In this context, the Soil and Vegetation Reflection Simulator (SAVERS) was employed to simulate GNSS-R data in order to analyze the GNSS-R sensitivity to biomass, to create a Look Up Table (LUT) of GNSS-R reflectivity and to carry out a biomass retrieval test. SAVERS simulated the mean power of the reflected GNSS-R signals by applying the integral bistatic equation taking into account the scattering contributions from soil and vegetation. It was shown that the GNSS-R sensitivity to forest biomass can be enhanced by filtering out the incoherent component of the signal, i.e. by using a long coherent integration time. A LUT was created by running SAVERS for a realistic range of the input parameters. Moreover, a GNSS-R synthetic dataset was generated for the case of a cork oak open forest in Portugal and a retrieval test was carried out. A neural network with two hidden layers was trained on the LUT and the forest biomass was estimated from the synthetic data. The biomass mean retrieval error was approximately 10 t/ha. |
|