1 Overview

The SExI forest simulator focuses on tree-tree interactions in a mixed multi-species agroforest. The high level of structural complexity of such traditional agroforestry systems defies classical forestry approaches when it comes to optimizing management practices. To cope with this complexity, farmers have adopted a tree-by-tree management approach, which is closer to gardening than to any usual tropical forestry or estate crop management model. Individual tree care and regular tending takes the form of seedlings transplanting, selective cleaning and felling, adjusted harvesting intensity.

Farmers’ approach appears to be in line with two basic tenets of biology: first, individuals are all different with behavior and physiology that result from a unique combination of genetic and environmental influence, and second, interactions are inherently local. Based on the same premises a computer model was developed to explore different management scenarios. The model uses an object-oriented approach where each tree is represented by an instance of a generic class of tree. The simulated object trees, mimicking real trees, interact through modifying their neighbors' environment. These modifications are mediated through two major resources: space and light. A 3D representation of a one-hectare plot of forest serves as the grounds for the simulation of this competition. The major objective of such a model is to get a coherent dynamic representation of a complex system, where complexity refers here to the assemblage of locally interacting individuals with different properties more specifically to the degree of interconnection between individual trees. The model provides insight on what are the critical processes and parameters of the dynamic of the system. It should also allow exploring prospective management scenarios, help assessing the relevance of present management techniques etc. Model sensitivity tests confirm the importance of the parameters related to tree geometry. This directly stems from the fact that competition is simulated by means of spatial interactions, so that anything that alters either the shape, the size, or the relative position of the trees have direct impact on the outcome of the competition and therefore on the growth dynamic. These elementary influences are straightforward but their effect at different times and scales are difficult to predict without simulating because of the numerous feedback loops at work and the non-linear dynamics of the system. To illustrate this, let’s examine very simple cases. By simulating growth in a mono-specific stand of regularly spaced trees planted at increasing densities, we observe the following response. Planting at medium density translates into growth in height of the trees in the center of the plot being superior to that of border trees, which is a response to the increasingly limited access to light of the trees in the center of the plot. When planting density is increased further though, growth in height of the trees in the center of the plot becomes less than their neighbors: the level of competition is so high that these trees get overtopped and suppressed by border trees in more favorable position with respect to access to light. Another simple test shows that ability to respond to low light availability by enhanced growth in height (a response, which occurs at the expense of growth in diameter) appears to be advantageous under specific conditions and disadvantageous under others. If all species in the mixture share the same ability and the same sensitivity to light level then this potential competitive advantage turns out to be disadvantageous both for individual tree growth and for overall plot productivity. But when trees with different sensitivity to light level or different ability to alter their allocation of growth between height and diameter occur in a mixture then this capacity proves to be an effective competitive advantage for individual species. By accelerating the establishment of a multi-strata structure it also increases the overall productivity of the plot through better allocation of spatial resources. Similarly, rather counter intuitively, an increased growth rate for a given crown size appears to be an advantage for a species under certain circumstances but not all: under very crowded conditions large crowns (showing low efficiency in terms of light and space utilization) can show competitive advantage by suffering less from crown encroachment and shading out competitor more efficiently. These are but a few examples of the insight such generic models can bring. More direct applications of the model include comparing alternative scenarios in terms of financial return for instance involving rotational versus permanent agroforests, etc.