cryogrid / pyshore

Detect waterbodies and derive their shorelines

PyShore

PyShore is a python based tool that allows you to detect changes in lake shorelines derived from satellite imagery.

Description

The script accepts ortho-ready (very high resolution) multi-temporal satellite imagery as input and automatically

1. removes image noise
2. detects waterbodies
3. removes potential misclassifications (originating from infrastructural elements)
4. derives lake shoreline geometries and
5. retrieves shore movement rates and direction

It was applied to analyze the shoreline changes of thermokarst lakes that strongly influence Arctic periglacial ecosystems (publication in prep.).

Prerequisites

Technical

• Python 2.7+ with the following dependecies (via pip):
• gdal
• scikit-image
• numpy
• scipy
• matplotlib
• geopandas

Formal

Assemble your data in one input folder. It should contain your

• (multi-temporal) satellite imagery in GeoTIFF format
• infrastructure data (e.g. OpenStreetMap) in Shapefile format
• study area image extent in Shapefile format

The file names of the satellite imagery must begin with the acquisition date, starting with the year, e.g. 16JUL22xxxx.TIF, 20060823xxxx.TIF. etc. Your folder content should look similar to this:

├── ORDER_SHAPE_32606.dbf 	# study area image extent to prevent NoData values.
├── ORDER_SHAPE_32606.prj
├── ORDER_SHAPE_32606.shp
├── ORDER_SHAPE_32606.shx
├── 06AUG15222517-M2AS-058878563040_01_P001_GS_pansharpened_cubic_0.5.TIF	# multi-temporal imagery
├── 10JUL09221426-M2AS-058878563030_01_P001_GS_pansharpened_cubic_0.5.TIF
├── 13JUL16225401-M2AS-058878563020_01_P001_GS_pansharpened_cubic_0.5.TIF
├── 16JUL10222531-M2AS-058878563010_01_P001_GS_pansharpened_cubic_0.5.TIF
├── gis_osm_buildings_a_free_1.cpg	# OSM infrastructure data, both line and polygon features
├── gis_osm_buildings_a_free_1.dbf
├── gis_osm_buildings_a_free_1.prj
├── gis_osm_buildings_a_free_1.shp
├── gis_osm_buildings_a_free_1.shx
├── gis_osm_landuse_a_free_1.cpg
├── gis_osm_landuse_a_free_1.dbf
├── gis_osm_landuse_a_free_1.prj
├── gis_osm_landuse_a_free_1.shp
├── gis_osm_landuse_a_free_1.shx
├── gis_osm_roads_free_1.cpg
├── gis_osm_roads_free_1.dbf
├── gis_osm_roads_free_1.prj
├── gis_osm_roads_free_1.shp
└── gis_osm_roads_free_1.shx

Usage

The provided python scripts work as follows:

SmmothAndPlot.py

Usage: python SmoothAndPlot.py <path_in> <order-shape> <sitename>>

• <path_in> name of folder containing your OSM infrastructure Shapefiles and multi-temporal satellite images

• <order-shape> path to your image extent Shapefile

• <sitename> user chosen string for naming the output files.

The script accepts mutli-temporal images (you might need to do some manual pansharpening to enhance the spatial resolution of the data) stored in your input folder. The waterbody detection is based on an adaptive thresholding (Otsu) on the near infrared band. If your bands are not in the following order: B, G, R, NIR make sure to change line 74. The binary information (water vs. non-water) is subsequently vectorized and saved to your input folder. After the extraction of waterbodies, infrastructure elements (from your given infrastructure data) and waterbodies < 0.1 ha (size can be changed in line 253) are removed. An intermediate output is stored in "1_PyShore_ProcData", containing the derived waterbodies for every time step and the merged infrastructure map as Shapefiles. In folder 2_PyShore_Output you will find a .csv file for every lakeID and time step that contains the movement rate and direction for each shoreline point (a spatial resolution of 50 cm of your input image leads to the definition of a shoreline vertex each 50 cm. If you want to define an other segmentation, make sure to change line 178.

Visualization

BatchCSVtoSHP.py

Converts every lakeID.csv file to an ESRI Shapefile for visualisation.

batchloadcsv_qgs.py

Allows batch import to QGIS 2.18+