.

  Zarco-Tejada, P. J., Miller, J. R., Noland, T. L., Mohammed, G. H., & Sampson, P. H. (2001). Scaling-up and model inversion methods with narrowband optical indices for chlorophyll content estimation in closed forest canopies with hyperspectral data.  IEEE Transactions on Geoscience and Remote Sensing ,  39 (7), 1491-1507. In this paper, the author uses vegetation index as a merit function. The method is more stable than using all channels. Integrating empirical method, physics based model inversion could be improved. 

Asner, G. P., Martin, R. E., Anderson, C. B., & Knapp, D. E. (2015). Quantifying forest canopy traits: Imaging spectroscopy versus field survey.  Remote Sensing of Environment ,  158 , 15-27. They use canopy sunlit reflectance at plot level and the trait samples from sunlit. The plot averaged refletance minimize canopy architectural effect. However actual field samples cover only 5% of a plot, the plot reflectance well explains canopy traits.

Combal, B., et al. "Retrieval of canopy biophysical variables from bidirectional reflectance: Using prior information to solve the ill-posed inverse problem."  Remote sensing of environment  84.1 (2003): 1-15.

Bousquet, L., Lachérade, S., Jacquemoud, S. and Moya, I., 2005. Leaf BRDF measurements and model for specular and diffuse components differentiation. Remote Sensing of Environment, 98(2-3): 201-211. Specular reflectance originates from leaf surface. On the other hand, diffuse reflectance originates from leaf intrastructure. To focus on the target traits, the two reflectance needs to be separated.

Kim, H. S., et al. (2008). "Actual and potential transpiration and carbon assimilation in an irrigated poplar plantation." Tree Physiol 28(4): 559-577. As for environment, its mechanisms and their causing factors have internal relationship. Data with diverse condition can appear with the boundary which limited by causing factors.

Kim, H. S., Palmroth, S., Thérézien, M., Stenberg, P., & Oren, R. (2011). Analysis of the sensitivity of absorbed light and incident light profile to various canopy architecture and stand conditions.  Tree Physiology ,  31 (1), 30-47. In this paper, they analyzed canopy light absorption, considering leaf angle distribution, leaf clumping and stand density. To compare canopy properties, they set models and compare how well each model estimates measured canopy light absorption. Canopy properties are controlled in 7 models (v1 – v7). The simplest model (v1), only considering Beer-Lambertian gap fraction, was developed to most complex model (v7) by gradually adding another properties. They are compared in respect of various LAI and sky condition.

Asner, Gregory P., and Roberta E. Martin. Spectral and chemical analysis of tropical forests Scaling from leaf to canopy levels. Remote Sensing of Environment 112.10 (2008) 3958-3970. When up-scaling leaf level to canopy level, canopy structure may obstruct leaf traits to fully appear as the form of spectra. Except LAI, however, most canopy structural features had a negligible effect on the spectra created by canopy modeling.

Gara, Tawanda W., et al. Leaf to canopy upscaling approach affects the estimation of canopy traits. GIScience & remote sensing 56.4 (2019) 554-575. In remotely sensing, upper-most leaves are dominantly detected so they are usually measured for upscaling, assuming that sunlit leaves are representing the whole canopy. Short plants or plants with sparse leaves, however, cannot be seen dominated by sunlit leaves because of canopy vertical diversity. In that case, it is significantly better to measure sunlit and mid-layer leaves(approach B,D and E) more than to measure only sunlit leaves(C). Though r square in remotely sensing canopy chlorophyll was higher at C, there was no significant difference between 2-leaf and 1-leaf approach.