RiM-FIM floodplain inundation modelling for the Edward-Wakool, Lower Murrumbidgee and Lower Darling River systems

[摘要] This report accompanies a series of floodplain inundation models for the Edward-Wakool, Lower Murrumbidgee and Lower Darling River Systems in the southern Murray Darling Basin.These models were made by the CSIRO Water for a Healthy Country Flagship for the Murray Darling Basin Authority.The River Murray Floodplain Inundation Model (RiM-FIM) method links measurements of water height at flow gauges to the extent of inundation mapped from Landsat satellite images and interpolates the distribution of inundation across the landscape using a lidar digital elevation model (DEM).RIM-FIM enables the distribution of inundation to be predicted at 5m spatial resolution over large areas, and at flow levels that were not observed in the Landsat images.The RIM-FIM models in this project show flood extents at 1GL increments in total daily flow ranging from the smallest flood shown in the images to the largest recorded flow with a measured height. The RiM-FIM process involves seven main steps:1.Identify gauges and define zone boundaries– Identify flow gauges that are best associated with the extent of inundation throughout the region, and define the region into zones representing the extent of inundation associated with each flow gauge2.Select flows and Landsat images – Identify flows at a range of flood levels for each gauge and interrogate the Landsat image archive for suitable images for those floods 3.Map flood extents from Landsat images –Classify the Landsat images to show the extent of inundation for each flow level4.Create a Digital Elevation Model (DEM) – Obtain lidar data for each region and convert it into a digital elevation model (DEM) including correcting errors, edge matching, correcting for dense reed beds, etc5.Calculate flood extents and water heights from the DEM – Link satellite image extents to the DEM and derive water heights from each flood image6.Produce the RiM-FIM – Generate the RiM-FIM model from the interpolation of inundation between satellite observations based on the DEM and remove disconnected inundated patches7.Sensitivity assessment – Conduct error checking, validation and accuracy assessment of the modelsA range of factors influence the accuracy of the RiM-FIM models for each region, including rainfall and river flow conditions before capture of the Landsat images, the spatial accuracy of inputs including the DEM and Landsat images, the accuracy of flood maps made from the Landsat images, the size and visibility of patterns of inundation relative to the pixel size of the Landsat images, the quality of the hydrological data and land and water use activities.Each of these factors can influence the extent and distribution of flooding relative to the flow measured at the stream gauge.Sensitivity analysis in this study, however, indicates that the depth and extent of inundation shown in these RiM-FIM models are consistent with the available hydrological data.This project has included a comprehensive review of the RIM-FIM method, including automation of several processes that were formerly labour intensive, time consuming and subjective. These include the use of a band ratio for inundation detection in the Landsat images to improve the consistency and objectivity of inundation detection, automating the detection of water heights from the DEM and improving the image filtering techniques.These improvements enable RiM-FIM models to be created more accurately, consistently and efficiently in future.

[发布日期] 2014-06-01 [发布机构] CSIRO

[学科分类] 地球科学(综合)