Calibração do modelo SiB2 para o Cerrado no Sudeste do Brasil

• Ano de defesa: 2016
• Autor(a) principal: Valdés, Roilan Hernández
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Santa Maria; BR; Meteorologia; UFSM; Programa de Pós-Graduação em Meteorologia
• Palavras-chave em Português: Calibração; SiB2; Cerrado sensu-stricto; Fluxos energéticos; Umidade do solo; Calibration; Woodland savanna; Energy fluxes; Soil humidity; CNPQ::CIENCIAS EXATAS E DA TERRA::GEOCIENCIAS::METEOROLOGIA
• Link de acesso: http://repositorio.ufsm.br/handle/1/10292

The land surface models (LSM) are the component of the numerical weather prediction and climate models that represents the processes of interaction between the biosphere and atmosphere. The need to improve the representation of bio-geophysical process stimulated the development of sophisticated parametrization. This development increased the complexity of LSMs and the number of parameters involved. Some of these parameters can be measured, but it does not ensure that the best results will be produced. Therefore, a general strategy is to use field experiments (local) to calibrate these parameters for different vegetation types, minimizing the differences between the simulated and observed value(s) of variable(s) of interest. LSMs are usually calibrated using observation of the sensible (H) and latent (LE) heat flux . Studies including soil moisture (�) in the calibration are less frequent (or restricted to the surface layer), but in forest ecosystems, under seasonal water stress, vertical soil moisture profile in vadose zone is essential for simulating transpiration, CO2 assimilation and the partition between the surface and underground flows. This paper describes the calibration of the Simple Biosphere Model (SiB2) for the Cerrado sensu-stricto, using flux measurements, soil moisture and atmospheric forcings collected in a micrometeorological tower located in Gleba Pé de Gigante, SP, in the period July 2009 to July 2012. For calibration, the SiB2 model was separated into modules that included the radiative, aerodynamic and finally, soil water processes and conductance and photosynthesis. The multi-objective calibration algorithm AMALGAM was applied to each of these modules using as objective functions: the mean square error (RMSE), Nash-Sutcliffe efficiency (NSE), the error of amplitude of the mean cycle annual (ACAM) or hourly (ACH). The calibrated variables were: albedo PAR (�PAR), global albedo (�), friction velocity (u�), net radiation (Rn), latent and sensible heat flux and total water storage (Az) up to two meters deep. Nonetheless, the modular calibration was compared with a global calibration in which only variables LE, H and Az were optimized. The calibration for radiative module allowed to reproduce the seasonal cycle and amplitude for albedo PAR, while the global albedo was lagged temporally and a slightly smaller amplitude than the observation but had considerable improvement compared with that simulated with the original parameters. The balance of radiation was reasonably simulated, with overestimation in winter and spring and it proved to be fundamentally sensitive to downward longwave radiation. The u� was slightly underestimated in the average daily cycle against to observed but had less error than the original parameters. On the other hand, the model discretization in three soil layers failed to represent the hydrological processes in the soil and surface simultaneously for Cerrado. Therefore, the soil structure was changed by introducing vertical root distribution profile, the process of hydraulic redistribution and updating the Green-Ampt infiltration scheme. These schemes were essential to the modeling of hydrologic processes of Cerrado vegetation, which is applicable to other deep root system. The global calibration fairly represented LE, H and Az, but resulted in �PAR anti-correlated, considerable underestimation of the � and u�, also in inconsistent partition for evapotranspiration components.