Parameter identification, estimation and optimization are indispensable process for model development and application, especially for complex time-consuming hydrological models or environmental models. The common-used methods need thousands upon thousands of model runs to obtain parameter samples space. In order to accurately assess the sensitivity of parameters and efficiently optimize the parameters of complex time-consuming models, Dr. ZHAN Chesheng’s group of the Institute of Geographical Sciences and Natural Resources Research (IGSNRR), CAS developed a high-speed, effective and efficient global sensitivity analysis and optimization approach.

Researchers found that the meta-modeling approach can be used to construct a surrogate model to replace the original complex dynamic models, and then the surrogate model can be integrated with the variance-based sensitivity analysis method and heuristic search optimization method to reduce the computational cost. They also discovered that the qualitative Morris screening method was particularly useful for computationally expensive models.

In addition, the surrogate model integrated with variance-based sensitivity analysis method (RSMSobol’ method) was used to reduce the computational cost with less model runs for the original model. The integration method based on the surrogate model and the classical optimization algorithm (e.g. SCE-UA) can complete the multi-parameter optimization process with a good accuracy and better efficiency compared with the classical optimization method and single surrogate modeling approach.

Their study concluded that the proposed approach has shown a powerful role or function in the parameter calibration, estimation and optimization processes, especially for those large and complex models. It will be a useful tool to access these parameters or input factors for multi-parameters or multi-processes complex models.

The above work was financially supported by National Basic Program of China (973 Program, 2010CB428403) and the National Natural Sciences Foundation of China (41271003 and 40901023).

The main findings have published on Environmental Modelling & Software (Zhan et al, An efficient integrated approach for global sensitivity analysis of hydrological model parameters. Environmental Modelling & Software, 2013, 41: 39-52), Chinese Science Bulletin (Song et al.,Integration of a statistical emulator approach with the SCE-UA method for parameter optimization of a hydrological model. Chinese Science Bulletin, 2012, 57(26): 3397-3403), and Journal of Geographical Sciences (Song et al, An efficient global sensitivity analysis approach for distributed hydrological model. Journal of Geographical Sciences, 2012, 22(2): 209-222).