Change Matrix Operation

On this page, you will find general help for the ProRaster product family including links to documentation, instructional videos, and training videos.
Previous Topic: Change Matrix Operation
Next Topic: Publish to MapInfo Pro
Back to: ProRaster Scientific Help
Back to: ProRaster Help
Back to: ProRaster
Go to: ProRaster Essential help page, ProRaster Premium help page, ProRaster Scientific help page.
The ProRaster User Guide is available for download as a PDF.

The Change Matrix Operation detects, quantifies and classifies changes between events in a raster with a temporal dimension.

The input is a single raster that must contain a temporal dimension and two or more temporal events. It is likely that you will create this source raster with the Raster Time Sequence operation which combines raster “frames” into a virtual raster with a temporal sequence.

The operation will consider all coincident data in pairs of events. It will classify the data (by assigning it to a bin you define, or by using the natural class of the data) and then record what bin the data was, and then what bin the data changed to. In other words, given two events, it counts all the cells that have changed from one “class” to another “class” for all combinations of classes. The result is a 2×2 matrix of size Bin Count by Bin Count. For any cell in the matrix, the X position indicates the bin the data was and the Y position indicates the bin the data changed to.

This operation is not virtual. When you hit OK it will proceed to process the input data at base resolution level (0) and, depending on the size or complexity of the input raster or the number of temporal events, it may take some time. It may consume all your CPU and RAM capacity whilst it is operating.

The operation creates three output files. The first is the change matrix output as a raster. The raster records the number of cells that transitioned from bin A to bin B. It also records the bin titles as strings and these can be reflected to the user in the map using the tooltip. The raster will have a temporal dimension so that you can flip through the processing results for each event pair.

For each event pair, an additional comma delimited ASCII (.csv) files is written. The filename is constructed using the from and to event times. The CSV file contains a mirror of the data written into the output raster file. Each row in the file will contain the From and To bin indices and bin labels, and the number of cells that made that change transition.

Finally, a single CSV file is written with a “FromTimeThreads” suffix. Like the CSV file written for each event, it contains a row for each from and to bin combination. But this file then records the change count for that bin combination for all event pairs in the processing operation. This allows you to track the change count for a single bin combination through time.

Select an input raster that has a temporal dimension and two or more temporal events. It will almost certainly be in either an MVR or MRR format.

Select an output raster that will contain the change matrix for each event pair in the input raster. The output raster will be in MRR format. Note that the CSV files will all use the output raster filename and location as the root name.

Select the field and band that will be processed and specify how to classify the data into a set of bins. This operation naturally operates on rasters containing a classified field. In this case you would choose the index band as the target and the bins mirror the classification table. For all other rasters, define the bins by supplying the base value of the first bin, the width of the bins, and the number of bins. Values that are >= the base value of the bin and < the base value of the next bin will be assigned to that bin.

Specify how the event pairs will be acquired. This is either “moving”, where it cycles through each event to the next event, or “Fixed” where it cycles from a named key event to every other event. It either case, the number of output events will be one less than the number of input events, and the time of each output event will reflect the time of the to event.

Tips

  • Display the MRR, color modulating by band 0 (the changed cell count), using a LUT Color layer.
  • Use a Log data transform.
  • Disable the Intensity component and use Nearest Neighbour interpolation.
  • Use tooltip mode “Key + All data table” to display the primary value (changed cell count) as well as the additional band values. Band 1 and 2 contain strings for the From class and To class.
  • Enable swapping by time using “Flip” to cycle through all the temporal events.
  • Note that cells that don’t change class are not counted, so from the bottom left to the top right there is a diagonal line of cells (where X=Y) that will have a changed cell count of zero.
  • Depending on the projection employed, you may be able to directly convert the cell count into an area. Simply multiply the cell count by the cell width and height for the base resolution level (0). To convert to a suitable projection prior to running he operation, use the Reproject operation.
  • To clip the input raster to a smaller region of interest, use the “Clip to polygon” processing operation prior to running the operation. 

The image below shows the raster generated by the Change Matrix operation for the National Land Cover Dataset (NLCD) for the USA. Color reflects the number of cells that have changed from a class to another class between the 1985 dataset and the 2024 dataset. This information is reflected in the table as the user moves the mouse pointer over the raster.