Improved uncertainty assessment of hydrologic models using data assimilation and stochastic filtering (Mississippi).

摘要

Hydrologic models are two-fold: models for understanding physical processes and models for prediction. This dissertation addresses the latter, which modelers use to predict, for example, streamflow at some future time given knowledge of the current state of the system and model parameters. Therefore, two elementary issues in contemporary earth system science and engineering are (1) the specification of model parameters (static states) values which characterize a system, (2) the estimation of dynamic states (prognostic) variables which express the system dynamic. Estimation of these components is needed to enable the model to generate the forecasts as accurate as possible. Methods for batch calibration, despite their recent advances, appear to lack the flexibility required to treat uncertainties in the current system as new information is received. Methods based in sequential Bayesian estimation seems better able to take advantage of the temporal organization and structure of information, so that better compliance of the model output with observations can be achieved. In this dissertation two approaches of sequential hydrologic data assimilation, having their origin in Bayesian estimation, for estimating model parameters, state variables and their uncertainties are explored. Providing a comprehensive review on the various aspects of estimation theory and the state-of-the-art of sequential data assimilation, the two algorithms: Ensemble Kalman Filter (EnKF) and sequential Monte Carlo or Particle Filter (PF) are employed and discussed thoroughly. The two filters have originally been developed to estimate the uncertain dynamic states in a system when incorporation of other sources of uncertainties including input (forcing data) and output observation (diagnostic variables) are possible. In this study the applicability of two aforementioned filters is extended to combined state-parameter estimation. The power, applicability and usefulness of the developed procedures for adaptive inference of posterior distribution of state-parameters which finally results to the ensemble streamflow forecasting are examined over a parsimonious conceptual hydrologic model (HyMOD) in Leaf River Basin located north of Collins, Mississippi.