T. Fei1, Y. Liu2, M. Bian1

1 - ITC, Natural Resource Management, Enschede, The Netherlands

2 - School of Resource and Environmental Science, Wuhan University, Wuhan, China


In this paper, we present a methodology for monitoring the underwater light climate for aquatic vegetation based on spectral reflectance above the water surface. This method is composed by three consequences steps:


First, based on the semi-physical bio-optical model, we built a water reflectance model. This model relates the inherent optical properties (IOP) of water to the apparent optical properties of the lake water and thus can simulate different spectral remote sensing reflectances above the water surface by different combinations of water constituents. This step is called forward process. In this study, the water reflectance model gives satisfactory simulation on most cases.


Then in the second step, an artificial neural network is trained by those simulations, and thus can

retrieve concentrations of water constituents by inverse the water reflectance model. This step is

called inverse process. By this method, the ANN-based algorithm can retrieve the concentrations of SPM, CHL and CDOM at the same time. And the accuracies are acceptable (R2=0.758, 0.741 and 0.389 for SPM, CHL and CDOM concentrations respectively).


In the third step, an underwater light climate model was built to predict the light climate at the bottom of the lake, where aquatic vegetation grows, according to the concentrations of water constituents retrieved in step two and the water depth.


The advantage of this method, comparing with empirical methods, is we dont need to collect a lot of water samples to build the regression formula for retrieving the concentrations of water constituents. And it is more site-independent.