CTRL LOGICTesela · User manual

Chapter 5

Change detection

Change detection compares the state of an area in two different periods and delineates the polygons where the terrain changed according to criteria that you set. It uses Sentinel-2 imagery and mathematical calculations (indices and thresholds); it does not use artificial intelligence. The result is indications that require verification, never a determination about their cause or legality.

There are two entry points: the “Monitoring” tool in the viewer (“Changes” and “Burns” tabs) and the “Change detection” page in the side menu, “Monitoring” group. Both request the analysis from the same engine. Burns are covered in chapter 6.

Requesting an analysis from the viewer

“Monitoring · change detection” panel with area, name, Changes and Burns tabs, and the “Same season in two years” block
“Monitoring · change detection” panel with the “Changes” tab active.
  1. Click “Monitoring” on the toolbar. The “Monitoring · change detection” panel opens.
  2. Define the area. You can choose an area you have already drawn, a polygon selected through “Identify → Detect changes here”, or click “Draw a new area”: click on the map for each vertex and double-click to close.
  3. Enter a name so that you can recognize the analysis later.
  4. Leave the “Changes” tab active (the “Burns” tab is covered in chapter 6).
  5. Set the “Before” and “After” periods (see below).
  6. If necessary, adjust the thresholds under “Advanced settings: thresholds”.
  7. Click “Detect changes”. The job is queued and takes a few minutes.

Periods and “Same season in two years”

Each period (start and end) is summarized in a single cloud-free image, calculated as the median of the available scenes. Three or four months of dry season work best; with less, there may be no clean image.

Comparing any two dates mixes real change with the effect of the rains. The “Same season in two years” block avoids this: you choose the months (“From” and “To”), the “Year before” and the “Year after”, and click “Fill in the periods”. The initial months are June to August; adjust them to your region. If the season has not started yet in one of the years, the panel asks you to choose an earlier year.

If your institution has more than one imagery provider configured, the “Imagery from” selector appears.

Advanced thresholds

A polygon counts as a change when the “after − before” difference meets all the conditions that are ticked. The initial values are:

IndexWhat it measuresInitial condition
NDVIVegetation vigorActive: drops by at least 0.2
BSIBare soilInactive: rises by at least 0.1
NBRBurning and vegetation moistureInactive: drops by at least 0.15
NDWISurface waterInactive: rises by at least 0.15

Where there is vegetation, a drop in NDVI is enough. In areas without vegetation (highlands, desert) it is better to use bare soil (BSI rises) or NBR (drops). The “Minimum patch (ha)” is also set, which discards smaller polygons; the initial value is 0.5 ha.

Following your analyses

Below the form, “My analyses” appears. Each card shows the status (Queued, Processing, Completed, Failed or Cancelled), the number of polygons and the hectares with indications, or “No changes” when no polygon met the rule. You can search by name and filter by status. A failed analysis can be retried (“Retry”) and a queued one can be cancelled (“Cancel”).

“My analyses” list with cards of completed analyses and the Open sheet, Before and after on the map and Archive buttons
The “My analyses” list with its actions. The example is a demonstration burn analysis; the actions are the same in a change analysis.

The analysis sheet

“Open sheet” shows the whole analysis in a window:

The types of change that can be assigned are “Vegetation loss”, “Exposed soil”, “New water” and “Unclassified”. They are descriptive categories of the type of alteration, not assessments of its origin.

From the sheet you can click:

Scenes by period

For the “before” and the “after” period, each analysis publishes several scenes that help interpret what was detected:

SceneWhat it is for
Natural colorWhat the eye would see
False colorHighlights healthy vegetation in bright red
SWIRShows soil and moisture; bare soil stands out
NDVIVegetation vigor
NDWISurface water
NDSISnow and ice

An urban false color (built-up surfaces and disturbed soil) is also generated. If an analysis ends with “No changes”, the scenes are still available: compare them by eye and, if you see differences, repeat the analysis with other thresholds.

Archiving and deleting

If you only want to tidy your list, use “Archive”. “Delete” is final.

Limits

Summary