Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Australie → 2013 → On the predictability of hydrology using land surface models and field soil moisture data

University of Newcastle (2013)

On the predictability of hydrology using land surface models and field soil moisture data

Chen, Min

Titre : On the predictability of hydrology using land surface models and field soil moisture data

Auteur : Chen, Min

Université de soutenance : University of Newcastle

Grade : Doctor of Philosophy (PhD) 2013

Soil moisture is a key variable in the hydrologic cycle and climate system. Climate change has been exerting a significant impact on the land surface hydrology and water resources, especially for water-limited areas. However, predictions of soil moisture and hydrology (i.e. surface runoff and deep drainage) from climate models have exhibited large uncertainties. Good observational soil moisture data (e.g. large spatial coverage, long-term and high temporal resolution) for the root zone are still rare, which strongly limits model evaluation and our understanding of ongoing changes. This thesis advances our ability to predict soil moisture and hydrology, especially by focusing on the potential usefulness of continuous time, point-scale soil moisture data at the field sites. This study contains two broad parts : (1) predictions of spatial patterns of soil moisture and (2) predictions of temporal dynamics of soil moisture and hydrology. In the first part of this thesis, an approach to predicting soil moisture at catchment scale using the HYDRUS model and a simple data assimilation method was tested. The model was satisfactorily calibrated on the Stanley microcatchment sites (in the Merriwa area, NSW, Australia) for a three-year period (2005-2007) with a single point rainfall record from this microcatchment for both surface 30 cm and full-profile (90cm) soil moisture measurements. Calibrated soil and vegetation parameters at one Stanley site were used to predict soil moisture at nearby Stanley sites using the same rainfall, and also for more distant Krui catchment sites using the local rainfall measurements at these sites. Correlations obtained between the site predictions and its field observations were as good as the calibration of the site itself, despite some biases which varied from site to site. These biases could be effectively reduced by simply adjusting the time series up or down by the difference in the means of the measured and predicted time series. This adjustment could be done with 4-7 local point measurements at Stanley, and 11-25 at Krui. Results show that it is possible to use a calibrated soil moisture model with measurements at a single site to extrapolate the soil moisture to other sites for a catchment with an area of up to 1000 km2 given similar soils and vegetation and local rainfall data. In the second part of this thesis, predictions of the three-year daily soil moisture and hydrology from land surface models (LSMs) were investigated. The choice between a constant leaf area index (LAI) and a temporally varying LAI for predicting daily soil moisture dynamics were examined with two different LSMs (IBIS and HYDRUS). The analyses found that with prescribed soil parameters, it is necessary to examine the model’s sensitivity to vegetation (e.g. LAI) before considering the vegetation dynamics. In this study, a temporal mean, rather an annual cycle of LAI, was adequate to reproduce daily soil moisture dynamics from 2005 to 2007 at the Stanley semi-arid site. To obtain LAI values (either a constant or a time series), a strong correlation between antecedent rainfall and LAI suggests a simple and effective way to do this via linear regressions. The two LSMs were also evaluated and compared within the Monte-Carlo GLUE framework using daily soil moisture data at the Stanley catchment. The IBIS soil moisture calibration was better for the surface 30 cm, because its soil hydraulic model is more appropriate for the surface condition. The HYDRUS calibration was superior for the full 90 cm profile, due to stronger vegetation sensitivity and a better fitting transpiration component, when compared to IBIS. Both models performed satisfactorily under average semi-arid hydroclimatic conditions. Model deficiencies only became significant when simulating the soil moisture response to extreme precipitation events. For both the surface 30 cm and full profile, better calibrations of soil moisture resulted in fewer uncertainties in the three-year total water partitioning among evapotranspiration, surface runoff and drainage. For the surface, the total evapotranspiration and surface runoff plus drainage from the best-performing soil moisture simulations were quite consistent between IBIS and HYDRUS. The consistency implies that LSMs are not able to predict hydrology unless soil moisture is accurately estimated, even though it proved inadequate to partition between surface runoff and drainage.

Mots clés : land surface model ; soil moisture ; SASMAS ; HYDRUS ; IBIS ; GLUE ; semi-arid


Version intégrale (10 Mb)

Page publiée le 5 février 2018