used to classify carbon emission and sequestration densities into four levels (Figs. 7-8). Carbon emissions in the BTH region showed significant spatial differentiation, with a distribution pattern of high in the southeast and low in the northwest (Fig. 7). Emissions were most prominent in the economic core of each city, forming spatial agglomerations, which were mainly distributed in the central urban areas of Beijing, Tianjin, Shijiazhuang, and Baoding. From 2000 to 2015, high-emission areas expanded outward from urban centers, with emissions increasing significantly. By 2020, emission density grew slowly, and high-value areas contracted, concentrating in central and southeastern cities. Spatial distribution of PLES carbon emissions in the BTH region (2000-2020) Spatial distribution of PLES carbon sequestration in the BTH region (2000-2020) Carbon sequestration showed an opposite spatial pattern to emissions (Fig. 8), increasing gradually from southeast to northwest. High-sequestration areas were primarily in the Taihang Mountains, Yanshan Mountains, and Bashang Plateau, while low-sequestration areas were in the low-lying North China Plain, such as Tianjin, Cangzhou, and Hengshui, which was consistent with the spatial distribution of woodlands and grasslands in the region. Analysis of future carbon metabolism simulation Analysis of PLES optimization precision To verify the effectiveness and accuracy of the Dinamica-EGO model, PLES data from 2010 to 2020 in the BTH region were simulated using three scenarios: (1) without global, policy, or random variables; (2) with global and policy variables only; (3) with global, policy, and random variables. Using 2010 PLES data as the baseline, 2020 PLES was simulated, and accuracy was verified by comparing simulated and actual values using Kappa coefficients in ENVI. Results showed Scenario (1) achieved 74.32% accuracy, with a quantity Kappaₙ of 0.71, and a location Kappaₘ of 0.68, with most errors originating from living and production spaces (ecological space simulation was more accurate). Scenario (2) improved accuracy by 3.95%, with a quantity Kappaₙ of 0.75, and a location Kappaₘ of 0.73, showing notable gains in living space simulation. Scenario (3) achieved 82.24% accuracy (a 7.92% improvement over Scenario 1), with a quantity Kappaₙ of 0.73, and a location Kappaₘ of 0.79 (Table 5, Fig. 9). Compared to traditional models (Scenario 1), the proposed model, which integrates macro-prediction and micro-allocation, was more accurate. Hence, it was reasonable and reliable to employ this model to simulate the future pattern of the PLES within the study area. Table 5Comparison of actual and simulated values of PLESScenarioScenario (1)Scenario (2)Scenario (3)Accuracy (%)74.3278.2782.24Kappaₙ0.710.750.79Kappaₘ0.680.730.79 Comparison of actual and simulated values of PLES Comparison of actual and simulated PLES patterns in the BTH region Simulation of PLES carbon emission reduction based on entropy weight-TOPSIS method The entropy weight-TOPSIS method was used to evaluate PLES solutions and identify optimal carbon reduction strategies across five scenarios (Table 6). MATLAB was used for iterative calculations until convergence, with 100 iterations to avoid local optima. Table 6Five scenarios and optimal solutions under different target preferencesTarget preferencesPreference weights(%)Optimal solutionMaxEBMinCEMinACCMinCOMEB priority1000001CE priority0100002ACC priority00100036COM priority00010019Comprehensive (TO)422621118 Five scenarios and optimal solutions under different target preferences To ensure the rationality of parameter settings and improve the reproducibility of the algorithm, a comprehensive sensitivity analysis was conducted for the core parameters of NSGA-II. Referring to relevant studies [9, 63] and combining the characteristics of PLES carbon reduction optimization in the BTH region, orthogonal experiments were designed to test the impacts of four key parameters (crossover rate, mutation rate, population size, and maximum iterations) on algorithm performance. Three evaluation metrics were adopted: Pareto solution count, hypervolume (HV), and spacing [19, 58], which reflect the quantity, diversity, and uniformity of optimal solutions, respectively. After multiple rounds of debugging, the parameters of the NSGA-II algorithm were set with a population size of 200, crossover rate of 0.8, mutation rate of 0.05, and maximum iterations of 100. This parameter setting ensured optimal solution diversity while aligning with established research standards. This process generated 59 Pareto-optimal solutions, from which the top five with the highest comprehensive benefits were selected. Detailed results of the sensitivity analysis are presented in the Supplementary Material (Appendix B-C). Optimized 2035 PLES solutions were compared to baseline scenarios to assess emissions (Table 7). Solution 19 and Solution 1 showed the greatest increase in carbon emissions compared to 2020, by 1.32 times and 1.31 times, respectively, while Solution 8 only increased by 1.19 times compared to 2020. For MaxEB, all solutions outperformed the baseline with optimization rates exceeding 10%. For MinCE, all solutions except Solution 19 achieved 4.0-7.9% optimization, with Solutions 8, 2, and 36 performing best. For MinCOM, optimization rates were low (1.9-3.0%), indicating stable land patterns. For MinACC, Solution 19 (traffic accessibility objective) and Solution 8 (comprehensive objective) performed best, achieving improvement rates of 8.8% and 8.3%, respectively. Notably, Solution 8 integrated PLES quantitative and spatial optimization. Compared to the baseline, the MaxEB, MinCE, MinCOM, and MinACC goals increased by 15.8%, 7.9%, 2.5%, and 8.3%, respectively, while reducing carbon emissions by 19.04% compared to the 2035 baseline scenario. These results confirm Solution 8 as the optimal PLES-based carbon reduction strategy. Table 7Comparison of optimal solutions with the baseline scenarioSolutionMaxEBMinCEMinCOMMinACCCarbon emissions (104t)Compared with 2020Target value (104 yuan)Optimization rate (%)Target value (104t/hm2)Optimization rate (%)Target value(km)Optimization rate (%)Target value(km)Optimization rate (%)Solution 1133543.2812.62306.894.0325.253.0320.397.780085.031.31 timesSolution 2129822.9010.02235.467.5327.362.3319.298.175194.341.23 timesSolution 36129966.2110.12287.615.1326.202.7322.996.878886.551.29 timesSolution 19130251.6510.32329.923.1328.991.9317.298.880696.361.32 timesSolution 8138782.6615.82226.297.9326.862.5318.528.372748.991.19 timesBaseline116782.89-2403.89-332.11-345.21-89866.401.49 times Comparison of optimal solutions with the baseline scenario Spatial distributions of the five scenarios were similar to the baseline but differed in specific regions (Fig. 10). The baseline showed shrinking ecological space in northern Chengde and western Zhangjiakou, which have since been converted into production and living areas, accompanied by an expansion of construction land. The MaxEB scenario expanded production/living space in southeastern Beijing, western Hebei, and northeastern Hebei at the expense of ecological space. The MinCE scenario reduced production and living space, with modest ecological expansion. The MinCOM scenario increased land use intensity, reduced living space, and expanded ecological space. The MinACC scenario expanded transportation land, adjusted production and living space proportions, and reduced ecological space. Spatial distribution of different PLES scenarios The comprehensive optimization scenario (Solution 8) adjusted PLES scales by expanding ecological space in northern/western areas, reducing industrial and residential land, and moderately increasing construction land. It strengthened southeast, south, and northeast axis layouts to promote urban-rural industrial integration, reduce commuting emissions, and align with urban development plans. Analysis of the optimal PLES optimization scheme Based on the non-dominated sorting genetic algorithm and the entropy weight-TOPSIS model, the proposed optimization scheme can effectively support the carbon emission reduction and planning for PLES. Table 8 shows the integrated optimization framework as the optimal solution: by 2035, PLES is projected to achieve 727.49 ×104 tons of carbon emissions and 138.78 ×108 yuan in economic benefits. In comparison to the 2035 baseline (unoptimized scenario), the optimized scheme strikes a balance between MaxEB and MinCE. Specifically, the reduction in production space under the optimized scheme (compared to 2020) stands at 5.95%, a smaller decrease than the 7.14% (compared to 2020) observed in the baseline scenario, reflecting the BTH region’s industrial structure adjustment of phasing out energy-intensive industries. For living space, the optimized scheme shows a 7.40% increase, significantly lower than the 16.32% growth in the baseline. In terms of ecological space, the optimized scheme achieves a 3.21% increase, outperforming the 1.27% growth in the baseline. Additionally, carbon emissions under the optimized scheme rise by 18.99%, a much smaller increment compared to the 46.91% increase in the baseline, achieving the 2035 carbon intensity target for the BTH region. Table 8Comparison of PLES optimization schemesPLES typeBaseline scenario (unoptimized)Comprehensive optimization scenario20202035Change range (%)2035Change range(%)Production space (104 ha)842.58782.38-7.14792.41-5.95Living space (104 ha)247.91288.3616.32266.257.40Ecological space (104 ha)1079.511093.231.271114.233.21Carbon emissions(104t)61133.6189866.4046.9172748.9918.99 Comparison of PLES optimization schemes Discussion Carbon metabolism analysis in the BTH region reveals crucial insights. The threefold increase in total carbon emissions from 2000 to 2020 underscores the urgency of implementing targeted carbon reduction strategies. Regional carbon sequestration capacity remains weak, failing to offset emissions and declining by 1.24% over the study period, highlighting the fragility of the ecological balance. This is consistent with Wang et al. [55], who emphasized that rapid urbanization in the BTH region has led to ecological space compression and weakened carbon sink capacity. However, our study further quantifies the contribution of different PLES types to carbon sequestration, with forest FES contributing 67.34%, which is higher than the 58% reported by Hu et al. [24] for northern China, reflecting the unique ecological structure of the Taihang-Yanshan Mountains in the BTH region. Spatially, carbon emissions and sequestration exhibit opposite distribution patterns shaped by PLES functional differentiation and key driving factors. The "high emissions in the southeast and low emissions in the northwest" pattern is jointly driven by economic concentration, population agglomeration, and land use conversion: the southeast, as the core economic zone, has a per capita GDP 2.3 times that of the northwest and a population density nearly four times higher, concentrating high-emission IPS and TPS while having a 25% ecological land conversion rate. In contrast, the "high sequestration in the northwest and low sequestration in the southeast" pattern is determined by the distribution of ecological spaces: the northwest is dominated by FES and GES with high sequestration coefficients and less human disturbance, while the southeast is dominated by cropland and construction land, resulting in weak sequestration capacity [50]. Simulations and optimizations via the Dinamica-EGO model coupled with the entropy weight-TOPSIS method yield actionable insights. Notably, the simulation model demonstrates high accuracy, particularly in Scenario 3 with an 82.24% accuracy rate, validating its reliability for predicting future PLES patterns. Among all designed scenarios, the comprehensive optimization scheme (Solution 8) performs best. This outperforms the 12.3% emission reduction rate achieved by Cui et al. [12] through land use optimization alone, highlighting the effectiveness of integrating PLES functional positioning and multi-objective optimization. Further comparison of this optimized scheme with the baseline PLES structure reveals clear directions for future PLES utilization and carbon reduction pathways. By 2035, production space is projected to decrease by 50.17×104 ha, which is a smaller decrease than the 7.14% in the baseline scenario. This reflects the industrial restructuring and "farmland-to-forest" ecological projects in the BTH region, specifically, reducing energy-intensive industrial land while retaining high-tech industrial clusters, which explains why economic benefits can increase by 15.8% despite the reduction in production space. For living space, the optimized scheme shows a 7.40% increase, significantly lower than the 16.32% growth in the baseline, aligning with the "reduction and optimization" planning for urban-rural construction land. This controlled growth helps reduce commuting emissions, which is consistent with the low-carbon urban planning concept proposed by Song et al. [48]. In terms of ecological space, the optimized scheme achieves a 3.21% increase, outperforming the 1.27% growth in the baseline and meeting the target of approximately 40% forest coverage specified in the BTH Ecological Plan. This expansion focuses on restoring carbon-rich ecosystems (FES and GES) in key sink areas, leading to a 4.8% increase in carbon sequestration compared to the baseline. Additionally, carbon emissions under the optimized scheme rise by 18.99%, a much smaller increment compared to the 46.91% increase in the baseline, achieving the 2035 carbon intensity target for the BTH region. This synergistic balance of economic growth, emission deceleration, and carbon sink enhancement. To evaluate computational efficiency, we compared the NSGA-II with the NSGA using four key indicators (Table 9). While the spacing values of the two algorithms are comparable, NSGA-II demonstrates distinct advantages over the traditional NSGA algorithm: it generates 11 more Pareto solutions and achieves a 27.8% higher HV, consistent with the superiority of NSGA-II in handling multi-objective spatial optimization reported by Zuo et al. [63]. Notably, our study integrates NSGA-II with the Dinamica-EGO model, addressing the inherent limitation of traditional land use optimization that focuses solely on quantitative allocation while lacking spatial simulation [34]. This integration enables the synergistic optimization of both the quantity and spatial structure of the PLES. Compared to other multi-objective optimization frameworks [2, 18], the proposed integrated framework not only achieves an optimized accuracy of 82.24% but also incorporates additional modules for carbon metabolism accounting and spatial pattern analysis, making the optimization results more aligned with low-carbon development requirements. Table 9Comparison of operation efficiency between NSGA-II and traditional NSGAAlgorithmPareto solution countHVSpacingEB optimization rate (%)CE optimization rate (%)COM optimization rate (%)ACC optimization rate (%)NSGA4815.120.005731.2630.223.266.28NSGA-II5919.330.005351.3353.194.348.85 Comparison of operation efficiency between NSGA-II and traditional NSGA Despite these findings, several limitations of this study merit attention for future research. First, the spatially fixed coefficients of different natural components are inherently uncertain, which impairs the accuracy of simulation outputs and input parameterization. Future research should focus on the spatiotemporal dynamic evolution of carbon storage across diverse regions, and enhance the dynamic tracking and mechanistic analysis of regional carbon metabolic processes, thereby improving the model’s adaptability and interpretability for characterizing carbon metabolism evolution in complex regions. Second, the inconsistent statistical standards at the urban level have become a key source of bias in data collection, directly leading to heterogeneity and incompleteness of regional datasets during the construction of the PLES carbon metabolism model. To address this issue, researchers can integrate multi-source big data, including point of interest (POI) data, high-resolution remote sensing images and cloud-native socioeconomic datasets. Cross-validation and data supplementation can then be adopted to effectively mitigate the systematic bias caused by inconsistent statistical standards, while refining the spatiotemporal granularity of PLES carbon metabolism analysis to significantly improve the accuracy and reliability of research findings. Additionally, although this study constructs a carbon metabolism and simulation framework at the PLES scale, the core analytical unit is still confined to different land use types within PLES. This perspective may overlook the disparities in carbon metabolic efficiency and the differentiation of emission reduction potential induced by regional heterogeneity (e.g., geographical location, climatic conditions, and land use intensity). It also fails to establish an effective link between refined spatial optimization schemes and the division of emission reduction responsibilities at the administrative level, thus limiting the practical applicability of the framework for multi-scale carbon management practices. Future research can further expand the analytical dimensions to explore carbon metabolic patterns and regulatory pathways at finer scales (e.g., production units and plot units), so as to enhance the practicality and targeting of the framework for carbon management decision-making across multiple scenarios. Conclusions and suggestions Conclusions The main contributions of this study are as follows: (1) It deeply couples carbon metabolism with the functional positioning of PLES, establishing a PLES classification-based carbon metabolism accounting system that clarifies the quantitative and spatial correlation between PLES structure and carbon source-sink dynamics. (2) It introduces spatial simulation models for carbon scenarios and constructs an integrated multi-model framework, realizing multi-scenario simulation and optimization of future PLES spatial evolution under carbon emission reduction goals. The established framework provides a reliable alternative for spatial carbon function assessment in regional integration areas and offers systematic support for the simulation prediction and targeted policy-making of land-based carbon emission reduction. The proposed method has demonstrated good applicability in the case study of the BTH region. The results indicate that the BTH region is a typical area with imbalanced carbon metabolism, characterized by weak and declining carbon sequestration capacity. Total carbon emissions have increased by 3.27 times over 20 years, with industrial and transportation production spaces as the core carbon sources. The spatial pattern of carbon emissions presents an agglomeration feature of "high in the southeast and low in the northwest", driven by industrial agglomeration and high population density, while carbon sequestration shows an opposite distribution of "high in the northwest and low in the southeast", mainly relying on the forest ecological space of the Taihang-Yanshan Mountains. Spatial carbon simulation results show that under the optimized scenario in 2035, the PLES structure of the BTH region will achieve a transformation of moderate reduction in production space and steady expansion of ecological space, with a carbon emission reduction potential of 19.04%, forming a spatial governance pattern of "carbon sequestration in the north and emission control in the south". Suggestions Based on the findings of this study, the following policy recommendations are proposed. First, establish a cross-regional carbon collaborative governance system in the BTH region and strengthen inter-provincial linkage mechanisms. Led by the BTH Coordinated Development Leading Group, formulate unified regional carbon emission reduction implementation rules, and clarify the division of responsibilities, supervision, and assessment standards for BTH in industries, transportation, and ecological protection. Establish a "BTH Carbon Sink Enhancement Special Fund" to prioritize supporting ecological restoration in high-carbon-sink areas (e.g., the Taihang-Yanshan Mountains), industrial low-carbon technological transformation, and public transportation network optimization, with preferential allocation of financial and technological resources to industrial-intensive high-emission cities such as Tangshan and Baoding [53]. In addition, promote the in-depth integration of carbon emission reduction targets with regional territorial spatial planning and industrial development planning, and incorporate the "PLES carbon metabolism multi-objective comprehensive optimization framework" constructed in this study into local governments’ performance evaluation systems, thereby driving the transformation of regional development models towards low carbon. Second, to break regional resource allocation barriers and path dependence, construct a carbon emission reduction cooperation corridor in the BTH region and optimize the carbon source-sink pattern of production, living, and ecological spaces. Given that IPS and TPS are core carbon sources, strictly prohibit the blind expansion of high-energy-consumption and high-emission projects. Promote circular economy technologies and new energy equipment to accelerate the green transformation of traditional industries (e.g., iron and steel, chemical engineering). Meanwhile, densify the intercity rail transit network, implement public transportation priority policies, encourage new energy vehicle popularization, and reduce the carbon emission intensity of commuting and cargo transportation. To address the potential impact of environmental pressure on regional carbon balance, all regions should abandon the extensive development model and transform ecological constraints into innovative momentum. Build a "BTH Carbon Metabolism Dynamic Monitoring Data Platform" integrating multi-source remote sensing, ground observation, and socio-economic data to strengthen real-time disclosure and collaborative assessment of carbon emissions and sink capacity, providing data support for precise emission reduction. Third, transform the government’s role from direct intervention to rule-making and collaborative services, releasing space for market entities and social forces to participate in carbon reduction. In response to the optimized reduction of production space, incentivize enterprises to improve energy and resource efficiency through tax preferences and financial discounts, promote cross-regional industrial transfer and value chain integration, and break administrative barriers to the flow of production factors. For the reasonable growth of living space, focus on the "reduction and optimization" of urban-rural construction, improve green building standards and supporting infrastructure, and enhance the convenience of low-carbon living for residents. Centering on ecological space expansion, prioritize land reclamation and ecological restoration in resource-exploited degraded areas (e.g., Mentougou District of Beijing, Tangshan City, Qian’an City of Hebei Province). Expand carbon sink capacity through projects such as "converting farmland to forests and grasslands" and "mine greening" to consolidate the regional ecological security barrier [49]. Establish a cross-regional carbon reduction benefit-sharing mechanism, encourage the realization of ecological product value, and enable ecological protection areas to obtain reasonable economic returns from enhanced carbon sinks, forming a virtuous cycle of mutual promotion between emission reduction impetus and ecological benefits. Appendix A See Table 10. Table 10Accounting parameters of carbon dioxide emissionsEnergy typeAverage low-level heat generationa(kJ·kg-1)Carbon content per unit calorific value b(GJ·T-1)Carbon oxidation rateb%Effective carbon emission factorb(tCO2·tce-1)raw coal2090826.370.942.67coke2843529.420.932.94crude oil4181620.080.982.11gasoline4307018.90.981.99kerosene4307019.60.982.06diesel oil4265220.20.982.12fuel oil4181621.10.982.22liquid gas5017917.20.981.81natural gas3223815.30.990.13biomass energy1672620.50.981.311 kgce = 1 kg standard coal (low calorific value = 29307 kJ); Abbreviations: tCO₂·tce⁻1 = tons of CO₂ per ton of standard coal; kJ·kg⁻1 = kilojoules per kilogram; GJ·t⁻1 = gigajoules per ton. This table provides energy-specific emission factors for converting energy consumption to carbon emissions, which are further standardized to CE (t/hm2) in Table 3 based on PLES area. Fuel with a low calorific value equal to 29307 kilojoules (KJ) is called 1kg standard coal (1 kgce); A is derived from the "General Principles for Comprehensive Energy Consumption Calculation" (GB/T 2589-2008); Source: "Guidelines for Compilation of Provincial Greenhouse Gas Inventories" (NDRC Climate [2011] No. 1041) Accounting parameters of carbon dioxide emissions 1 kgce = 1 kg standard coal (low calorific value = 29307 kJ); Abbreviations: tCO₂·tce⁻1 = tons of CO₂ per ton of standard coal; kJ·kg⁻1 = kilojoules per kilogram; GJ·t⁻1 = gigajoules per ton. This table provides energy-specific emission factors for converting energy consumption to carbon emissions, which are further standardized to CE (t/hm2) in Table 3 based on PLES area. Fuel with a low calorific value equal to 29307 kilojoules (KJ) is called 1kg standard coal (1 kgce); A is derived from the "General Principles for Comprehensive Energy Consumption Calculation" (GB/T 2589-2008); Source: "Guidelines for Compilation of Provincial Greenhouse Gas Inventories" (NDRC Climate [2011] No. 1041) Appendix B See Table 11. Table 11Parameter levels for sensitivity analysis Parameters Level 1 Level 2 Level 3 Crossover rate0.70.80.9Mutation rate0.030.050.07Population size150200250Maximum iterations80100120 Parameter levels for sensitivity analysis Appendix C See Table 12. Table 12Sensitivity analysis results of NSGA-II parametersParameter numberCrossover rateMutation ratePopulation sizeMaximum iterationsPareto solution countHVSpacing10.70.03150803815.270.007220.80.052001005919.330.005330.90.072501205519.560.0057 Sensitivity analysis results of NSGA-II parameters Author contributions Chao Tian: Conceptualization, Methodology, Software, Writing-Original Draft; WenHui Wang: Methodology, Formal Analysis, Writing Original; JinTao Zhao: Writing-Review and Editing. All authors reviewed the manuscript. Funding This research was supported by National Natural Science Foundation of China (No.42475177); Education Department Foundation of Hebei Province (No. QN2025034); the University-Level Foundation of Scientific Research (No. XBQ202519); Natural Science Foundation of Hebei Province (No. C2023408010). Data availability Data sharing is not applicable to this article as no datasets were generated or analysed during the current study. Declarations Ethics approval and consent to participate Not applicable. Consent to publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.