Exploring Complexity in a Human–Environment System: An Agent-Based Spatial Model for Multidisciplinary and Multiscale Integration

Abstract

Traditional approaches to studying human–environment interactions often ignore individual-level information, do not account for complexities, or fail to integrate cross-scale or cross-discipline data and methods, thus, in many situations, resulting in a great loss in predictive or explanatory power. This article reports on the development, implementation, validation, and results of an agent-based spatial model that addresses such issues. Using data from Wolong Nature Reserve for giant pandas (China), the model simulates the impact of the growing rural population on the forests and panda habitat. The households in Wolong follow a traditional rural lifestyle, in which fuelwood consumption has been shown to cause panda habitat degradation. By tracking the life history of individual persons and the dynamics of households, this model equips household agents with “knowledge” about themselves, other agents, and the environment and allows individual agents to interact with each other and the environment through their activities in accordance with a set of artificial-intelligence rules. The households and environment coevolve over time and space, resulting in macroscopic human and habitat dynamics. The results from the model may have value for understanding the roles of socioeconomic and demographic factors, for identifying particular areas of special concern, and for conservation policy making. In addition to the specific results of the study, the general approach described here may provide researchers with a useful general framework to capture complex human–environment interactions, to incorporate individual-level information, and to help integrate multidisciplinary research efforts, theories, data, and methods across varying spatial and temporal scales.

Key Words:

• Agent-based modeling
• complexity theory
• multidisciplinary and multiscale integration
• household socioeconomics and demographics
• giant panda conservation

Acknowledgements

We thank Guangming He, Jinyan Huang, Shiqiang Zhou, Zhiyun Ouyang, and Hemin Zhang for their assistance in data acquisition. We appreciate the valuable comments from Dr. Michael Goodchild and the four anonymous reviewers. We are also indebted to financial support from the National Science Foundation (NSF), National Institute of Child Health and Human Development (NICHD), American Association for Advancement of Sciences, the John D. and Catherine T. MacArthur Foundation, and Michigan State University. Special thanks go to the online help from Swarm Development Group, in particular Dr. Paul Johnson at the University of Kansas.

Notes

Notes: 

(1) The selection of the variables for the combination tests depends on the results of the sensitivity analysis: the most sensitive factor in each of the four categories in Table 3 is selected.

Notes:  

(1) House buffer distance (see Table 2 for its definition).

(2) The standard error is in the parentheses following each average value.

(3) The letters “s” and “b” stand for “small” and “big”, respectively, corresponding to the minimal and maximal values of each parameter.

Notes:  

(1) The first numbers in the spaces are the default values in the model, and the second values are those used in the associated scenarios.

Notes: 

¹The default values for the associated variables based on field observations. Habitat area (km²) under the default values at year 2016 is 575.49 (0.69). Minimal and maximal values for the associated variables used to test the model.

²The habitat area (km²) under minimal and maximal values for the associated parameter. The numbers in the parentheses are standard errors.

³An adult child who remains in his/her parental home after marriage.

Notes:  

¹The perturbation range of 50 percent is determined in consideration of: (1) it should be relatively small (otherwise we can use extreme tests as in Table 2); and (2) the response magnitude of the habitat change should be large enough for our calculation. An alternative of –50 percent perturbation is not included simply for space consideration.

²The standard error is in the parentheses following each average value.

³We use two-sample paired t-test at the 0.05 level to test whether the predicted habitat area is different from the baseline value at year 20.

⁴Here only 20 percent perturbation because an upper birth age of 75 years old (a 50 percent increase) does not make sense in the real world.

⁵All the 4 parameters under the category “Migration” have insignificant impact on panda habitat over 20 years. The numbers reported here are simulation results over 30 years, and the sensitivity index is calculated using the amount of habitat over 30 years (568.20 km²).

1. For link to these tools, go to http://wiki.swarm.org/wiki/Main_Page (last accessed on 17 June 2004).

2. Electricity is the readily available substitute for fuelwood in the Reserve, subject to government price control and some quality problems (An et al. 2002). Other energy sources such as coal, charcoal, biogas, and sun/wind power are not used and no market exists for them.

3. Though we observed some people who took temporary jobs in outside areas (primarily big cities), they still had their residence registration (known as Hukou) license in Wolong. More importantly, they often come back to Wolong during busy agricultural seasons and Chinese spring festivals and conduct resource-related activities such as fuelwood collection. Thus, they are not treated as out-migrants.

4. Shiqiang Zhou from the Wolong Nature Reserve is an experienced researcher with extensive knowledge in local biology, ecology, and socioeconomic and demographic situations. The authors have closely worked with him to collect the data, build the models (including earlier models as published by An et al. 2001, 2002,;2003), and discuss the model outcomes during 1998–2004.