Comparación de modelos empíricos y de transferencia radiativa para estimar contenido de humedad en pastizalespoder de generalización

Abstract

Vegetation water content (FMC) is a key variable in several applications, many of which require a global knowledge. Because of that, the calibration of highly generalizing power models to estimate that variable is required. Remote sensing is one of the sources of information more suitable in that respect. There exist mainly two kinds of models who relate remote sensed data to FMC: empirical and theoretical. The empirical models are highly dependent of the data used in the calibration phase and the conditions under which those data were taken. Therefore, these models have low generalizing power. The RTM have a strong physic (generally the radiative transfer theory, that is why they are so-called Radiative Transfer Models, RTM) so they can be applied to more diverse conditions. However, if accurate estimations want to be obtained, they must be properly parameterizated with field data and include auxiliary information, what may affect their generalizing potential. The objective of this paper is to compare the performance of the empirical model and the RTM proposed by Yebra et al. (2008) in terms of generalizing power. For doing that, a validation sample of 92 observations of field measured FMC and their corresponding MODIS data was used. These data correspond to seven grassland plots located in Spain and eight in Australia. The results show that both models offer similar accuracy when applying to grassland with analogous types of vegetation to the calibration site (RMSE=41.39 y 43.44% against 38.23 and 33.83%, empirical and RTM, respectively). Nevertheless, the RTM offer greater accuracy than the empirical when the models are applied to grassland with different characteristic to those of the calibration site (RMSE=38.93 y 61.66% against 11.27 y 19.37%, empirical and RTM, respectively), what support that the RTM have more generalizing power.