Accessibility navigation

Ensemble generation in land surface models for soil moisture data assimilation

Ejigu, A. A. (2020) Ensemble generation in land surface models for soil moisture data assimilation. PhD thesis, University of Reading

Text - Thesis
· Please see our End User Agreement before downloading.

[img] Text - Thesis Deposit Form
· Restricted to Repository staff only


It is advisable to refer to the publisher's version if you intend to cite from this work. See Guidance on citing.

To link to this item DOI: 10.48683/1926.00107577


Soil moisture is a crucial meteorological variable to understand land surface and atmospheric processes like the water cycle, the carbon cycle and the energy balance. However, its link with those processes makes the measurement and modelling difficult. Data assimilation is a mechanism to combine observations with modelled estimates and uncertainties to provide the best prediction for the state or parameter of a system. Proper uncertainty representation is an essential procedure to get a skilful result from the data assimilation. In this thesis, we demonstrate uncertainty representation techniques in the forcing data, numerical models and the parameters for soil moisture data assimilation. We use the Joint UK Land Environment Simulator land surface model to estimate soil moisture and the Ensemble Transform Kalman Filter and the four-Dimensional Ensemble Variational data assimilation methods to combine soil moisture estimates with observations. Satellite observed, synthetic and in-situ soil moisture data are assimilated. When in-situ soil moisture observations for three soil layers are assimilated, employing stochastic forcing via generated rainfall to account for errors in observed rainfall has shown substantial improvement for ensemble spread as well as forecast skill of posterior surface soil moisture. However, additional stochastic forcing via model error is needed to improve forecast skills for the deeper layers. For parameter estimation, prior soil texture parameter errors are represented by the Dirichlet distribution where both share positivity and boundedness. Synthetic data assimilation results show that truth parameters can be recovered even though prior parameters are less informed. The advantage over the Gaussian distribution is that the Dirichlet distribution automatically assigns correlations for the prior covariance matrix. The robustness of the method is tested for different soil types. Posterior parameters obtained from assimilating in-situ and satellite observations showed improvement in soil moisture forecast skills beyond the assimilation window. It is also shown that satellite observations are representative of the state of soil moisture for areas with no or less woody vegetation cover.

Item Type:Thesis (PhD)
Thesis Supervisor:Quaife, T. and Lawless, A.
Thesis/Report Department:Department of Meteorology
Identification Number/DOI:
Divisions:Science > School of Mathematical, Physical and Computational Sciences > Department of Meteorology
ID Code:107577


Downloads per month over past year

University Staff: Request a correction | Centaur Editors: Update this record

Page navigation