Assessing the effects of the clumping phenomenon on BRDF of a maize crop based on 3D numerical scenes using DART model
详细信息 查看全文
- 作者：Sylvie Duthoit ; Valé ; rie Demarez ; Jean-Philippe Gastellu-Etchegorry ; Emmanuel Martin ; Jean-Louis Roujean
- 关键词：Maize canopy ; Clumping ; Bidirectional Reflectance Distribution Function (BRDF) ; Discrete Anisotropic Radiative Transfer model ; Leaf area index (LAI)
- 刊名：Agricultural and Forest Meteorology
- 出版年：2008
- 出版时间：4 July 2008
- 年：2008
- 卷：148
- 期：8-9
- 页码：1341-1352
- 全文大小：1457 K
文摘
Inverting radiative transfer (R-T) models against remote sensing observations to retrieve key biogeophysical parameters such as leaf area index (LAI) is a common approach. Even if new inversion techniques allow the use of three-dimensional (3D) models for that purpose, one-dimensional (1D) models are still widely used because of their ease of implementation and computational efficiency. Nevertheless, they assume a random distribution of foliage elements whereas most canopies show a clumped organization. Due to that crude simplification in the representation of the canopy structure, sizeable discrepancies can occur between 1D simulations and real canopy reflectance, which may further lead to false LAI values. The present investigation aims to appraise to which extent the incorporation of a clumping index (noted λ) into 1D R-T model could improve the simulations of Bidirectional Reflectance Distribution Function (BRDF). Canopy BRDF is simulated here for three growth stages of a maize crop with the Discrete Anisotropic Radiative Transfer (DART) model in the visible and near infrared spectral bands, for two contrasted soil types (dark and bright) and different levels of heterogeneity to represent the canopy structure. 3D numerical scenes are based on in-situ structural measurements and associated BRDF simulations are thus considered as references. 1D scenarios assume either that leaves are randomly distributed (λ = 1) or clumped (λ < 1). If BRDF simulations seem globally reliable under the assumption of a random distribution in near infrared, it can also lead to relative errors on the total BRDF up to 30 % in the red spectral band. It comes out that the use of a clumping index in a 1D reflectance model generally improves BRDF simulations in the red considering a bright soil, which seems relatively independent of LAI. In the near infrared, best results are usually obtained with homogeneous canopies, except with the dark soil. Clearly, influent factors are mainly the LAI and the spectral contrast between soil and leaves.