Use of Regression Trees to link Uromycladium acaciae symptoms to Acacia Mearnsii growth variates

Payn R.G. and Little K.M.
Publication Type
Technical Note
Publication Year
TN 05-2018

From 2012/13 Uromycladium acaciae (wattle rust) has spread throughout the Acacia mearnsii (black wattle) growing areas of South Africa. The newly emerged disease affects trees of all age classes and causes growth reductions and mortalities with severe infestations. Fungicides have been tested and found to be effective for managing the disease, with the timing of application necessary for optimal control. If fungicide applications could be linked to the emergence of key wattle rust symptoms, then this would assist with the timing of applications only when necessary, rather than as a preventative spray. Wattle rust symptoms were analysed from the untreated Control plots of two trials, one in the KwaZulu-Natal Midlands and one in southern KwaZulu-Natal, to determine whether wattle rust disease expression could be linked to black wattle tree growth. Regression trees were used for the analysis, as linear and multiple regression techniques would be unsuitable for the type of data collected. Regression trees were overfitted and attempts at testing the robustness of the model by cross-validation were unsuccessful. No individual symptom emerged as a significant predictor of tree growth, indicating that fungicide application should take place with the onset of any of the wattle rust symptoms tested.