Predicting Diameter Distributions of Longleaf Pine Plantations: A Comparison Between Artificial Neural Networks and Other Accepted Methodologies
Artificial neural networks (NN) are becoming a popular estimation tool. Because they require no assumptions about the form of a fitting function, they can free the modeler from reliance on parametric approximating functions that may or may not satisfactorily fit the observed data. To date there have been few applications in forestry science, but as better NN software and fitting algorithms become available, they may be used to solve a wide variety of problems-particularly problems where the underlying relationship between predicted and predictors is unknown. We benchmark tested an aitemative to the traditional Weibull probability distribution function, diameter-at-breast-height moment, and direct parameter prediction models for approximating stand-diameter distributions. Using a feedforward backpropagation network, we demonstrated that NN are a somewhat better option. Unlike Weibull approximations, NN solutions cannot easily be mathematically constrained to match known reality constraints, but this difficulty is easy to overcome in practice.