Validation of forward and inverse modes of a homogeneous canopy reflectance model

Abstract

The homogeneous canopy reflectance model ACRM was used to simulate forest reflectance and was compared with hyperspectral data of the topographically complex experimental forest Rosalia of the University of Applied Life Sciences and Natural Resources (BOKU) in Austria. Forward and inverse modes of the ACRM model were validated. Ground truth data were taken (1) from experiments performed at 17 pure beech plots and (2) from model simulations performed for 21 pure beech plots using an ecosystem model. The validations were performed separately for these two types of reference data. The ground reflectance obtained from the HyMap data was compared with simulations performed with ACRM. In addition to the correction of the data to remove the atmospheric effects, corrections had to be applied to remove the effects of the complex topography of the area of Rosalia. The simulated reflectance showed an offset to the HyMap retrieved reflectance between +4% and +6% in the visible (extending from 400 nm to 700 nm), +28% to +30% in the near infrared (NIR) (extending from 700 nm to 1400 nm) +53% to +77% in the middle‐infrared (MIR) (extending from 1400 nm to 3000 nm) using the modelled and the experimental ground truth, respectively. The correlation coefficient varied between 0.35 and 0.45 in the visible, 0.6 and 0.76 in the NIR and between 0.37 and 0.64 in the MIR. This correlation may be improved, if within canopy fluctuations of chlorophyll and water content were available. The leaf area index (LAI) was retrieved using the ACRM model. The estimated LAI was in good agreement with the LAI ground measurements and systematically higher by 0.1 compared to the simulated LAI. The correlation was 0.49 and 0.82, respectively. Altogether, the ACRM model showed a large offset to the HyMap retrieved reflectance in the NIR and MIR wavelength ranges. A precision around 33% to 74% may be expected after correction of the offset. The LAI may be determined with a precision between 0.32 and 0.5. The ACRM model is a useful tool to predict LAI. Care should be taken for the forward modelling of forest canopy reflectance.

Acknowledgments

These investigations were performed within the scope of the project ‘Crop Drought Stress Monitoring by Remote Sensing (DROSMON)’, supported by the Austrian Science foundation (FWF), grant number P17647‐N04. We thank the reviewers for their most valuable suggestions.