Continuous soil moisture accounting and routing modelling to support short lead-time streamflow forecasting

[摘要] This report is a preliminary investigation of Continuous soil moisture accounting and routing models for short-term forecasting of streamflow. The use of these models can improve streamflow forecasts because they objectively quantify the effect of soil moisture on runoff efficiency (i.e., rainfall on dry soils can produce less runoff, but it is difficult to subjectively anticipate what the losses will be).Such models commonly contain three components: conversion of rainfall to runoff, gathering runoff to river channels, and movement of streamflow through the channel network to the catchment outlet (“rainfall-runoff”, “catchment routing”, and “channel routing”, respectively).The current research uses the Probability Distributed Moisture model (PDM) for rainfall-runoff and catchment routing. Versions of the Bureau operational methods for catchment and Muskingum-type channel routing were also studied. The sequence of routing streamflow through the river network is the routing topology; the Bureau’s current method is contrasted with a Node-link method. The former is more flexible in the networks it can represent but the latter may be more user-friendly and less prone to unphysical configurations. A utility was created to translate control files from one method to the other.This modelling effort is a first step in a process to evaluate possible candidates for operational adoption. New systems can gain value if they are compatible with existing operational resources. At the same time, operational assets are the result of many years of effort by hydrologists, are refined through forecasting experience, and have much to offer researchers. For example, this research reused existing model parameter values, model algorithms, and hourly spatially-distributed forcing datasets for two catchments.In one experiment, operational channel routing parameters were fixed and rainfall-runoff and catchment routing parameters were optimized using the Shuffled Complex Evolution method. In other experiments, all three were optimized. The simulation of streamflow for one catchment (the Tully) was largely successful, although simulation for the other (the Ovens) was very poor, similar to past studies using the same data. The poor result may be due to rainfall data-quality issues.A factor limiting research and adoption of Continuous soil moisture accounting models is the relative lack of sub-daily (e.g., hourly) data. An opportunity exists to calibrate some model water balance parameters using daily forcing data and using individual sub-daily events to calibrate hydrograph timing parameters. Models calibrated using daily data were not appreciably different (in parameter values and skill in sub-daily simulations) from those calibrated on sub-daily data, suggesting this may be a fruitful approach.Daily time-step spatially-lumped optimizations and simulations were also done for 331 catchments using the PDM model for rainfall-runoff and catchment routing. Geographic patterns of simulation skill (e.g., lower skill in western flowing streams in New South Wales and higher skill in central Victoria) were similar to the results of other studies of the same data using other rainfall-runoff models. Maps and typical ranges of parameter values are given, highlighting those PDM model parameters that are difficult to reliably calibrate.Future directions of this research activity are outlined, with particular emphasis on improved calibration and parameterization techniques, forecast evaluation methods, use of Numerical Weather Prediction output, and characterization of forecast uncertainty. Extensive appendices (1) describe the implementation of the PDM model and the interpretation of its parameters, (2) describe catchment and channel routing methods and (3) compare channel routing topologies.

[发布日期] 2009-07-01 [发布机构] CSIRO

[学科分类] 地球科学(综合)