README.md

Landscape Classification (CoPe COMET sprint)

Zhendong Cao, Phil Wernette, Jaap Nienhuis

Objective

Apply image classification (see https://developers.google.com/earth-engine/guides/classification) and image segmentation (see https://en.wikipedia.org/wiki/Image_segmentation) to obtain landform characteristics of river deltas using Google Earth Engine code and available data.

Approach

Use the Google Earth Engine (EE) Python API with Jupyter notebooks running in Google Colab to subset EE datasets of interest, and (1) train a machine learning model (Supervised Classification) to obtain pixel-level probabilities of river delta deposits. (2) Apply connected components algorithms to group neighbouring pixels of high delta likelihood together retrieve delta landform statistics.

1. Image classification Use supervised ML techniques to train a classifier to identify deltas https://colab.research.google.com/github/giswqs/geemap/blob/master/examples/notebooks/32_supervised_classification.ipynb Needs training dataset (pick cells we know are deltas) Needs data to source (collect rasters that we think will predict delta occurrence)

2. Image segmentation Use connected components object-based methods to group delta pixels https://developers.google.com/earth-engine/guides/image_objects Needs output from image classification

Area of Interest

Input Datasets (subject to change)

• Elevation: MERIT DEM: Multi-Error-Removed Improved-Terrain DEM (https://developers.google.com/earth-engine/datasets/catalog/MERIT_DEM_v1_0_3)

• Imagery: Landsat 8 Tier 1 (https://developers.google.com/earth-engine/datasets/catalog/LANDSAT_LC08_C01_T1_SR)
  • Includes ALL available bands

• Land Cover: NLCD: USGS National Land Cover Database (https://developers.google.com/earth-engine/datasets/catalog/USGS_NLCD_RELEASES_2016_REL)

• Soil Great Groups: OpenLandMap USDA soil taxonomy great groups (https://developers.google.com/earth-engine/datasets/catalog/OpenLandMap_SOL_SOL_GRTGROUP_USDA-SOILTAX_C_v01)

• Soil Organic Carbon Content: OpenLandMap Soil organic carbon content (https://developers.google.com/earth-engine/datasets/catalog/OpenLandMap_SOL_SOL_ORGANIC-CARBON_USDA-6A1C_M_v02)
• Soil Water Content: OpenLandMap Soil water content at 33kPa (field capacity) (https://developers.google.com/earth-engine/datasets/catalog/OpenLandMap_SOL_SOL_WATERCONTENT-33KPA_USDA-4B1C_M_v01)
  • Available Depths: 0cm, 30 cm, 60 cm, 100cm, 200cm

• Topographic Diversity: Global SRTM Topographic Diversity (https://developers.google.com/earth-engine/datasets/catalog/CSP_ERGo_1_0_Global_SRTM_topoDiversity)

• Segmented deltas based on preliminary classification data

• Vectorized delta outlines

Programs/Languages

• Google Colab Notebooks
• Google Earth Engine
• Python

TASKS

• Develop code to pull available datasets for Gulf of Mexico relevant to coastal deltas. (option is available to download the data to your Google Drive, if desired)
• Create training dataset of known/identified deltas and non-deltas.
• Create segmentation algorithm based on binary delta-nondelta pixels
• Reduce segmented deltas to vectors
• Develop ML script to (1) train segmentation using 70/30 training/validation split, and (2) segment new imagery to identify deltas.
  • EE classifier?
  • TF classifier?
• Join resulting data polygons with feeder river data