HR: 10:35h
AN: H21H-02 [PDF]
TI: Assimilating MODIS Snow Areal Extent Data Using an Ensemble Kalman Filter
AU: * Andreadis, K M
EM: kostas@hydro.washington.edu
AF: UW Land Surface Hydrology Research Group, Department of Civil and Environmental Engineering Box 352700,
University of Washington, Seattle, WA 98195-2700 United States
AU: Lettenmaier, D P
EM: dennisl@u.washington.edu
AF: UW Land Surface Hydrology Research Group, Department of Civil and Environmental Engineering Box 352700,
University of Washington, Seattle, WA 98195-2700 United States
AB:
The importance of snow to hydrologic prediction and water resources management in the West has long been recognized. Current
model-based approaches to hydrologic prediction are limited by model shortcomings, while more empirical approaches are
inevitably limited by the temporal and spatial sparseness of observations. Remote sensing offers an opportunity to augment
the hydroclimatic information provided by in situ sensors and models. On the other hand, remote sensing data, and especially
visible band snow extent estimates, are problematic as well because of discontinuities in coverage due to cloud cover, and
the absence of coincident information about snow water storage. Data assimilation offers a framework for the optimal
combination of observations and models for estimation of hydrologic state variables and fluxes. A major advantage of data
assimilation is its capability to account for modeling and measurement errors. The ensemble Kalman filter (enKF), is one
promising data assimilation technique that appears to be appropriate for many hydrologic applications. The enKF is different
from the traditional Kalman filter in that it uses a Monte Carlo approach to propagate the error matrices. We describe an
enKF-based approach to assimilating MODIS snow cover data into a macro scale hydrology model. The study area is the Snake
River basin, where about 70% of runoff originates as snow. The dynamic modeling construct is based on the Variable
Infiltration Capacity (VIC) model, applied at 1/8 degree spatial resolution, with subgrid partitioning into a maximum of five
elevation bands. Results showed that the enKF is an effective and computationally attractive solution for the assimilation
of remotely sensed data. The model was able to improve snow water equivalent prediction when compared to benchmark
simulations. In addition to a theoretical evaluation using a simulation approach, the enKF-based estimates of snow areal
extent and associated streamflow predictions are compared with a more ad hoc probability anomaly approach that adjusts the
model's snow water equivalent toward anomalies of SNOTEL station data relative to the SNOTEL climatologies. Finally, specific
limitations having to do with assumptions made about model errors and model sub-optimality are discussed.
DE: 1833 Hydroclimatology
DE: 1863 Snow and ice (1827)
DE: 3337 Numerical modeling and data assimilation
DE: 3360 Remote sensing
SC: Hydrology [H]
MN: 2003 Fall Meeting