Comparing Artificial Neurla Network with Least Square Regression Techniques for Lai Retrieval from Remtoe Sensing Data

Leaf area index (LAI) is an important input in spatially distributed modeling of surface energy balance, evapotranspiration, and vegetation productivity. Remote sensing can facilitate the rapid collection of LAI information on individual fields over large areas in a timely and cost-effective manner. In this study, we developed a set of LAI models using least square regression (LSR) and artificial neural network (ANN) techniques, and evaluated them for their ability to retrieve LAI from Landsat Thematic Mapper (TM) data. The LAI measurements were made in 47 randomly selected commercial fields in Moore and Ochiltree Counties located in the Texas High Plains. Ground measurements were coincided with the Landsat 5 overpasses over the study area during the 2005 growing season. Numerous spectral vegetation indices were examined for retrieving LAI with both LSR and ANN models. The structure independent pigment index (SIPI) showed the highest correlation with LAI. Results showed that the best ANN and LSR models gave R values of 0.91 and 0.84, respectively, indicating that ANN-based models were superior to LSR-based LAI models. A further evaluation of the methods is planned with an independent dataset from the 2007 Bushland Evapotranspiration and Agricultural Remote Sensing Experiment.