Gap Identification and Damage Classification of Deciduous Vegetation by 2009 Ice Storm - Process and Workflow by Tina Rotenbury

Using ERDAS Imagine, the district wanted to determine deciduous forest conditions before the 2009 Ice Event, how much loss of deciduous vegetation after the event and what gain/loss of vegetation has occurred since the event. Incorporated various scripts and models as well as Raster Calculator in ArcMap to provide a map product for the district to use to field verify the data.

Transcript

1. Tina Rotenbury Supervisor’s Office – Russellville, AR
2. Area of Interest Sylamore Ranger District
3.  Map ice storm damage from January 2009 event  Assess current regeneration (green-up) status in storm damaged areas 2009 to 2011 regeneration change.  District is also wanting to look into Indiana Bat Habitat Restoration. Near Hwy 341 Mountain View, AR
4. Roads were completely blocked in some areas.
5.  Locate and download Landsat Scenes from GloVis  2007 Leaf-on Pre-event (August)  2009 Leaf-on Post-event (August)  2011 Leaf-on Current (July)  Ran the LPGS (Level 1 Product Generation System) script for each image  Gathered Necessary GIS Data  DEMs  Shapefiles  Models
6.  Created Subset of Images  Used Models to create Normalized Difference Vegetation Indices (NDVI) & Normalized Difference Moisture Indices (NDMI) Rasters  Created Layer Stack Image of the 2007 image  Created Mask of 2009 image  Used Raster Calculator in ArcMap to refine areas of interest above the 800 ft elevation  Used From Raster command to convert grids to polygons
7. The output file name Aug2007_rfl_subset_ndvi_ndmi_20clas s_12iterations_95pct_recode.tif is going to be used to create my mask.
8. Unsupervised Classification Used to create the pine mask.
9. Creating the Mask files – under Raster, Subset & Chip select Mask Input File = file to mask * Input Mask File = the recoded unsupervised classification file Setup Recode all values to 0 except the value you want to keep Output file = New Masked file Be sure to check Ignore Zero in Output Stats *Input files should be your pre/post images that are going to be used for the change highlight difference. Also – had to do an unsupervised classification to mask out the clouds for the 2009/2011 Landsat Images. Then had to apply all Masked images to the original images to be able to get a more accurate change highlight/difference image.
10. Used the Image Difference tool in ERDAS to pull out a 10, 15 and 20 percent change threshold of the NDVI and NDMI for each Pre/Post event scene. This sample shows the NDVI 2007 image (Before Image) and the NDVI 2009 (After Image) which will produce a 20% NDVI difference and a 20% NDVI Change Highlight.
12. Raster Calculator shows a conditional statement to simplify NDMI image to only 2 categories to show decrease moisture values in areas of potential damage.
13. Raster Calculator: conditional statement selecting from DEM where elevation is >800 and where Sylamore NDVI 20pct is within that elevation and all other values = 0.
14. Conditional Statement to pull out only the decrease values from the NDVI image. Values of 1 keep as 1 make all other values 0.
15. Used the Reclassify tool from Arc tool box, Spatial Analyst to remove all values from the NDMI/NDVI except the Decrease/Increase values.
16. Decreased NDVI Polygon Shapefile with areas >= 20 Acres
17. Map shows the Decrease NDVI of 25 pct above the 800ft elevation threshold.
21.  Staff wanted something by the end of FY2011  Due to budgets and other time constraints – need to revisit with Forest Silviculturist to determine new time line and make adjustments where necessary.
22.  Accuracy Assessment?  Visit with resource staff and plan a trip to the field for verification of data, using map product for comparison.  See if district staff have visited sketch flight areas and have verified the data and get their information, if available.
25. Tina Rotenbury Ozark-St. Francis National Forests Supervisor’s Office 605 West Main Street, Russellville, AR 72801 trotenbury@fs.fed.us 479-964-7217