Elsevier

Agricultural Systems

Volume 164, July 2018, Pages 264-276
Agricultural Systems

Development of an agent-based model for estimation of agricultural land preservation in rural Japan

https://doi.org/10.1016/j.agsy.2018.05.004Get rights and content

Highlights

  • MAS model in this research aims not only for virtual experiments but also for application to actual agricultural plan formulation.

  • It is expected as a tool for constructing a simulation model from the result of simple questionnaire survey and analyzing it.

  • Especially in Japan it is decided that subsidies to rice farmers are decided to be abolished and it is urgent to streamline management by grouping farmers.

  • The data used for the analysis is somewhat old, but contents to be provided are covered, including model construction, simulation by scenario analysis, and verification.

  • Future developments such as linkage with GIS and promotion of introduction to farmers are greatly expected.

Abstract

In rural Japan, the lack of successors for aging farmers has become a serious problem, given that these areas experience a population outflow as well. In response, national authorities have promoted reconsideration and strengthening of regional agricultural management systems. In order to achieve consensus for such a transition, it is important to streamline this agricultural management. In this study, we constructed an analytical simulation model based on multi-agent simulations to support such changes. With this model, we investigated the effectiveness of deliberate organization of agricultural management. First, we collected data on farmer behavioral patterns and intentions. In addition, we gathered data at individual farm level with a field survey, and predicted an initial trend (Trend_Simulation). In order to compare simulations with the Trend_Simulation, we assumed that the future labor force in the model settlement was centralized and performed the work as an agricultural organization (Systematic_Simulation). The results from Trend_Simulation predicted that farmland degradation would occur from 2010 onwards, after which the amount of abandoned cultivated and fallow land would increase rapidly. In contrast, for the Systematic_Simulation, as a result of increased management efficiency through labor force accumulation and joint use of agricultural machines, abandonment of cultivated land would not occur for at least 20 more years. Finally, expansion of management scale per individual farm through land leasing between farms was predicted to decrease gradually in the Trend_Simulation, but to increase in the Systematic_Simulation.

Introduction

In rural Japan, a lack of successors for aging farmers has become a serious problem, given the population outflow from rural areas. In response to this shift in population composition, national authorities have undertaken a structural reform of regional agricultural areas, and promoted a systematization of individual farm management systems [i.e., collective farming systems (CFS)] as a way to ensure agricultural succession. To organize CFS successfully and efficiently, and to consolidate similar attitudes with the farm landowners, predictions about future agricultural prospects would be a useful in assisting such a reform (Van Ittersum et al., 1998).

Most previously reported prediction models for evaluating regional agricultural planning at community level have been based on a linear programming theory, similar to mathematical programming. For example, Nanseki (1991) proposed a model that applied multiple objective linear programming (MOLP) to a farm management plan in a small area in Japan, and validated its decision-making capacity. In addition, land use valuation modeling, which unifies mathematical programming and geographic information systems (GIS) to determine a normative optimal solution, has been developed. For example, Campbell et al. (1992) presented an optimal land use planning model that considered the export and import of agricultural products in Antigua, a small country in the West Indies. Likewise, Chuvieco (1993) developed an urban land use planning model for Castellón, off the coast of Spain. Sante and Crecente (2007) validated MOLP for the management plan of the Terra Cha district in Spain. They assumed different situations in which the prior land value was based on the intended land use (e.g., social, economic or environmental purposes). However, none of these regional agricultural models have direct farmland leases and farm-work trusts, which both include interactions among farms within a farm management system.

In light of the above studies, it is important to capture the behavior of farms as a disaggregate model. In other words, we have to regard farming activities being managed by several farms with different management resources and intentions, in a multi-agent system (MAS) within a rural community. In this study, we constructed a prototype simulation model [i.e., agent-based simulation model for agricultural planning (ASMAP)1], based on the MAS, to predict future regional agriculture scenarios. Data collected from an actual agricultural settlement were used to parameterize the ASMAP, and the results were verified for the accuracy and validity of the model.

Farmer behaviors are diverse and vary from country to country. Japan's current policy for supporting small farmers is changing dramatically. Therefore, farmer behavior has become extremely diversified. Nevertheless, although there are many articles on the methodologies used to predict ideal farmland conservation based on the optimization of management resources, there are few studies that model farmer behavior that incorporate diversity and bounded rationality as variables. Although the data used in this study is less recent, from the time of the survey to the present, there has been little progress made in the field of modeling research. Although there are many aspects to be clarified such as the generalization of the interfaces, social value can be found in the spread of simulation technologies applied to this type of research.

Section snippets

What are “multi-agent systems”?

Multi-agent systems were developed as a method for analyzing and predicting social phenomena or human activities using an artificial social model. This method has been extensively applied since Schelling first developed the self-forming neighborhood model (Schelling, 1978). The model predicts a segregation of many individual entities, influencing individual actions based on the degree of satisfaction with an allocation of attributes to the neighborhood. Axelrod (1997) developed various

Model outline

The ASMAP outline based on the ODD (Overview, Design concepts, Details) protocol is described in Table 1. The ODD protocol is a universal format guideline designed for the MAS model (Grimm et al., 2010; Müller et al., 2013). A brief explanation of each element of the ODD on ASMAP is presented in Table 1, while additional details are described after this subchapter.

Study area and data collection

As the study area we used the settlement Yamada (from here on abbreviated as settlement Y) in Kamikawa, which is a small town in the

Analysis framework

The performance of the ASMAP was validated using a comparison analysis based on future predictions. First, we added data about the farmer behavioral patterns and intentions, as well as capital levels of individual farms that were obtained by a field survey (data are omitted), and predicted an initial trend (Trend_Simulation). In order to compare simulation options with the Trend_Simulation, we assumed that the future labor force in the model settlement was centralized as an FCG, and performed

Validation method

To determine the accuracy of the simulation results from the ASMAP we used the following methods. An uncertainty analysis was conducted to verify the validity of the model. Then, by analyzing the relationship between the outcome and the main variables the sensitivity analysis was conducted to understand the structure of the model more deeply. The uncertainty analysis was used to evaluate the extent in which the outcome fluctuated by predicting the survey time point after setting the model with

Conclusion

In this study, we used an ASMAP to predict future progress in agricultural-work trusts, land lease areas, as well as agricultural structural changes, for a small area. And we confirmed that farmland preservation during 20 years will be enough possible by promoting the agricultural land lease and allocating agricultural labor force optimally of even if the future agricultural labor force is decreasing. The effects of CFS are clearly observable, and the forecast map of farmland conservation

Acknowledgements

We thank the work of all collaborators involved in this research. The Kozo-Keikaku Multi-Agent Simulator (KK-MAS), used to develop ASMAP, was kindly provided by Kozo Keikaku Engineering Inc., Tokyo, Japan. And we appreciate all farmers in kamikawa town helping our laborious suevey.

The research was supported by JSPS KAKENHI Grant Numbers 26304034, 16H03311.

References (33)

Cited by (23)

  • Reveal the severe spatial and temporal patterns of abandoned cropland in China over the past 30 years

    2023, Science of the Total Environment
    Citation Excerpt :

    Cropland abandonment has positive and negative impacts on the environment. The negative impact mainly refers to reducing the contribution of agriculture to GDP (Baumann et al., 2011), which harms the development of agriculture and food security (Ito et al., 2016; Yamashita and Hoshino, 2018). Objectively, the abandonment of cropland reduces the frequency of land use, but can protect soil fertility, and can better guarantee the comprehensive production capacity of cropland (He and Ma, 2002).

  • Technological innovation, industrial structural change and carbon emission transferring via trade-------An agent-based modeling approach

    2022, Technovation
    Citation Excerpt :

    Compared to statistical models, econometric regression models, and CGE models, by modeling the emergent phenomena resulting from the interactions of abundant heterogeneous agents to exhibit dynamic changes in a complex macroeconomic structural system, ABM can be better used to examine the effect of changes on external conditions, like technological innovation policies on individual entities at the micro level (Ausloos et al., 2015). Therefore, applications of the ABM method have spanned a broad range of areas, such as modeling residents' behavior in the stock market and financial marketplace (Paulin et al., 2018; Petrović et al., 2020), modeling farmers' investment decisions in an agricultural region (Maes and Van Passel, 2017; Yamashita and Hoshino, 2018; Mack et al., 2020), and understanding consumers’ purchasing behavior (Rai and Henry, 2016; Sturley et al., 2018; Westphal and Sornette, 2020). To address the two aforementioned limitations, using an ABM approach within an IO analytical framework, this paper proposes the diffusion of technological innovation behaviors in simulation models.

View all citing articles on Scopus
View full text