MODIS-based standing water detection for flood and large reservoir mapping: Algorithm development and applications for the Australian continent

Select | Print


Guerschman, Juan; Warren, Garth; Byrne, Guy; Lymburner, Leo; Mueller, Norman; Van Dijk, Albert

Guerschman, Juan; Warren, Garth; Byrne, Guy; Lymburner, Leo; Mueller, Norman; Van Dijk, Albert


2011-01


Report


86 p.


Accurate and timely monitoring of natural and man-made water bodies is a critical information input for water accounting, groundwater recharge estimation and flood response and forecasting. Remotely sensed information from satellites and airborne instruments can be used for estimating the extent and dynamics of standing water for large areas and complement in-situ observations Different uses of information on standing water extent lead to different requirements regarding spatial scale, data latency and temporal repetition these drive the selection of the most appropriate sensor/s and methods. For example, flood monitoring and warning systems need rapid access to processed data but optical imagery tends to be affected by cloud contamination. Retrospective analysis of flood recurrence, in contrast, does not require immediacy in the data suply and can rely on temporal compositing methods to overcome cloud contamination. Similarly daily data is not needed when monitoring water use in natural and human-made dams, but spatial resolution becomes critical particularly regarding the ability to detect water in small (subpixel) reservoirs. This report presents a methodology for estimating standing water using the Moderate Resolution Imaging Spectroradiometer (MODIS), an optical sensor with a spatial resolution (pixel size) of 250 to 500 meters. The methodology was developed by investigating the spectral properties of standing water, particularly when it occupies a MODIS pixel only partially. The analysis was based on the use of a series of simultaneous classifications performed using higher resolution Landsat TM data using a image segmentation algorithm as described by Mueller and Lymburner (2010). Several empirical models using MODIS surface reflectance and ancillary variables derived from a Digital Elevation Model (DEM) were analysed. The model that had the best performance included surface reflectance from MODIS bands 6 (~1600 nm) and 7 (~2100 nm), the Normalised Difference Vegetation Index, the Normalised Difference Water Index and the Multi-resolution Valley Bottom Flatness index. The MODIS sensor is mounted in two satellites which overpass daily in the morning (Terra, ~10.30 am) and in the afternoon (Aqua, ~1.30 pm). Using data from both satellites allows taking full advantage of the temporal repetition and provides the best chance of obtaining cloud-free data to monitor floods, particularly in near-real time applications. Analysing flood and reservoir dynamics for large areas (e.g. the Australian continent) and large periods of time (e.g. one decade) with sub-daily MODIS data represents a logistical problem as the data volumes involved would be large. To assess this issue a comparison was performed between the use of daily data (MODIS standard product MOD09GA), 8-day temporal composites using the “best pixel” selection technique (MOD09A1) and 16-day composites corrected by that model the bidirectional reflectance distribution function (MCD43A4). The results showed that in flood events part of the water is “missed” when temporal composites are used due to the incomplete use of all the cloud-free imagery (MOD09A1) or the “averaging” between dry and wet dates (MCD43A1). In the case of reservoirs, which water levels vary on weekly rather than daily steps, these issues tend to be less relevant. Despite this, using 8-day temporal composites from MOD09A1 (or MYD09A1) for retrospective analysis of floods and reservoir standing water recurrence is an acceptable approximation, although an underestimation of total area flooded, when large areas and long time-spans are required. The algorithm developed was applied to 10 ½ years of MOD09A1 data for the Australian continent, to generate a description of flood recurrence at 500m pixel resolution. A classification scheme was adopted to characterise each 500m grid cell in the continent with respect to the recurrence of inundation and the fraction of the area affected. The classification res...


CSIRO


Canberra


Natural Resource Management


Published Version (pdf) (12.66MB)


https://doi.org/10.4225/08/58518bd176131


© 2011 CSIRO To the extent permitted by law, all rights are reserved and no part of this publication covered by copyright may be reproduced or copied in any form or by any means except with the written permission of CSIRO.


Water for a Healthy Country Flagship Report Series


EP105358


Technical Report (Author)


English


1835095X


Guerschman, Juan; Warren, Garth; Byrne, Guy; Lymburner, Leo; Mueller, Norman; Van Dijk, Albert. MODIS-based standing water detection for flood and large reservoir mapping: Algorithm development and applications for the Australian continent. Canberra: CSIRO; 2011. csiro:EP105358. https://doi.org/10.4225/08/58518bd176131



Loading citation data...

Citation counts
(Requires subscription to view)