This raster dataset covers all of NSW and is a raw count of inundated pixel observations from all available Landsat acquisitions from mid 1984 to mid 2016. The dataset was produced by applying a water index to each Landsat scene using the technique developed by Fisher and Danaher (2016). Water indexed images were classified into inundated and not inundated classes using a threshold value of -10. Masking of cloud, cloud shadow and other erroneous pixels resulting from sensor anomalies was undertaken using the F-mask technique (Zhu and Woodcock 2012) and these pixels were allocated a 'no data' value . The classified images (with pixels allocated to 'inundated' or 'not inundated' or 'no data' classes) were then stacked and the number of inundated observations were counted for each pixel in available Landsat scenes. Known commission errors include areas of terrain shadow, building shadow especially in urban areas, and tall dense forest such as some pine plantations. Known omission errors include areas of greater vegetation cover. Potential users should note that inundated observations are only recorded for cloud free observation times and locations, thus inundation events on cloudy days may not have been detected.

This raster dataset covers all of NSW and is a raw count of inundated pixel observations from all available Landsat acquisitions from mid 1984 to mid 2016. The dataset was produced by applying a …Show full descriptionThis raster dataset covers all of NSW and is a raw count of inundated pixel observations from all available Landsat acquisitions from mid 1984 to mid 2016. The dataset was produced by applying a water index to each Landsat scene using the technique developed by Fisher and Danaher (2016). Water indexed images were classified into inundated and not inundated classes using a threshold value of -10. Masking of cloud, cloud shadow and other erroneous pixels resulting from sensor anomalies was undertaken using the F-mask technique (Zhu and Woodcock 2012) and these pixels were allocated a 'no data' value . The classified images (with pixels allocated to 'inundated' or 'not inundated' or 'no data' classes) were then stacked and the number of inundated observations were counted for each pixel in available Landsat scenes. Known commission errors include areas of terrain shadow, building shadow especially in urban areas, and tall dense forest such as some pine plantations. Known omission errors include areas of greater vegetation cover. Potential users should note that inundated observations are only recorded for cloud free observation times and locations, thus inundation events on cloudy days may not have been detected.

This raster dataset covers all of NSW and is a raw count of inundated pixel observations from all available Landsat acquisitions from mid 1984 to mid 2016. The dataset was produced by applying a water index to each Landsat scene using the technique developed by Fisher and Danaher (2016). Water indexed images were classified into inundated and not inundated classes using a threshold value of -10. Masking of cloud, cloud shadow and other erroneous pixels resulting from sensor anomalies was undertaken using the F-mask technique (Zhu and Woodcock 2012) and these pixels were allocated a 'no data' value . The classified images (with pixels allocated to 'inundated' or 'not inundated' or 'no data' classes) were then stacked and the number of inundated observations were counted for each pixel in available Landsat scenes. Known commission errors include areas of terrain shadow, building shadow especially in urban areas, and tall dense forest such as some pine plantations. Known omission errors include areas of greater vegetation cover. Potential users should note that inundated observations are only recorded for cloud free observation times and locations, thus inundation events on cloudy days may not have been detected.