I am not sure if you have heard of the invasive estuarine weed rice grass (spartina anglica), but it has an unique spectral signature even in the RGB colour bands and I am using this to map rice grass and determine its rate of spread (hectares per year) in far North West Tasmania, Australia.
I have been provided with 3 datasets; April 2011, Jan 2014 and Dec 2015 and have so far managed to undertake this workflow in ArcMap;
- Supervised classification to identify rice grass and mudflats (field verification completed);
- Using the Image Analysis Window, compute the different between 2011 and 2014, 2014 and 2015 and 2011 and 2015;
- Create 3 study areas to reduce analysis times and which are rice grass only sites;
- Clip study areas within intertidal zone layer as this is where rice grass occurs;
- Currently 3 raster datasets outputs; Diff_2011_2014, Diff_2014_2015 and Diff_2011_2015.
I am not sure of the next steps to take for analysis and am aware of some errors still in the data. Should I convert the above raster data sets to vector polygons and then I can manually edit and remove the errors? Or should I continue process in raster? Do I need to normalise the 3 data sets? Any suggestions are welcomed.
There is the project webmap in you are interested; map.
Thanks for your time and I look forward to your answer,