betaFeedback

WorksComputer Science

A Cooperative Game Theoretic Model for Sustainable, Smallholder Carbon Farming

Abstract

Carbon farming offers a promising path to improve farmer livelihoods, mitigate climate change, and make farming sustainable. However, smallholder farmers, who dominate agriculture in developing economies, face structural barriers to participating in carbon farming due to high certification costs, limited landholdings, and limited budgets available to them. In this paper, we develop a cooperative game-theoretic framework to study how coalition formation can enable viable and profitable participation of smallholder farmers in carbon farming. Our approach explores three components: (i) Solo Portfolio Optimization (SPO), which determines an optimal portfolio of carbon farming practices that individual farmers can adopt under budget constraints; (ii) Joint Portfolio Optimization (JPO), which computes the maximum total economic surplus achievable through the formation of farmer coalitions; and (iii) surplus allocation among participating farmers in a principled way using cooperative game-theoretic solution concepts, such as Shapley value, core, least core, and nucleolus. We show that the underlying cooperative game is superadditive under an affordability condition and non-convex in general. As part of this study, we release CarbonFARM-Bench, a carbon farming dataset constructed by experts, and conduct controlled computational experiments on this dataset. We observe that solo participation in carbon farming is largely infeasible for smallholder farmers under baseline parameters, whereas coalition formation creates substantial value driven by economies of scale. In addition, we demonstrate that stability-aware allocation mechanisms are required for sustaining cooperation. Together, our results highlight that enabling smallholder carbon farming is fundamentally a challenge of institutional design, and we offer several policy recommendations to make it attractive to smallholder farmers.

Introduction

Agriculture forms the economic backbone of many developing economies, supporting the livelihoods of a large share of the population. However, smallholder farmers, who typically cultivate fragmented landholdings (of less than 2 hectares), operate with limited capital and often face distress. Erratic rainfall, volatile commodity prices, and rising input costs translate directly into income shocks, leaving small and marginal farmers perpetually exposed to distress. At the same time, agriculture accounts for approximately 12% of global greenhouse gas (GHG) emissions [23], placing the sector at the center of a fundamental tension: how to sustain and enhance agricultural productivity while meaningfully reducing its carbon footprint.

Against this backdrop, carbon farming has emerged as a promising path towards simultaneously addressing sustainability and livelihood concerns [78]. It encompasses crop cultivation practices such as conservation tillage, reduced fertilizer and pesticide application, improved irrigation, and need-based organic or bio-fertilization, that strengthen the capacity of agricultural systems to sequester atmospheric CO2CO_2 or reduce net GHG emissions [87]. The benefits extend well beyond climate mitigation: these practices can lower input costs, enrich agro-biodiversity, and improve water and nutrient efficiency, thus generating tangible co-benefits for the farmers who adopt them [103]. Crucially, verified emission reductions and removals can be certified and traded as carbon credits in voluntary carbon markets, thereby opening an additional, potentially stable revenue stream for smallholders [66].

However, realizing this potential is not straightforward for smallholder farmers. Carbon credit programs incur substantial fixed costs, such as measurement, reporting, and verification (MRV), third-party auditing, and registry fees, that are often independent of farm size [94]. For a farmer cultivating one or two hectares, these costs can easily eclipse the value of carbon credits generated, making solo participation economically non-viable [82]. Coalition formation emerges as a natural institutional alternative. By pooling land and coordinating practices, groups of smallholders can collectively generate credit volumes that justify shared certification costs and present a credible counterparty to buyers and registries.

Applying cooperative action to carbon farming, however, introduces many challenges. The fundamental tension lies between the heterogeneity of carbon farming practices and the stability of any coalition formed around them. Even within a single village, smallholders vary considerably in farm size, soil quality, and access to capital. As a result, the portfolio of practices that maximizes net profit for one farmer may or may not be workable for another farmer. Moreover, even when two farmers adopt the same portfolios of practices, their actual sequestration outputs might differ. So, within a coalition, the members will rationally pursue different practice portfolios, and these individual choices jointly determine both the total carbon revenue the coalition can earn and the structure of the shared MRV costs, which scale with aggregate acreage. Crucially, in practice farmers do not pool their own input costs, yield outcomes, or finances: these remain private, and farmers are unwilling to bear another’s agronomic costs. This raises a problem that goes beyond the question of whether a coalition is collectively worthwhile. For a coalition to be viable, its shared surplus must be distributed so that, after each farmer’s private costs and benefits are accounted for, every subgroup has a compelling reason to stay rather than go solo or form a rival group.

In this paper, we model each farmer’s optimal carbon-practice portfolio as a constrained optimization problem, which we call the solo portfolio optimization problem. Next, we model coalition formation as a cooperative game in which the value of any coalition is the maximum total economic surplus (pooled carbon surplus plus members’ private gains) achievable through joint portfolio optimization, subject to individual feasibility constraints and cost-sharing rules. Finally, we model surplus allocation as a linear program over the pooled carbon surplus, comparing core-based allocations (core, least core, and the nucleolus) against the Shapley value and standard baselines.

Contributions

This paper makes the following contributions.

  • A Cooperative Game-Theoretic Framework for Carbon Farming. We develop a three-phase framework that integrates individual decision-making, coalition formation, and revenue allocation. Through this, we provide a principled approach to modeling and exploring smallholder participation in carbon farming and carbon markets.

  • CarbonFARM-Bench Dataset. We release CarbonFARM-Bench (Fair Allocation and Revenue Maximization for Smallholders in Carbon Farming Benchmark), an expert-constructed, literature-informed benchmark for smallholder carbon farming, covering eight practices, pairwise interaction matrices, MRV and transaction costs, and synthetic farmer panels, to support reproducible research on coalition formation and surplus allocation.

  • Structural Characterization of Coalition Game. We formally prove that the underlying coalition game is superadditive under an affordability condition, and non-convex in general.

  • Evaluation of Revenue Allocation Mechanisms. We systematically compare six revenue allocation mechanisms: four based on solution concepts in cooperative game theory (core, least core, nucleolus, Shapley value) and two standard baselines (area-proportional allocation and attributable-revenue allocation). Our results show that stability-driven approaches are required to ensure participation and prevent coalition deviations.

  • Policy-Relevant Insights for Inclusion of Smallholder Farmers. Based on empirical observations, we provide policy recommendations that help smallholder farmers access the benefits of carbon farming.

Review of Relevant Work

Smallholder Carbon Farming Economics. A longstanding literature examines the costs that make carbon projects difficult for smallholders to implement. Cacho et al. [15] provide a typology of transaction costs for carbon-sink projects in developing countries. In contrast, Cacho & Lipper [14] derive a feasibility frontier that identifies the minimum viable project size at a given carbon price. Cacho et al. [16] extend this analysis to offset exchange between a project developer and a group of landholders. However, by treating the group as a single seller, they leave the distribution of internal surplus unmodeled. Shames et al. [82] discuss ways to reduce costs and improve benefits to support the scaling up of smallholder carbon projects. More recent work focuses on monitoring, reporting, and verification (MRV). Smith et al. [86] review soil-carbon MRV approaches and propose a global MRV framework, highlighting why MRV costs can be high and partly fixed. Oldfield et al. [57] emphasize the trade-off between measurement rigor and cost in soil-carbon crediting. Vanderheyden et al. [94] develop a cost model for four MRV systems and find that costs vary substantially across monitoring approaches and project designs, suggesting that policy support may be needed to make participation viable for smaller farms.

Field evidence further illustrates how these barriers can translate into unequal participation and benefits. Cariappa & Krishna [19] surveyed 841 farmers in seven carbon-farming villages in Haryana and Madhya Pradesh, India. They find that participants were predominantly large landholders, only 4% were women, and about 99% had received no payment. Moreover, 28% had already discontinued the practices, indicating that participation does not necessarily translate into sustained adoption. Aggarwal & Brockington [2] find that participation in a forest-carbon project in India involved foregone crop, fodder, and fuelwood benefits, with small and marginal farmers particularly affected. Together, these findings suggest that carbon farming’s distributional outcomes depend not only on potential carbon revenue but also on participation, payment delivery, and the opportunity costs borne by farmers.

On the agronomic side, Li et al. [44] evaluate paddy-field practices in terms of their carbon sequestration capacity and contribution to household income, while Piscitelli et al. [65] develop bottom-up approaches to identify strategies for increasing soil carbon stocks. Pham-Kieu et al. [64] show that combining practices can yield greater abatement and profit than adopting any single practice, highlighting the potential value of considering practice portfolios rather than isolated interventions. Our Solo and Joint Portfolio Optimization problems build on this perspective by incorporating farmer-level budget constraints and shared certification costs. Taken together, these studies establish the economic and agronomic rationale for collective carbon farming, but leave open how the surplus generated by a pooled project should be distributed among heterogeneous farmers.

Cooperatives and Collective Action. Markelova et al. [51] review how collective organization helps smallholders reduce transaction costs and access markets, while Ostrom [58] document how groups can sustain cooperation without external enforcement. However, the benefits of collective action are not uniform across settings. Agarwal [1] find that group farms in Kerala outperformed individual family farms in output per hectare and net returns, whereas those in Telangana performed worse. Bikkina et al. [10] show that farmer producer organizations (FPOs) in Gujarat can lower transaction costs through collective action, but remain constrained by limited capital. Participation is also heterogeneous: a meta-analysis of 168 studies finds that farm size, access to credit, and distance to markets are significantly associated with cooperative membership, although the estimated effect sizes are generally small (Grashuis, [34]). Together, these findings suggest that collective organization can create economic benefits, but its effectiveness and accessibility depend on local conditions and farmers’ resources.

Cooperatives have also been designed to address specific agricultural activities. Bhardwaj et al. [9] use farmer collectives to design fair and cost-optimal auctions for procuring agricultural inputs. Sarkar et al. [79] study agricultural cooperatives using multi-agent coalitional models and show that similarity-based coalition formation can generate positive revenue for smallholders. Zheng et al. [106] develop cooperative game-based incentive mechanisms to support biopesticide adoption through FPOs. In the carbon setting, empirical studies associate cooperatives with lower agricultural emissions, mainly through reduced fertilizer use [98, 105], while Mantey et al. [50] examine the role of dairy cooperatives in smallholder carbon projects. Although these studies highlight the potential of cooperation for agricultural and environmental outcomes, they do not examine how the resulting carbon surplus is distributed among participating farmers. In practice, carbon projects are also increasingly moving toward aggregation. The VCS methodology VM0042 supports grouped projects [96], and an AWD-based project in Pakistan pooled smallholders under a single certification umbrella to obtain the first rice methane credits under the Gold Standard [56]. These developments illustrate how aggregation can facilitate certification, while leaving the question of internal surplus allocation open.

Incentives and Collective Payments. Raina et al. [69] map incentive mechanisms in carbon-farming contracts and distinguish results-based, action-based, and hybrid payments, highlighting how payment design, monitoring, and risk allocation shape farmer participation. Suich [89] find in Mozambique and Namibia that incentives may fail to sustain participation when they are poorly communicated, unsuitable for participants, or inadequately delivered. Chu et al. [22] analyze a tripartite evolutionary game among government, farmer cooperatives, and farmers under carbon trading, finding that a higher share of carbon revenue for farmers increases participation. These studies highlight the importance of payment design for participation, but Chu et al. [22] assume a fixed sharing ratio within a non-cooperative framework, whereas our approach derives the sharing rule from coalition-based allocation.

Collective environmental payments have also been examined in the agri-environment literature. The agglomeration bonus in Parkhurst et al. [61] provides landowners with an additional payment when enrolled parcels are spatially connected, thereby rewarding coordinated participation. Kuhfuss et al. [41] show in a choice experiment that a bonus conditional on a group enrollment threshold increases farmers’ willingness to join. Wätzold & Drechsler [99] find that the relative performance of homogeneous payments, agglomeration bonuses, and agglomeration payments depends on whether participants can make side payments. This finding is relevant to our transferable-utility assumption, under which coalition members can redistribute surplus among themselves.

Cooperative Games and Allocation. Cooperative game theory provides a framework for analyzing profit allocation in agriculture and natural-resource management. Bekolli et al. [7] study multidistributor-farmer games in which farmers sell through a cooperative, establish the existence of stable allocations, and characterize a rule that compensates farmers. Bekolli et al. [6] introduce totally balanced games for firms sharing crop-processing technology and derive core-stable compensation rules. Parrachino et al. [62] review cooperative games in environmental and water-resource problems, while Ambec & Sprumont [3] provide a canonical analysis of river sharing among heterogeneous agents, characterizing a core-stable distribution. Together, these studies illustrate how cooperative game theory can be used to analyze surplus distribution and allocation stability.

Frisk et al. [31] allocate the savings from pooled transport planning among eight Swedish forest companies. They compute coalition values using a linear program and compare the Shapley value, nucleolus, separable costs, shadow prices, and volume weights. Guajardo & Rönnqvist [35] survey allocation rules in collaborative transportation, including how frequently these rules yield core allocations. In linear production games, coalition values are obtained from linear programs, and LP duality guarantees a core allocation under the model’s assumptions [59]. By contrast, Caprara & Letchford [18] study integer minimization games in which coalition values are defined by integer programs, and compute cost shares using row and column generation. Petrosjan & Zaccour [63] extend cooperative allocation to dynamic settings by developing a time-consistent Shapley allocation for joint emission reduction.

Solution Concepts and Computation. The core [33], least core [52], nucleolus [81], and Shapley value [83] are the standard tools for allocating the surplus of a coalition. Non-emptiness of the core is characterized by balancedness [12, 84], and convex games have a non-empty core containing the Shapley value [85]. Exact computation requires evaluating all 2n2^n coalitions, each of which is an optimization problem in our setting, so cost grows quickly with the number of farmers. The nucleolus can be computed by a sequence of nested linear programs [40], and Benedek et al. [8] give a faster descent algorithm with a verification test. Yan & Procaccia [104] show that arbitrarily good approximations of the least core can be computed efficiently, and they make a case for the least core as an alternative to the Shapley value in credit assignment, while proving an impossibility result for the nucleolus. For larger coalitions, sampling-based approximations exist for the Shapley value [53], and core-like stability can be guaranteed with high probability from sampled coalition values [5].

Gap. Existing studies either examine the cost barriers to smallholder carbon farming without addressing how the surplus from collective participation is distributed, or investigate cooperative allocation in agriculture without incorporating the carbon-specific features of practice portfolios, farmer-level budget constraints, heterogeneous agronomic payoffs, and economies of scale in certification. The interaction between individual practice choices, the gains from collective certification, and the allocation of coalition surplus therefore remains insufficiently explored. This study addresses this gap by integrating Solo and Joint Portfolio Optimization with cooperative game-theoretic allocation rules.

Preliminaries

TU Game

A cooperative game with transferable utilities (TU) is defined by a pair (N,v)(N, v), where N={1,2,…,n}N = \{1, 2, \ldots, n\} is the set of players and v:2N→Rv : 2^N \to\mathbb{R} is the characteristic function satisfying v(∅)=0v(\emptyset) = 0. For every coalition S⊆NS \subseteq N, the quantity v(S)v(S) denotes the total utility that the members of SS can jointly generate and redistribute among themselves without loss.

In our setting, the players are farmers, and the characteristic function v(S)v(S) represents the maximum jointly attainable economic surplus of the coalition SS subject to the relevant constraints. Once the coalition values are determined, the key question becomes how to distribute the pooled carbon surplus among the farmers in a way that is fair, interpretable, and resistant to defection. We address this through the cooperative game-theoretic solution concepts introduced in the following sections.

The Core

Let x=(xi)i∈N\mathbf{x} = (x_i)_{i \in N} be an allocation vector, where xix_i denotes the payoff assigned to player ii. The core [80] is the set of all allocations satisfying

∑i∈Nxi=v(N)(1)\sum_{i \in N} x_i = v(N) \tag*{(1)}

and

∑i∈Sxi≥v(S),∀S⊆N.(2)\sum_{i \in S} x_i \ge v(S), \qquad\forall S \subseteq N. \tag*{(2)}

The first condition is efficiency: the entire value of the grand coalition is distributed among the players. The second condition is coalitional rationality: every coalition receives at least as much as it could secure on its own. Hence, an allocation in the core is stable against deviations by any subgroup of players.

The Least Core

When the core is empty, the coalitional stability constraints can be relaxed by allowing a uniform slack ε≥0\varepsilon\ge0. The least core [52] is defined through the optimization problem

min⁡x,εεs.t.∑i∈Nxi=v(N),∑i∈Sxi≥v(S)−ε∀S⊆N.(3)\begin{aligned} \min_{\mathbf{x},\varepsilon}\quad& \varepsilon\\ \text{s.t.}\quad& \sum_{i\in N} x_i = v(N),\\ & \sum_{i\in S} x_i \ge v(S)-\varepsilon\quad\forall S \subseteq N. \tag*{(3)} \end{aligned}

The parameter ε\varepsilon measures the maximum amount by which any coalition’s value claim may be violated. Thus, the least core identifies allocations that minimize the largest dissatisfaction across all coalitions.

The least core is especially useful when exact core stability is unattainable. Rather than discarding stability altogether, it finds the ”closest” possible approximation to the core.

The Nucleolus

For an allocation x\mathbf{x}, define the excess of coalition SS as

e(S,x)=v(S)−∑i∈Sxi.(4)e(S,\mathbf{x}) = v(S) - \sum_{i\in S} x_i. \tag*{(4)}

The excess measures the dissatisfaction of coalition SS under allocation x\mathbf{x}: a larger value of e(S,x)e(S,\mathbf{x}) means that coalition SS is further from receiving its worth. The nucleolus [81] is the unique allocation that lexicographically minimizes the ordered vector of coalition excesses, beginning with the largest excess, then the second largest, and so on.

Intuitively, the nucleolus refines the least core by not only minimizing the maximum dissatisfaction, but also resolving remaining conflicts as evenly as possible across coalitions. In games with a non-empty core, the nucleolus always belongs to the core.

The Shapley Value

The Shapley value [83] allocates the worth of the grand coalition according to each player’s expected marginal contribution across all possible coalition formation orders. Formally, for each player i∈Ni \in N, Shapley value is defined as

ϕi(v)=∑S⊆N∖{i}∣S∣!(∣N∣−∣S∣−1)!∣N∣![v(S∪{i})−v(S)].(5)\phi_i(v) = \sum_{S\subseteq N\setminus\{i\}} \frac{|S|!(|N|-|S|-1)!}{|N|!}\left[v(S\cup\{i\})-v(S)\right]. \tag*{(5)}

The Shapley value is widely used as it satisfies several desirable axioms, including efficiency, symmetry, dummy-player treatment, and additivity. It therefore provides a natural fairness benchmark based on contribution. However, unlike the core, it is not defined in terms of coalitional stability, and a Shapley allocation need not belong to the core. In applications such as coalition formation among farmers, the Shapley value is therefore best viewed as a marginal-contribution benchmark rather than a guaranteed stable allocation rule.

Methodology

Notation

Let F={1,…,n}F = \{1,\ldots,n\} be the set of farmers and P={1,…,m}P = \{1,\ldots,m\} the set of carbon-farming practices. For farmer FiF_i and practice PjP_j we use:

  • FSiFS_i: farm size of FiF_i (ha);

  • BiB_i: budget available to FiF_i ($);

  • OCjOC_j: operational cost of PjP_j relative to baseline ($/ha);

  • CMPijCMP_{ij}: farmer-specific carbon mitigation potential of PjP_j on FiF_i’s land (tCO2_2e/ha);

  • YjY_j: percentage yield change of PjP_j relative to conventional cultivation;

  • Y0Y^0: conventional yield (t/ha);

  • CP: crop price ($/t);

  • CCP: carbon credit price ($/tCO2_2e).

Due to variations in soil composition, geography, topography, and agro-climatic conditions, we assume CMP varies across farmers. Whereas operational costs and yield impacts are assumed to be uniform across farmers. This is a modeling simplification; the framework can be extended to farmer-specific OCijOC_{ij} and YijY_{ij} without altering the model’s overall structure. A complete notation list is provided in Appendix A, though each symbol is also introduced in the text when it first appears.

We emphasize that CMP in this study is an economic modeling quantity representing potential CO2_2-equivalent mitigation, rather than a claim that this amount would qualify as tradable carbon credits under a particular registry or methodology. It should be noted that project-level additionality, permanence, leakage, and uncertainty deductions, as well as registry-specific eligibility requirements, are outside the scope of the present model.

Practice Portfolios and Interaction Effects

A portfolio is a subset of practices (Π⊆P\Pi\subseteq P) that a particular farmer takes up for carbon farming. Let I⊆{(j,k):j<k}I \subseteq\{(j,k):j<k\} be the set of hard-incompatible practice pairs that represent combinations of practices that cannot be done together on the field. Therefore, a portfolio is admissible if it contains no pair from II.

Joint adoption of practices can produce biological and agronomic interactions that are not captured by summing individual effects [11]. For example, agroforestry and cover cropping may jointly sequester more carbon than the sum of their separate contributions (synergy), and two reduced-tillage practices may share machinery costs (cost complementarity). We model these effects pairwise, using symmetric parameters αjk=αkj\alpha_{jk}=\alpha_{kj}, βjk=βkj\beta_{jk}=\beta_{kj}, and γjk=γkj\gamma_{jk}=\gamma_{kj} for each pair (Pj,Pk)(P_j,P_k):

  • αjk\alpha_{jk}: synergistic or antagonistic effect on sequestration when PjP_j and PkP_k are jointly adopted. αjk>0\alpha_{jk}>0 indicates a sequestration synergy and αjk<0\alpha_{jk}<0 indicates an antagonism.

  • βjk\beta_{jk}: cost interaction. βjk<0\beta_{jk}<0 captures shared infrastructure or input savings and βjk>0\beta_{jk}>0 captures additional management complexity.

  • γjk\gamma_{jk}: yield interaction. γjk>0\gamma_{jk}>0 indicates a compounding yield benefit from the combination and γjk<0\gamma_{jk}<0 a yield penalty.

A portfolio of size ∣Π∣|\Pi| contains ∣Π∣(∣Π∣−1)/2|\Pi|(|\Pi|-1)/2 pairs. To keep the total interaction from growing with portfolio size [4], we scale the pairwise terms by the interaction weight

w(Π)={2∣Π∣(∣Π∣−1),∣Π∣≥2,0,∣Π∣≤1.(6)w(\Pi)= \begin{cases} \frac{2}{|\Pi|(|\Pi|-1)}, & |\Pi|\geq2,\\ 0, & |\Pi|\leq1. \end{cases} \tag*{(6)}

so that the total interaction is the average of the pairwise effects. The portfolio-level carbon mitigation, operational cost, and yield revenue change are then

CMPi(Π)=∑j∈ΠCMPij+w(Π)∑j<kj,k∈Παjk,(7)\mathrm{CMP}_i(\Pi)=\sum_{j\in\Pi}\mathrm{CMP}_{ij}+w(\Pi)\sum_{\substack{j<k\\j,k\in\Pi}}\alpha_{jk}, \tag*{(7)}
OC(Π)=∑j∈ΠOCj+w(Π)∑j<kj,k∈Πβjk,(8)\mathrm{OC}(\Pi)=\sum_{j\in\Pi}\mathrm{OC}_j+w(\Pi)\sum_{\substack{j<k\\j,k\in\Pi}}\beta_{jk}, \tag*{(8)}
ΔY(Π)=Y0CP100(∑j∈ΠYj+w(Π)∑j<kj,k∈Πγjk).(9)\Delta Y(\Pi)=\frac{Y^0\mathrm{CP}}{100}\left(\sum_{j\in\Pi}Y_j+w(\Pi)\sum_{\substack{j<k\\j,k\in\Pi}}\gamma_{jk}\right). \tag*{(9)}

Here ΔY(Π)\Delta Y(\Pi) is the yield revenue change ($/ha), positive for a gain and negative for a loss, and OC(Π)\mathrm{OC}(\Pi) may be negative (a saving).

Reliable field estimates of higher-order practice interactions are scarce. The parameters αjk\alpha_{jk}, βjk\beta_{jk}, and γjk\gamma_{jk} should therefore be read as scenario parameters, not as universally estimated agronomic effects.

Certification Costs

Carbon credit certification involves two categories of costs: Monitoring, Reporting, and Verification (MRV) costs and Transaction costs (T). For a coalition SS with total area A(S)=∑i∈SFSiA(S)=\sum_{i\in S}\mathrm{FS}_i, we define

CMRV(S)=FixMRV+VarMRVA(S)δ,δ∈(0,1),(10)C_{\mathrm{MRV}}(S)=\mathrm{Fix}_{\mathrm{MRV}}+\mathrm{Var}_{\mathrm{MRV}}A(S)^\delta,\qquad\delta\in(0,1), \tag*{(10)}
CT(S)=FixT+VarT∣S∣,(11)C_T(S)=\mathrm{Fix}_T+\mathrm{Var}_T|S|, \tag*{(11)}
C(S)=CMRV(S)+CT(S).(12)C(S)=C_{\mathrm{MRV}}(S)+C_T(S). \tag*{(12)}

FixMRV\mathrm{Fix}_{\mathrm{MRV}} ($) covers the base administrative and auditing overhead incurred one time per certified unit, regardless of size, VarMRV\mathrm{Var}_{\mathrm{MRV}} ($/haδ^\delta) scales with certified area, and δ∈(0,1)\delta\in(0,1) is a sublinearity parameter capturing economies of scale in monitoring ([46]). Because δ<1\delta<1, the MRV cost is concave in total area, so pooling farms under a single certification umbrella strictly reduces per-hectare MRV costs relative to individual certification. FixT\mathrm{Fix}_T ($) is a one-time registry and contracting overhead, and VarT\mathrm{Var}_T ($/farmer) is a per-farmer administration fee ([28]). For a singleton,

C({i})=FixMRV+VarMRVFSiδ+FixT+VarT.C(\{i\})=\mathrm{Fix}_{\mathrm{MRV}}+\mathrm{Var}_{\mathrm{MRV}}\mathrm{FS}_i^\delta+\mathrm{Fix}_T+\mathrm{Var}_T.

Shared and Private Quantities

Farmers in a coalition file jointly and generate credits together, but they do not pool agronomic costs or outcomes. No farmer bears another’s operational costs or yield losses, and none shares another’s yield gains. We therefore partition the economics into two groups.

Shared (pooled) quantities. The carbon revenue and the certification cost,

R(S,ΠS)=∑i∈SFSi⋅CMPi(Π)⋅CCP,(13)R(S,\Pi^S)=\sum_{i\in S}\mathrm{FS}_i\cdot\mathrm{CMP}_i(\Pi)\cdot\mathrm{CCP}, \tag*{(13)}
C(S)=CMRV(S)+CT(S).C(S)=C_{\mathrm{MRV}}(S)+C_T(S).

Here, R(S,ΠS)R(S,\Pi^S) is the coalition’s pooled revenue, where each farmer FiF_i’s revenue is the sequestration per hectare, scaled by farm size and the prevailing carbon credit price. C(S)C(S) is the joint certification cost from equation (12), which arises only from filing together and does not depend on any farmer’s chosen portfolio. These two are the only quantities that are pooled; everything else is private to each farmer.

Private quantities. Operational cost and yield effects, which belong to each farmer alone:

gi(Π)=FSi(ΔY(Π)−OC(Π)).(14)g_i(\Pi) = \mathrm{FS}_i\bigl(\Delta Y(\Pi) - \mathrm{OC}(\Pi)\bigr). \tag*{(14)}

The term FSiΔY(Π)\mathrm{FS}_i\Delta Y(\Pi) is the revenue from yield gain (or loss), positive when practices improve yield and negative when they reduce it, and FSiOC(Π)\mathrm{FS}_i\mathrm{OC}(\Pi) is the operational cost. Because ΔY(Π)\Delta Y(\Pi) and OC(Π)\mathrm{OC}(\Pi) can each change sign, they jointly govern the net agronomic payoff gi(Π)g_i(\Pi): a yield loss enters as a phantom cost, while a cost-reducing practice enters as a phantom revenue.

Two surpluses. Throughout the paper we distinguish two quantities for a coalition SS with selected profile ΠS\Pi^S:

p(S)=R(S,ΠS)−C(S)(pooled, transferable carbon-market surplus),(15)p(S) = R(S,\Pi^S) - C(S) \qquad\text{(pooled, transferable carbon-market surplus),} \tag*{(15)}
G(S)=∑i∈Sgi(ΠiS)(private, non-transferable gains),(16)G(S) = \sum_{i\in S} g_i(\Pi_i^S) \qquad\text{(private, non-transferable gains),} \tag*{(16)}
v(S)=p(S)+G(S)(total economic surplus).(17)v(S) = p(S) + G(S) \qquad\text{(total economic surplus).} \tag*{(17)}

Only p(S)p(S) can be moved between farmers. The characteristic function v(S)v(S) defined measures total economic surplus, because the coalition selects its portfolios to maximize what its members actually earn.

Non-pooling premise. Non-pooling means that private costs and yields are never transferred, never charged to another farmer, and never used to fund another farmer’s participation. It does not mean that private effects are invisible to portfolio selection. We encode it through three rules:

(NP1) Objective. Portfolios are selected to maximize total economic surplus. Private quantities enter the objective as welfare, but only through each farmer’s own portfolio.

(NP2) Constraints. Each farmer’s private economics constrain that farmer’s own portfolio through a farmer-specific constraint, never through an aggregate one.

(NP3) Transfers. Only the pooled surplus p(F)p(F) is distributed. A farmer’s final surplus is

ui=xi+gi(Πi),(18)u_i = x_i + g_i(\Pi_i), \tag*{(18)}

where xix_i is her share of the pooled surplus and gig_i is retained privately.

This is a quasi-linear utility structure: only xix_i is transferable, while gi(Πi)g_i(\Pi_i) stays with the farmer. It formalizes our assumption that farmers do not pool their input costs or yield outcomes, but still care about them.

Per-farmer constraints. Let SS be a coalition and let

κi(S)=FSiA(S)C(S)\kappa_i(S) = \frac{\mathrm{FS}_i}{A(S)} C(S)

be the area-proportional certification outlay of farmer ii. A portfolio Πi\Pi_i is individually admissible in SS if it is admissible with respect to II and

(Budget)Bi+gi(Πi)≥κi(S),(19)\text{(Budget)}\qquad B_i + g_i(\Pi_i) \ge\kappa_i(S), \tag*{(19)}
(Self-reliance)Ri(Πi)+gi(Πi)−κi(S)≥0.(20)\text{(Self-reliance)}\qquad R_i(\Pi_i) + g_i(\Pi_i) - \kappa_i(S) \ge0. \tag*{(20)}

The budget constraint requires that the farmer’s funds, adjusted for her own yield and operating effects, cover her certification outlay. Self-reliance requires that the benefit generated on her own land covers her own certification outlay.

Solo Portfolio Optimization (SPO)

A farmer FiF_i acting alone either adopts a carbon-farming portfolio and certifies, or does not participate and earns nothing. If she participates, she selects a portfolio Πiso⊆P\Pi_i^{\mathrm{so}} \subseteq P by solving

Πiso=arg max⁡Π admissible Ri(Π)+gi(Π)−C({i})s.t.Bi+gi(Π)≥C({i}),Ri(Π)+gi(Π)≥C({i}).(21)\Pi_i^{\mathrm{so}} = \underset{\Pi\ \mathrm{admissible}}{\operatorname{arg\,max}}\ R_i(\Pi) + g_i(\Pi) - C(\{i\}) \quad \begin{aligned} \text{s.t.}\quad&B_i + g_i(\Pi) \ge C(\{i\}),\\ &R_i(\Pi) + g_i(\Pi) \ge C(\{i\}). \end{aligned} \tag*{(21)}

The objective is the carbon revenue Ri(Π)R_i(\Pi) plus the private net agronomic payoff gi(Π)g_i(\Pi), minus the certification cost C({i})C(\{i\}), which collects the monitoring, reporting, and verification (MRV) and transaction costs. The budget constraint ensures that net costs do not exceed the farmer’s available capital BiB_i, and the self-reliance constraint ensures that the benefit generated on her own land covers her certification outlay.

Standalone value. Let OPT⁡i\operatorname{OPT}_i denote the optimal value of equation 21, with OPT⁡i=−∞\operatorname{OPT}_i = -\infty if the problem is infeasible. The standalone value of farmer FiF_i, which is her total surplus from acting alone, is

v({i})=max⁡{0,OPT⁡i}.(22)v(\{i\}) = \max\{0,\operatorname{OPT}_i\}. \tag*{(22)}

Solo participation is feasible when OPT⁡i>0\operatorname{OPT}_i > 0. Otherwise FiF_i does not participate, Πiso=∅\Pi_i^{\mathrm{so}} = \varnothing, and v({i})=0v(\{i\}) = 0. The value v({i})v(\{i\}) is the surplus farmer FiF_i retains after all costs are settled, and it serves as her outside-option payoff. Every allocation rule must guarantee each farmer at least v({i})v(\{i\}) for individual rationality to hold.

Joint Portfolio Optimization (JPO)

When a coalition S⊆FS \subseteq F with ∣S∣≥2|S| \ge2 forms, its members file for certification jointly and share MRV and transaction costs C(S)C(S). Each farmer Fi∈SF_i \in S selects a coalition portfolio ΠiS⊆P\Pi_i^S \subseteq P. These portfolios may differ across farmers, and may differ from the farmer’s solo portfolio Πiso\Pi_i^{\mathrm{so}}, because the reduced certification burden can make additional or different practices financially viable.

Operational costs and yield-related revenue are private to each farmer; only certification costs are pooled. Pooling these farmer-specific quantities would be unfair. Farms differ across farmers in size, soil, climate, and other local conditions, and portfolios differ as well, so the practices adopted, the operational costs incurred, and the yield gains or losses realized all vary from one farmer to another. A farmer who adopts a costly practice, who operates a farm with less favorable conditions, or who experiences a yield loss would otherwise have part of that burden shifted onto others. In contrast, a farmer with low costs or high yield gains would have to share the benefit. Certification costs, by contrast, arise from the joint filing itself and are not tied to any one farmer’s choices or farm conditions so that they can be shared in proportion to area. Each farmer’s own economics, therefore, enter the coalition’s problem only through her portfolio and constraints equation 19–equation 20.

The coalition selects portfolios to maximize its members’ total economic surplus:

OPT⁡(S)=max⁡{Πi}i∈S∑i∈S[Ri(Πi)+gi(Πi)]−C(S)s.t.Πi individually admissible in S∀i∈S.(23)\operatorname{OPT}(S) = \underset{\{\Pi_i\}_{i\in S}}{\max} \sum_{i\in S}\left[R_i(\Pi_i) + g_i(\Pi_i)\right] - C(S) \quad \begin{aligned} \text{s.t.}\quad&\Pi_i\text{ individually admissible in }S\quad\forall i\in S. \end{aligned} \tag*{(23)}

There are ∣S∣|S| pairs of constraints, one per farmer. That is, each farmer’s budget, adjusted for yield gains or losses and operational costs, must cover her area-proportional share of the coalition’s certification costs, and the benefit generated on her own land must cover the same share. For ∣S∣=1|S| = 1, equation 23 coincides with equation 21, so SPO is the singleton case of JPO.

Selected profile. When several profiles attain the optimum, ties are broken in favor of the profile that maximizes

min⁡i∈S[Ri(Πi)+gi(Πi)−κi(S)],\min_{i\in S}\left[R_i(\Pi_i) + g_i(\Pi_i) - \kappa_i(S)\right],

and any remaining tie by a fixed ordering of portfolios. This makes the selected profile ΠS={ΠiS}i∈S\Pi^{S}=\{\Pi^{S}_{i}\}_{i\in S} unique, and favors the profile that leaves the worst-off member with the largest net benefit. If OPT⁡(S)≤0\operatorname{OPT}(S)\le0 or no admissible profile exists, SS does not certify jointly, and we set ΠiS=∅\Pi^{S}_{i}=\varnothing for all i∈Si\in S.

Characteristic function.

v(S)=max⁡{0,OPT⁡(S)}(∣S∣≥1),v(∅)=0,(24)v(S)=\max\{0,\operatorname{OPT}(S)\}\quad(|S|\ge1),\qquad v(\varnothing)=0, \tag*{(24)}

so v(S)≥0v(S)\ge0 for all SS and no coalition has value −∞-\infty. The characteristic function measures the maximum total economic surplus of SS, namely v(S)=p(S)+G(S)v(S)=p(S)+G(S) as in equation 17. Only the pooled component p(S)p(S) is transferable (NP3).

Revenue Allocation

Once the grand coalition is formed, the next question is how to divide its pooled surplus among the farmers. An allocation x=(xi)i∈Fx=(x_i)_{i\in F} distributes the pooled surplus of the grand coalition. We write xi=ρi−κi(F)x_i=\rho_i-\kappa_i(F), where ρi\rho_i is the share of gross carbon revenue received by farmer ii and κi(F)\kappa_i(F) is the certification outlay she actually pays. The outlay used in the budget constraint equation 19 is therefore the payment that farmers actually make, and xix_i is the resulting net share.

Let ΠF\Pi^{F} be the profile selected by JPO for the grand coalition. Then ∑i∈Fxi=p(F)\sum_{i\in F}x_i=p(F). After the payout, each farmer adds her private term, so her final surplus is

ui=xi+gi(ΠiF),∑i∈Fui=p(F)+G(F)=v(F).(25)u_i=x_i+g_i(\Pi^{F}_{i}),\qquad\sum_{i\in F}u_i=p(F)+G(F)=v(F). \tag*{(25)}

Only xix_i is transferable, and gi(ΠiF)g_i(\Pi^{F}_{i}) always stays with the farmer. The self-reliance test equation 20 uses the revenue generated on FiF_i’s own land, Ri(Πi)R_i(\Pi_i), only as a screen for admissible portfolios.

Deviation behavior. A subgroup T⊊FT\subsetneq F may leave the grand coalition and file for certification on its own. It can choose any profile that is individually admissible in TT, including ∅\varnothing, which means not certifying. The members of TT can share the pooled surplus of their chosen profile freely among themselves, but they cannot transfer private gains. A deviating singleton {i}\{i\} has nobody to pool with and behaves as in SPO. A farmer who leaves can also remain a conventional farmer with no portfolio and no carbon revenue, which is why v({i})v(\{i\}) is floored at zero in equation 22.

Deviation value. A farmer who stays in the grand coalition keeps the private gain gi(ΠiF)g_i(\Pi^{F}_{i}). A deviating group must therefore be compared against this gain rather than the raw coalition value. For ∅≠T⊆F\varnothing\ne T\subseteq F we define

v^(T)=v(T)−∑i∈Tgi(ΠiF),v^(∅)=0.(26)\hat{v}(T)=v(T)-\sum_{i\in T}g_i(\Pi^{F}_{i}),\qquad\hat{v}(\varnothing)=0. \tag*{(26)}

Equivalently,

v^(T)=p(T)+∑i∈T[gi(ΠiT)−gi(ΠiF)].\hat{v}(T)=p(T)+\sum_{i\in T}\left[g_i(\Pi^{T}_{i})-g_i(\Pi^{F}_{i})\right].

It is the amount of transferable surplus that TT must receive for deviation to be unattractive, given that its members keep their private gains under the grand-coalition profile if they stay. For the grand coalition, v^(F)=v(F)−G(F)=p(F)\hat{v}(F)=v(F)-G(F)=p(F). For a singleton, v^({i})=v({i})−gi(ΠiF)\hat{v}(\{i\})=v(\{i\})-g_i(\Pi^{F}_{i}), and when solo participation is infeasible, v({i})=0v(\{i\})=0, so v^({i})=−gi(ΠiF)\hat{v}(\{i\})=-g_i(\Pi^{F}_{i}). If ΠiT=ΠiF\Pi^{T}_{i}=\Pi^{F}_{i} for every i∈Ti\in T, then v^(T)=p(T)\hat{v}(T)=p(T).

Proposition 1 (Blocking condition). Let xx be an allocation with ∑i∈Fxi=p(F)\sum_{i\in F}x_i=p(F) and let T⊊FT\subsetneq F be a deviating group as described above. There exists a profile and a split of the transferable surplus of TT that make every member of TT strictly better off than under xx if and only if

v^(T)>∑i∈Txi.\hat{v}(T)>\sum_{i\in T}x_i.

Proof. Under xx, member ii obtains ui=xi+gi(ΠiF)u_i = x_i + g_i(\Pi_i^F). Suppose TT deviates to an admissible profile Π\Pi. Its members share the pooled surplus

pT(Π)=∑i∈TRi(Πi)−C(T)p_T(\Pi) = \sum_{i\in T} R_i(\Pi_i) - C(T)

through transfers yiy_i with ∑i∈Tyi=pT(Π)\sum_{i\in T} y_i = p_T(\Pi), and member ii obtains yi+gi(Πi)y_i + g_i(\Pi_i). For a singleton, yi=pT(Π)y_i = p_T(\Pi). The yiy_i can be transferred freely inside TT, whereas the gig_i cannot. So every member can be made strictly better off under Π\Pi if and only if the totals satisfy

pT(Π)+∑i∈Tgi(Πi)>∑i∈Tui.p_T(\Pi) + \sum_{i\in T} g_i(\Pi_i) > \sum_{i\in T} u_i.

The left-hand side is the total surplus of TT under Π\Pi, and its maximum over admissible profiles is v(T)v(T) by equation 23 and equation 24. Hence, a suitable profile and split exist if and only if

v(T)>∑i∈T[xi+gi(ΠiF)].v(T) > \sum_{i\in T} [x_i + g_i(\Pi_i^F)].

which is the same as v^(T)>∑i∈Txi\hat{v}(T) > \sum_{i\in T} x_i.

For sufficiency, take Π=ΠT\Pi= \Pi^T and set

yi=xi+gi(ΠiF)−gi(ΠiT)+η,η=v(T)−∑i∈T[xi+gi(ΠiF)]∣T∣>0.y_i = x_i + g_i(\Pi_i^F) - g_i(\Pi_i^T) + \eta,\qquad\eta= \frac{v(T) - \sum_{i\in T} [x_i + g_i(\Pi_i^F)]}{|T|} > 0.

Then

∑i∈Tyi=v(T)−∑i∈Tgi(ΠiT)=p(T),\sum_{i\in T} y_i = v(T) - \sum_{i\in T} g_i(\Pi_i^T) = p(T),

as required.

Only the totals matter here. The transferable shares can be redistributed inside TT, while the private gains cannot, so they enter the test as fixed offsets.

Feasibility Constraints

An allocation is admissible if it satisfies two conditions. The first is efficiency, which requires that the entire pooled carbon-market surplus is distributed and nothing is left unallocated:

∑i∈Fxi=p(F)=v^(F).(27)\sum_{i\in F} x_i = p(F) = \hat{v}(F). \tag*{(27)}

The second is stability, which ensures that no subgroup of farmers can be made better off, in terms of total surplus equation 25, by leaving and forming an independent coalition TT:

∑i∈Txi≥v^(T)∀ ∅≠T⊊F.(28)\sum_{i\in T} x_i \ge\hat{v}(T) \qquad\forall\ \varnothing\ne T \subsetneq F. \tag*{(28)}

By Proposition 1, equation 28 is exactly the no-blocking condition for the class of deviations described above. Individual rationality is the special case ∣T∣=1|T| = 1, where the constraint reads xi≥v({i})−gi(ΠiF)x_i \ge v(\{i\}) - g_i(\Pi_i^F), or equivalently ui≥v({i})u_i \ge v(\{i\}).

Cooperative Game-Theoretic Solutions

We apply the solution concepts from Section 3 to the farmer coalition game, replacing the coalition worths in the stability constraints with the deviation values v^(T)\hat{v}(T). Every mechanism satisfies efficiency equation 27. When the core is non-empty, we look for allocations inside it. Otherwise, we fall back on approximate stability concepts.

Least core. We solve the LP from § 3.3 using the deviation values v^(T)\hat{v}(T) from equation (26):

min⁡x,εεs.t.∑i∈Fxi=p(F),∑i∈Txi≥v^(T)−ε∀ ∅≠T⊊F.(29)\begin{aligned} \min_{x,\varepsilon}\quad&\varepsilon\\ \text{s.t.}\quad&\sum_{i\in F}x_i=p(F),\\ &\sum_{i\in T}x_i\ge\hat{v}(T)-\varepsilon\quad\forall\ \varnothing\ne T\subsetneq F. \tag*{(29)} \end{aligned}

If the optimal slack ε∗≤0\varepsilon^* \le0, the core is non-empty. If ε∗>0\varepsilon^*>0, no allocation is fully stable, and ε∗\varepsilon^* is the smallest uniform violation. The least core minimizes the maximum coalitional dissatisfaction through a single LP, but it need not be unique, since several allocations can achieve the same optimal slack ε∗\varepsilon^*. Among them, we select the allocation that also minimizes ∑ixi2\sum_i x_i^2.

Nucleolus. The nucleolus refines the least core to a unique allocation by lexicographically minimizing the coalition excesses e(T,x)=v^(T)−∑i∈Txie(T,x)=\hat{v}(T)-\sum_{i\in T}x_i. It always lies within the least core, and when the core is non-empty, it also lies within the core. We compute it through a sequence of LPs. At each stage, the constraints that bind at the previous optimum are fixed, and ε\varepsilon is minimized over the remaining coalitions until the allocation is unique.

Shapley value. We compute the Shapley value on the deviation game, with v^(∅)=0\hat{v}(\varnothing)=0:

ϕi(v^)=∑T⊆F∖{i}∣T∣!(∣F∣−∣T∣−1)!∣F∣![v^(T∪{i})−v^(T)].(30)\phi_i(\hat{v})=\sum_{T\subseteq F\setminus\{i\}}\frac{|T|!(|F|-|T|-1)!}{|F|!}\left[\hat{v}(T\cup\{i\})-\hat{v}(T)\right]. \tag*{(30)}

The Shapley value serves as a marginal-contribution benchmark for total farmer surplus. It satisfies efficiency, symmetry, the dummy axiom, and additivity, but it is not guaranteed to satisfy equation (28). We therefore check core membership after computing it, using equation (27)–equation 28.

Attributable-revenue split. In this baseline, each farmer receives the revenue generated on her own land and pays her area-proportional certification outlay:

xiAR=Ri(ΠiF)−κi(F),∀i∈F.(31)x_i^{\mathrm{AR}}=R_i(\Pi_i^F)-\kappa_i(F),\qquad\forall i\in F. \tag*{(31)}

This is efficient, since ∑ixiAR=R(F,ΠF)−C(F)=p(F)\sum_i x_i^{\mathrm{AR}}=R(F,\Pi^F)-C(F)=p(F). It is consistent with the self-reliance screen, but it does not account for coalitional stability.

Area-proportional split. In this baseline, the pooled surplus is divided in proportion to farm area:

xiPS=FSiA(F)p(F),∀i∈F.(32)x_i^{\mathrm{PS}}=\frac{FS_i}{A(F)}p(F),\qquad\forall i\in F. \tag*{(32)}

It is a simple land-based benchmark that ignores differences in carbon mitigation potential across farms.

Modeling Assumptions and Limitations

  • Coalition-selected portfolios. The coalition selects the profile that maximizes total surplus, which need not coincide with the portfolio each member would pick if choosing only for herself.

  • Implementation. xix_i does not depend on realized practices, so we assume members implement the selected JPO portfolio, enforced by contract or by result-based verification through MRV.

  • Single period. Costs and payments are modeled for one season, without recurring certification or payment lags.

  • Farmer heterogeneity. Private economics differ across farmers through area, portfolio, and CMPij\mathrm{CMP}_{ij}. Currently OCij\mathrm{OC}_{ij} and YijY_{ij} are not considered but the model can support it.

Superadditivity and Convexity

In this section, we study two structural properties of the total-surplus game (F,v)(F,v). We show that the game is superadditive on feasible coalitions under an explicit affordability condition, and that it is not convex in general.

Notation and Lemmas

For a coalition SS, we denote its total area, total certification cost, and average certification cost per hectare by

A(S)=∑i∈SFSi,A(S) = \sum_{i \in S} \mathrm{FS}_{i},
C(S)=FixMRV+VarMRVA(S)δ+FixT+VarT∣S∣,C(S) = \mathrm{Fix}_{\mathrm{MRV}} + \mathrm{Var}_{\mathrm{MRV}} A(S)^{\delta} + \mathrm{Fix}_{T} + \mathrm{Var}_{T}|S|,
cˉ(S)=C(S)A(S).\bar{c}(S) = \frac{C(S)}{A(S)}.

For a farmer ii with portfolio Πi\Pi_i, we define two per-hectare quantities:

θi(Πi)=BiFSi+ΔY(Πi)−OC(Πi)\theta_i(\Pi_i) = \frac{B_i}{\mathrm{FS}_i} + \Delta Y(\Pi_i) - \mathrm{OC}(\Pi_i)

(net funds per hectare),

ηi(Πi)=CMPi(Πi)CCP+ΔY(Πi)−OC(Πi)\eta_i(\Pi_i) = \mathrm{CMP}_i(\Pi_i)\mathrm{CCP} + \Delta Y(\Pi_i) - \mathrm{OC}(\Pi_i)

(gross economic benefit per hectare).

Here ηi(Πi)=(Ri(Πi)+gi(Πi))/FSi\eta_i(\Pi_i) = (R_i(\Pi_i) + g_i(\Pi_i))/\mathrm{FS}_i. The affordability margin of farmer ii is

mi(Πi)=min⁡{θi(Πi),ηi(Πi)}.m_i(\Pi_i) = \min\{\theta_i(\Pi_i), \eta_i(\Pi_i)\}.

Since κi(S)=FSicˉ(S)\kappa_i(S) = \mathrm{FS}_i\bar{c}(S), dividing the budget constraint equation (19) by FSi\mathrm{FS}_i gives θi(Πi)≥cˉ(S)\theta_i(\Pi_i) \ge\bar{c}(S), and dividing the self-reliance constraint equation (20) by FSi\mathrm{FS}_i gives ηi(Πi)≥cˉ(S)\eta_i(\Pi_i) \ge\bar{c}(S). Together, the two constraints are equivalent to the per-hectare condition

mi(Πi)≥cˉ(S)for every i∈S.(33)m_i(\Pi_i) \ge\bar{c}(S) \qquad\text{for every } i \in S. \tag*{(33)}

In other words, every member of the coalition must have, both in available funds and in the benefit generated on her own land, at least as much per hectare as the average certification cost per hectare of the coalition.

A coalition SS is called feasible if some admissible profile satisfies equation (33). For an infeasible coalition, OPT⁡(S)=−∞\operatorname{OPT}(S) = -\infty and, by equation (24), v(S)=0v(S) = 0, which means that it does not certify. Since cˉ(S)>0\bar{c}(S) > 0 and ηi(∅)=0\eta_i(\varnothing) = 0, the empty portfolio is never admissible for a member of a certifying coalition.

Lemma 1 (Concavity). For δ∈(0,1)\delta\in(0,1) and a,b>0a,b > 0,

(a+b)δ<aδ+bδ.(a+b)^{\delta} < a^{\delta} + b^{\delta}.

Proof. Let φ(t)=tδ\varphi(t) = t^{\delta}. It is strictly concave because

φ′′(t)=δ(δ−1)tδ−2<0,\varphi''(t) = \delta(\delta- 1)t^{\delta-2} < 0,

and φ(0)=0\varphi(0) = 0. Put λ=a/(a+b)\lambda= a/(a+b). Concavity gives

φ(a)>λφ(a+b),\varphi(a) > \lambda\varphi(a+b),
φ(b)>(1−λ)φ(a+b).\varphi(b) > (1-\lambda)\varphi(a+b).

Adding the two inequalities gives the claim.

Lemma 2 (Merging reduces total and average certification cost). Let S,TS,T be disjoint coalitions, Q=S∪TQ = S \cup T, a=A(S)a = A(S), b=A(T)b = A(T) and F0=FixMRV+FixTF_0 = \mathrm{Fix}_{\mathrm{MRV}} + \mathrm{Fix}_{T}. Then

Γ(S,T):=C(S)+C(T)−C(Q)=F0+VarMRV[aδ+bδ−(a+b)δ]>0,\Gamma(S,T) := C(S) + C(T) - C(Q) = F_0 + \mathrm{Var}_{\mathrm{MRV}}\left[a^{\delta} + b^{\delta} - (a+b)^{\delta}\right] > 0,

and consequently

cˉ(Q)<acˉ(S)+bcˉ(T)a+b≤max⁡{cˉ(S),cˉ(T)}.(34)\bar{c}(Q) < \frac{a\bar{c}(S) + b\bar{c}(T)}{a+b} \le\max\{\bar{c}(S),\bar{c}(T)\}. \tag*{(34)}

Proof. The saving. The MRV terms contribute

2FixMRV+VarMRV(aδ+bδ)−FixMRV−VarMRV(a+b)δ.2\mathrm{Fix}_{\mathrm{MRV}}+\mathrm{Var}_{\mathrm{MRV}}(a^\delta+b^\delta)-\mathrm{Fix}_{\mathrm{MRV}}-\mathrm{Var}_{\mathrm{MRV}}(a+b)^\delta.

For the transaction terms, ∣Q∣=∣S∣+∣T∣|Q|=|S|+|T|, so exactly one FixT\mathrm{Fix}_{T} is saved. Adding the two contributions gives the formula for Γ\Gamma.

Positivity. By Lemma 1, the bracket is positive. We assume that VarMRV>0\mathrm{Var}_{\mathrm{MRV}}>0 or F0>0F_0>0, so Γ>0\Gamma>0.

The average cost. Since Γ>0\Gamma>0,

C(Q)<C(S)+C(T)=acˉ(S)+bcˉ(T).C(Q)<C(S)+C(T)=a\bar{c}(S)+b\bar{c}(T).

Dividing by a+ba+b gives the first inequality in equation 37. The second holds because a weighted average never exceeds the maximum.

Intuitively, after a merger, the average cost per hectare falls below that of the more expensive side, so that side can always afford the merged coalition. Only the cheaper side may lose affordability, because it now pays a higher average cost than before.

Lemma 3 (Feasible coalitions have non-negative surplus). If SS is feasible and {Πi}i∈S\{\Pi_i\}_{i\in S} is any admissible profile satisfying equation 36, then

∑i∈S[Ri(Πi)+gi(Πi)]−C(S)≥0.\sum_{i\in S}\left[R_i(\Pi_i)+g_i(\Pi_i)\right]-C(S)\geq0.

Hence, for every feasible SS,

v(S)=OPT⁡(S)=∑i∈S[Ri(ΠiS)+gi(ΠiS)]−C(S).v(S)=\operatorname{OPT}(S)=\sum_{i\in S}\left[R_i(\Pi_i^S)+g_i(\Pi_i^S)\right]-C(S).

Proof. The self-reliance constraint equation 20 gives Ri(Πi)+gi(Πi)≥κi(S)R_i(\Pi_i)+g_i(\Pi_i)\geq\kappa_i(S) for each i∈Si\in S. Summing over i∈Si\in S and using ∑i∈Sκi(S)=C(S)\sum_{i\in S}\kappa_i(S)=C(S) gives the first claim. The second follows from equation 24.

Superadditivity

Definition 1 (Superadditivity). (F,v)(F,v) is superadditive on a family of coalitions if v(S∪T)≥v(S)+v(T)v(S\cup T)\geq v(S)+v(T) for all disjoint S,TS,T in the family such that S∪TS\cup T also belongs to it.

Theorem 1 (Superadditivity). Let S,TS,T be disjoint, non-empty, feasible coalitions with optimal profiles ΠS∗,ΠT∗\Pi_S^*,\Pi_T^*, labelled so that cˉ(S)≤cˉ(T)\bar{c}(S)\leq\bar{c}(T). Define

mS=min⁡i∈Smi(ΠS,i∗),ΔS(T)=C(S∪T)−C(S).m_S=\min_{i\in S}m_i(\Pi_{S,i}^*),\qquad\Delta_S(T)=C(S\cup T)-C(S).

If

(H)cˉ(S∪T)≤mS,(35)(H)\qquad\bar{c}(S\cup T)\leq m_S, \tag*{(35)}

then the concatenated profile Π~Q\widetilde{\Pi}_Q is feasible for Q=S∪TQ=S\cup T, and

v(S∪T)≥v(S)+v(T)+Γ(S,T)>v(S)+v(T).v(S\cup T)\geq v(S)+v(T)+\Gamma(S,T)>v(S)+v(T).

Condition (H)(H) holds in particular in either of the following cases:

  1. cˉ(S)=cˉ(T)\bar{c}(S)=\bar{c}(T);

  2. the condition

(H′)ΔS(T)≤A(T)mS,(36)(H')\qquad\Delta_S(T)\leq A(T)m_S, \tag*{(36)}

that is, the extra cost of absorbing TT, per hectare of TT, is at most the tightest per-hectare affordability margin in SS.

The idea of the proof is simple. We keep every farmer’s portfolio unchanged. The merged coalition is cheaper per hectare than its more expensive part, so the farmers of TT remain feasible, and (H) is exactly what the farmers of SS need.

Proof. Let every farmer keep her own portfolio, so that every mim_i is unchanged. By equation (33), it suffices to show that mi≥cˉ(Q)m_i \ge\bar{c}(Q) for all i∈Qi \in Q.

Step 1: farmers of TT. Since ΠT∗\Pi_T^* is feasible for TT, we have mi≥cˉ(T)m_i \ge\bar{c}(T). By equation (34) and the labelling,

cˉ(T)=max⁡{cˉ(S),cˉ(T)}>cˉ(Q).\bar{c}(T) = \max\{\bar{c}(S), \bar{c}(T)\} > \bar{c}(Q).

Hence mi>cˉ(Q)m_i > \bar{c}(Q).

Step 2: farmers of SS. For i∈Si \in S, the definition of mSm_S and condition (H) give

mi≥mS≥cˉ(Q).m_i \ge m_S \ge\bar{c}(Q).

Step 3: value. The per-farmer benefit Ri(Πi)+gi(Πi)R_i(\Pi_i) + g_i(\Pi_i) is additive across farmers and depends only on farmer ii’s own portfolio, which is unchanged in Π~Q\widetilde{\Pi}_Q. The certification cost satisfies C(Q)=C(S)+C(T)−ΓC(Q) = C(S) + C(T) - \Gamma. By Lemma 3, v(S)=∑i∈S[Ri(ΠS,i∗)+gi(ΠS,i∗)]−C(S)v(S) = \sum_{i \in S}[R_i(\Pi_{S,i}^*) + g_i(\Pi_{S,i}^*)] - C(S), and likewise for TT. Therefore

v(Q)≥∑i∈Q[Ri(Π~Q,i)+gi(Π~Q,i)]−C(Q)=v(S)+v(T)+Γ.\begin{aligned} v(Q) &\ge\sum_{i \in Q}[R_i(\widetilde{\Pi}_{Q,i}) + g_i(\widetilde{\Pi}_{Q,i})] - C(Q)\\ &= v(S) + v(T) + \Gamma. \end{aligned}

Special case (a): cˉ(S)=cˉ(T)\bar{c}(S) = \bar{c}(T). Then equation (34) gives cˉ(Q)<cˉ(S)\bar{c}(Q) < \bar{c}(S). Feasibility of SS gives cˉ(S)≤mS\bar{c}(S) \le m_S, so cˉ(Q)<mS\bar{c}(Q) < m_S, which is (H).

Special case (b): (H′)(H') implies (H). Feasibility of SS gives mS≥cˉ(S)m_S \ge\bar{c}(S), hence C(S)≤A(S)mSC(S) \le A(S)m_S. Then

C(Q)=C(S)+ΔS(T)≤A(S)mS+A(T)mS=A(Q)mS,\begin{aligned} C(Q) &= C(S) + \Delta_S(T)\\ &\le A(S)m_S + A(T)m_S\\ &= A(Q)m_S, \end{aligned}

and dividing by A(Q)A(Q) gives (H).

Remark 1 (Why a condition is needed). Without (H), the conclusion can fail even when QQ is feasible. We illustrate this with a small instance. Take δ=12\delta= \frac{1}{2}, FixMRV=FixT=0\mathrm{Fix}_{\mathrm{MRV}} = \mathrm{Fix}_{T} = 0, VarMRV=VarT=10\mathrm{Var}_{\mathrm{MRV}} = \mathrm{Var}_{T} = 10, and no yield effect. There are two mutually exclusive practices: P2P_2 with OC=0\mathrm{OC} = 0 and P1P_1 with OC=0.5\mathrm{OC} = 0.5 $/ha.

  • S={F0}S = \{F_0\} with FS=100\mathrm{FS} = 100 ha and B/FS=2.5B/\mathrm{FS} = 2.5. Carbon revenue is 3 $/ha under P2P_2 and 10 $/ha under P1P_1.

  • TT consists of ten farmers with FS=0.1\mathrm{FS} = 0.1 ha each, B/FS=200B/\mathrm{FS} = 200, and carbon revenue of 1000 $/ha under either practice.

coalitionAACCcˉ\bar{c}outcome
SS1001101.1F0F_0 uses P1P_1 (θ=2.0≥1.1\theta= 2.0 \ge1.1); v=950−110=840v = 950 - 110 = 840
TT1110110all use P2P_2 (same revenue, lower cost); v=1000−110=890v = 1000 - 110 = 890
Q=S∪TQ = S \cup T101≈210.50\approx210.50≈2.084\approx2.084P1P_1 unaffordable for F0F_0 (2.0<2.0842.0 < 2.084), so F0F_0 uses P2P_2

Table 1.

In QQ, the total benefit ∑i∈Q[Ri+gi]\sum_{i\in Q}[R_i+g_i] is 300+1000=1300300+1000=1300, so

v(Q)=1300−210.50=1089.50<1730=v(S)+v(T).v(Q)=1300-210.50=1089.50<1730=v(S)+v(T).

The reason is that the many tiny farms raise the average cost per hectare and push the large farmer below the affordability threshold. Here mS=min⁡{2.0,9.5}=2.0<cˉ(Q)m_S=\min\{2.0,9.5\}=2.0<\bar{c}(Q), so (H)(H) fails.

Convexity

Definition 2 (Convexity). (F,v)(F,v) is convex if for all S⊆T⊆FS\subseteq T\subseteq F and i∉Ti\notin T,

v(T∪{i})−v(T)≥v(S∪{i})−v(S).v(T\cup\{i\})-v(T)\ge v(S\cup\{i\})-v(S).

Convexity says that a farmer’s marginal contribution never decreases as the coalition grows. Writing Δi(X)=v(X∪{i})−v(X)\Delta_i(X)=v(X\cup\{i\})-v(X), adding a farmer ii to a coalition has two effects on Δi\Delta_i:

  • a cost saving, which grows with the coalition because of the concave MRV term;

  • a benefit effect: a larger coalition lowers cˉ\bar{c} and can make costly, high-sequestration practices affordable for other members.

The benefit effect can shrink as the coalition grows. For example, a poor farmer who cannot afford a costly practice alone may be unlocked by either of two neighbors, but once one neighbor has joined, the second neighbor unlocks nothing new. This is the mechanism behind the following result.

Theorem 2 (The game is not convex in general). For δ=12\delta=\frac{1}{2} there is an instance within the model in which (F,v)(F,v) is not convex.

Proof. Instance. Let FixMRV=FixT=0\mathrm{Fix}_{\mathrm{MRV}}=\mathrm{Fix}_{T}=0, VarMRV=VarT=10\mathrm{Var}_{\mathrm{MRV}}=\mathrm{Var}_{T}=10, δ=12\delta=\frac{1}{2} and ΔY≡0\Delta Y\equiv0. There are two mutually exclusive practices:

  • P2P_2 earns 40 $/ha in carbon revenue with OC=0\mathrm{OC}=0, so its net benefit is 40 $/ha;

  • P1P_1 earns 100 $/ha with OC=10\mathrm{OC}=10 $/ha, so its net benefit is 90 $/ha.

The farmers (FSi,Bi)(\mathrm{FS}_i,B_i) are

F1=(1,25),F2=(24,1920),F3=(24,480).F_1=(1,25),\qquad F_2=(24,1920),\qquad F_3=(24,480).

By equation 36, farmer ii may use practice pp in coalition XX if and only if

BiFSi−OCp≥cˉ(X)andηp≥cˉ(X),\frac{B_i}{\mathrm{FS}_i}-\mathrm{OC}_p\ge\bar{c}(X)\quad\text{and}\quad\eta_p\ge\bar{c}(X),

where ηP1=90\eta_{P_1}=90, ηP2=40\eta_{P_2}=40.

We take

S={F1}⊂T={F1,F2},i=F3∉T.S=\{F_1\}\subset T=\{F_1,F_2\},\qquad i=F_3\notin T.

Values. Here

C(X)=10A(X)+10∣X∣.C(X)=10\sqrt{A(X)}+10|X|.

Since P1P_1 has the higher net benefit per hectare, the optimal profile is all-P1P_1 whenever that is feasible.

coalitionAACCcˉ\bar{c}optimal profilevv
S={F1}S=\{F_1\}12020F1F_1 uses P2P_2 (P1P_1: θ=25−10=15<20\theta=25-10=15<20)40−20=2040-20=20
S∪{F3}S\cup\{F_3\}25702.8all P1P_1 (θ=15\theta=15, 10≥2.810\ge2.8)2250−70=21802250-70=2180
T={F1,F2}T=\{F_1,F_2\}25702.8all P1P_1 (θ=15\theta=15, 70≥2.870\ge2.8)2250−70=21802250-70=2180
T∪{F3}T\cup\{F_3\}49100≈2.04\approx2.04all P1P_14410−100=43104410-100=4310

Table 19 (PDF p. 17).

In the last three rows, all-P1P_1 is feasible and has the largest net benefit, so it is optimal. In the first row, P1P_1 is infeasible for F1F_1, so F1F_1 must use P2P_2.

Violation.

Δ3(S)=2180−20=2160,Δ3(T)=4310−2180=2130.\Delta_3(S) = 2180 - 20 = 2160, \qquad \Delta_3(T) = 4310 - 2180 = 2130.

Since S⊆TS \subseteq T and F3∉TF_3 \notin T but Δ3(T)<Δ3(S)\Delta_3(T) < \Delta_3(S), convexity fails.

Experimental Results

This section evaluates the proposed framework through controlled computational experiments. We first describe CarbonFARM-Bench, the expert-constructed dataset on which all experiments are run, and then the experimental setup. We then present results on the structural barriers to solo participation; the effects of certification costs, farm scale, and carbon prices on coalition viability; the stability of alternative allocation mechanisms; and the resulting policy implications.

Dataset

All experiments in this paper use CarbonFARM-Bench (Fair Allocation and Revenue Maximization for Smallholders in Carbon Farming Benchmark), an expert-constructed carbon farming dataset for smallholder systems, which we release alongside this work.1 CarbonFARM-Bench was assembled by domain experts who consolidated agronomic, economic, and carbon-market evidence from the literature into a single, internally consistent benchmark for studying coalition formation among smallholder farmers. Because published sources report these quantities in heterogeneous units, regions, and contexts, expert curation was needed to reconcile them into one coherent parameter set. The full construction procedure, including parameter tables, interaction matrices, sampling distributions, and farmer panels, is described in Appendix § 12. The main components are summarized below.

  • Practice-level data. The dataset covers m=8m = 8 rice-specific and common carbon sequestration practices. Each practice is characterized by its sequestration potential, operational cost, and yield impact, with values grounded in [25, 93, 24].

  • Pairwise interaction matrices. Three symmetric 8×88 \times8 matrices (α,β,γ)(\alpha, \beta, \gamma) encode the effects of co-adopting practices on sequestration, cost, and yield, respectively. They capture agronomic synergies and antagonisms that cannot be inferred from practice-level values alone.

  • Farmer-level data. The base panel contains n=15n = 15 smallholder farmers. Farm sizes are drawn from a truncated log-normal distribution over 0.10–2.00 ha, following observed smallholder farm-size distributions [47], and budgets are positively correlated with farm size. We use n=15n = 15 because it brings out the conceptual findings most clearly. To show that these findings scale, we also report results on a larger panel of n=250n = 250 farmers.

  • Economic and MRV parameters. Carbon credit prices and the costs of monitoring, reporting, and verification (MRV) and of transactions are set to the baseline values in Table 1.

Experimental Setup

All experiments use n=15n = 15 farmers (with the base configuration parameters in Table 1) and are averaged over 10 independently sampled farmer panels. We use n=15n = 15 as an exact-computation benchmark because exhaustive evaluation of all 2n−12^n - 1 non-empty coalitions is tractable at this scale. This enables exact computation and verification of core-related properties and of all solution concepts.

SymbolDescriptionBaselineUnit
CCPCarbon credit price60$/tCO2_2e
FixMRV_{\mathrm{MRV}}Fixed cost of MRV2000$
VarMRV_{\mathrm{MRV}}Variable cost of MRV80$/haδ^{\delta}
δ\deltaMRV scale exponent0.70—
FixT_{\mathrm{T}}Fixed transaction cost1000$
VarT_{\mathrm{T}}Variable transaction cost10$/farmer
bˉ\bar{b}Mean farmer budget140$/ha

Table 1 (PDF p. 19). Baseline configuration parameters in CarbonFARM-Bench.

The numerical experiments are intended as controlled computational experiments for evaluating the structural properties of the proposed coalition model, and for obtaining empirical estimates of carbon-farming profitability under the stated assumptions.

We report the following quantities throughout. The solo value ∑iv({i})\sum_i v(\{i\}) is the non-cooperative benchmark which tells whether farmers can participate individually. The grand-coalition value v(N)v(N) is the total surplus created when everyone certifies together. It splits into two parts: one, the pooled component p(N)p(N), which is carbon revenue net of certification costs and is the transferable surplus that the allocation rules divide among members, and second, the private component G(N)G(N), which collects the agronomic gains (yield and input-cost effects) that each farmer keeps irrespective of the allocation.

The cooperative gain Δv=v(N)−∑iv({i})\Delta v = v(N) - \sum_i v(\{i\}) measures what is created by cooperating over and above what farmers can obtain alone, and its relative size Δv/v(N)\Delta v/v(N) shows how much of the surplus is due to coalition formation. Finally, since the grand coalition may itself fail to meet the certification and affordability constraints for some parameter values, we report the fraction of panels on which it certifies, so that v(N)v(N) and p(N)p(N) are always read together with the probability that the coalition exists.

Structural Barriers to Solo Participation

Table 2 varies the mean budget per hectare bˉ\bar{b} from 0 to 1000 $/ha, and Table 3 multiplies all farm areas by a factor kk while holding the budget per hectare fixed.

Insights

Budget. No farmer can certify alone at any budget level, even at more than seven times the baseline. At the baseline certification cost, a farm of under two hectares cannot generate enough carbon credits to satisfy the self-reliance constraint (equation (20)), making this constraint the primary reason that rules out solo participation. Whereas the coalition is insensitive to the budget once bˉ≥75\bar{b} \ge75 $/ha. The budget constraint (19) binds on the grand coalition only at very small budgets; when bˉ≤25\bar{b} \le25, it fails to certify in 10% of panels.

Farm scale. Solo feasibility emerges only once the average farm size grows severalfold. The grand coalition creates more value than the sum of solo values at every scale, but the relative gain(Δv/v(N)\Delta v/v(N)) shrinks: It falls from 100% for k≤2k \le2 to 26% at k=20k = 20. Conversely, halving farm sizes makes even the grand coalition unviable in 80% of panels. Coalition formation is thus most valuable where farms are smallest, yet its gain persists even for large farms.

Together, Table 2 and Table 3 point to a structural barrier: budget increases do not relax the binding constraint on solo farmers, farm size does. This gives a clear economic rationale for coalition formation as an institutional necessity for smallholders.

bˉ\bar{b} ($/ha) | Solo-feasible | ∑iv({i})\sum_i v(\{i\}) | v(N)v(N) | p(N)p(N) | Grand feasible
0005,6001,3560.90
25005,6001,3560.90
50005,9181,3451.00
75005,9291,3581.00
140*005,9291,3581.00
400005,9291,3581.00
1000005,9291,3581.00

Table 2. Effect of the mean budget bˉ\bar{b}, * represents the base configuration parameter value.

kkAvg. size (ha)Solo-feasible∑iv({i})\sum_i v(\{i\})v(N)v(N)Δv\Delta vGrand feasible
0.50.34004714710.20
1∗1^*0.67005,9295,9291.00
21.340015,15915,1591.00
32.010.254224,42623,8831.00
53.353.09,38943,01433,6261.00
106.708.146,35389,64343,2901.00
2013.4012.4135,822183,19147,3691.00

Table 3. Effect of farm scale kk (areas multiplied by kk), ∗^* represents the base configuration parameter value.

Effect of Certification Costs

ParameterValuev(N)v(N)p(N)p(N)Grand feasible
FixMRV_{\mathrm{MRV}} ($) | 0 | 7,929 | 3,358 | 1.00
10006,9292,3581.00
2000∗2000^*5,9291,3581.00
30004,0865730.70
40002,3211120.40
6000000.00
FixT_{\mathrm{T}} ($) | 0 | 6,929 | 2,358 | 1.00
1000∗1000^*5,9291,3581.00
20004,0865730.70
40001,580−144-1440.30
6000000.00
VarMRV_{\mathrm{MRV}} (/haδ/ha^\delta)06,3301,7591.00
80∗80^*5,9291,3581.00
1605,5199461.00
2404,8445980.90
VarT_{\mathrm{T}} ($/farmer) | 0 | 6,079 | 1,508 | 1.00
10∗10^*5,9291,3581.00
405,1959510.90
804,0515380.70
1602,199−13-130.40
320000.00

Table 4. Effect of certification-cost parameters, ∗^* represents the base configuration parameter value.

Insights

Fixed costs. Around the baseline, v(N)v(N) moves one-for-one with both FixMRV\mathrm{Fix}_{\mathrm{MRV}} and FixT\mathrm{Fix}_{\mathrm{T}} (each $1000 change in either shifts v(N)v(N) and p(N)p(N) by $1000). Above the baseline the effect becomes non-linear because the affordability constraints start to bind and also effects the grand coalition feasibility.

Variable MRV. Tripling VarMRV\mathrm{Var}_{\mathrm{MRV}} from 80 to 240 lowers v(N)v(N) by 18% and begins to threaten feasibility. Whereas setting VarMRV=0\mathrm{Var}_{\mathrm{MRV}} = 0 raises v(N)v(N) by 7%.

Variable transaction cost. Around the baseline, each extra dollar of VarT\mathrm{Var}_{\mathrm{T}} lowers v(N)v(N) by $15, one dollar for each of the 15 members. Because this cost is paid per farmer, it does not dilute as the coalition grows, and feasibility starts to erode at VarT≥40\mathrm{Var}_{\mathrm{T}} \ge40.

Together we observe that fixed costs and the variable transaction cost VarT\mathrm{Var}_{\mathrm{T}} drive coalition viability, while VarMRV\mathrm{Var}_{\mathrm{MRV}} is the mildest lever. Subsidizing fixed and per-farmer transaction costs is the most effective way to keep certification viable for smallholder coalitions.

Effect of the MRV Scale Exponent

δ\deltav(N)v(N)p(N)p(N)
0.06,2501,679
0.36,1711,600
0.56,0781,506
0.7*5,9291,358
0.95,6931,121
1.05,526955

Table 5. Effect of the MRV scale exponent δ\delta, * represents the base configuration parameter value.

Insights

As δ\delta approaches 1, the economies of scale in monitoring disappear, and the value of the grand coalition decreases monotonically. Even at δ=1\delta= 1, the grand coalition retains most of its value, since pooling still removes duplicated fixed costs. The effect is much larger on the transferable surplus p(N)p(N) that is actually divided among farmers, which falls by 43% over the same range. Preserving sub-linear MRV costs therefore matters most when the pooled surplus is thin. This makes a case for public support for MRV technologies and institutions, such as shared remote-sensing infrastructure, that keep monitoring costs sublinear.

Effect of the Number of Farmers

nnv(N)v(N)v(N)/nv(N)/np(N)/np(N)/nCert. cost/haSolo-feasible
50———0
102,46424627432.20
155,92939591365.80
2511,775471168236.50
5027,702554243131.50
10058,53358527781.90
250149,09559629450.30
500307,75861630837.60
1000615,55361631030.70

Table 6. Effect of Coalition Size.

Insights

As the number of farmers increases the certification cost per hectare decreases ($432 at n=10n = 10 to $31 at n=1000n = 1000), whereas the total and the per farmer surplus rise and then flatten: v(N)/nv(N)/n goes from $246 (n=10n = 10) to $554 (n=50n = 50), $596 (n=250n = 250) and $616 (n=1000n = 1000). Most of the scale benefit is thus realised with a few dozen to a few hundred members, which is consistent with the size of typical farmer producer organisations.

Effect of the Carbon Credit Price

CCP ($/tCO2_2e) | v(N)v(N) | p(N)p(N)
101,842−2,750-2{,}750
303,475−1,096-1{,}096
505,111540
60*5,9291,358
1009,2024,631
15013,2938,721
20017,38312,812
30025,56520,993
50041,92737,356

Table 7. Effect of the carbon credit price, * represents the base configuration parameter value.

Insights

The total surplus v(N)v(N) rises roughly linearly with the price. The pooled surplus p(N)p(N) becomes positive only above about $43/tCO2_2e. Below that price p(N)<0p(N) < 0: carbon revenue does not cover certification, and the coalition is kept feasible by the farmers’ private agronomic gains G(N)G(N). The feasibility of the grand coalition at low prices should therefore be read with care as it can be driven by yield and cost co-benefits of the practices as much as by carbon revenue. A programme that relies on carbon income alone would need a price of roughly $43 or more at baseline costs.

Comparison of Revenue Allocation Mechanisms

We compare six allocation mechanisms: the Shapley value, the core (represented by its centroid, estimated by hit-and-run sampling), the least core, the nucleolus, and the two baselines, namely the area-proportional split and the attributable-revenue split. Allocations are reported as final surplus ui=xi+gi(ΠN)u_i = x_i + g_i(\Pi_N). Table 8 reports stability at the baseline, table reports 9 the final payoffs for one panel and its equity indicators.

Table 9. Final payoffs ui=xi+giu_i = x_i + g_i ($) for the panel with seed 42, sorted by farm size. The last row gives the Gini coefficient of uu under each mechanism for this panel.

FarmerSize (ha)Avg CMPShapleyCoreLeast coreNucleolusArea-prop.Attrib. rev.
F10.161.7786.866.961.561.5101.191.0
F20.211.78130.4101.4128.895.2132.7145.2
F30.261.88165.2134.1179.7120.6164.3182.9
F40.262.03156.4130.9177.9119.7164.3181.2
F50.441.90272.7229.5328.1194.8278.0284.2
F60.601.97391.4393.1467.8329.2379.1436.3
F70.652.00415.7380.5490.5339.1410.7433.1
F80.691.76424.4361.5508.7340.0436.0423.6
F90.821.71522.1488.8567.9482.0518.1531.5
F100.971.93621.3607.2636.1617.7612.9627.8
F111.281.73814.0844.2777.1873.6808.8802.7
F121.321.95840.6861.4795.2905.9834.0824.5
F131.461.71948.4993.1858.91,055.6922.5937.5
F141.551.68983.51,100.8899.81,112.2979.4970.7
F151.871.781,150.81,230.11,045.41,276.41,181.61,051.3
Gini(uu)0.3470.3850.3090.4070.3260.333

Table 10. Superadditivity and convexity checked over all coalitions (n=15n = 15).

Panel seedv(N)v(N)Feasible coalitionsSuperadd. violationsConvexity violations
427,92317,610 / 32,767013.65% (234,854)
436,2216,559 / 32,76706.89% (118,484)
448,57719,824 / 32,767015.55% (267,506)
455,3774,760 / 32,76705.51% (94,778)
463,8211,216 / 32,76701.28% (22,028)
474,318779 / 32,76700.80% (13,766)
488,39417,278 / 32,767013.46% (231,576)
496,31911,341 / 32,767011.14% (191,568)
505,0524,075 / 32,76704.24% (72,984)
513,29187 / 32,76700.05% (916)

Insights

Stability. On all 10 baseline panels the core is non-empty and in fact large, the least-core slack is ε∗=−74.6\varepsilon^* = -74.6 on average. Consequently all six allocation rules are stable on every panel, with no blocking coalition and no individual-rationality violation. Also on average the worst-off farmer gains $74.6 over her solo outcome. The size of the core is consistent with strong economies of scale: most proper subgroups cannot certify at all, so no subgroup can come close to what the grand coalition delivers.

Allocation. The least core is the most egalitarian of the core-based rules and shifts surplus towards small farms relative to the nucleolus and the core centroid. Its Gini coefficient of final surplus is 0.31, against 0.41 for the nucleolus and 0.39 for the core centroid. The nucleolus and the core centroid concentrate surplus on the largest farms.

Superadditivity and Convexity. We observe no violation of superadditivity is found in the panels of study. Whereas non-convexity is found across all the panels. The violation rate rises with the number of feasible coalitions (from 916 violations when only 87 coalitions are feasible to over 267,000 when 19,824 are), so non-convexity is most pronounced in panels where the certification threshold is crossed by many coalitions.

When Do Allocation Rules Become Unstable?

At the baseline, all six allocation rules are stable (Table 8). This is a property of the baseline parameters as no coalition can credibly threaten to leave. We therefore ask under what conditions stability breaks down. By Proposition 1, an allocation xx is blocked by a group T⊊FT \subsetneq F if and only if v^(T)>∑i∈Txi\hat{v}(T) > \sum_{i \in T} x_i, so we count the coalitions that satisfy this inequality under each rule.

MechanismStable (frac.)ViolationsMean max excess ($) | Mean min gain ($)
Shapley1.000−53.6-53.685.5
Core (centroid)1.000−70.2-70.278.6
Least core1.000−74.6-74.674.6
Nucleolus1.000−74.6-74.674.6
Area-proportional1.000−43.9-43.9101.8
Attributable-revenue1.000−47.1-47.1100.5

Table 8. Stability at the baseline. Max excess is max⁡T[v^(T)−x(T)]\max_T[\hat{v}(T)-x(T)] (negative implies stability); min gain is the smallest improvement over solo.

We vary two parameters that govern how strongly a subgroup can outperform its share of the grand coalition. The first is the carbon credit price CCP, which scales the pooled revenue R(S,ΠS)R(S,\Pi^S) relative to the certification cost C(S)C(S). The second is the farmer-specific heterogeneity ν\nu in carbon mitigation potential, where

CMPij=CMPj(1+U(−ν,ν)),\mathrm{CMP}_{ij} = \mathrm{CMP}_{j}\left(1 + U(-\nu,\nu)\right),

so that a larger ν\nu means the carbon yield per hectare is more unequal across farms. We compare the area-proportional split and the Shapley value. These are the two rules that do not enforce the stability constraints equation (28) by construction.

CCPν\nuArea-proportional # blockArea-proportional stableShapley # blockShapley stable
600.2001.0001.00
600.6001.0001.00
600.950.90.7001.00
1500.2001.0001.00
1500.6025.00.5001.00
1500.95256.70.102.80.50
3000.201.30.7001.00
3000.60524.00.0014.90.10
3000.951,278.70.00184.60.00

Table 11. Instability under CMP heterogeneity ν\nu and carbon price. “# block” is the mean number of blocking coalitions; “stable” is the fraction of panels with no blocking coalition.

Insights

Area-proportional split. The area-proportional rule is the first to fail, and it fails as either heterogeneity or price rises. Since xiPSx_i^{\mathrm{PS}} depends only on farm size equation (32), it pays every hectare the same amount regardless of how much carbon that hectare generates. It thus over-rewards low-CMP land and under-rewards high-CMP land, which gives subgroups with above-average CMP a reason to leave. At CCP =60= 60 the rule is stable on all panels up to ν=0.6\nu= 0.6. At CCP =150= 150 it is stable on only half of the panels at ν=0.6\nu= 0.6, and on only 10% at ν=0.95\nu= 0.95. At CCP =300= 300 it is already unstable on 30% of panels at ν=0.2\nu= 0.2, and on every panel once ν≥0.6\nu\ge0.6. Raising the price enlarges the stakes of misallocation, because a larger share of v(N)v(N) is pooled carbon revenue.

Shapley value. The Shapley value is considerably more robust, because it rewards marginal contribution and therefore partly tracks carbon heterogeneity, but it is not immune. It remains stable on every panel at the baseline price for all ν\nu, and at CCP =150= 150 for ν≤0.6\nu\le0.6. It loses core membership when high prices coincide with high heterogeneity: at (CCP,ν)=(150,0.95)(\mathrm{CCP},\nu) = (150,0.95) it is stable on 50% of panels, and at (300,0.6)(300,0.6) on only 10%. This is consistent with the non-convexity result of Theorem 2, since the Shapley value of a non-convex game need not lie in the core.

Structure of blocking coalitions. In the panels we inspected, the most violated coalitions are very large, typically 12–14 of the 15 farmers holding roughly 84–96% of the total area, and have a slightly higher mean CMP than the panel average. Instability therefore takes the form of a high-carbon majority leaving one or two low-carbon farmers behind.

These results show that a rule which ignores carbon heterogeneity is fragile exactly where it over-pays the farmers who contribute least. Simple rules are adequate when farms are similar and prices are modest, but as ν\nu or CCP rises, allocations must be verified against the stability constraints equation (28), and core-based rules such as the least core or nucleolus are the safer default.

Policy Implications

Our findings show that smallholder farmers cannot derive the benefits of carbon farming on their own, because certification (MRV) and transaction costs are too high relative to the credits their farms can generate. Support aimed at individual farmers, such as credit or cash transfers, is therefore unlikely to open participation, whereas aggregation does. Governments could promote farmer collectives to enable group participation. Introducing carbon farming as an additional activity within existing farmer producer organisations may substantially reduce the institutional cost of coalition formation by leveraging structures that already exist, since groups of a few dozen to a few hundred members are already large enough to capture most of the scale benefit. It may also provide a supplementary income stream for member farmers and give governments a ready-made channel for smallholder participation in carbon markets.

Public support is most effective when it targets the costs that decide whether a group can certify at all. Fixed MRV and fixed transaction costs are the main entry barrier, and initial financial support, such as grants, subsidies, and low-interest loans for one-time registry, baseline, and auditing costs, will help smallholder collectives overcome it. Per-farmer administrative costs also deserve attention, as they do not dilute as a coalition grows, and standardised digital enrolment can keep them low. Government support for shared digital MRV systems, through remote-sensing-based tools, automated reporting tools, AI-based techniques, and standardised protocols, can significantly reduce costs and encourage farmers to adopt carbon farming. Keeping monitoring costs sub-linear in farm area matters most when the surplus available to share is thin.

The results also highlight the importance of principled revenue sharing within groups. Simple rules, such as an area-proportional split, are adequate when farms are similar and carbon prices are modest, but they become fragile as carbon heterogeneity or prices rise, and the Shapley value follows. Core-based rules, such as the least core and the nucleolus, remain stable whenever any stable allocation exists and could serve as the default for heterogeneous groups or higher-price settings. Revenue-sharing mechanisms should strike a balance between stability and fairness, ensuring that all farmers have an incentive to remain in the coalition.

Conclusion and Future Work

This paper presented a cooperative game-theoretic framework for enabling smallholder participation in carbon farming through coalition formation. It links three stages: solo portfolio optimization, joint portfolio optimization, and surplus allocation. The framework respects the fact that farmers pool only certification costs and carbon revenue, and keep their agronomic costs and yields private. Coalition formation offers smallholders a channel to additional income and the sustainability co-benefits of carbon farming.

Our results show that individual participation becomes economically infeasible due to farm size and fixed certification costs, while participating in a coalition of farmers generates non-trivial value through economies of scale. We have demonstrated that coalition value is structurally driven by MRV cost dynamics rather than by available budget levels alone. While land size is a natural and important criterion for allocation, value in carbon farming is created not merely by owning more land but by combining land, technology, infrastructure (MRV), and management.

On the structure of the game, we proved that it is superadditive under an affordability condition on merging coalitions. We also showed by construction that it is non-convex in general. The Shapley value is therefore not guaranteed to lie in the core, and the core itself may be empty in general. We compared the Shapley value, the core centroid, the least core, and the nucleolus against area-proportional and attributable-revenue baselines. At baseline, the core was non-empty and large, and all six rules were stable. Stability broke down as carbon heterogeneity and carbon prices rose. Among the core-based rules, the least core was the most egalitarian, while the nucleolus and core centroid concentrated surplus on larger farms. Stability is thus the property that matters most for sustained participation. The Shapley value is a sound choice when it lies in the core, and the least core or nucleolus is the safer default for heterogeneous groups or higher-price settings.

Overall, the study establishes that carbon farming for smallholders is fundamentally an institutional design problem, requiring aggregation, cost-sharing flexibility, and stability-aware incentive mechanisms for scalable and equitable adoption. Our study yields several policy recommendations for the successful adoption of carbon farming practices by smallholder farmers.

Future Work

Several promising directions emerge naturally from this work. First, while we have used expert-constructed data for clear reasons, empirical validation using real-world datasets from farmer cooperatives or pilot carbon farming projects would significantly strengthen the model’s practical relevance. Second, extending the model to dynamic, multi-season settings would allow the study of trust, learning, and coalition stability over time. Third, incorporating uncertainty in carbon prices, yields, and sequestration could enable analysis of risk-sharing mechanisms and robust coalition design.

Fourth, our exact computations rely on enumerating all 2n2^n coalitions, which limits the allocation analysis to small panels. Scaling to realistic group sizes calls for sampling-based approximations of the Shapley value, least core, and nucleolus, with probabilistic stability guarantees. Fifth, we studied only the grand coalition, which is justified under our affordability condition but not unconditionally, since superadditivity can fail when coalitions are infeasible or merging raises some members’ outlays. Examining alternative coalition structures, such as clustered or regional groupings, and cost-sharing rules that guarantee superadditivity would give a fuller picture. Sixth, future work should develop simple, transparent allocation rules that approximate core stability and are easy to deploy in practice. Finally, we assumed that farmers are honest and implement the coalition-selected portfolios, which is enforced by contract or MRV. Relaxing this assumption, including strategic misreporting and portfolio deviations, would motivate mechanism-design approaches to the problem.

Acknowledgments

The authors thank Prof. Soma Dhawala, Prof. Lalith Achoth, Prof. Vedamurthy, Mr. Ravi Trivedi and Dr. Soumya K for their valuable suggestions and agronomic insights. We have utilized various AI language models and assistants, such as GPT and Claude, to enhance the writing quality and provide summaries of source materials. However, the authors diligently verified all information provided by these AI tools to ensure accuracy and factual integrity.

Notation Summary

  • F={1,2,…,n}F = \{1,2,\ldots,n\}: set of farmers.

  • P={1,2,…,m}P = \{1,2,\ldots,m\}: set of carbon sequestration practices.

  • FiF_i: farmer ii.

  • PjP_j: practice jj.

  • S,T⊆NS,T \subseteq N: coalitions of farmers.

  • ∣S∣|S|: size of coalition SS.

  • FSi\mathrm{FS}_i: farm size of farmer ii (ha).

  • BiB_i: seasonal budget of farmer ii ($).

  • ha: hectare.

  • tCO2e\mathrm{tCO}_{2}\mathrm{e}: tonne of CO2\mathrm{CO}_{2} equivalent.

  • CCP\mathrm{CCP}: carbon credit price ($/tCO2_2e).

  • OCj\mathrm{OC}_j: per-hectare operational cost of practice jj.

  • CMPj\mathrm{CMP}_j: baseline additionality-adjusted per-hectare sequestration of practice jj.

  • CMPij\mathrm{CMP}_{ij}: farmer-specific additionality-adjusted per-hectare sequestration of practice jj for farmer ii.

  • YjY_j: per-hectare yield revenue effect of practice jj.

  • Π⊆P\Pi\subseteq P: portfolio of practices.

  • Πiso\Pi_i^{\mathrm{so}}: optimal solo portfolio of farmer ii.

  • Πico\Pi_i^{\mathrm{co}}: coalition portfolio adopted by farmer ii.

  • ΠS={Πico}i∈S\Pi_S = \{\Pi_i^{\mathrm{co}}\}_{i \in S}: coalition practice profile.

  • αjk\alpha_{jk}: pairwise interaction effect of practices jj and kk on sequestration.

  • βjk\beta_{jk}: pairwise interaction effect of practices jj and kk on operational cost.

  • γjk\gamma_{jk}: pairwise interaction effect of practices jj and kk on yield.

  • CMRV(S,ΠS)C_{\mathrm{MRV}}(S,\Pi_S): MRV cost of coalition SS.

  • CT(S)C_T(S): transaction cost of coalition SS.

  • FixMRV,VarMRV\mathrm{Fix}_{\mathrm{MRV}},\mathrm{Var}_{\mathrm{MRV}}: fixed and variable MRV cost parameters.

  • FixT,VarT\mathrm{Fix}_T,\mathrm{Var}_T: fixed and variable transaction-cost parameters.

  • δ∈(0,1)\delta\in(0,1): MRV scale exponent.

  • v~({i})\tilde{v}(\{i\}): optimized standalone net value of farmer ii.

  • v(S)v(S): characteristic-function value of coalition SS.

  • v^({i})\hat{v}(\{i\}): adjusted solo value used in allocation constraints.

  • x=(xi)i∈S\mathbf{x} = (x_i)_{i \in S}: allocation vector for coalition SS.

  • xix_i: share of allocable carbon-market surplus assigned to farmer ii.

  • e(S,x)e(S,\mathbf{x}): excess of coalition SS under allocation x\mathbf{x}.

  • ε\varepsilon: least-core relaxation parameter.

  • ϕi(v)\phi_i(v): Shapley value of farmer ii.

Dataset

This section documents the construction, calibration, and verification of the CarbonFARM-Bench dataset used in the coalition formation experiments in the main paper, covering the numerical parameters, modeling assumptions, and data construction procedures adopted.

Section B.2 derives the three base parameters (Carbon Mitigation Potential (CMP), Net Operating Cost, and Yield Change) for each of the eight carbon farming practices. Section B.3 explains how combining two practices can produce effects that are different from just adding them up, and how each pair is classified as having a positive, negative, or negligible interaction on carbon, cost, and yield. Section B.4 documents the MRV and transaction costs. Section B.5 explains the synthesis of farm sizes and operating budgets for the farmer groups.

Carbon Practices

We consider the following eight carbon practices for our study. These practices are a combination of rice-specific and crop-agnostic practices. It should be noted that this list is agnostic and can be replaced with any set of practices.

  • Crop Residue Retention (CRR)

  • Reduced / Minimum Tillage (RMT)

  • Alternate Wetting and Drying (AWD)

  • System of Rice Intensification (SRI)

  • Direct Seeded Rice (DSR)

  • Integrated Nutrient Management (INM)

  • Manure Application (MA)

  • Biochar Application (BA)

Each practice jj is characterized by three literature-derived base parameters: a carbon mitigation potential CMPj\mathrm{CMP}_{j} (tCO2e ha−1 yr−1\mathrm{tCO}_{2}\mathrm{e}\,\mathrm{ha}^{-1}\,\mathrm{yr}^{-1}), a net incremental operating cost ΔOCj\Delta OC_{j} (USD ha−1\mathrm{ha}^{-1}, PPP-adjusted), and a percentage change in grain yield YjY_{j} relative to conventional cultivation. When practices are adopted jointly, the model additionally applies pairwise interaction terms αjk\alpha_{jk} (carbon), βjk\beta_{jk} (cost), and γjk\gamma_{jk} (yield), plus a binary hard incompatibility indicator Δjk\Delta_{jk}, so that joint outcomes need not equal the naive sum of individual effects. Coalition-level enrollment additionally incurs MRV and transaction costs that scale with enrolled area and farmer count.

Base Parameter Derivation

Carbon Mitigation Potential (CMP)

Carbon Mitigation Potential (CMP) quantifies the annual climate mitigation benefit achieved by adopting a carbon farming practice relative to conventional continuously flooded, puddled transplanted rice cultivation. The mitigation benefit arises through two complementary mechanisms:

  1. reduction in greenhouse gas (GHG) emissions resulting from improved crop and water management; and

  1. additional soil organic carbon (SOC) stored in the soil.

These two mitigation pathways are integrated into a single CO2\mathrm{CO}_2-equivalent metric, referred to as Carbon Mitigation Potential (CMP), which forms the basis for evaluating the climate benefit of each practice throughout this study.

The CMP of practice jj is therefore defined as

CMPj=E0rj+4412ΔSOCj,(37)\mathrm{CMP}_j = E_0 r_j + \frac{44}{12}\Delta SOC_j, \tag*{(37)}

where

  • E0=7.8E_0 = 7.8 tCO2\mathrm{CO}_2e ha−1^{-1} yr−1^{-1} denotes the baseline annual GHG emission from conventional rice cultivation [68]. This values varies based on the crop.

  • rjr_j denotes the fractional reduction in Global Warming Potential (GWP) attributable to practice jj, where positive values represent lower GHG emissions relative to conventional management;

  • ΔSOCj\Delta SOC_j denotes the additional annual SOC sequestration (Mg C ha−1^{-1} yr−1^{-1}) associated with practice jj;

  • the factor 44/12=3.66744/12 = 3.667 converts stored carbon into its CO2\mathrm{CO}_2 equivalent following the IPCC Guidelines (Intergovernmental Panel on Climate Change (IPCC), 2019).

The formulation consists of two additive components corresponding to the mitigation pathways introduced above,

E0rj⏟Avoided GHG emissions+3.667ΔSOCj⏟SOC sequestration,\underbrace{E_0r_j}_{\text{Avoided GHG emissions}} + \underbrace{3.667\Delta SOC_j}_{\text{SOC sequestration}},

allowing avoided emissions and soil carbon sequestration to be evaluated independently. This distinction is important because some practices increase SOC while simultaneously increasing methane emissions under flooded conditions. Combining these effects into a single empirical value would obscure the individual contribution of each mitigation pathway.

Figure 1 summarizes the procedure used to estimate CMP for each carbon farming practice. Literature-derived evidence is first used to determine the GWP reduction (rjr_j) and SOC sequestration (ΔSOCj\Delta SOC_j) associated with each practice. These parameters are then substituted into (37) to obtain the benchmark CMP values used throughout the optimization framework.

Workflow for estimating Carbon Mitigation Potential (CMP) for each carbon farming practice

Figure 1. Workflow for estimating Carbon Mitigation Potential (CMP) for each carbon farming practice.

Biochar represents a special case within this framework because its carbon benefit originates primarily from a one-time addition of stable carbon rather than a continuous annual increase in native SOC. Consequently, ΔOCBC\Delta OC_{\mathrm{BC}} is treated as an annualized stable-carbon credit calculated over a specified accounting horizon (Section B.2.1).

Applying the above methodology yields the CMP parameter values summarized in Table 12. The scientific rationale, parameter selection, and detailed derivation for each practice are presented in the following subsections.

Practicerjr_jΔSOCj\Delta SOC_j Mg C ha−1^{-1} yr−1^{-1}CMP tCO2_2e ha−1^{-1} yr−1^{-1}
Crop Residue Retention−0.050-0.0500.270.270.600.60
Reduced / Minimum Tillage0.1500.1500.000.001.171.17
Alternate Wetting and Drying0.36840.36840.000.002.872.87
System of Rice Intensification0.2800.2800.000.002.182.18
Direct Seeded Rice0.3700.3700.000.002.892.89
Integrated Nutrient Management0.0000.0000.500.501.831.83
Manure Application0.0000.0000.460.461.691.69
Biochar Application0.0860.0860.42∗0.42^{*}2.202.20

Table 12. Literature-informed carbon mitigation parameters used in the benchmark dataset.

Crop Residue Retention

Crop residue retention returns organic biomass to agricultural soils, increasing soil organic carbon (SOC). Under flooded rice cultivation, however, decomposition of incorporated residues can stimulate methane production because of enhanced anaerobic microbial activity. The practice therefore affects both SOC sequestration and greenhouse gas emissions at the same time [101] (Lu et al. 2015).

Parameter Selection.

Based on reported SOC sequestration rates, an annual SOC gain of

ΔSOCCRR=0.27 Mg C ha−1 yr−1\Delta SOC_{\mathrm{CRR}} = 0.27\ \mathrm{Mg\ C\ ha^{-1}\ yr^{-1}}

is adopted.

To account for the increased methane emissions reported in flooded paddy systems, a conservative GWP penalty of

rCRR=−0.05r_{\mathrm{CRR}} = -0.05

is assigned.

The GWP penalty is treated as a modeling assumption informed by the literature rather than a value directly reported by any single study.

CMP Calculation.

Substituting the selected parameters into Equation (37),

CMPCRR=7.8⋅(−0.05)+3.667⋅(0.27)=0.60.\mathrm{CMP}_{\mathrm{CRR}} = 7.8 \cdot(-0.05) + 3.667 \cdot(0.27) = 0.60.

Interpretation. Crop residue retention provides a positive overall mitigation benefit because increased SOC sequestration outweighs the assumed increase in methane emissions, giving the practice a moderate CMP value of 0.60 tCO2_2e ha−1^{-1} yr−1^{-1}.

Reduced / Minimum Tillage

Reduced or minimum tillage decreases soil disturbance and machinery operations, lowering fossil-fuel consumption and greenhouse gas emissions compared with conventional tillage. Literature syntheses consistently report reduced GWP under conservation tillage in rice-based production systems [30].

Parameter Selection.

A representative GWP reduction of

rRMT=0.15r_{\mathrm{RMT}} = 0.15

is adopted directly from the reported literature synthesis.

No independent SOC contribution is assigned,

ΔSOCRMT=0,\Delta SOC_{\mathrm{RMT}} = 0,

because SOC gains reported under conservation agriculture are often jointly attributed to reduced tillage and residue retention together. Assigning an additional SOC benefit here would therefore lead to double counting.

CMP Calculation.

Using Equation (7),

CMPRMT=7.8⋅(0.15)=1.17.CMP_{\mathrm{RMT}} = 7.8 \cdot(0.15) = 1.17.

Interpretation.

The mitigation benefit of reduced tillage arises entirely from avoided greenhouse gas emissions. Since no independent SOC credit is assigned, the resulting CMP reflects only the reduction in GWP.

Alternate Wetting and Drying (AWD)

Alternate Wetting and Drying (AWD) periodically drains paddy fields, interrupting the anaerobic soil conditions responsible for methane production. As a result, AWD is widely recognized as one of the most effective water-management practices for reducing methane emissions from irrigated rice cultivation.

Parameter Selection.

A recent global meta-analysis reports approximately 43.2% lower methane emissions and 36.8% lower overall GWP under AWD relative to continuously flooded rice systems [43]. Accordingly,

rAWD=0.3684.r_{\mathrm{AWD}} = 0.3684.

Because the principal mitigation pathway of AWD is methane reduction, no additional SOC sequestration benefit is assigned,

ΔSOCAWD=0.\Delta SOC_{\mathrm{AWD}} = 0.

CMP Calculation.

Applying Equation (37),

CMPAWD=7.8⋅(0.3684)=2.87.CMP_{\mathrm{AWD}} = 7.8 \cdot(0.3684) = 2.87.

Interpretation. AWD achieves one of the highest CMP values among the evaluated practices because its mitigation benefit is dominated by substantial reductions in methane emissions rather than by SOC sequestration.

System of Rice Intensification (SRI)

The System of Rice Intensification (SRI) integrates improved crop, water, and soil management practices to enhance resource-use efficiency. Compared with conventional puddled transplanted rice, intermittent irrigation under SRI reduces the duration of anaerobic conditions, lowering methane emissions while maintaining or improving crop productivity.

Parameter Selection.

Field experiments conducted in the Indo-Gangetic Plains report approximately 28% lower net GWP under SRI compared with conventional rice cultivation [39]. Since rjr_j represents the fractional reduction in GWP, the reported value is directly adopted as

rSRI=0.28.r_{\mathrm{SRI}} = 0.28.

No independent SOC sequestration benefit is assigned,

ΔSOCSRI=0,\Delta SOC_{\mathrm{SRI}} = 0,

because the reported mitigation primarily arises from improved water management rather than measurable increases in soil carbon storage.

CMP Calculation.

Applying Equation (37),

CMPSRI=7.8⋅(0.28)=2.18.CMP_{\mathrm{SRI}} = 7.8 \cdot(0.28) = 2.18.

Interpretation.

The mitigation benefit of SRI is driven almost entirely by reduced GHG emissions, so SRI exhibits one of the highest CMP values among the considered practices despite receiving no additional SOC credit.

Direct Seeded Rice (DSR)

Direct Seeded Rice (DSR) eliminates puddling and transplanting, reducing the duration of flooded conditions during crop establishment. This substantially suppresses methane production while also reducing labour and water requirements compared with conventional transplanted rice.

Parameter Selection.

A recent meta-analysis reports approximately 69.9% lower methane emissions and approximately 37% lower overall GWP under DSR relative to puddled transplanted rice [73]. Accordingly,

rDSR=0.37.r_{\mathrm{DSR}} = 0.37.

No separate SOC benefit is assigned,

ΔSOCDSR=0,\Delta SOC_{\mathrm{DSR}} = 0,

because the reported climate benefit originates primarily from reduced methane emissions rather than enhanced soil carbon sequestration.

CMP Calculation.

Using Equation (37),

CMPDSR=7.8⋅(0.37)=2.89.CMP_{\mathrm{DSR}} = 7.8 \cdot(0.37) = 2.89.

Interpretation. Among all evaluated practices, DSR provides one of the largest climate mitigation benefits because its substantial reduction in methane emissions directly translates into a high CMP value.

Integrated Nutrient Management (INM)

Integrated Nutrient Management (INM) combines inorganic fertilizers with organic nutrient sources such as farmyard manure or compost. Long-term application of integrated nutrient management enhances soil organic matter inputs, improves soil health, and promotes gradual SOC accumulation.

Parameter Selection.

Long-term rice-based experiments consistently report increased SOC under integrated nutrient management [55, 70]. Based on these studies, an annual SOC sequestration rate of

ΔSOCINM=0.50 Mg C ha−1 yr−1\Delta SOC_{\mathrm{INM}} = 0.50\ \mathrm{Mg\ C\ ha^{-1}\ yr^{-1}}

is adopted.

The net GWP response of INM varies considerably depending on the type of organic amendment, fertilizer application rate, and irrigation regime. No universal reduction in GWP is therefore imposed, and

rINM=0.r_{\mathrm{INM}} = 0.

CMP Calculation.

Substituting these values into Equation (37),

CMPINM=3.667⋅(0.50)=1.83.\mathrm{CMP}_{\mathrm{INM}} = 3.667 \cdot(0.50) = 1.83.

Interpretation.

The mitigation benefit of INM is attributed entirely to increased SOC sequestration. Although no direct GHG reduction is assumed, long-term carbon storage results in a substantial CMP value.

Manure Application

Application of farmyard manure directly adds organic carbon to the soil, increasing soil organic matter and promoting long-term SOC sequestration. Numerous long-term rice–wheat experiments have demonstrated sustained increases in SOC following repeated manure application [13].

Parameter Selection.

Based on reported long-term sequestration rates, the annual SOC gain is taken as

ΔSOCmanure=0.46 Mg C ha−1 yr−1.\Delta SOC_{\mathrm{manure}} = 0.46\ \mathrm{Mg\ C\ ha^{-1}\ yr^{-1}}.

Although manure improves soil fertility, it can also stimulate methane production under flooded rice cultivation. Since the overall GWP response depends strongly on field conditions and manure management, no independent GWP reduction is assumed:

rmanure=0.r_{\mathrm{manure}} = 0.

CMP Calculation.

Applying Equation (37),

CMPmanure=3.667⋅(0.46)=1.69.\mathrm{CMP}_{\mathrm{manure}} = 3.667 \cdot(0.46) = 1.69.

Interpretation.

The climate benefit of manure application is therefore derived solely from additional SOC sequestration. Any potential reduction in GHG emissions is conservatively excluded to avoid overstating the mitigation benefit.

Biochar Application

Biochar contributes to climate change mitigation through two distinct mechanisms. First, biochar amendments can reduce greenhouse gas emissions from rice cultivation by improving soil physical and chemical properties, suppressing methane production. Second, biochar contains highly stable carbon that remains in the soil for many years, providing a long-term carbon storage benefit. Biochar is therefore the only practice considered in this study whose CMP includes both an avoided-emissions component and a stable-carbon sequestration component.

Parameter Selection.

A recent global meta-analysis of rice production systems reports an average reduction of approximately 8.6% in Global Warming Potential (GWP) following biochar application [100]. Since rjr_j represents the fractional reduction in GWP, the reported value is adopted directly as

rBC=0.086.r_{\mathrm{BC}} = 0.086.

The corresponding avoided-emissions component is therefore

E0rBC=7.8⋅(0.086)=0.671 tCO2e ha−1 yr−1.E_0 r_{\mathrm{BC}} = 7.8 \cdot(0.086) = 0.671\ \mathrm{tCO_2e\ ha^{-1}\ yr^{-1}}.

Unlike the remaining practices, the carbon benefit of biochar results from a one-time application of stable carbon rather than a continuous annual increase in native SOC. Accordingly, the annual SOC credit is derived from representative biochar characteristics.

A representative application rate of

ABC=10 Mg ha−1A_{\mathrm{BC}} = 10\ \mathrm{Mg\ ha^{-1}}

is assumed, together with

fC=0.60,fstable=0.70,f_C = 0.60,\qquad f_{\mathrm{stable}} = 0.70,

where fCf_C denotes the carbon fraction of biochar and fstablef_{\mathrm{stable}} denotes the fraction of carbon assumed to remain stable over the accounting period.

The total stable carbon added by one application is therefore

Cstable=ABC⋅fC⋅fstable=10⋅(0.60)⋅(0.70)=4.20 Mg C ha−1.C_{\mathrm{stable}} = A_{\mathrm{BC}} \cdot f_C \cdot f_{\mathrm{stable}} = 10 \cdot(0.60) \cdot(0.70) = 4.20\ \mathrm{Mg\ C\ ha^{-1}}.

Because CMP is expressed on an annual basis, this stable-carbon stock is annualized over a 10-year accounting horizon,

TBC=10 yr,T_{\mathrm{BC}} = 10\ \mathrm{yr},

giving

ΔSOCBC=4.2010=0.42 Mg C ha−1 yr−1.\Delta SOC_{\mathrm{BC}} = \frac{4.20}{10} = 0.42\ \mathrm{Mg\ C\ ha^{-1}\ yr^{-1}}.

The corresponding CO2_2-equivalent contribution becomes

3.667⋅ΔSOCBC=3.667⋅(0.42)=1.540 tCO2e ha−1 yr−1.3.667 \cdot\Delta SOC_{\mathrm{BC}} = 3.667 \cdot(0.42) = 1.540\ \mathrm{tCO_2e\ ha^{-1}\ yr^{-1}}.

The assumed application rate, carbon fraction, stable-carbon fraction, and accounting horizon are explicit modeling assumptions adopted to construct a representative benchmark dataset rather than universally applicable experimental constants.

CMP Calculation.

Substituting both mitigation components into Equation (37),

CMPBC=7.8⋅(0.086)+3.667⋅(0.42)=0.671+1.540=2.2\mathrm{CMP}_{\mathrm{BC}} = 7.8 \cdot(0.086) + 3.667 \cdot(0.42) = 0.671 + 1.540 = 2.2

Interpretation.

Biochar achieves one of the highest CMP values among the evaluated practices because it combines two complementary mitigation pathways: direct reductions in GHG emissions and long-term storage of stable carbon in soil. Unlike the remaining practices, a substantial fraction of its mitigation benefit originates from annualized stable-carbon storage rather than changes in annual SOC dynamics alone.

Summary of CMP Estimation The CMP values presented above provide a consistent estimate of the annual climate mitigation potential of the eight carbon farming practices considered in this study. Although all practices are evaluated using the common framework of Equation (7), the dominant mitigation pathway differs across practices. AWD, SRI, and DSR derive most of their mitigation benefit from reductions in methane emissions, whereas INM and manure application primarily contribute through enhanced SOC sequestration. Crop residue retention represents a balance between increased SOC storage and higher methane emissions, whereas biochar combines avoided emissions with long-term, stable carbon storage.

The resulting CMP dataset therefore represents a literature-informed benchmark that integrates published experimental evidence with transparent modeling assumptions wherever direct estimates are unavailable. This consistent parameterization provides the basis for evaluating the climate benefits of alternative carbon farming portfolios within the cooperative optimization framework developed in this study.

Net Operating Cost

For practice pp, net operating cost is the incremental cost of adopting pp relative to conventional cultivation,

ΔOCp=OCp−OCconv,(38)\Delta OC_{p} = OC_{p} - OC_{\mathrm{conv}}, \tag*{(38)}

where ΔOCp<0\Delta OC_{p} < 0 denotes a saving and ΔOCp>0\Delta OC_{p} > 0 denotes additional expenditure. The operating-cost boundary follows standard cost-of-cultivation accounting practice: hired and machine labour, land preparation, irrigation/pumping, fuel/energy, seed/nursery operations, fertilizer and manure, crop protection, and transport/handling of organic amendments. All local-currency figures are converted to PPP-adjusted USD using a fixed exchange assumption of 1 USDPPP_{\mathrm{PPP}} = 25 local-currency units (LCU).

Table 13 summarizes the resulting values.

PracticeNet OC (LCU /ha)Net OC (USD/ha, PPP)
Crop Residue Retention−395−15.80
Reduced / Minimum Tillage−1,250−50.00
Alternate Wetting and Drying−550−22.00
System of Rice Intensification+1,500+60.00
Direct Seeded Rice00.00
Integrated Nutrient Management+1,200+48.00
Manure Application+3,000+120.00
Biochar Application+20,000+800.00

Table 13. Net operating-cost assumptions relative to conventional cultivation.

Each value is a literature-informed representative parameter, not a directly reported empirical point estimate for the exact vector of practices, since no single study reports all eight incremental costs for a common location, season, and production system. The literature is instead used to establish which cost components are affected, the direction of the effect, and a plausible order of magnitude; a conservative representative value is then selected.

  • Crop Residue Retention is assigned a small saving (−395 LCU/ha) reflecting avoided residue disposal net of any additional spreading/management requirement [77]; the magnitude is deliberately conservative because it depends strongly on the residue-management baseline.

  • Reduced/Minimum Tillage saves on machinery, fuel, and tillage labour; against published cost-of-cultivation benchmarks of 12,342 LCU/ha machine labour and 31,770 LCU/ha hired labour [27, 17], the assumed 1,250 LCU/ha saving is a conservative ∼10% of the machine-labour line item.

  • AWD saves on irrigation/pumping, supported by reports of 23–33% lower water use [75, 88], positive economic returns from reduced pumping [54], and up to ∼30% irrigation savings [74]; only a small conservative saving (−550 LCU/ha) is monetized.

  • SRI nets to a modest positive cost because assumed additional labour/management (transplanting, spacing, weeding) is taken to marginally exceed seed/nursery/irrigation savings (Tamil Nadu Agricultural University, n.d.; Pyasi et al., 2024).

  • DSR is assigned exactly 0 LCU/ha as a deliberately conservative assumption: reported labour/establishment savings are large ([42, 21] reports ∼97% labour and ∼80% cost savings on the sowing/transplanting operation specifically) but are assumed to be offset by additional weed-management and herbicide costs, and no literature source directly supplies a net figure.

  • INM and Manure incur positive net costs reflecting procurement, transport, and application labour for organic inputs, offset only partially by fertilizer substitution [92, 49].

  • Biochar receives the largest incremental cost (+20,000 LCU/ha) because it additionally requires pyrolysis/production and transport [95]; this figure is explicitly a scenario/calibration parameter rather than a value observed in the cited paddy-cultivation studies, and the report recommends sensitivity analysis around it.

Yield Change

For practice jj, percentage yield change is

Yj=Ypractice,j−YbaselineYbaseline×100,(39)Y_j = \frac{Y_{\mathrm{practice},j} - Y_{\mathrm{baseline}}}{Y_{\mathrm{baseline}}} \times100, \tag*{(39)}

with Yj>0Y_j > 0 an increase, Yj<0Y_j < 0 a penalty, and Yj≈0Y_j \approx0 yield-neutral. Table 14 summarizes the values used.

PracticeYjY_j (%)Basis
Crop Residue Retention+5.2Rice-specific meta-analysis ([37])
Reduced / Minimum Tillage+4.8Conservative estimate based on South Asian rice systems reported by [76, 97]
Alternate Wetting and Drying0.0Mild AWD maintains yield ([20, 32])
System of Rice Intensification+15.0Regional field comparison ([72])
Direct Seeded Rice−12.0Rice-specific meta-analysis (Xu et al., 2019)
Integrated Nutrient Management+1.4Regional meta-analysis vs. 100% NPK (Padbhushan et al., 2021)
Manure Application+3.9Representative meta-analytic response ([36])
Biochar Application+10.7Rice-specific meta-analysis ([45])

Table 14. Yield-change parameters for the eight final practices.

  • Crop Residue Retention: a rice-specific meta-analysis of in-situ residue retention reports a 5.2% yield increase [37].

  • Reduced / Minimum Tillage: A conservative estimate based on the 19–22% yield gain reported for strip tillage in South Asian rice systems and 2.7% in rice-rapseed cultivation [76, 97].

  • AWD: a meta-analysis of 56 studies/528 comparisons found an average 5.4% yield reduction overall, but this was driven by severe water stress; under properly managed mild AWD (the practice represented here), yield was not significantly reduced, so YAWD=0%Y_{\mathrm{AWD}} = 0\% is adopted [20].

  • SRI: a regional field comparison reported 76.56 vs. 64.76 q/ha (≈15% increase) under SRI with recommended NPK [72]; this conservative regional value is retained even though a broader 78-study, 27-country review reports larger average gains (24% vs. recommended practice, 56% vs. farmer practice) [91].

  • DSR: a meta-analysis of 53 studies/440 paired comparisons reports a 12% average yield penalty for DSR relative to transplanted rice (Xu et al., 2019).

  • INM: a regional meta-analysis reports a 1.4% yield increase relative to the 100% chemical-NPK comparison (the fully-fertilized baseline is used in preference to an unfertilized-control comparison, which would overstate the benefit) (Padbhushan et al., 2021).

  • Manure: synthesis evidence places partial manure-substitution yield gains at roughly 1.4–5.9% and a broader grain-crop synthesis at roughly 3.3–3.9%; a representative central value of +3.9% is adopted [36].

  • Biochar: a rice-specific meta-analysis reports a 10.73% yield increase [45], rounded to +10.7%.

Pairwise Interactions

Adopting two practices together need not produce outcomes equal to the sum of their individual effects. For example, SRI already bundles AWD-like irrigation and refined nutrient handling, so simply adding SRI’s +15% yield term to Biochar’s +10.7% would double-count the nutrient-management pathway both partly rely on.

We introduce pairwise interaction terms αjk\alpha_{jk} (carbon), βjk\beta_{jk} (cost), and γjk\gamma_{jk} (yield) for every unordered pair of the eight practices (82)=28\binom{8}{2} = 28 pairs, plus a binary hard-incompatibility indicator Δjk∈{0,1}\Delta_{jk} \in\{0, 1\}. All four matrices are symmetric with zero diagonal.

Categorical Classification and Sampling

Each pair, in each dimension, is first assigned a category strong positive, positive, negative, strongly negative, or zero-like based on mechanistic overlap or conflict (Section B.3.2). A numeric value is then drawn once, with a fixed random seed (42), from a category-specific normal distribution and clipped to a category-independent bound.

Pairwise Classification

Two principles govern the classification:

Categoryα\alpha (carbon)γ\gamma (yield, %)β\beta (cost, USD/ha)
strong_pos0.12 (0.03)0.750 (0.375)−48.0 (10.0)
pos0.08 (0.03)0.375 (0.250)−20.0 (6.0)
neg−0.15 (0.04)−2.250 (0.75)48.0 (12.0)
neg_strong−0.25 (0.05)−7.000 (0.50)80.0 (16.0)
zero_like0.00 (0.015)0.000 (0.25)0.0 (4.0)
Clip bound[−0.30, 0.30][−7.50, 3.75][−80.0, 60.0]

Table 15. Sampling distributions (mean, sd) and clip bounds. Beta (cost) is shown in PPP-adjusted USD (divided by the PPP rate of 25 from the underlying INR distributions).

  1. A shared mechanism is penalized in every dimension it touches. SRI already bundles AWD-like irrigation and refined nutrient handling; pairing it with INM, Manure, or Biochar re-exercises that same mechanism rather than adding an independent one, so these three pairs carry a strongly negative penalty (S−S-) in carbon, cost, and yield larger than an ordinary two-practice overlap, reflecting SRI’s near-ceiling as a management package. INM and Manure likewise draw on the same organic-nutrient pathway, so this pair is negative in all three dimensions.

  1. Otherwise, carbon-adding practices are treated as complementary, not competing. Crop Residue Retention, INM, Manure, and Biochar add carbon through distinct pathways (residue-C, organic/inorganic nutrients, stabilized biochar-C), so pairings among them other than INM+Manure are positive or strongly positive in carbon and yield, not diminishing-returns.

Table 16 lists every pair with a notable interaction. Pairs not listed default to zero-like.

Pairα\alpha (C)γ\gamma (Y)β\beta (cost)Mechanism
BA + CRRS+S+++−-
BA + MAS+S+S+S+−-Additive material/labour cost; stabilization and supply functions remain complementary
CRR + RMTS+S+++++Preserving carbon (RMT) and adding carbon (CRR) are complementary, not competing for the same pool
CRR + MA++++−-
CRR + INM++S+S+−-
CRR + DSR++−-−-Different mechanism per dimension: residue-carbon input (carbon) vs. drilling/weed interference (yield) vs. specialized equipment/weed control (cost)
AWD + DSRS+S+++S+S+
AWD + RMT−-−-++
AWD + BA++S+S+−-
BA + INM++S+S+−-Additive material/labour cost; stabilization and supply functions remain complementary
DSR + RMT−-−-S+S+Shared reduced-tillage operations save cost; water-management/establishment combination still costs yield and carbon
DSR + MA−-−-−-
INM + MA−-−-−-Same organic-nutrient-input pathway
INM + SRIS−S-S−S-−-Nutrient-management overlap with SRI’s own package (near-ceiling penalty)
MA + SRIS−S-S−S-−-Nutrient-management overlap with SRI’s own package (near-ceiling penalty)
BA + SRIS−S-S−S-−-Nutrient-management overlap with SRI’s own package (near-ceiling penalty)

Table 16. Pairwise classification. S+S+ strongly positive, ++ positive, −- negative, S−S- strongly negative, −- zero-like (unclassified). Hard-excluded pairs (Δ=1\Delta= 1, Table 17) are omitted here. Mechanism notes are given only where documented in the generator code; blank cells indicate a category assignment with no accompanying rationale in the code.

Hard Incompatibilities

Four pairs are treated as hard-infeasible (Δjk=1\Delta_{jk}=1), all involving SRI (Table 17).

PairReason
DSR + SRIContradictory crop-establishment mechanisms (direct seeding vs. transplanting).
AWD + SRIAWD’s intermittent-irrigation water management is already embedded within the SRI package.
CRR + SRIThe only non-Δ\Delta pair negative in all three matrices simultaneously – the same “triple-negative” fingerprint as the confirmed DSR+SRI exclusion.
RMT + SRITreated as hard-excluded in the current model, consistent with the other SRI incompatibilities above.

Table 17. Hard incompatibilities in the final model.

Two further pairs are left as soft, implementable interactions with intentionally mixed signs across dimensions (Table 18): each has a plausible mixed-objective explanation, such as a real cost saving coexisting with a real yield/carbon penalty.

PairRationale
RMT + DSRShared reduced-tillage operations produce a real cost saving; the weed and water-management/establishment combination still carries a yield and carbon penalty.
INM + BA, MA + BAAdditive material/labour cost, but the underlying mechanisms (stabilization/retention vs. nutrient supply) remain complementary, so carbon and yield stay positive.
CRR + DSRDifferent mechanism per dimension: residue-carbon input (carbon, positive) vs. drilling/weed interference (yield, negative) vs. specialized equipment and weed control (cost, negative).

Table 18. Pairs with intentionally different signs across interaction dimensions.

MRV and Transaction Cost Calibration

Coalition enrollment incurs Monitoring, Reporting, and Verification (MRV) costs in addition to the practice-level operating costs described earlier. Following the carbon-farming MRV framework of Vanderheyden et al. [94], we distinguish between (i) technical MRV costs associated with monitoring and quantifying carbon outcomes and (ii) transaction and administrative costs associated with certification, reporting, verification, and farmer coordination. Accordingly, for a coalition SS with enrolled area A(S)A(S) (ha) and N=∣S∣N=|S| participating farmers, the certification cost is modeled as

Ccert(S)=FMRV+VMRV⋅A(S)δ+FT+VT⋅N.(40)C_{\mathrm{cert}}(S)=F_{\mathrm{MRV}}+V_{\mathrm{MRV}}\cdot A(S)^{\delta}+F_T+V_T\cdot N. \tag*{(40)}

The baseline parameterization used in the numerical experiments is

FMRV=$2000,VMRV=$80/ha0.70,δ=0.70,FT=$1000,VT=$10/farmer.(41)F_{\mathrm{MRV}}=\$2000,\qquad V_{\mathrm{MRV}}=\$80/\mathrm{ha}^{0.70},\qquad\delta=0.70,\qquad F_T=\$1000,\qquad V_T=\$10/\mathrm{farmer}. \tag*{(41)}
  • Fixed technical MRV cost (FMRV=$2000F_{\mathrm{MRV}}=\$2000) represents the project-level infrastructure required for carbon-credit certification, including MRV setup, monitoring system configuration, data management, model deployment, and technical coordination. The European Commission reports an indicative upfront MRV expenditure of approximately EUR 1000 for soil-carbon certification together

with substantial recurring monitoring and verification costs over the project lifetime [29]. Since the proposed formulation represents the overall project-level technical MRV infrastructure rather than only the initial setup cost, a conservative baseline value of $2000 is adopted.

  • Variable technical MRV cost (VMRV=$80/ha0.70V_{\mathrm{MRV}} = \$80/\mathrm{ha}^{0.70}) captures the area-dependent component of monitoring expenditure. [94] identify activity-data collection and soil sampling as the dominant monitoring cost drivers and demonstrate that monitoring costs increase with monitored area while exhibiting significant economies of scale. As the literature does not provide a coefficient directly corresponding to the functional form VMRVAδV_{\mathrm{MRV}}A^{\delta}, the value VMRV=80V_{\mathrm{MRV}} = 80 is adopted as a calibrated modeling parameter representing the order of magnitude of variable monitoring costs.

  • Economies of scale are represented through the exponent δ=0.70\delta= 0.70. This sub-linear specification reflects the empirical observation that average MRV costs decrease with increasing project size because several monitoring activities, such as image processing, reporting, verification, and technical support, can be shared across participating farms.

  • Fixed transaction cost (FT=$1000F_T = \$1000) represents project-level certification, reporting, quality assurance, and third-party verification activities. This value is consistent with the average third-party verification cost of approximately EUR 1250 per project reported by Vanderheyden et al. [94].

  • Variable transaction cost (VT=$10V_T = \$10/farmer) captures residual farmer-level administrative activities including documentation, communication, record maintenance, and coordination with the certification agency. Since the major monitoring activities are already represented within the technical MRV component, only the remaining administrative burden is included in the transaction cost. Accordingly, VTV_T is treated as a calibrated modeling parameter rather than a directly estimated empirical quantity.

The proposed decomposition is motivated by the empirical findings of Vanderheyden et al. [94], who report that monitoring accounts for approximately 77% of total MRV expenditure, while reporting and verification together contribute approximately 23%. Accordingly, monitoring-related activities are represented through the technical MRV component, whereas reporting- and verification-related activities are represented through the transaction cost component. Because technical MRV depends primarily on certified area whereas transaction costs depend primarily on the number of participating farmers, the relative contribution of these two components varies naturally with coalition composition. Therefore, the reported 77:23 decomposition serves as the conceptual motivation for the proposed cost structure rather than as an exact calibration target.

Farmer Dataset

A panel of n=15n=15 farmers is used for most of the experiments. Each farmer FiF_i is characterized by a farm size, FSiFS_i (ha), and an operating budget, BiB_i. Farm sizes are generated from a truncated log-normal distribution [47], log⁡(FSi)∼N(−0.5,1.02)\log(FS_i) \sim\mathcal{N}(-0.5, 1.0^2), truncated to the range [0.10, 2.00] ha [71] to represent smallholder farms. The generated values are rounded to two decimal places. This parameterisation reflects the smallholder structure, placing most farmers in the 0.2–1.0 ha range while admitting a small fraction of holdings near 2 ha [26]. Each farmer’s budget is set as

Bi=FSi×bˉ×(1+ϵi),ϵi∼N(0,0.202),ϵi∈[−0.20,+0.20],(42)B_i = FS_i \times\bar{b} \times(1 + \epsilon_i), \qquad\epsilon_i \sim\mathcal{N}(0, 0.20^2), \qquad\epsilon_i \in[-0.20, +0.20], \tag*{(42)}

where bˉ=$140/ha\bar{b} = \$140/ha is the base budget rate. The budget is thus positively correlated with farm size but introduces farmer-level heterogeneity in per-hectare capital availability, ensuring that budget constraints bind differentially across the panel.

References

  1. [1]Bina Agarwal. Can group farms outperform individual family farms? Empirical insights from India. World Development, 108:57–73, 2018. doi: 10.1016/j.worlddev.2018.03.010.DOI
  2. [2]Ashish Aggarwal and Dan Brockington. Reducing or creating poverty? analyzing livelihood impacts of forest carbon projects with evidence from india. Land Use Policy, 95:104608, 2020. doi: 10.1016/j.landusepol.2020.104608.
  3. [3]Stefan Ambec and Yves Sprumont. Sharing a river. Journal of Economic Theory, 107(2):453–462, 2002. doi: 10.1006/jeth.2001.2949.DOI
  4. [4]Eyal Bairey, Eric D. Kelsic, and Roy Kishony. High-order species interactions shape ecosystem diversity. Nature Communications, 7(1):12285, 2016. ISSN 2041-1723. doi: 10.1038/ncomms12285. URL https://doi.org/10.1038/ncomms12285.
  5. [5]Maria-Florina Balcan, Ariel D. Procaccia, and Yair Zick. Learning cooperative games. In Proceedings of the 24th International Joint Conference on Artificial Intelligence (IJCAI), pp. 475–481, 2015.
  6. [6]Anjeza Bekolli, Luis A. Guardiola, and Ana Meca. HarvestTech agriculture cooperatives: Beneficiaries and compensations. Journal of Industrial and Management Optimization, 2025a. doi: 10.3934/jimo.2025079.
  7. [7]Anjeza Bekolli, Luis A. Guardiola, and Ana Meca. Profit allocation in agricultural supply chains: exploring the nexus of cooperation and compensation. TOP, 33(3):580–610, 2025b. doi: 10.1007/s11750-024-00692-w.DOI
  8. [8]Márton Benedek, Jörg Fliege, and Tri-Dung Nguyen. Finding and verifying the nucleolus of cooperative games. Mathematical Programming, 190:135–170, 2021. doi: 10.1007/s10107-020-01527-9.DOI
  9. [9]Mayank Ratan Bhardwaj, Bazil Ahmed, Prathik Diwakar, Ganesh Ghalme, and Y Narahari. Designing fair, cost-optimal auctions based on deep learning for procuring agricultural inputs through farmer collectives. In 2023 IEEE 19th International Conference on Automation Science and Engineering (CASE), pp. 1–8. IEEE, 2023.arxiv.org/abs/2304.07341
  10. [10]Nalini Bikkina, Rama Mohana R. Turaga, and Vaibhav Bhamoriya. Farmer producer organizations as farmer collectives: A case study from India. Development Policy Review, 36:O669–O687, 2018. doi: 10.1111/dpr.12274.DOI
  11. [11]Ian Billick and Ted J Case. Higher order interactions in ecological communities: what are they and how can they be detected? Ecology, 75(6):1529–1543, 1994.
  12. [12]Olga N. Bondareva. Some applications of linear programming methods to the theory of cooperative games. Problemy Kibernetiki, 10:119–139, 1963. In Russian.
  13. [13]B. S. Brar, K. Singh, G. S. Dheri, and B. Kumar. Carbon sequestration and soil carbon pools in a rice-wheat cropping system: Effect of long-term use of inorganic fertilizers and organic manure. Soil and Tillage Research, 128:30–36, 2013. doi: 10.1016/j.still.2012.10.001.DOI
  14. [14]Oscar Cacho and Leslie Lipper. Abatement and transaction costs of carbon-sink projects involving small-holders, 2007.
  15. [15]Oscar J Cacho, Graham R Marshall, and Mary Milne. Transaction and abatement costs of carbon-sink projects in developing countries. Environment and Development Economics, 10(5):597–614, 2005.DOI
  16. [16]Oscar J. Cacho, Leslie Lipper, and Jonathan Moss. Transaction costs of carbon offset projects: A comparative study. Ecological Economics, 88:232–243, 2013.DOI
  17. [17]Aldo Calcante and Roberto Oberti. A technical-economic comparison between conventional tillage and conservative techniques in paddy-rice production practice in northern italy. Agronomy, 9(12), 2019. ISSN 2073-4395. doi: 10.3390/agronomy9120886. URL https://www.mdpi.com/2073-4395/9/12/886.DOI
  18. [18]Alberto Caprara and Adam N. Letchford. New techniques for cost sharing in combinatorial optimization games. Mathematical Programming, 124(1–2):93–118, 2010. doi: 10.1007/s10107-010-0357-7.DOI
  19. [19]Adeeт AG Cariappa and Vijesh V Krishna. Carbon farming in india: are the existing projects inclusive, additional, and permanent? Climate Policy, 25(5):756–771, 2025.
  20. [20]D. R. Carrijo, M. E. Lundy, and B. A. Linquist. Rice yields and water use under alternate wetting and drying irrigation: A meta-analysis. Field Crops Research, 203:173–180, 2017. doi: 10.1016/j.fcr.2016.12.002.DOI
  21. [21]Anjali Chaudhary, V. Venkatramanan, Ajay Kumar Mishra, and Sheetal Sharma. Agronomic and environmental determinants of direct seeded rice in south asia. Circular Economy and Sustainability, 3(1):253–290, 2023. ISSN 2730-597X. doi: 10.1007/s43615-022-00173-x. URL https://doi.org/10.1007/s43615-022-00173-x. Epub 2022-05-06.
  22. [22]Qianqian Chu, Haoyang Li, Nicola Cannon, Xianmin Chang, and Jian Feng. An evolutionary game analysis of carbon trading mechanisms for governments, farmer professional cooperatives and farmers. Systems, 13(6):413, 2025.DOI
  23. [23]Climate Watch. Historical ghg emissions. https://www.climatewatchdata.org/ghg-emissions, 2025. Accessed: 2026-03-21.
  24. [24]James Dahlgreen and Adam Parr. Exploring the impact of alternate wetting and drying and the system of rice intensification on greenhouse gas emissions: A review of rice cultivation practices. Agronomy, 14(2), 2024. ISSN 2073-4395. doi: 10.3390/agronomy14020378. URL https://www.mdpi.com/2073-4395/14/2/378.DOI
  25. [25]Saikat Ranjan Das, Dibyendu Chatterjee, Saurav Saha, Dibyendu Sarkar, Rounak Alam, Souvik Dey, Samrat Ghosh, Bitish Kumar Nayak, Pete Smith, and Himanshu Pathak. Enhancing carbon sequestration potential of lowland rice agroecosystems for environmentally clean production system: A review. Climate Smart Agriculture, 2(2):100054, 2025. doi: https://doi.org/10.1016/j.csag.2025.100054.
  26. [26]Department of Agriculture, Cooperation & Farmers Welfare. Agriculture census 2015–16 (phase-i): Provisional results – all india report on number and area of operational holdings. Technical report, Department of Agriculture, Cooperation & Farmers Welfare, Ministry of Agriculture & Farmers Welfare, Government of India, 2018. Section 3.10.
  27. [27]Department of Economics and Statistics, Government of Kerala. Report on cost of cultivation of important crops in kerala 2012–13, 2013. URL https://ecostat.kerala.gov.in/publication-detail/report-on-cost-of-cultivation-of-important-crops-in-kerala-2012-13.
  28. [28]Geraldine Ducos, Pierre Dupraz, and Francois Bonnieux. Agri-environment contract adoption under fixed and variable compliance costs. Journal of environmental planning and management, 52(5):669–687, 2009.DOI
  29. [29]European Commission. Impact assessment report accompanying the proposal for a regulation establishing a Union certification framework for carbon removals. Commission Staff Working Document SWD(2022) 377 final, European Commission, Brussels, 2022. URL https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52022SC0377.
  30. [30]Xiaoru Fan, Haoyuan Kong, Ying Zhao, Hongyan Hu, Zhenqiang Jiang, and Xiaolong Wang. Global synthesis of agricultural practices on crop yield, soil organic carbon and greenhouse gas emissions. Journal of Cleaner Production, 540:147529, 2026.
  31. [31]Mikael Frisk, Maud Göthe-Lundgren, Kurt Jörnsten, and Mikael Rönnqvist. Cost allocation in collaborative forest transportation. European Journal of Operational Research, 205(2):448–458, 2010.
  32. [32]Rong Gao, La Zhuo, Yiduo Duan, Chenjian Yan, Zhiwei Yue, Zikun Zhao, and Pute Wu. Effects of alternate wetting and drying irrigation on yield, water-saving, and emission reduction in rice fields: A global meta-analysis. Agricultural and Forest Meteorology, 353:110075, 2024. ISSN 0168-1923. doi: https://doi.org/10.1016/j.agrformet.2024.110075. URL https://www.sciencedirect.com/science/article/pii/S0168192324001904.
  33. [33]Donald B Gillies. Solutions to general non-zero-sum games. Contributions to the Theory of Games, 4(40):47–85, 1959.
  34. [34]Jasper Grashuis. Determinants of collective action by farm producers: A meta-analysis of the likelihood of co-operative membership. Journal of Rural Studies, 117:103639, 2025. doi: 10.1016/j.jrurstud.2025.103639.DOI
  35. [35]Mario Guajardo and Mikael Rönnqvist. A review on cost allocation methods in collaborative transportation. International Transactions in Operational Research, 23(3):371–392, 2016. doi: 10.1111/itor.12205.DOI
  36. [36]Qiong Hou, Yuemin Ni, Shan Huang, Ting Zuo, Jian Wang, and Wuzhong Ni. Effects of substituting chemical fertilizers with manure on rice yield and soil labile nitrogen in paddy fields of China: A meta-analysis. Pedosphere, 33:172–184, 2023. doi: 10.1016/j.pedsph.2022.09.003.DOI
  37. [37]S. Huang, Y. Zeng, J. Wu, Q. Shi, and X. Pan. Effect of crop residue retention on rice yield in China: A meta-analysis. Field Crops Research, 154:188–194, 2013. doi: 10.1016/j.fcr.2013.08.013.DOI
  38. [39]N. Jain, R. Dubey, D. S. Dubey, J. Singh, M. Khanna, H. Pathak, and A. Bhatia. Mitigation of greenhouse gas emission with system of rice intensification in the Indo-Gangetic plains. Paddy and Water Environment, 12(3):355–363, 2014. doi: 10.1007/s10333-013-0390-2.DOI
  39. [40]Alexander Kopelowitz. Computation of the kernels of simple games and the nucleolus of n-person games. Research Memorandum 31, Hebrew University of Jerusalem, 1967.
  40. [41]Laure Kuhfuss, Raphaële Prégét, Sophie Thoyer, and Nick Hanley. Nudging farmers to enrol land into agri-environmental schemes: the role of a collective bonus. European Review of Agricultural Economics, 43(4):609–636, 2016. doi: 10.1093/erae/jbv031.DOI
  41. [42]Anil Kumar, Suresh Kumar, Kuldeep Dahiya, Sundeep Kumar, and Mukesh Kumar. Productivity and economics of direct seeded rice (Oryza sativa L.). Journal of Applied and Natural Science, 7(1):410–416, 2015. doi: 10.31018/jans.v7i1.625. URL https://journals.ansfoundation.org/index.php/jans/article/view/625.DOI
  42. [43]Lin Li, Zheng Huang, Yixue Mu, Shaokun Song, Yicheng Zhang, Ye Tao, and Lixiao Nie. Alternate wetting and drying maintains rice yield and reduces global warming potential: A global meta-analysis. Field Crops Research, 318:109603, 2024. doi: 10.1016/j.fcr.2024.109603.DOI
  43. [44]You Li, Sakae Shibusawa, and Masakazu Kodaira. Carbon sequestration potential and farming income: Identifying the optimal carbon farming practices in japanese paddy fields. Engineering in Agriculture, Environment and Food, 6(2):68–76, 2013.
  44. [45]Y. Liu, H. Li, T. Hu, A. Mahmoud, J. Li, R. Zhu, X. Jiao, and P. Jing. A quantitative review of the effects of biochar application on rice yield and nitrogen use efficiency in paddy fields: A meta-analysis. Science of the Total Environment, 830:154792, 2022. doi: 10.1016/j.scitotenv.2022.154792.DOI
  45. [46]Trevon D Logan. Economies of scale in the household: Puzzles and patterns from the american past. Economic Inquiry, 49(4):1008–1028, 2011.
  46. [47]Sarah K Lowder, Jakob Skoet, and Terri Raney. The number, size, and distribution of farms, smallholder farms, and family farms worldwide. World development, 87:16–29, 2016.DOI
  47. [49]K. Manjappa. Integrated nutrient-management practices effect on productivity of rice (Oryza sativa) and soil properties under rainfed lowland rice situation—a long-term study. Indian Journal of Agronomy, 68(2), 2023. doi: 10.59797/ija.v68i2.339.DOI
  48. [50]Vida Mantey, Arnold Missiamé, Christine Bosch, Regina Birner, Athena Birkenberg, Viviane Guesbeogo Yameogo, and John Mburu. Carbon projects, cooperative membership and technical efficiency of small-holder dairy farmers in kenya. Environmental Development, pp. 101373, 2025.DOI
  49. [51]Helen Markelova, Ruth Meinzen-Dick, Jon Hellin, and Stephan Dohrn. Collective action for smallholder market access. Food Policy, 34(1):1–7, 2009.
  50. [52]Michael Maschler, Bezalel Peleg, and Lloyd S Shapley. Geometric properties of the kernel, nucleolus, and related solution concepts. Mathematics of operations research, 4(4):303–338, 1979.
  51. [53]Rory Mitchell, Joshua Cooper, Eibe Frank, and Geoffrey Holmes. Sampling permutations for shapley value estimation. Journal of Machine Learning Research, 23(43):1–46, 2022.
  52. [54]K. Mote, V. P. Rao, and V. Anitha. Alternate wetting and drying irrigation technology in rice. Indian Farming, 70(4), 2020. URL https://epubs.icar.org.in/index.php/IndFarm/article/view/105779.
  53. [55]A. K. Nayak, B. Gangwar, A. K. Shukla, S. P. Mazumdar, A. Kumar, R. Raja, A. Kumar, V. Kumar, P. K. Rai, and U. Mohan. Long-term effect of different integrated nutrient management on soil organic carbon and its fractions and sustainability of rice-wheat system in Indo Gangetic plains of India. Field Crops Research, 127:129–139, 2012.DOI
  54. [56]NetZeroAg. First rice methane reduction credits issued by gold standard. https://www.netzero-ag.com/news/first-rice-methane-reduction-credits-issued-by-gold-standard, 2025. Accessed: October 13, 2026.
  55. [57]Emily E. Oldfield, Alison J. Eagle, Rachel L. Rubin, Joseph Rudek, Jonathan Sanderman, and Doria R. Gordon. Crediting agricultural soil carbon sequestration. Science, 375(6586):1222–1225, 2022. doi: 10.1126/science.abl7991.DOI
  56. [58]Elinor Ostrom. Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press, Cambridge, 1990. doi: 10.1017/CBO9780511807763.
  57. [59]Guillermo Owen. On the core of linear production games. Mathematical Programming, 9:358–370, 1975. doi: 10.1007/BF01681356.DOI
  58. [61]Gregory M. Parkhurst, Jason F. Shogren, Chris Bastian, Paul Kivi, Jennifer Donner, and Rodney B. W. Smith. Agglomeration bonus: an incentive mechanism to reunite fragmented habitat for biodiversity conservation. Ecological Economics, 41(2):305–328, 2002.DOI
  59. [62]Irene Parrachino, Stefano Zara, and Fioravante Patrone. Cooperative game theory and its application to natural, environmental, and water resource issues: 1. basic theory, 2006.DOI
  60. [63]Leon Petrosjan and Georges Zaccour. Time-consistent Shapley value allocation of pollution cost reduction. Journal of Economic Dynamics and Control, 27(3):381–398, 2003.
  61. [64]My Pham-Kieu, Stephen Ives, Warwick Badgery, and Matthew Tom Harrison. Stacking interventions enhances carbon removals and profitability of livestock production systems. Advanced Science, 12(33):e03382, 2025.DOI
  62. [65]Lea Piscitelli, Annalisa De Boni, Rocco Roma, and Giovanni Ottomano Palmisano. Carbon farming: How to support farmers in choosing the best management strategies for low-impact food production. Land, 13(1):5, 2023.DOI
  63. [66]V Priyanka, Geetha Charan, Rohit P Suresh, Thandava Sunkara, Manojkumar Patil, Kartik Sagar, Aashman Trivedi, K Soumya, Subir Paul, Parashuram Hadimani, Ganesh Babu, Ravi Trivedi, and Y Narahari. Carbon farming: An expository, inter-disciplinary survey. Journal of the Indian Institute of Science, pp. 1–63, 2026.DOI
  64. [68]Haoyu Qian, Xiangchen Zhu, Shan Huang, Bruce Linquist, Yakov Kuzyakov, Reiner Wassmann, Kazunori Minamikawa, Maite Martinez-Eixarch, Xiaoyuan Yan, Feng Zhou, et al. Greenhouse gas emissions and mitigation in rice agriculture. Nature Reviews Earth & Environment, 4(10):716–732, 2023.
  65. [69]Nidhi Raina, Matteo Zavalloni, and Davide Viaggi. Incentive mechanisms of carbon farming contracts: A systematic mapping study. Journal of Environmental Management, 352:120126, 2024. doi: 10.1016/j.jenvman.2024.120126.DOI
  66. [70]Shivani Ranjan, Sanjay Kumar, Swaraj Kumar Dutta, Smruti Ranjan Padhan, Parmeswar Dayal, Sumit Sow, Dhirendra Kumar Roy, Dibyajyoti Nath, Kirttiranjan Baral, and Vikram Bharati. Influence of 36 years of integrated nutrient management on soil carbon sequestration, environmental footprint and agronomic productivity of wheat under rice-wheat cropping system. Frontiers in Environmental Science, 11:1222909, 2023. doi: 10.3389/fenvs.2023.1222909.DOI
  67. [71]George Rapsomanikis. The economic lives of smallholder farmers, 2015.
  68. [72]B. G. M. Reddy, Ravishankar, S. Balaganvhi, and V. R. Joshi. Nutrient management of rice (Oryza sativa) under system of rice intensification (SRI). The Indian Journal of Agricultural Sciences, 83(6), 2013. doi: 10.56093/ijas.v83i6.30318.
  69. [73]K. Srikanth Reddy, C. M. Parihar, P. Panneerselvam, Ayan Sarkar, Kiranmoy Patra, Sneha Bharadwaj, D. R. Sena, G. Sreeja Reddy, Alok Sinha, Rajkumar Dhakar, Virender Kumar, and Hari Sankar Nayak. Meta-analysis of yield-emission trade-off in direct seeded vs. puddled transplanted rice: Towards a cleaner and sustainable production. Cleaner Environmental Systems, 16:100238, 2025. Published online Nov. 2024.DOI
  70. [74]M. Richards and B. O. Sander. Alternate wetting and drying in irrigated rice. Climate-smart agriculture practice brief, CGIAR Research Program on Climate Change, Agriculture and Food Security, Copenhagen, Denmark, 2014. URL https://ccafs.cgiar.org/resources/publications/alternate-wetting-and-drying-irrigated-rice.
  71. [75]R. Sahu, A. Chaurasiya, R. Kumar, and R. K. Sohane. Alternate wetting and drying technology for rice production. Indian Farming, 73(3):11–13, 2023. URL https://epubs.icar.org.in/index.php/IndFarm/article/view/133179.
  72. [76]Nazmus Salahin, Mohammad Jahiruddin, Mohammad Rafiqul Islam, Md Khairul Alam, M Enamul Haque, Sharif Ahmed, Alaa Baazeeem, Adel Hadifa, Ayman EL Sabagh, and Richard W Bell. Establishment of crops under minimal soil disturbance and crop residue retention in rice-based cropping system: Yield advantage, soil health improvement, and economic benefit. Land, 10(6):581, 2021. doi: 10.3390/land10060581.DOI
  73. [77]Sukanta K. Sarangi, Sudhanshu Singh, Ashish K. Srivastava, Madhu Choudhary, Uttam K. Mandal, Tashi D. Lama, Kshirendra K. Mahanta, Virender Kumar, Parbodh C. Sharma, and Abdelbagi M. Ismail. Crop and residue management improves productivity and profitability of rice–maize system in salt-affected rainfed lowlands of east india. Agronomy, 10(12), 2020. ISSN 2073-4395. doi: 10.3390/agronomy10122019. URL https://www.mdpi.com/2073-4395/10/12/2019.DOI
  74. [78]JD Sarkar and AK Sarkar. Carbon farming: Pathways to sustainable agriculture. Agri Articles, 2025.
  75. [79]Samriddhi Sarkar, Tuhin Biswas, Mariana Curado Malta, Deolinda Meira, and Animesh Dutta. A coalition formation framework of smallholder farmers in an agricultural cooperative. Expert Systems with Applications, 221:119781, 2023.DOI
  76. [80]Herbert E Scarf. The core of an n person game. Econometrica: Journal of the Econometric Society, pp. 50–69, 1967.
  77. [81]David Schmeidler. The nucleolus of a characteristic function game. SIAM Journal on applied mathematics, 17(6):1163–1170, 1969.
  78. [82]Seth Shames, Louise E Buck, and Sara J Scherr. Reducing costs and improving benefits in smallholder agriculture carbon projects: implications for going to scale. In Climate change mitigation and agriculture, pp. 69–77. Routledge, 2013.
  79. [83]Lloyd S. Shapley. A value for n-person games. In Harold W. Kuhn and Albert W. Tucker (eds.). Contributions to the Theory of Games II, number 28 in Annals of Mathematics Studies, pp. 307–317. Princeton University Press, 1953. doi: 10.1515/9781400881970-018.
  80. [84]Lloyd S. Shapley. On balanced sets and cores. Naval Research Logistics Quarterly, 14(4):453–460, 1967. doi: 10.1002/nav.3800140404.DOI
  81. [85]Lloyd S. Shapley. Cores of convex games. International Journal of Game Theory, 1(1):11–26, 1971. doi: 10.1007/BF01753431.DOI
  82. [86]Pete Smith, Jean-François Soussana, Denis Angers, Louis Schipper, Claire Chenu, Daniel P. Rasse, Niels H. Batjes, Fenny van Egmond, Stephen McNeill, Matthias Kuhnert, Cristina Arias-Navarro, Jørgen E. Olesen, Ngonidzashe Chirinda, Dario Fornara, Eva Wollenberg, Jorge Álvaro-Fuentes, Alberto Sanz-Cobena, and Katja Klumpp. How to measure, report and verify soil carbon change to realize the potential of soil carbon sequestration for atmospheric greenhouse gas removal. Global Change Biology, 26(1):219–241, 2020. doi: 10.1111/gcb.14815.DOI
  83. [87]Rajiv Kumar Srivastava, Sanju Purohit, Edris Alam, and Md Kamrul Islam. Advancements in soil management: Optimizing crop production through interdisciplinary approaches. Journal of Agriculture and Food Research, 18:101528, 2024.DOI
  84. [88]JS Sudharani, N Praveen, and K Avilkumar. Alternate wetting and drying (awd) a promising water saving technology in rice production system for farmers of telangana, india. Extended Summaries, pp. 434, 2020.
  85. [89]Helen Suich. The effectiveness of economic incentives for sustaining community based natural resource management. Land Use Policy, 31:441–449, 2013. doi: 10.1016/j.landusepol.2012.08.008.DOI
  86. [91]A. K. Thakur, K. G. Mandal, O. P. Verma, and R. K. Mohanty. Do system of rice intensification practices produce rice plants phenotypically and physiologically superior to conventional practice? Agronomy, 13 (4):1098, 2023a. doi: 10.3390/agronomy13041098.
  87. [92]R. K. Thakur, N. K. Bisen, A. K. Shrivastava, S. K. Rai, and S. Sarvade. Impact of integrated nutrient management on crop productivity and soil fertility under rice (Oryza sativa)–chickpea (Cicer arietinum) cropping system in Chhattisgarh plain agro-climatic zone. Indian Journal of Agronomy, 68(1):9–13, 2023b. doi: 10.59797/ija.v68i1.195.DOI
  88. [93]Anshul Tyagi and AK Haritash. Climate-smart agriculture, enhanced agroproduction, and carbon sequestration potential of agroecosystems in india: A meta-analysis. Journal of Environmental Studies and Sciences, 15(1):167–185, 2025. doi: https://doi.org/10.1007/s13412-024-00917-1.
  89. [94]Lisa Vanderheyden, Nathan Gayot, and Goedele Van den Broeck. What monitoring, reporting & verification (mrv) systems can reduce costs and enhance scalability of carbon farming? Journal of Environmental Management, 401:128885, 2026.DOI
  90. [95]Rajeev Varma, Anand Naik, Mallikarjun Reddy, M. A. Bellakki, S. R. Balanagoudar, and Siddaram. Impact of various levels of biochar on growth and yield of chickpea (cicer arietinum l.). International Journal of Plant & Soil Science, 37(15), 2025. doi: 10.9734/IJPSS/2025/v37i15247. URL https://doi.org/10.9734/IJPSS/2025/v37i15247.
  91. [96]Verra. VM0042 improved agricultural land management. Verified Carbon Standard methodology, versions 2.1 and 2.2, 2024. URL https://verra.org/methodologies/vm0042-improved-agricultural-land-management-v2-1. Accessed 2026-10-13.
  92. [97]Chunyun Wang, Zongkai Wang, Mengzhen Liu, Maria Batool, Ali Mahmoud El-Badri, Xianling Wang, Hongxiang Lou, Dongli Shao, Xiaoqiang Tan, Zhen Li, Jie Kuai, Bo Wang, Jing Wang, Zhenghua Xu, Guangsheng Zhou, Daohong Jiang, and Jie Zhao. Optimizing tillage regimes in rice-rapeseed rotation system to enhance crop yield and environmental sustainability. Field Crops Research, 2024.DOI
  93. [98]Hongji Wang and Tongwei Qiu. The effects of farmer cooperatives on agricultural carbon emissions reduction: Evidence from rural china. Journal of Cleaner Production, 450:141881, 2024.
  94. [99]Frank Wätzold and Martin Drechsler. Agglomeration payment, agglomeration bonus or homogeneous payment? Resource and Energy Economics, 37:85–101, 2014.DOI
  95. [100]Shenglin Wen, Ningbo Cui, Yaosheng Wang, Daozhi Gong, Zhihui Wang, Liwen Xing, Zongjun Wu, and Yixuan Zhang. Effect of biochar application on yield, soil carbon pools and greenhouse gas emission in rice fields: A global meta-analysis. Agricultural and Forest Meteorology, 369:110571, 2025.
  96. [101]L. Xia et al. Meta-analysis of straw return effects on greenhouse-gas emissions and global warming potential in agricultural soils. Field Crops Research, 2024.
  97. [103]Devideen Yadav, DV Singh, Subhash Babu, M Madhu, and Anita Kumawat. Carbon farming: Innovations, potential, and challenges for sustainable future. Indian Journal of Agronomy, 70:20–30, 2025.
  98. [104]Tom Yan and Ariel D Procaccia. If you like shapely then you’ll love the core. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pp. 5751–5759, 2021.
  99. [105]Hong Zhang, Fulin Wei, Jixiang Lai, Han Xiao, and Kuan Li. How can cooperatives drive small-scale farmers to achieve a “carbon reduction effect” in the planting industry: Evidence from china. Sustainability, 17(18):8479, 2025.
  100. [106]Xiao-Xue Zheng, Jingwen Guo, Fu Jia, and Shiyuan Zhang. Cooperative game theory approach to develop an incentive mechanism for biopesticide adoption through farmer producer organizations. Journal of Environmental Management, 319:115696, 2022.

Paper details

vlnc:W_q6vk95yn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

GPT (multiple versions), Claude

View statement
Looking for related papers…
Contents