Persistent waterholes derived from Landsat archive for NAWRA

Persistent waterholes provide important refuge habitats. Waterhole persistence data shows the percentage of years where surface water has been identified along the stream channels during the driest period of the dry seasons from 1988 to 2015 and is derived from Landsat imagery. This dataset was created for the CSIRO Northern Australia Water Resource Assessment (NAWRA) project.\nThese data are maps of persistent waterholes derived from the Landsat archive from 1988 to 2015 (i.e. 27 years). The Landsat archive data were extracted from Geoscience Australia’s Surface Reflectance NBAR product available on the National Computing Infrastructure. The method used to derive the persistent waterhole maps is described in the NAWRA technical report (Sims et al., 2016). The data are provided as percentage of (valid) years that a pixel contained water at the end of the dry season. A percent value of 100 means that the pixel was wet at the end of every dry season for all available years. Those pixels containing water in them at the end of the dry season for at least 90% of the years are provided as shape files. More information is provided in the supporting attachment.\nLineage: The following steps were performed to derive the persistent waterholes maps:\n•\tFive to eight streamflow gauges, distributed across each catchment, were used to identify the end of the dry season as well as the length of the dry season from 1988 to 2015. No-flow was defined as zero ML/day. \n•\tThe end- of- dry- season date was used to identify the last Landsat scene before the start of the wet season (this was based on available imagery with minimal cloud cover) for each year from 1988 to 2015. In some cases, potential waterholes were obscured by cloud. Other times there were nulls in the ETM data; in May 2003 the ETM sensor had a Scan Line Corrector (SLC) error resulting in strips of null data across much of the scene. Hence a persistent waterhole was identified based on the number of cloud-free (and null-free) observations at the end of the dry season.\n•\tA selection of images (three to five) from the end of the dry season (from 1988–2015) from each Landsat row/path scene were visually examined to identify the best threshold to use for the NDWIXu (Xu 2006) to create a water mask (a value of -0.3 was found to be suitable for all areas).\n•\tThe Landsat NDWIXu water-masked images from the end of each dry season from 1988 to 2015 were used to identify the waterholes persisting at the end of every dry season (based on available and cloud-free observations). Those pixels identified as water at the end of most of the dry seasons are potentially persistent waterholes.\nIn-channel masks were used to remove surface water identified outside the watercourse using a distance of 500 m buffer from streamlines. Large dams were manually removed when necessary.\n