Simulating Land-Use Entelechy using Multi Agent-based Behavioral Economic Landscape (MABEL) Model

Abstract

Land-Use Entelechy can be described as the process in which function transforms in form. Multi Agent-based Behavioral Economic Landscape (MABEL) model, simulates complex decision-making of agents in a parcel-based framework. The model architecture is based in the Swarm modeling environment, and extends its capabilities via utilization of multiple software components (GIS, statistical and Bayesian Modeling, decision-modeling), and a client-server communication protocol framework. MABEL agents are corresponding to a coupled socio-economic and level-3 GIS classification of the landscape, and conceptualize individuals as land-owners in residential, commercial, agricultural, forested, and other land-use classifications. Parcel-based identification of agents through a GIS-digitized database, is combined with a socio-economic attribute database (representing for each agent the demographic, economic and social characteristics related to its ownership), unique for each initial agent in the simulation. Using a finite-horizon Markov decision-making iterative framework and adaptive Bayesian Belief Networks (BBN's), the agents maximize their expected utility related to land-use change in a land-market framework of land acquisition. The primary set of questions that MABEL simulation addresses is how individual preferences and utility assumptions are related to land-use acquisition and changes observed over time. The model observes the adaptive changes in these preferences and how are related with trends and changes in land-use characteristics and social processes associated with them such as sprawl, loss of agriculture, population-changes, employment and housing characteristics, etc. Interactive decision-making scenarios are utilized to simulate different stream of policy-making rules, and community-based approaches to sustainable futures. The results indicate the variability and versatility of different key decisions associated with planning and policy-making. Inferences for and from the future related to decisions to be made in the present allows for exploring a decision-making horizon, capable to assist policy-makers to achieve their forward-planning goals, identify potential problem areas related to land-use and spatial interactions, and explore the uncertainty and variability of suggested sustainable futures. In the model, the decision-theoretic control of the agents' actions is achieved through a combination of multi-attribute economic utility framework, and a behavioral probabilistic change framework, such as a robust Bayesian classifier for Belief Networks. The transition-decision-action sequence converges to a dual property setting of optimal action - optimal policy, using a non-linear extended Kalman-filtering (EKF) estimation of spatial and temporal noise covariance from observations. The adaptability of agents' through temporal and spatial interactions thus, does not depend on the assumption of fixed (or constant) utility preferences over time, but assesses their decision-theoretic environment through an observation - correction process. Evidence entering the agents' perceptional environment as observations through the landscape, other agents' action-space, and decision-based rules by policy-makers, are capable of modifying or re-evaluating future agent decisions. A critical question to be assessed in this paper is whether a series of sequential local optimal decisions (in a coupled temporal - spatial scale) of the agents, can, and under which policy-making and uncertainty conditions provide the basis of achieving a global optimum (resilient) across these scales.