PREDICTION OF LEAF AREA INDEX FOR SOUTHERN PINE PLANTATIONS FROM SATELLITE IMAGERY USING REGRESSION AND ARTIFICIAL NEURAL NETWORKS

Leaf Area Index (LAI) is an important predictor of southern pine forest productivity. In this study several models for estimating LAI values using remote sensing were developed and tested. Of these, a best model was selected based on performance and potential for operational application. The generalized southern pine LAI predictive model (GSP-LAI) was developed using artificial neural network (ANN) multivariate regression that incorporated important local information including phenological and climatic data. In validation tests the model explained> 75 % of variance (R2 = 0.77) with an RMSE < 0.50. The GSP-LAI model was applied to Landsat ETM+ image of the Bradford forest, north of Waldo, FL. Within the extent are substantial slash (Pinus elliottii) and loblolly (Pinus taeda) pine plantations. Based on image and stand data LAI values for 10,797 ha were estimated to range between 0 and 3.93 m2 m-2 with a mean of 1.53 m2 m-2