- Title
- Enhancement of water storage estimates using GRACE data assimilation with particle filter framework
- Creator
- Tangdamrongsub, N.; Han, S-C.; Yeo, I. Y.
- Relation
- 22nd International Congress on Modelling and Simulation (MODSIM2017). MODSIM2017: 22nd International Congress on Modelling and Simulation (Hobart, Tas. 3-8 December, 2017) p. 1041-1047
- Relation
- https://www.mssanz.org.au/modsim2017/
- Publisher
- Modelling and Simulation Society of Australia and New Zealand (MODSIM)
- Resource Type
- conference paper
- Date
- 2017
- Description
- An accurate knowledge of soil moisture and groundwater storage is crucial to understand hydrological process and extreme climate events. The model outputs of the terrestrial water storage are biased by inaccurate forcing data, inefficacious model physics, and improper model parameter calibration. To mitigate the model uncertainty, the observation (e.g., from remote sensing as well as ground in-situ data) are often integrated into the model to improve the simulation result via data assimilation (DA). This study intends to enhance the estimation of soil moisture and groundwater storage by assimilating the Gravity Recovery And Climate Experiment (GRACE) observation into the Community Atmosphere Biosphere Land Exchange (CABLE) land surface model using the particle filter (PF) framework. The PF is developed for GRACE DA in order to accommodate different types of posterior error distribution and thus allow the realistic system representation where the distribution of model and observation errors are usually unknown. The early development of PF commonly suffered from the particle degeneracy and impoverishment problems, mainly caused by the insufficient number of particles. This study uses the sequential importance resampling (SIR) approach to reduce the problems. The simulation conducted to evaluate the filter performance and determine the effective number of particles shows that the SIR approach can deliver the accurate water storage estimates with the usage of only 100 particles. Moreover, the uncertainty of GRACE observation is obtained directly from the full error variance-covariance matrix provided as a part of the GRACE data product. This method demonstrates the use of a realistic representative of GRACE uncertainty, which is spatially correlated in nature, leads to an improvement of storage computation. The developed GRACE DA scheme is demonstrated over the Goulburn catchment located in the Upper Hunter region, NSW, where the ground observations (surface soil moisture, root-zone soil moisture, and groundwater level) from the Scaling and Assimilation of Soil Moisture and Streamflow (SASMAS) network and the Department of Primary Industries, Office of Water, New South Wales are available for evaluation of our DA results. This study is the first time the GRACE-PF is exploited in a small catchment size (≤ 6,540 km²), proving an important insight about the potential of GRACE over a smaller region beyond its limit of spatial resolution at ~250 km. Preliminary results show that our developed technique successfully disaggregates the catchment-scale GRACE information into finer vertical and spatial scale (~25 km), leading to a significant improvement particularly in groundwater and, marginally in deep soil moisture components. On average, GRACE DA improves the groundwater storage computation in terms of correlation coefficient (ρ) by approximately ~47 % (from 0.38 to 0.56). The ρ value changes from 0.535 to 0.543 by only 1.4 % for the deeper soil moisture (beneath 60 cm) computation. The improvement is found mainly from deeper layers with slower temporal variations, which is consistent with the interannual time scale of the GRACE signals being most characteristic over that catchment. However, GRACE DA slightly degrades the computation of the near surface soil moisture by approximately 2.2 % (in ρ). The coarse temporal and spatial resolution of GRACE is attributed to the less impact of the GRACE DA on surface soil moisture estimation. In conclusion, it is apparent that GRACE DA provides a crucial benefit to deep storage computation. Further development will incorporate satellite soil moisture observations from Soil Moisture Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) missions with GRACE in the assimilation scheme to simultaneously improve different storage components, including surface soil moisture. Comprehensive evaluation of PF’s results in comparison to EnKF results will also be conducted to understand the filter’s performance with regard to accuracy of water storage estimates.
- Subject
- GRACE; data assimilation; particle filter; CABLE; Goulburn catchment; SASMAS; groundwater; soil moisture
- Identifier
- http://hdl.handle.net/1959.13/1386263
- Identifier
- uon:32388
- Identifier
- ISBN:9780987214379
- Language
- eng
- Full Text
- Reviewed
- Hits: 2133
- Visitors: 2559
- Downloads: 224
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | ATTACHMENT02 | Publisher version (open access) | 1 MB | Adobe Acrobat PDF | View Details Download |