geeet

image image image

Evapotranspiration (ET) models for use in python and with integration into Google Earth Engine.

geeet provides hybrid evapotranspiration (ET) models that run with numerical values and with Google Earth Engine images.

Introduction

geeet is a Python package providing a common set of building blocks for estimating evapotranspiration (ET) from remote sensing observations. It also features complete ET models such as PT-JPL and TSEB. All modules in geeet are designed to work with the input data provided in two formats: (1) as numpy ndarrays and (2) as Google Earth Engine (GEE) images. GEE is a cloud-based platform dedicated to Earth observation research that offers a multi-petabyte catalogue of geospatial data. Importantly, GEE offers cloud computing capabilities, which means that a user can interact with this geospatial data directly without having to download or process any data on premises. Access to these cloud-based services requires signing up for a GEE account. The geeet Python package was created to offer ET modeling tools that work for any user, whether they have a GEE account or not. For this reason, numpy is the only requirement for geeet.

Installation

geeet is available on the Python Package Index (pip). To install geeet using pip, run this command:

pip install geeet

If you want to install the latest development version hosted on github, run this command:

pip install git+https://github.com/kaust-halo/geeet

geeet is alsao available on conda-forge. To install geeet using conda, run this command:

conda install -c conda-forge geeet

The only requirement is a modern python installation (e.g. >3.6) and numpy. However, to test any of the GEE capabilities of geeet you will need to install the Python earthengine API (available through pip and conda), and have a GEE account.

First time use: running a toy example

If you have a GEE account and the earthengine API installed, we recommend first taking a look at this notebook demonstrating the basic use of the hybrid ET models with a simple toy example. In a nutshell, running one of the pre-built models can be done in two lines of code, e.g.:

from geeet.tseb import import tseb_series
et_tseb = tseb_series(img = sample_tseb_inputs)

where sample_tseb_inputs is a ee.Image containing all the necessary inputs for the TSEB model. You can use the same function to run the same model with numpy arrays:

et_tseb_out = tseb_series(\
    Tr = Tr, Alb = albedo, NDVI = NDVI, P = P, Ta = Ta, U = U, \
    Sdn = Sdn, Ldn = Ldn, doy = doy, time = Time, Vza = Vza,\
    longitude = lon, latitude = lat, zU = zU, zT = zT)

where in this case you supply all the inputs as numpy arrays through keyword arguments.

The keyword argument img is reserved only for GEE images. If img is supplied and is an instance of ee.Image, all other keyword arguments are ignored (with exceptions for some scalar parameters. Check the docstrings for each model).

geeet models can also be mapped to an ee.ImageCollection, e.g.:

from geeet.ptjpl import ptjpl_arid
et_outputs = et_inputs.map(ptjpl_arid)

where et_inputs is an ee.ImageCollection with the required inputs.

PT-JPL model for arid environments (as described in Aragon et al., 2018)

For a simple on-premises test of this PT-JPL model, run:

from geeet.ptjpl import ptjpl_arid
ptjpl_arid(Ta=25+273, P = 95500, NDVI = 0.75, F_aparmax=0.73, Rn=487.87, RH=25, doy=1, time=11, longitude=38.25)

The second notebook example is a self-contained, more realistic example that demonstrates the use of real GEE datasets with this PT-JPL model.

You can preview a pre-processed output of this example here (requires a GEE account).

Two-source Energy Balance model (TSEB)

geeet includes a two-source energy balance model mostly based on the original parameterizations of Norman et al., 1995. Specifically, it initializes the estimates of the temperatures of the soil and the canopy layers using a Priestley-Taylor equation. It then iteratively updates the temperatures, energy fluxes, and resistance values using the in-series resistance network parameterization.

For a simple on-premises test of this TSEB model, run:

from geeet.tseb import tseb_series
tseb_series(Tr = 26+273, Alb = 0.22, NDVI = 0.75, P=95500, Ta = 25+273, U=2, Sdn=745, Ldn=345, doy=1, time=11, Vza=0, longitude=38.25, latitude=30.25, zU=10, zT=2)

For a more realistic example using GEE data, see the following:

  1. This notebook demonstrates the inputs data setup using Landsat-8 satellite imagery and meteorological data from the ECMWF ERA5 reanalysis. The resulting image contains all the inputs necessary to run the TSEB model and is available as a public GEE asset and can be visualized with this sample GEE code editor script.
  2. This notebook shows how to run the TSEB model using the sample GEE TSEB inputs asset. The resulting image contains all the inputs as well as the outputs of the TSEB model and is available as a public GEE asset. It can also be visualized with this sample GEE code editor script.
  3. Finally, this notebook demonstrates running the TSEB model using the same data but as numpy arrays. To do this, the sample inputs image was exported to a GeoTIFF file. It also compares the output of the “on-premises” model to the output from the GEE model run (the outputs image was also exported as a GeoTIFF file).

References

References for each model are found in REFERENCES.txt. The source code for each module contains references for each function as well. Finally, each model contains two functions to display the references: cite() shows the main citation for the model, while cite_all() shows all the references for that model.

If you use this package for research, please cite the relevant model references.

Contributions

Contributions are welcome. Please open an issue to:

You can also submit a pull request if you have another working ET model - at least with numpy arrays as inputs.

Credits

This package was created with Cookiecutter and the giswqs/pypackage project template.

This site is open source. Improve this page.