Skip to content

Land Cover Change Analysis Workflow

Jump to bottom Edit New page
Trisha Singh edited this page May 6, 2018 · 32 revisions

Step 1: Obtain Remote Sensing Data

Detailed instructions are in Lab Five in the course manual. Summary:

  • Save a kml file of the area of interest and load it in Earthdata to look for images (spatial filter).
  • To find appropriate tiles in Earthdata, apply collection filter (landsat) and temporal filter (years of interest). To further refine the area of interest, apply a spatial filter by grid code derived from a relevant Landsat tile.
  • Download granules and geotiff products.
  • Uncompress .tar.gz files.
  • Note: Remember that most Landsat images will require patching, so download more than one granule.

Step 2: Write up Metadata for Granules

  • Use tips from Lab 4 and end of Lab 5 for metadata format.

Step 3: Correct the Images

Landsat 8

These steps are outlined in Lab 7.

Before Landsat 8

Requires the following: concatenate band images, optical calibration, and manage noData. These steps can be carried out in Orfeo and are outlined in Lab 8 and in the videos L8a_Concatenate, L8b_OpticalCalibration, L8h_RecapOpticalCalibrationMaskingPatching. Summary:

  • Image concatenation: Launch OTB applications browser and use Image Manipulation -> ConcatenateImages. Set output to uint 8. Set different RGB composites to check. (L8a_Concatenate)

  • Optical calibration: Open the Calibration -> Optical Calibration tool. Save output as float, calibration level is Image to Top of Atmosphere, fill out the rest of the parameters using the scrapeMetaData sheet (may need to alter the formulae in the sheet depending on landsat year). (L8b_OpticalCalibration)

  • NoData management: (L8h_RecapOpticalCalibrationMaskingPatching, 6:45 onwards)

    • Resample to match grid system. Use Geometry -> Superimpose tool.
    • Create data mask (=1 if value in both images, 0 if no data in either image). Use Miscellaneous -> Band Math tool and expression <im1b1 && im1b2 && im1b3 && im1b4 && im1b5 && im2b1 && im2b2 && im2b3 && im2b4 && im2b5>.
    • Apply this mask to trim images. Use Conversion -> ManageNoData tool -> Apply a mask as no data.

Step 4: Unsupervised Classification

  • Use SAGA and trimmed images from previous step.
  • Tools -> Imagery -> Classification -> K-Means Clustering for Grids. (L8i_KMeans)
  • If helpful, reclassify clusters. (L8i2_KMeansClasses)
Add a custom footer
Add a custom sidebar
Clone this wiki locally