Urban tree growth modelling with artificial neural network

Abstract

Municipal administrations devote large budgets in order to preserve the integrity of their urban forests. However, urban ecological conditions are extreme for tree growth and survival: recent downtown districts plantings suffer unusually high mortality rates. Strict management is therefore mandatory but critical information is missing on procedures that would provide for successful planting and growth. The empirically acquired knowledge of practitioners must be captured into computerised models to improve the efficiency of urban forest management and existing tree data banks. A back-propagation ANN model was built to provide assistance in deciding urban planting sites and to predict tree parameters. The first expected outputs from the ANN study were the accurate predictions of tree diameter at breast height (DBH measurement), tree diameter growth index (DBH increment) and total crown growth index. Two levels of modelling were performed; a general prediction model for all species under study and specific species by species models. Results are consistent for both levels. DBH measurement and total crown index are predictable using artificial neural networks and should be retained as significant parameters to model tree growth. DBH increment projection is more variable and is species dependant. Despite the interest of designing a growth projection model on DBH increment, use of this parameter is probably unrealistic.