Model Impervious Surfaces Using Sentinel Imagery
Description
Characterizes a given watershed as pervious, impervious, or open water using deep learning and Sentinel-2 imagery. Users must be logged into an AGOL organization to run this tool, unless they provide their own Sentinel-2 imagery and local copy of the CORINE deep learning model.
Usage
Characterizes a given watershed as pervious, impervious, or open water using deep learning and Sentinel-2 imagery. Users must be logged into an AGOL organization to run this tool, unless they provide their own Sentinel-2 imagery and local copy of the CORINE deep learning model. Output rasters include 1) the deep learning predicted CORINE landcover classes, and 2) the reclassified version of these classes.
This tool performs the following actions:
- Exports the most recent and cloud free Sentinel-2 image from the Living Atlas using the input watershed boundary as the image extents
- Saves the Sentinel-2 image locally, including a full Sentinel image (13 bands) and the natural color representation of the image (RGB bands)
- Classifies the full Sentinel image into CORINE landcover classes using the Esri pre-trained deep learning model from Living Atlas
- Reclassifies the CORINE classes to categories of impervious, pervious, and open water
If users wish to model the CORINE landcover classes in their area for a specific time period, they can manually export a Sentinel-2 image with their time filter and use this image instead. Also, users can point the tool directly to a local version of the Esri deep learning model if it was previously downloaded.
Parameters
| Parameter Name | Type | Direction | Data Type | Dialog Reference |
|---|---|---|---|---|
| Input Boundary Feature | Required | Input | Feature Layer | Feature class to use as the export extents for the sentinel image |
| Impervious Classes | Required | Input | Multiple Value | Classes that will be reclassified as impervious |
| Output CORINE Classified Raster | Required | Output | Raster Layer | Name for the output predicted CORINE classes raster |
| Output Reclassified Impervious Raster | Required | Output | Raster Layer | Name for the output reclassified CORINE classes raster |
| (Optional) Input Sentinel Raster | Optional | Input | Raster Layer | Optional input Sentinel image. Use this option if there is a specific timeframe for which CORINE landcover classes should be predicted. The image used must encompass the input boundary feature and include all 13 bands. |
| (Optional) CORINE Deep Learning Model | Optional | Input | File | Optional input CORINE deep learning model. Use this option if the CORINE Deep Learning Model file (*.dlpk or *.emd) has already been saved locally and you do not want to download it from the Living Atlas. |