HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation

Variarional data assimilation in the land surface model ORCHIDEE using YAO

Abstract : A land surface model (LSM) is a numerical model describing the exchange of water and energy between the land surface and the atmosphere. Land surface physics includes an extensive collection of complex processes. The balance between model complexity and resolution, subject to computational limitations, represents a fundamental query in the development of a LSM. With the purpose of adapting the value of the model parameters to values that reproduces results in the real world, measurements are necessary in order to compare to our estimations to the real world. The calibration process consists in an optimization of model parameters for a better agreement between model results and a set of observations, reducing the gap between the model and the available measurements. Here we show how variational data assimilation is applied to the energy and water budgets modules of the ORCHIDEE land surface model in order to constrain the model internal parameters. This part of the model is denoted SECHIBA. The adjoint semi-generator software denoted YAO is used as a framework to implement the 4DVAR assimilation. A sensitivity analysis was performed in order to identify the most influent parameters to temperature. With the parameter hierarchy resolved, twin experiments using synthetic observations were implemented for controlling the most sensitive parameters. Results obtained suggest that land surface temperature assimilation has the potential of improving the output estimations by adjusting properly the control parameters. Finally, several assimilations were made using observational meteorology dataset from the SMOSREX site in Toulouse, France. The experiments implemented, using different prior values for the parameters, show the limits of the temperature assimilation to constrain control parameters. Even though variable estimation is slightly improved, this is due to final parameter values are at the edge of a variation interval in the cost function. Effectively reaching a minimum would require allowing the parameters to visit unrealistic values. SECHIBA does not correctly simulates simultaneously temperature and fluxes and the relationship between the two is not always consistent according to the regime (or parameter values that are used). We must therefore work on the physical aspects to better simulate the temperature. Likewise, the parameter sensitivity to temperature is not always sufficient, giving as a result a flat cost function. Our results show that the assimilation system implemented is robust, since performances results in twin experiments are satisfactory. The coupling between the hydrology and the thermodynamics in SECHIBA must be reviewed in order to improve variable estimation. An exhaustive study of the prior errors in the measurements must be conducted in order to retrieve more adapted weighing terms in the cost function. Finally, the assimilation of other variables such as soil moisture should be performed to evaluate the impacts in constraining control parameters
Document type :
Complete list of metadata

Cited literature [86 references]  Display  Hide  Download

Contributor : Abes Star :  Contact
Submitted on : Monday, April 27, 2015 - 12:57:16 PM
Last modification on : Monday, December 13, 2021 - 9:14:07 AM
Long-term archiving on: : Monday, September 14, 2015 - 1:51:51 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01145923, version 1


Hector Simon Benavides Pinjosovsky. Variarional data assimilation in the land surface model ORCHIDEE using YAO. Earth Sciences. Université Pierre et Marie Curie - Paris VI, 2014. English. ⟨NNT : 2014PA066590⟩. ⟨tel-01145923⟩



Record views


Files downloads