Modeling Avena fatua seedling emergence dynamics: An artificial neural network approach

CHANTRE GUILERMO; BLANCO ANÍBAL M.; LODOVICHI, MARIELA V.; BANDONI J. ALBERTO; SABBATINI, M. RICARDO; LOPEZ RICARDO; VIGNA MARIO; GIGÓN RAMÓN

COMPUTERS AND ELETRONICS IN AGRICULTURE

ELSEVIER SCI LTD

Lugar: Amsterdam; Año: 2012 vol. 88 p. 95 - 102

0168-1699

Avena fatua is an invasive weed of the semiarid region of Argentina. Seedling emergence patterns are veryirregular along the season showing a great year-to-year variability mainly due to a highly unpredictableprecipitation regime. Non-linear regression techniques are usually unable to accurately predict fieldemergence under such environmental conditions. Artificial Neural Networks (ANNs) are known for theircapacity to describe highly non-linear relationships among variables thus showing a high potential appli-cability in ecological systems. The objectives of the present work were to develop different ANN modelsfor A. fatua seedling emergence prediction and to compare their predictive capability against non-linearregression techniques. Classical hydrothermal-time indices were used as input variable for the develop-ment of univariate models, while thermal-time and hydro-time were used as independent input variablesfor developing bivariate models. The accumulated proportion of seedling emergence was the output var-iable in all cases. A total of 528 input/output data pairs corresponding to 11 years of data collection wereused in this study. Obtained results indicate a higher accuracy and generalization performance of theoptimal ANN model in comparison to non-linear regression approaches. It is also demonstrated thatthe use of thermal-time and hydro-time as independent explanatory variables in ANN models yields bet-ter prediction than using combined hydrothermal-time indices in classical NLR models. The best obtainedANN model outperformed in 43.3% the best NLR model in terms of RMSE of the test set. Moreover, thebest obtained ANN predicted accumulated emergence within the first 50% of total emergence 48.3% bet-ter in average than the best developed NLR model. These outcomes suggest the potential applicability ofthe proposed modeling approach in weed management decision support systems design.