Aqua Monitor shows how the Earth's surface water has changed during the last 30 years.
http://aqua-monitor.appspot.com
See also the following Nature Climate Change paper: http://nature.com/nclimate/journal/v6/n9/full/nclimate3111.html
The website shows static map by default, visualizing surface water changes which occurred during 1985-2016 as explained in the above paper (see supplementary materials).
A dynamic mode can be turned on for a more detailed analysis. In this mode, the surface water changes are computed on-the-fly using parameters provided by the user. Additionally, percentile composite images are generated for two selected years.
The following Google Earth Engine script can be used to generate surface water changes on-the-fly: https://code.earthengine.google.com/ca025b5b16b195c18793936cb56b6a69. And the following was used to generate static maps: https://code.earthengine.google.com/ca025b5b16b195c18793936cb56b6a69.
The results presented in the paper include a few additional clean-up steps (noise clean-up in the mountains using HAND mask, deforestation-like changes). But they were excluded during static map generation.
Install Node.js, which is used by build scripts to compile sources and prepare everything for deployment.
Then run the following commands:
- npm install - downloads and installs all packages required to build the app (see packages.json)
- bower install - downloads and installs all client-side packages (see bower.json)
After that, run:
- gulp build - compiles everything (minifies code, generates styles, etc.) and places results in dist/
The resulting dist/ directory can be used to deploy everything as a Google App Engine app.
Use "gulp watch" to monitor sources continuously during development - they will be automatically compiled into dist/ on every change. See gulpfile.babel.js for the rules used to do this.
To deploy the Aqua Monitor under Google App Engine. The following files need to be added / modified:
- app/privatekey.json - add your service account key, this is used by Python backend.
- app/privatekey-web.json - add your client id, secret, and a refresh token, used at runtime to generate access to GEE for the JavaScript code.
Deploy using gcloud:
- gcloud app deploy --project=
Since we're using Python 2.7 here, we need old version of GCS tools here, such as dev_appserver.py
Installing this on Windows for development purposes is not easy. One way to do this is via https://pypi.org/project/gae_installer/:
- pip install gae_installer
(Google App Engine Standard environment, Python 2.7)
- cd dist
- python dev_appserver.py app.yaml
On Windows use (assumin virtualenv with Python 2.7 named aqua-monitor):
- cd dist
- python <replace_with_python_installation_dir>\envs\aqua-monitor\Lib\site-packages\google_appengine\dev_appserver.py app.yaml
On Linux (after you upgrade gcloud with the relevant components) gcloud components install app-engine-python-extras
gulp serve:gae
Open the aqua monitor at port 8081
This will work only at high zoom levels, the following (experimental) arguments can be used to tune the algorithm:
Format: http://aqua-monitor.appspot.com?min_year=1990&max_year=1995
- averaging_months1 = 36 - first filtering period, in months, all images from that period are used to compute percentile
- averaging_months2 = 18 - second filtering period, in months, all images from that period are used to compute percentile
- all_years = false - when true - linear regression in time will be computed using ALL years (moving average), >> can be extremely slow and even crash <<, but also produce much more accurate results.
- all_years_step - step in years
- min_year = 2000 - default value is 2000, but currently up to 1985 is supported. 1972 (60m resolution) will be added later
- max_year = 2015
- min_year_selection - first year to select in the timeline
- max_year_selection - last year to select in the timeline
- min_doy = 0
- max_doy = 365
- water_slope_opacity = 0.4 - can be used to show all slope values, not only the largest ones, ~0.7 is nice
- percentile=20 - default percentile is 20, use smaller values to select darker ones (also higher chance of shadows)
- filter_count = 0 - minimum number of images required, otherwise the result will be empty
- slope_threshold = 0.03 - slope threshold used to filter large changes
- slope_threshold_sensitive = 0.02 - threshold for sensitive wetness slope, used after zoom in to refine change around larger change
- ndvi_filter = -99 - can be used to filter-out changes which are not due to surface water (vegetation cover, like deforestation). This can be x2 slower. Using 0.1 to exclude deforestation woks in most cases.
- smoothen=true - smoothen changes image or leave it blocky
- debug = false - currently only shows surface water change refinement regions at higher zoom levels