Predictions in Ungauged Basins: Promise and Progress (Proceedings of symposium S7 held during the Seventh IAHS Scientific Assembly at Foz do Iguaçu, Brazil, April 2005). IAHS Publ. 303, 2006, 164-176.

 

Multi-objective conditioning of a SVAT model for heat and CO2 fluxes prediction

 

XINGGUO MO1, SUXIA LIU2, ZHONGHUI LIN1, XIAOMIN SUN1 & ZHILIN ZHU1

1       Key Lab of Ecological Net Observation and Modelling,
Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China

moxg@igsnrr.ac.cn

2       Key Lab of Water Cycle and Related Land Surface Processes,
Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China

 

Abstract  The parameters of a SVAT model (VIP) are explored by a multi-objective likelihood measure using the Generalized Likelihood Uncertainty Estimation (GLUE) framework based on field data collec­ted in the North China Plain during the winter wheat growing season in 2001. Agreement indexes of latent, sensible, ground heat and CO2 fluxes and radiometric surface temperature between the observed and the modelled data are used to evaluate the model performance, in which 13 parameters were selected for calibration and model uncertainty esti­mation. Although the single objective approach effectively constrains the corresponding model response, the multiple objective technique, including both fluxes and state variables, presents a more efficient constraint. The outstanding effect of surface radiometric temperature for calibration suggests that thermal remote sensing might be a promising tool for distributed SVAT model calibration and evaluation over large areas. It is found that, although the model appears to have a serious equifinality problem, the interactions and compensation effects between the parameters are not strong, with both linear and nonlinear correla­tion coefficients being small. Sensitivity analyses using both scatter plots and partial correlation coefficients show that model responses are sensitive to half of 13 parameters.

 

Key words  GLUE; parameter calibration; SVAT model; uncertainty