Wednesday, 25 January 2017
4E (Washington State Convention Center )
Land surface model simulations are generally improved by three ways: enhancing model physical representations, providing better observations and assimilation of observations, and calibrating model parameters. In our study, parameter calibration is used to improve soil moisture and soil temperature simulations of high-resolution land data assimilation system (HRLDAS) in 9 different vegetation types in East Asia (60oE-160oE, 0oN-65oN). To decrease the total time consumed in parameter calibration, optimization is only performed on the most sensitive parameters which influence either soil moisture or soil temperature more than other parameters. Firstly, 19 tunable parameters in HRLDAS are selected. Then, parameter sensitivity analysis and parameter optimization are performed successively on 9 vegetation types one by one. Hourly forcing data with grid resolution of 0.625o×0.625o from 2008 to 2015 are used to drive HRLDAS. Hourly observations from 63 to 795 meteorological stations for 9 vegetation types from 2009 to 2013 are used to calibrate model parameters, and those from 2014 to 2015 are used to verify the robustness of optimal parameter values comparing with default parameter value. Results show that there are 3 to 6 most sensitive parameters for soil moisture, and 8 to 13 for soil temperature in 9 different vegetation types. Comparing with default parameter values, the errors between soil moisture simulations and observations decrease 12.3% to 25%, and errors between soil temperature and observations decrease 8.9% to 16.6%. Apparently, parameter calibration improves soil moisture and temperature simulations of HRLDAS in East Asia. Besides, parallelization of serial HRLDAS program is achieved in this study, which makes the running time of HRLDAS decrease 90% on parallel computer clusters comparing with serial HRLDAS running on a single machine so that the parameter calibration can be conducted within an acceptable consuming time of 60 days.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner