Predicting leaf area index in wheat using an improved empirical model

Leaf area index (LAI) is typically estimated from remote sensing data acquired at nadir position. Known issues of this mono-angle approach include a saturation limit at intermediate values of LAI and inadequacy to represent any structural characteristics of vegetation. In this study, we present an improved LAI estimation model that incorporates multiangle reflectance data exploring the feasibility of addressing these issues, especially for LAI estimation in winter wheat. The improved model takes advantage of angular information in the normalized difference between hotspot and darkspot for improving LAI estimation by better accounting for foliage clumping. Four vegetation indices were also considered for LAI estimation, including three versions of the normalized difference vegetation index (NDVI) and the normalized hotspot-signature vegetation index (NHVI). A geometric-optical canopy model named Five-Scale was used to simulate a range of bidirectional reflectance for sensitivity analysis. The results indicated that better accuracy in LAI prediction was observed from our improved model than from NHVI or any NDVI. A validation with in situ measurements of LAI and bidirectional reflectance in the growth cycle of wheat indicated that the improved model provided the best correlation (R-2 = 0.93) among all models, followed by the NHVI. (C) 2013 Society of Photo-Optical Instrumentation Engineers (SPIE) [DOI: 10.1117/1.JRS.7.073577]