Multi-Objective Crop and Livestock Allocation Modeling (MOCLAM) for Sustainable Agriculture: A Case Study of Semi-Arid India

1. Introduction

Worldwide, agricultural productivity is growing strongly, though at different levels and paces across countries, regions and sub-sectors (Ruttan 2002; FAO 2017). The productivity growth contributes to an affordable supply of food, feed, fuel and fiber to meet increasing demand. Nonetheless, this increase partly has come at the expense of the deterioration of scarce natural resources (Fuglie et al. 2020). Despite being home to 18% of the world’s population, India possesses only 4% of its water resources and 2.4% of its land resources (Bhattacharyya et al. 2015; Gulati and Juneja 2022). According to Falkenmark’s classification, India is currently labeled as a “water-stressed” country where annual per-capita water availability is less than 1700m³ (Falkenmark 1986; World Bank 2013). If the current trends in water use continue and steps are not taken, the country will face “severe water stress” (water availability <1000m³/capita/annum) by 2050 (Parish et al. 2012; OECD 2012).

Agriculture is the largest consumer of freshwater. Globally, 70% of water withdrawals and 90% of consumptive water use go for irrigation (UNESCO 2022). Therefore, the sustainability of natural resources, particularly land and water, depends on cropping patterns and agricultural management practices (Varade and Patel 2019). A cropping pattern skewed toward water-intensive crops may lead to the depletion of the groundwater table and pose a serious threat to food security and natural resource sustainability. Apart from natural resources, there are several criteria that need to be taken care of while deciding the cropping pattern. One of the important criteria is the implication of the resultant cropping pattern on allied enterprises such as livestock (Sekaran et al. 2021). The decision on allocation of scarce resources under different crops and allied enterprises like livestock, agroforestry and fisheries involves complex trade-offs between multiple objectives including food self-sufficiency, and income adequacy and stability. These complexities become even more intricate when considering the increasing risks posed by climate change.

In developing countries, agricultural production systems are often subjected to bio-physical and socio-economic constraints, risks and uncertainties. The sustainability of natural resources while ensuring food security and remunerative employment is one of the significant challenges. There are enormous negative externalities of sole and intensive cultivation of crops or livestock (Herrero et al. 2009; Groot et al. 2012) which can be mitigated by the physical reconnection of livestock to land (Naylor et al. 2005). A mixed farming system offers a promising alternative to the ongoing crop and livestock production system and helps develop sustainable agroecosystems (Hilimire 2011). Beyond a stable source of food, nutrition and livelihood for smallholders, especially for rural women, livestock also serves as a significant profit-making enterprise for millions of people across India. Through various sectors such as dairy farming, poultry and meat production, rural communities generate substantial income and improve their economic stability (Yadav and Sharma 2013). It also serves as a form of savings, which can easily be liquidated into cash and provide many other benefits both directly and indirectly (Banda and Tanganyika 2021). The objective of agricultural sustainability is unlikely to succeed unless a system approach in farming is followed (Gill et al. 2009; Gangwar et al. 2016). This necessitates designing a comprehensive model of optimal crop and livestock enterprise planning.

Budgeting and mathematical programming are the primary methods for optimal crop planning. Linear programming, first used for optimizing feed ratio by Waugh (1951), is a frequently used technique to solve single objective functions. However, addressing one objective function at a time is the major shortcoming of this technique (Zgajnar et al. 2010), making it less suitable for crop planning. Agriculture relies on multiple interdependent and conflicting objectives and thus needs a mathematical technique capable of simultaneously addressing multiple objectives. Various mathematical approaches have been developed for multi-criteria-decision-analysis (MCDA) in agriculture. The nonlinear programming model, fractional programming model and goal programming model are commonly used in the MCDA approach for obtaining optimal cropping patterns and are known for providing an ideal framework for the optimal solution under multiple conflicting objectives (Fasakhodi et al. 2010; Jayaraman et al. 2015; Mosleh et al. 2017). Amongst these, goal programming is the most widely used mathematical model for solving agricultural land allocation and farming problems and gives better solutions for diversified crops and enterprises and when there are conflicting objectives (Val-Arreola et al. 2006; Jafari et al. 2008; Manos et al. 2013; Prišenk et al. 2014; Yang et al. 2016; Galán-Martín et al. 2017; Nasikh et al. 2021).

In India, attempts have been made to optimize the limited resources in agriculture. Using Elitist–Jaya and Elitist-teaching learning-based optimization algorithms, a study optimizes cropping patterns in Gujarat with the single objective of maximizing net economic gains from crops (Kumar and Yadav 2019). Amarasinghe et al. (2010) attempted to include dairy for allocating area under crops in the Trans-Gangetic Plains (TGP) of India but used crop water productivity as the sole criteria for area allocation. Another study by Singh (2014) used linear programming to allocate land and water resources under different crops in another region of TGP for maximizing net annual returns and suggested reducing the area under rice in the region. Extending the scope, a more recent study applied an optimization model at a larger geographical scale for allocating district-wise area under 12 major crops with the objective of maximizing the aggregate national agricultural revenue under the constraints of cropped area, irrigation water (sub-surface), food security and nutritional requirement (Devineni et al. 2022).

A review of these studies suggests three major gaps in the literature on the allocation of scarce resources in agriculture. First, livestock, an integral part of agriculture, is either excluded from the planning or dealt with in isolation without integrating with crops. Second, most of the studies optimize with the single objective of maximizing returns. Third, surface water resources accounting for 37.5% of the irrigated area in India (GoI 2023a) have been overlooked in crop and allied enterprise planning.

The arid and semi-arid regions of the tropics are among the world’s most vulnerable areas to climate disasters and poverty (World Bank 2021; FAO 2015). Rainfall variability and water scarcity continue to affect smallholder farmers’ food and income security and displace them regionally and sectorally in search of alternative employment opportunities. Consequently, the efficient allocation of resources, particularly of land and water, becomes more complex in these areas. Therefore, using goal programming, this study develops optimum plans for a mixed farming system considering the conflicting objectives of increasing economic efficiency, social inclusiveness and environmental friendliness. To illustrate the model’s applicability, a case study of the drought-prone Bundelkhand region of India was conducted under different technology and policy scenarios. The methodology for developing the model and the results obtained are discussed hereunder.

2. Methodology

2.1. Model description

There are several multi-objective optimization programming techniques available in the literature (Udias et al. 2016; Gunantara 2018; Ortiz-Partida et al. 2023). Goal programming is one where the optimizations of multiple and conflicting goals are achieved within the given decision space. It has been widely used for resource optimization under conflicting objective functions (Jones and Tamiz 2010; Aksaraylı and Pala 2018; Zhuang and Hocine 2018; Hanks et al. 2017). Moreover, it provides an ideal framework to obtain optimal solutions when the number of constraints is large (Vivekanandan et al. 2009; Kaur et al. 2023; Ayyad et al. 2023; Yazdani et al. 2023). Typically, a multi-objective optimization model’s formulation involves three components: Specifying objective functions and goals, selecting decision variables and identifying real-world constraints.

2.1.1. Goals and objectives

The first step in the model building was identifying the competing objectives allowing the objective function to be specified. As the main aim of this study was to design sustainable agricultural plans, the objectives were based on three broad principles of sustainability, i.e., economic viability, social acceptability and environmental friendliness. The economic viability in the model was ensured by defining that the net economic margins per unit of area are positive and the use of labor, financial capital and fertilizer does not go beyond the availability (≤ constraints in Eq. (1)). For ensuring the social sustainability, it was defined that the combined economic returns from all the crops and enterprises are maximized (LHS of Eq. (1)), and the region is self-sufficient in food production (≥ constraints in Eq. (1)). The ecological dimension was captured by defining a separate objective function that the total water use of the region is minimized (LHS of Eq. (2)) subjected to the constraint that region is self-sufficient in food production (≥ constraints in Eq. (2)), the total economic returns are higher than that in existing cropping pattern (≥ constraints in Eq. (2)) and the other resource and non-resource constraints are satisfied. Additionally, the ecological dimension was captured by including an additional constraint of greenhouse gas emissions. See Sec. 2.1.3 for a more detailed definition of the constraints. Equations (1) and (2) were optimized using linear programming to arrive at the goal of maximum attainable net economic returns and the minimum volume of water to attend the economic returns arrived in Eq. (1), respectively. The description of various notations used in the linear programming models is presented in Table 1 while the algebraic expressions of the linear programming models are mentioned below.

The value of objective functions arrived by solving Eqs. (1) and (2) using linear programming formed the goals ($G_j$) for the Goal Programming model defined in Eq. (12). The remaining goals (interchangeably used as soft constraints in goal programming¹) were decided either based on resource limits or targets set by the Government. For example, for GHG emissions, we targeted not exceeding the current emissions level. Similarly, the self-sufficiency goals of cereal, pulses and oilseed production were fixed to the current production level, assuming perfect substitutability within the crop groups. It is important to note that the later mentioned goals were also included in linear programming models as constraints to ensure that the suggested plans do not deviate from reality. The process is given in the flow chart (Figure 1). As shown in the figure, three LPs were run to determine (1) the maximum attainable total economic margins for the region (LP1) with the constraints listed left side of the figure, (2) the minimum water use that can generate the net returns determined in LP1 worked out above (LP2) and (3) the maximum attainable returns with additional supply of water by harvesting rainwater within the given constraint space (LP3). Once the goals are determined, the objective function in the goal programming model (Eq. (12)) is defined. In this case, the objective function was to minimize the deviations from the set goals.

2.1.2. Decision variables

The decision variables in mathematical programming, technically called activities, are a set of quantities that need to be determined to solve the problem. In the case of agricultural-enterprise-planning, these could be the area to be allocated to crops and crop processes (way of producing), the head of livestock units to be maintained, the amount of additional capital inflow, etc. Defining the decision variables is one of the most crucial steps in formulating a mathematical problem. In our case, we classified the activities into three different types: (1) Real activities, (2) production processes and (3) intermediate activities. Real activities included 24 field crops (pigeon pea, pearl millet, barley, chilies, chickpea, groundnut, cluster bean, sorghum, lathyrus, lentil, linseed, maize, mentha, mesta, green gram, rapeseed and mustard, onion, sesame, soybean, sugarcane, tomato, black gram, rice and wheat), agroforestry (agri-horticulture with a combination of gooseberry+wheat +groundnut) and four livestock species (indigenous cattle, crossbred cattle, buffalo, sheep and goats). The production process covered in the model was the resource-saving and/or efficiency-enhancing technologies like the method of sowing, agronomic practices, crop cultivars, livestock breeds, etc. Crop activities producing intermediate products (livestock feed and fodders) were considered intermediate activities (berseem and sorghum fodder). The returns from intermediate activities were assumed to be zero. Another important class of activity is hiring or borrowed activities. This activity has special significance in sensitivity analysis and identifying investment and development needs, mainly when plans are developed at the aggregate level. Therefore, we parameterized these as deviation variables while formulating goal programming (Sec. 2.2).

2.1.3. Constraints

Identification and specification of resource constraints are crucial in formulating a mathematical problem. As mentioned in Sec. 2.1.1, these values were used as Right-Hand-Sides (RHS) of Eqs. (1) and (2). In the case of goal programming, these formed the goals. The perfectly substitutable resources in all the uses were combined and expressed by a single constraint equation. For example, the available capital resources are substitutable in different crops and enterprises throughout the year. Unlike capital, labor resources unused in a month cannot be accumulated to be used for peak season. Therefore, we expressed the capital constraint through a single equation and the labor constraint through 12 different equations. The details of the constraints are given below. We categorized the constraints into two categories, i.e., resource and non-resource constraints.

2.1.4. Resource constraints

Land: Rather than considering a fixed season period for all the crops, we followed the crop calendar approach by formulating 12 different land constraint equations for each month where total land in a particular month was constrained by total arable area excluding area under perennial crops. Based on the month of sowing and harvesting, a binary 0 or 1 coefficient of land occupancy (${\mathrm{\text{lo}}}_{\mathrm{\text{cm}}})$ was introduced where 1 means that crop ‘c’ will occupy the land in month m or in other words it is the growing season for the crop. On the other hand, zero means the crop is not in the field or the season for the crop is over or yet to start. The generalized land constraints are expressed in Eq. (3). The land constraint equations provide the scope for simulating the effect of the duration of crops and varieties. The crop calendar for different crops is given in Table S1.

Water: While considering the water availability constraint, most of the resource optimization studies in India have considered groundwater water as the only source of irrigation (Das and Datta 2001; Sondhi and Kaushal 2006; Rejani et al. 2009; Varade and Patel 2019; Nath et al. 2021; Devineni et al. 2022). However, surface water also constitutes a significant part of irrigation. Around 1/4th of the total irrigation in India is through surface sources and in some regions, particularly confined aquifers, the surface water contributes the majority of the irrigated area (GoI 2023a). Therefore, both groundwater and surface sources were considered for the water availability constraint. Thus, we assumed that the sum of irrigation or consumptive water use of cth crop or enterprise must be less than AWRA (Eq. (4)), where AWRA includes water availability from the surface and groundwater sources. Since the water use coefficients were estimated in terms of the actual evapotranspiration of irrigation water, it does not include conveyance losses. Therefore, the irrigation water availability was adjusted by applying an irrigation efficiency of 70% to groundwater and 40% to surface water (CWC 2014). Since the precise data on season-wise volumetric water availability were not available, an additional constraint of irrigated area in a season was included in the model using the following equation : where ${ir}_c$ is binary coefficient irrigation type (1 if crop is irrigated, 0 otherwise).

Labor: While the study region boasts abundant labor availability and grapples with the issue of out-migration (Agrawal et al. 2016), it is imperative to still consider the total labor pool in the region when undertaking planning and building scenarios. This ensures that the proposed plans remain within the constraints of available labor resources, preventing them from becoming unfeasible and unsustainable in the long term. It also highlights the month-wise potential for employment generation within the region, which could help curb migration. Furthermore, labor availability can pose a constraint, especially during peak harvesting seasons. Studies have indicated peak season agricultural labor scarcity in the region, particularly after the implementation of the Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA), a flagship program of the Government of India in 2005 (Nagaraj et al. 2016; Gulati et al. 2013).

The labor constraint period is generally considered on an annual or seasonal basis. However, it is unrealistic to assume that the labor force falling short of the requirement in the month of peak harvesting can be substituted by surplus labor of lean season. Therefore, the period should be sufficiently granular within which substitution is possible. Hence, we considered separate constraints of labor availability for each month (Eq. (6)). The total labor availability was estimated based on data of cultivators and agricultural laborers from population Census (GOI Undated) converted into man-days, assuming 8h of effective working in a day. Further, for a production system to be socially sustainable, workers in agriculture should get a sufficient number of off days from work. Therefore, the labor availability was estimated by assuming a day off per week (Chand et al. 2015). The per unit labor requirements coefficients ($L_{\text{cm}}$) were estimated using the data from the cost of cultivation as mentioned in Table 2.

Fertilizers: Three fertilizer constraints, one for each macronutrient (N, P, $K)$, were included in the model using constraint equation (7). where $F_{\mathrm{\text{cf}}}$ is the use coefficient of fertilizer type ‘f’(kg/ha) of cth crop and ${\mathrm{\text{AAF}}}_f$ is the annual availability of fth fertilizer. The use of fertilizer in the existing set of cropping patterns (status quo) was considered as availability. The rationale behind assuming the fertilizer supply constraint was (1) to align with environmental sustainability objectives and (2) India’s heavy reliance on imports and the substantial burden of fertilizer subsidies borne by the government. Given the logistical challenges involved, augmenting fertilizer supply in the short term for a specific region proves to be a daunting task.

Capital: Financial capital is one of the most limiting factors, particularly in the smallholder production systems of developing countries (Fan and Rue 2020; FAO 2015). Capital in this study implies operational expenses required for real activities, crop processes and intermediate activities. Therefore, the sum of capital used by all the crop or livestock enterprises must be equal to or less than the total available capital with the farmers (Eq. (8)). The availability of the capital was estimated as we did in the case of fertilizers mentioned above.

Livestock feed and fodder: The optimum number of livestock units a cropping pattern can sustain was parameterized endogenously using a set of following equations. For example, if the model allocates more area under sorghum which produces 10t of additional quantity of fodder on a dry matter basis, the model allocates ∼4 head of livestock as 10t of dry matter is sufficient to sustain four head of livestock. where ${\mathrm{\text{FU}}}_{\mathrm{\text{kc}}}$ is the fodder use or generation coefficient of kth type of fodder (dry, green and concentrates) by cth livestock or crop activity (kg/SAU or kg/ha) on a dry matter basis. Conceptually, crops do not use any kind of fodder. Hence, in the case of crops, the fodder requirement coefficient was considered with a minus sign and termed as fodder generation coefficient ($-{\mathrm{\text{FU}}}_{\mathrm{\text{kc}}}$). The fodder generation/requirement coefficients were estimated separately for dry (${\mathrm{\text{FU}}}_d$), green (${\mathrm{\text{FU}}}_g$) and concentrate (${\mathrm{\text{FU}}}_{\mathrm{\text{con}}}$) using the following expressions. where $Y_c$, ${\mathrm{\text{GSR}}}_c$, ${\mathrm{\text{GCR}}}_c$ and ${\mathrm{\text{DM}}}_c$ are grain yield (fodder yield in case of green fodders), grain to straw ratio, gain to concentrate or oil cake ratio, and dry matter contents, respectively of cth crop. The grain-to-straw ratios, grain-to-concentrate ratios and dry matter contents of different crops were taken from published studies (Jain et al. 1996; Ramachandra et al. 2007; Jain and Rane 2007) and also in discussion with experts on animal nutrition. Small ruminants (sheep and goats) were considered to be wholly dependent on the existing carrying capacity of pasture and grazing lands, cultivable wasteland, tree leaves from forests, farm forestry and other common lands for fodder (Shinde and Mahanta 2020).

2.1.5. Non-resource constraints

Three non-resource — self-sufficiency (cereals, pulses and oilseeds), maximum area and GHG emission — constraints were also included in the model. Prioritizing the self-sufficiency of a country or region’s essential agricultural consumption needs emerges as a pivotal focal point for bolstering the social sustainability of agriculture. Depending on the nation’s productive capacity, import capabilities and equitable domestic food distribution, the goal food self-sufficiency is absolutely important. Each country may pass through a distinctive array of circumstances influencing its ability to secure an adequate food supply for its population (Clapp 2014). Some nations, in pursuit of comprehensive economic development, actively advocate for food self-sufficiency as a strategic imperative. With this background, we explicitly imposed a constraint of self-sufficiency in the model (Eq. (10)). The self-sufficiency constraint was decided based on the region’s human and livestock population and past production trends. Similarly, a restriction was also imposed on the maximum area allocated to a specific crop type to prevent to model from allocating disproportionately excessive land under it. This restriction was based on an empirical study conducted by Jain et al. (2020). By adhering to these constraints, the model ensures a more balanced and realistic distribution of agricultural land, promoting sustainable land use practices. Besides this, expert opinions were also sought through a workshop for setting the goals/constraints and assigning weights to various goals. The workshop engaged agricultural policy analysts associated with the National Agricultural Research System of India, with expertise in agricultural economics and climate change. Additionally, inputs were gathered from professionals in the development department of Agriculture and Farmers Welfare of the Government of India. Furthermore, experts from the study area were consulted via telephone to contribute insights into defining the value of non-resource constraints/goals and determining appropriate weights for these goals.

where $Y_c$ is the grain yield of cth crop; $Z_{\mathrm{\text{ct}}}$ is the binary coefficient of a crop belonging to tth crop group (cereal/pulses/oilseed) (1 if the crop belongs to tth crop group and 0 otherwise), ${\mathrm{\text{CPL}}}_t$ is the current/self-sufficient level of production of tth crop group.

To ensure that the level of greenhouse gas emission from crops (paddy) and livestock is controlled, an additional constraint of GHG emission (Eq. (11)) was included in the model. Livestock and rice are the major contributors to the total GHG emissions from agriculture (Vetter et al. 2017). Therefore, we considered GHG emissions (methane and nitrous oxide on CO2_e) from paddy and livestock only.

where $E_i$ is the emission factor of ith crop or livestock taken from published studies (referred in Table 2).

2.2. Model formulation

Goal programming intends to minimize the weighted deviations of multiple objectives from desired goals, which can be achieved by transforming the objective function into a constrained equation and then converting these into equalities by allowing under and over-achievements (McGregor and Dent 1993). Accordingly, for each goal ($G_j$), a pair of deviation variables ($d_j^{-}$ & $d_j^{+}$) were defined: (i) Equating the amount by which the solution overachieves the goal ($d^{-}$), and (ii) variable equating the amount by which solution falls short of goal ($d^{+}$). These absolute deviation variables were then expressed as relative terms ($D_j$). Now, the goal programming aimed to minimize the sum of relative deviations. The general mathematical representation of the model is given as follows :

where $w_j$, $D_j$ and $G_j$ are assigned weights, deviations variables and desired levels of jth goal, respectively.

The equations were solved using the Simplex method. It is important to note that overachievement was permitted for ≥ constraints (mainly for non-resource constraints like production of cereals, pulses and oilseeds), while the reverse was the case for ≤ constraints (land, labor, fertilizer, water use, GHG emission). For example, area under the crops beyond the available land was not allowed and the production of pulses below the existing production level was not allowed. Both, over and underachievement, were allowed for ‘=’ constraints. Further, based on the nature of the goal, overachievement/underachievement was rewarded or panelized by applying the weights. The weights were based on the experts’ opinion as mentioned above.

3. Case Study

The model was applied to the Bundelkhand region of India located central part of the country (23${}^{\circ }10{}^{\prime }$ and 26${}^{\circ }27{}^{\prime }$ N latitude and 78${}^{\circ }4{}^{\prime }$ and 81${}^{\circ }34{}^{\prime }$ E longitude) stretching from Indo-Gangetic Plain to the north and the Vindhya mountain range to the south (Figure 2). The region covers 7.08 million hectares spread across 13 districts of Madhya Pradesh (six districts) and Uttar Pradesh (seven districts) of which nearly 58% is under cultivation (GoI 2023b). It is one of the most resource-constrained regions of the country and faces severe agroecosystem stresses (Singh et al. 2014). Frequent droughts aggravate the stresses further and lead to loss of income, increased migration and risks of conflicts (Gupta et al. 2014; Agrawal et al. 2016; Singh et al. 2021a). As per the existing cropping pattern in the region, irrigation water is short by 33% of the requirement (Chand et al. 2020). The frequency of droughts has increased in recent decades (IWMI 2015) and agriculture in the region is highly vulnerable to extreme weather conditions (Thomas et al. 2016; Srivastava et al. 2020). Plenty of rainwater losses are due to run-off exacerbated by the region’s undulating topography and poor soils (Gupta et al. 2014). Considering the changing climate and uncompetitive economic conditions, there are challenges in producing more and retaining the workforce in rural areas. In such situations, farmers always strive to sustain livelihoods and seek to wrest the livelihood in diversified ways.

Chickpeas, wheat, sorghum, paddy, maize, barley, lentil, sesame, mustard, groundnut, soybean, peas, black gram, green gram, vegetables and fruits are the important crops of the region (GoI 2023b). Except for the perennial horticultural crops that have longer lifespans and generally do not need to be replanted every year, all the crops were included in the model. Agroforestry, especially the agri-horticulture-based production systems, has been identified as a potential intervention for addressing challenges like moisture stress in the region (Chavan et al. 2016; Kumar et al. 2013). The case study also sought to determine whether the provisioning of ecosystem services from agroforestry is compelling enough to incentivize farmers to adopt this practice. If not, to what extent do farmers need to be incentivized for non-provisioning ecosystem services?

3.1. The data

The case study is based on the data collected from 250 representative households under a comprehensive survey of the Cost of Cultivation of Principal Crops (CCPC) of the Directorate of Economics and Statistics, Ministry of Agriculture as well as other secondary sources. Under the CCPC scheme, data are collected following three-stage stratified random sampling; sub-districts (administratively called tehsils or taluka), a cluster of villages and operational holding as the first, second and third stages, respectively. The data are collected from 25 villages or clusters of villages, one from each sub-district of the region. From each cluster, a sample of 10 operational holdings is selected randomly, thus representing all the climatic conditions, farmer categories and crops of the region. The sample plot level data in summary form for a limited number of parameters can be accessed from the Directorate of Economics and Statistics, Government of India (DES 2023). Besides CCPC, data collected under frontline demonstrations (FLDs) conducted by Krishi Vigyan Kendras (KVKs) were also used. The data from FLDs pertain to crop or livestock production and protection technologies and management practices in the farmers’ field collected by scientists and subject matter specialized of district-level farm science centers called as KVKs. Besides, we also used secondary data collected from published and unpublished sources. The details of the data source are presented in Table 2.

4. Results and Discussion

Excel Solver was used for solving the models. Our motivation in setting out a spectrum of scenarios was to simulate the potential impact of different technologies and policy choices in addressing the region’s problems. Therefore, we build four different policy scenarios by emphasizing (1) Resource optimization (S1), (2) technology-led agriculture (S2), (3) improving water use efficiency (S3) and (4) resource-efficient sustainable intensification (S4). Levels of technology intensity differentiated from S2 to S3. These scenarios were compared with the existing cropping pattern (S0) on parameters like net economic margins and resource use, particularly land and water. The details of the scenarios are given in Table S2.

The first scenario (S1) was formulated by optimizing the existing resources with technologies and input use efficiency status-quo. The optimal cropping pattern obtained from GP models is presented in Table 3. The findings indicate that the resources in the Bundelkhand region are currently utilized in a sub-optimal manner, presenting an opportunity to enhance income and conserve water through optimal resource allocation. This inefficiency stems from market imperfections attributable to farmers’ limited awareness about the prices, improved technologies and best practices including selection of crops and varieties, and environmental benefits thereof. Less than half of agricultural households in the country accessing technical advice from one or other sources and the access in this particular region is further low (GoI 2021; Managuli et al. 2019). Optimization of existing resources alone (S1) could increase returns per unit of net sown area by 42% along with a 6.52% reduction in water usage (Table 4). Farm credit emerged as the most limiting factor in the region as in all the scenarios the entire available working capital of INR 88.31 billion had been exhausted (Table 4). The recommended optimal plan (S1) suggests expanding the cultivation of pigeon pea, sorghum and vegetables (Table 3, Column 3) by utilizing kharif fallow in the region. Currently, almost half of the net sown area of 4.12 million hectares is cultivated during the kharif, while the remaining half is left fallow. Optimization can help in utilizing these fallow lands and increasing the cropping intensity from 135% to 154% inferring that the area cultivated more than once in a year increased by 19%. To a limited extent, these crops have also substituted areas under water-intensive crops like paddy and sugarcane. Expanding the area under kharif crops, particularly sorghum, also helps in increasing the livestock population by 70%, primarily comprising local cattle (81%) and small ruminants (Table 5). A significantly higher area under other crops, mainly kharif fodder (sorghum and cluster bean), is allocated. This indicates that by optimizing the existing resources, the area under kharif fallow can be brought under cultivation and can help in increasing the availability of fodder for livestock. These findings also imply that the current practice of leaving cattle free to survive (Anna Pratha in local parlance) can be controlled, which otherwise destroys almost 25–35% of kharif crop produce (Saran et al. 2000). Small ruminant rearers in the region generally belong to the socially backward class most having land holding <2 ha and derive significantly higher share of income from this enterprise (NITI Aayog 2015; Singh et al. 2013a). Their livelihood can be uplifted by adopting this plan. Optimization offers 51% additional man-days, potentially checking seasonal migration.

In the second scenario (S2), we conducted simulations to assess the impact of high-yielding varieties of sesame (TKG-306), improved cattle (Kenwariya, an improved indigenous) breed, and advanced sowing methods (Raised-bed planting). The results unveiled a promising prospect wherein the incorporation of these technological advancements within the region exhibited the potential to expand the gross cropped area by approximately 2 million hectares compared to the baseline scenario (S0). This expansion primarily comes from the increased cultivation of sesame and sorghum translating into incremental production of oilseed by 42% and cereals by 2.35%.

In the third scenario (S3), the effect of increased water efficiency was also added to the three technologies that were included in scenario S2. It was assumed to increase the water use efficiency by 10% in the all the crops. Results showed that 22.39% of water can be saved under this scenario without affecting the net returns compared to S0. This highlights the potential of application of microirrigation and improving the efficiency of canal irrigation through different methods. The significant deviations in the cropping pattern in S3 when compared to S1 were as follows: (1) A shift in the area from pigeon pea to green gram, (2) relatively less area allocation for sesame and (3) a higher area in other crops, mainly under fodder crops (cluster bean in kharifand berseem in rabi). However, implementing microirrigation systems will require additional investment and will increase the cost of cultivation. Although the Government of India and state governments offer subsidies covering up to 90% of the total installation cost, which range from INR 32,200 to INR 90,000 per hectare (Sherief 2024; PIB 2024; Surendran et al. 2016), this cost is still high, especially for smallholders. Therefore, there is a need to develop and promote low-cost microirrigation systems. Studies suggest that the overall cost of microirrigation systems can be reduced to INR 25,000, making them feasible for replication in agro-ecosystems worldwide (Surendran et al. 2016; Kumar et al. 2022). Besides saving water, these low-cost systems also save energy having positive environmental implications.

Despite its potential to significantly contribute to sustainable land management, area allocation for agri-horticulture remained minimal (<100 ha) across all scenarios. The Central Agroforestry Research Institute, the only research institute in the country focused on research in agroforestry and related domains, is located in the region and actively develops and demonstrates models both locally and nationwide. A study conducted by the Institute in the region suggests that, despite concerted efforts and watershed development, the actual adoption of agroforestry land use by farmers was unsatisfactory. This was primarily due to regulatory hurdles in harvesting and transporting trees from private lands, lower yields of field crops when grown with trees and low survivability of trees during summer (Palsaniya et al. 2010). Additionally, the economic returns from trees including agroforestry take a longer period to materialize, with a very low annuity of returns in the initial years and significant price risks (Samriti et al. 2020). This arises from the lack of market and price-discovery systems for timber and also mechanisms to compensate farmers for the non-tradable benefits associated with tree plantation. Notably, agroforestry systems possess the capacity to generate substantial value, estimated at US$69 (∼INR 5,750)–US$167 (∼INR 13,900) per hectare per year solely through below-ground carbon sequestration (Kiran Kumara et al. 2023). The findings underscore the need for policy interventions aimed at recognizing and incentivizing the multifaceted benefits of agri-horticulture. The Government of India has recently launched the Green Credits Programme to create a market-based incentive for sustainable agricultural practices beyond just carbon emission reductions and envisaged that corporate sector would buy these green credits as part of their social obligations. However, its implementation requires concerted efforts in quantification and valuation of environmental benefits from such practices.

Despite receiving a good amount of annual rainfall on average, the region has perennially grappled with water scarcity due to high run-off. The southern part of the region, particularly the Vindhyan plateau, presents challenges with high slopes and impermeable rocks, resulting in excessive run-off and limited underground groundwater storage (Prakash et al. 1998; Chaurasia and Chandra 2021). However, this adversity also presents an opportunity for effective rainwater management. Therefore, the fourth scenario (S4) was devised to augment water availability by 24% through rainwater harvesting. The rationale behind this augmentation is grounded in the findings of studies conducted by Singh (2012), Garg et al. (2020) and Singh et al. (2013b). These studies highlighted that the soil moisture content in the degraded hills can be increased up to 24% if different rainwater harvesting techniques are used. The comparison of results revealed that the net returns could be enhanced by 61.54%, with GHG emission less by 13.52% compared to the existing cropping pattern. There was a slight deviation in the cropping pattern (compared to S3) with higher area allocation under wheat and other crops (vegetables) but less area under sorghum and sesame. Considering the unit cost of INR 88/m³ of water harvested based on the average cost of harvesting rainwater worked out in various studies (Rao et al. 2010; Das et al. 2017; Kumari et al. 2017) also after adjusting the time value of money, implementation of this scenario will require INR 81 billion higher investments by the Government. This was calculated by multiplying the additional water harvested by the prices per unit of harvesting and arriving at current prices using discounting factors. The rate of discounting applied was the prevailing interest rate in the market, i.e., 6.5%/annum.

5. Conclusion and Policy Implications

Agricultural planning problems at various levels are complex decision-making challenges. They involve multiple objectives, diverse crops and allied enterprises, with interactions between these enterprises often leading to conflicting objectives. In this study, a multi-objective crop and livestock allocation model (MOCLAM) was developed for optimal land allocations for different crops and the corresponding head of livestock that can sustain in the given cropping pattern. The constraints incorporated in the model were the availability of irrigation water, land, labor (monthly), fertilizers (N, P &K) and capital. In addition, two non-resource constraints, self-sufficiency (food, edible oilseeds and fodder) and emission of greenhouse gases, were also incorporated into the model. The uniqueness of the proposed model lies in three key aspects. First, the model adopts a systemic view of agriculture by simultaneous optimization of both crops and livestock, recognizing their interdependence as a pivotal strategy for promoting sustainable agriculture in the region. Second, MOCLAM introduces a quasi-dynamic framework where livestock numbers are determined endogenously, based on the evolving cropping pattern which is uncommon in agriculture. This adaptive methodology enhances the model’s responsiveness to changing conditions, bolstering its robustness in addressing the dynamic nature of agriculture. Third, the model comprehensively incorporates water considerations from both surface and groundwater sources, a dimension previously overlooked in crop and enterprise planning.

MOCLAM effectively addresses the conflicting goals inherent in agricultural enterprise planning by simultaneously optimizing crop and livestock allocations, taking into account the complex interplay between these enterprises. By incorporating constraints such as irrigation water, land, labor, fertilizers and capital, as well as non-resource constraints like self-sufficiency and greenhouse gas emissions, MOCLAM ensures that diverse objectives are balanced. The model’s systemic approach recognizes the interdependence of crops and livestock, promoting sustainable agriculture. Its quasi-dynamic framework, which adjusts livestock numbers based on evolving cropping patterns and resultant fodder availability, enhances adaptability and responsiveness to changing agricultural conditions. Similarly, the broader definition of water resource constraint further increases the model’s practicality and acceptability. Overall, the comprehensive and adaptive methodology enables MOCLAM to manage and reconcile the conflicting objectives in agricultural planning effectively.

The applicability of MOCLAM was demonstrated with a case study of the Bundelkhand region of India, one of the most resource constraint regions of the country. The optimization results show that it is possible to increase the net economic margins of the farmers by up to 42%, save water by 6.5% and utilize huge fallow lands in the kharif without technological and policy interventions. These necessities educate farmers and strengthen agricultural extension system in the region. The availability of working capital appeared one of the major constraints in adopting alternative cropping pattern. Therefore, institutional credit at affordable interest rates could not only encourage farmers to use improved technologies in field crops but also adopt scientific livestock production and management practices that encourage utilizing kharif fallow. Water saving up to 22.39% without compromising economic margins is possible if technological interventions like micro-irrigation and improved sowing methods are adopted. The development of low-cost drip and sprinkler irrigation systems and easy availability of loans could encourage the farmers to accelerate the adoption of micro-irrigation in the region. Findings further revealed that farmers’ returns can be increased up to 62% if the water availability is increased by 24%. However, this will require an additional investment of around INR 81 billion for rainwater harvesting. The investment in agriculture and rural development to be strategically channeled to incentivize the revival of deep-rooted traditional water harvesting structures by involving local governments and the communities.

All five scenarios suggested diverting the area from the conventional rice–wheat cropping system towards pulses and oilseeds. Increasing the number of local cattle and small ruminants, particularly goats, was also recommended across all plans. However, none of the scenarios identified agri-horticulture as a feasible alternative to field crops in the region. Besides value chain development for fruit-based products from fruit crops, there is a need to develop better price discovery mechanisms for marketing timber products from these crops. Enhancing farmers’ skills in primary processing at the field level can help reduce the transportation costs of wood. These findings suggest that alternatives to crop husbandry, like agri-horticulture and agroforestry, are unlikely to be adopted by farmers unless they receive compensation for social gains such as carbon sequestration and groundwater recharge. Also, alternative high-value agroforestry practices like fast-growing teak-based need to be promoted and a database on costs, returns and social gains from such practices — tested both in experimental plots at research stations and on farmers’ fields — should be developed for evidence-based policy decisions. The social benefits make up a significant portion of the total value of ecosystem services in agroforestry, with carbon sequestration alone contributing 20–25% depending on the system type (Jose and Udawatta 2021; Singh et al. 2021b). It is estimated that non-tradable services contribute around Rs. 7,759/ha/year, primarily due to improvements in soil fertility, higher rates of carbon sequestration and nutrient retention in soil from reduced soil erosion (Kiran Kumara et al. 2024). Providing this value to farmers could perhaps motivate the farmers to increase the adoption of agroforestry. Farmers can also be compensated for lower crop yields, which save significant amounts of water. This calls for developing an institutional mechanism to incentivize farmers to adopt these practices.

Supplementary Material

ORCID

Prem Chand  https://orcid.org/0000-0001-8645-4107

Rajni Jain  https://orcid.org/0000-0002-8493-7858

Note

¹ Hard constraints are constraints that must be satisfied at all times, while a soft constraint is a want that to be satisfied as much as possible if the cost of doing so is not high (Kendall 1975) or in simple terms it is feasible to do.