Diversity deficits in modelled landscape mosaics Grace B. Villamor a, ⁎, Meine van Noordwijk b , Quang Bao Le c , Betha Lusiana d , Robin Matthews e , Paul L.G. Vlek a a Centre for Development Research (ZEF), University of Bonn, Germany b World Agroforestry Centre (ICRAF), Bogor, Indonesia c ETH Zürich, Institute of Environmental Decisions (IED), Zürich, Switzerland d Institute for Plant Production in the Tropics and Sub Tropics (380a), Hohenheim University, Germany e Macaulay Land Use Research Institute, Craigiebuckler, Aberdeen, Scotland, UK abstract article info Article history: Received 25 May 2010 Received in revised form 4 August 2010 Accepted 28 August 2010 Available online 7 September 2010 Keywords: Diversity deficit Landscape mosaics Agents' decision-making Hybrid models Multi-agent system models We outline several diversity factors that modellers and models can include directly or indirectly in order to improve the accuracy and usefulness of the model. Without considering these factors, different types of diversity deficit can arise. These deficits can be considered in three domains: 1) in the real world where actual diversity is less than a potential state that is deemed desirable (hence we worry about loss of biodiversity and cultural diversity); 2) in modelling of the real world (where ‘residual variance’ may represent a diversity deficit of the model); and 3) in our recognition of the driving forces that are used to construct a model (a diversity deficit due to oversimplification). The goal of this review is to use these three domains of diversity deficit to evaluate existing models, with a longer term goal of creating a more robust framework for assessing landscape models in the future. To that end, we evaluate the behaviour characteristics and routines of agents in some current models. We also address one of the fundamental challenges to modelling diversity, which is the integration of non-economic motivations in the decision making of human agents. © 2010 Elsevier B.V. All rights reserved. 1. Introduction Diversity in landscape mosaics influences the production of goods and services and the tradeoffs between them at multiple scales through the lateral flows of water, nutrients, soil, organisms, fire and information (van Noordwijk et al., 2004). Diversity of this sort often gets lost through decision making oriented toward homogenization, which is usually linked to economies of scale in production for markets, or a result of poor awareness of the potential relevance of diversity. Although most of the diversity may be seen as neutral to current or planned functions, and is tolerated rather than actively pursued, loss of diversity can reduce our options for the future. Diversity deficits can be considered in three domains: 1) in the real world where actual diversity is less than a potential state that is deemed desirable (hence we worry about loss of biodiversity and cultural diversity); 2) in modelling of the real world (where ‘residual variance’ may represent a diversity deficit of the model); and 3) in our recognition of the driving forces that are used to construct a model (a diversity deficit due to oversimplification). The hypotheses we pursue here is that diversity deficits of the third type contribute to those of the second, which does not allow us to stem the trend to increasing deficits of the first type. Land use planning has conventionally focused on ‘what is best’ for different parts of the map, taking into account the properties found throughout the landscape the landscape, the functions and intentions of stakeholders, their interrelationships between multiple constraints and tradeoffs. The resulting plans typically exhibit a higher degree of homogeneity. Land use planning therefore is associated with a coarsening of the landscape mosaic, even if management objectives include maintaining total diversity. Multi-functionality often seems to be retained at coarser scales, but is reduced at high resolution. Optimization routines of most models are set up to identify the single-best choice among an array of options, reducing diversity in the expectation of increased efficiency. In contrast, most people's decisions are about managing portfolios using ‘a-bit-more-of-this, a- bit-less-of-that’ in decisions of relative resource allocation (e.g. financial, land, labour, etc.). A large but poorly quantified part of landscape diversity is explaining why farms make divergent man- agement decisions given essentially the same set of options at any point in time (Swift et al., 2004). Technical opportunities to shift from single agent to multi-agent modelling have lead to a large number of modelling attempts that include agent diversity. Diversity is the variation between components of a system. The interpretation of ‘diversity’ thus depends on the system boundaries. For conservation strategies, diversity in landscape mosaics can result Ecological Informatics 6 (2011) 73–82 ⁎ Corresponding author. E-mail addresses: gracev@uni-bonn.de (G.B. Villamor), m.vannoordwijk@cgiar.org (M. van Noordwijk), quang.le@env.ethz.ch (Q.B. Le), b.lusiana@cgiar.org (B. Lusiana), R.Matthews@macaulay.ac.uk (R. Matthews), p.vlek@uni-bonn.de (P.L.G. Vlek). 1574-9541/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.ecoinf.2010.08.003 Contents lists available at ScienceDirect Ecological Informatics journal homepage: www.elsevier.com/locate/ecolinf