openet-ndvi 0.0.9

Earth Engine based NDVI ET Model

WARNING: This code is in development, is being provided without support, and is subject to change at any time without notification

This repository provides an Earth Engine Python API based implementation of a simple model for computing evapotranspiration (ET) as a linear function of the normalized difference vegetation index (NDVI). In this model, the fraction of reference ET (ETf) is computed as:

\begin{equation*} ETf = m * NDVI + b \end{equation*}

where m and b have defaults of 1.25 and 0.0 respectively. The actual ET is computed as:

\begin{equation*} ET = ETf * ETr \end{equation*}

Input Collections

The NDVI ET model is currently implemented for the following Earth Engine image collections:

Landsat TOA

• LANDSAT/LC08/C01/T1_TOA or LANDSAT/LC08/C01/T1_RT_TOA

• LANDSAT/LE07/C01/T1_TOA or LANDSAT/LE07/C01/T1_RT_TOA

• LANDSAT/LT05/C01/T1_TOA

• LANDSAT/LT04/C01/T1_TOA

Landsat SR

• LANDSAT/LC08/C01/T1_SR

• LANDSAT/LE07/C01/T1_SR

• LANDSAT/LT05/C01/T1_SR

• LANDSAT/LT04/C01/T1_SR

Sentinel 2 TOA

• COPERNICUS/S2

Model Structure

The primary way of interact with the NDVI-ET model are through the “Collection” and “Image” classes.

Collection

The Collection class should be used to generate image collections of ET (and and other model Variables). These collections can be for image “overpass” dates only or interpolated to daily, monthly, or annual time steps. The collections can be built for multiple input collections types, such as merging Landsat 8 and Sentinel 2.

The Collection class is built based on a list of input collections ID’s, a date range, and a study area geometry.

Required Inputs

collections

List of Earth Engine collection IDs (see Input Collections).

start_date

ISO format start date string (i.e. YYYY-MM-DD) that is passed directly to the collection .filterDate() calls.

end_date

ISO format end date string that is passed directly to .filterDate() calls. The end date must be exclusive (i.e. data will go up to this date but not include it).

geometry

ee.Geometry() that is passed to the collection .filterBounds() calls. All images with a footprint that intersects the geometry will be included.

Optional Inputs

etr_source

Reference ET source collection ID. Optional, the default is ‘IDAHO_EPSCOR/GRIDMET’.

etr_band

Reference ET source band name. Optional, the default is ‘etr’.

cloud_cover_max

Maximum cloud cover percentage. The input collections will be filtered to images with a cloud cover percentage less than this value. Optional, the default is 70 (%).

filter_args

Custom filter arguments for teh input collections. This parameter is not yet fully implemented.

model_args

A dictionary of argument to pass through to the Image class initialization. This parameter is not yet fully implemented.

Overpass Method

variables

List of variables to calculate/return.

Interpolate Method

variables

List of variables to calculate/return.

t_interval

Time interval over which to interpolate and aggregate values. Choices: ‘daily’, ‘monthly’, ‘annual’, ‘custom’ Optional, the default is ‘custom’.

interp_method

Interpolation method. Choices: ‘linear’ Optional, the default is ‘linear’.

interp_days

Number of extra days before the start date and after the end date to include in the interpolation calculation. Optional, the default is 32.

Collection Examples

import openet.ndvi as ndvi_et

overpass_coll = ndvi_et.Collection(
        collections=['LANDSAT/LC08/C01/T1_TOA'],
        start_date='2017-06-01',
        end_date='2017-09-01',
        geometry=ee.Geometry.Point(-121.5265, 38.7399),
        etr_source='IDAHO_EPSCOR/GRIDMET',
        etr_band='etr') \
    .overpass(variables=['et', 'etr', 'etf'])

monthly_coll = ndvi_et.Collection(
        collections=['LANDSAT/LC08/C01/T1_TOA'],
        start_date='2017-06-01',
        end_date='2017-09-01',
        geometry=ee.Geometry.Point(-121.5265, 38.7399),
        etr_source='IDAHO_EPSCOR/GRIDMET',
        etr_band='etr') \
    .interpolate(variables=['et', 'etr', 'etf'] t_interval='monthly')

Image

The Image class should be used to process a single image, an image collection with custom filtering, or to apply custom parameters to each image in a collection.

Typically the NDVI-ET Image is initialized using one of the collection/sensor specific helper methods listed below (see below). These methods rename the bands to a common naming scheme, apply basic cloud masking, and .

Image collections can be built by mapping one of the helper methods over an image collection. Please see the Image Mapping example notebook for more details.

The Image class can also be initialized using any Earth Engine image with an ‘ndvi’ band and a ‘system:time_start’ property.

Landsat Collection 1 Top-of-Atmosphere (TOA) Input Image

To instantiate the class for a Landsat Collection 1 TOA image, use the Image.from_landsat_c1_toa() method.

The input Landsat image must have the following bands and properties:

SPACECRAFT_ID	Band Names
LANDSAT_4	B1, B2, B3, B4, B5, B7, B6, BQA
LANDSAT_5	B1, B2, B3, B4, B5, B7, B6, BQA
LANDSAT_7	B1, B2, B3, B4, B5, B7, B6_VCID_1, BQA
LANDSAT_8	B2, B3, B4, B5, B6, B7, B10, BQA

Property	Description
system:index	Landsat Scene ID Must be in the Earth Engine format (e.g. LC08_044033_20170716)
system:time_start	Image datetime in milliseconds since 1970
SPACECRAFT_ID	Used to determine which Landsat type (for band renaming) Must be: LANDSAT_4, LANDSAT_5, LANDSAT_7, or LANDSAT_8

Landsat Collection 1 Surface Reflectance (SR) Input Image

To instantiate the class for a Landsat Collection 1 SR image, use the Image.from_landsat_c1_sr() method.

The input Landsat image must have the following bands and properties:

SATELLITE	Band Names
LANDSAT_4	B1, B2, B3, B4, B5, B7, B6, pixel_qa
LANDSAT_5	B1, B2, B3, B4, B5, B7, B6, pixel_qa
LANDSAT_7	B1, B2, B3, B4, B5, B7, B6, pixel_qa
LANDSAT_8	B2, B3, B4, B5, B6, B7, B10, pixel_qa

Property	Description
system:index	Landsat Scene ID Must be in the Earth Engine format (e.g. LC08_044033_20170716)
system:time_start	Image datetime in milliseconds since 1970
SATELLITE	Used to determine which Landsat type (for band renaming) Must be: LANDSAT_4, LANDSAT_5, LANDSAT_7, or LANDSAT_8

Sentinel 2 TOA Input Image

To instantiate the class for a Landsat Collection 1 TOA image, use the Image.from_sentinel2_toa() method.

The input Landsat image must have the following bands and properties:

SPACECRAFT_NAME	Band Names
Sentinel-2A	B2, B3, B4, B8, B11, B12, QA60
Sentinel-2B	B2, B3, B4, B8, B11, B12, QA60

Property	Description
system:index	Sentinel 2 Scene ID Must be in the Earth Engine format (e.g. 20180716T183929_20180716T185042_T10SGJ)
system:time_start	Image datetime in milliseconds since 1970
SPACECRAFT_NAME	Used to determine which Sentinel 2 type Must be: Sentinel-2A or Sentinel-2B Not currently used or checked

Image Example

import openet.ndvi as ndvi_et
landsat_img = ee.Image('LANDSAT/LC08/C01/T1_TOA/LC08_044033_20170716')
et_img = ndvi_et.Image.from_landsat_c1_toa(
    landsat_img, etr_source='IDAHO_EPSCOR/GRIDMET', etr_band='etr).et

Variables

The NDVI-ET model can compute the following variables:

ndvi

Normalized difference vegetation index [unitless]

etf

Fraction of reference ET [unitless]

etr

Reference ET (alfalfa) [mm]

et

Actual ET [mm]

There is also a more general “calculate” method that can be used to return a multiband image of multiple variables (see example…)

Reference ET

The reference ET data source is controlled using the “etr_source” and “etr_band” parameters.

The model is expecting an alfalfa reference ET (ETr) and will not return valid results if a grass reference ET (ETo) is used.

Reference ET Sources

GRIDMET

Collection ID: IDAHO_EPSCOR/GRIDMET

http://www.climatologylab.org/gridmet.html

Spatial CIMIS

Collection ID: projects/openet/cimis/daily

https://cimis.water.ca.gov/SpatialData.aspx

Example Notebooks

Detailed Jupyter Notebooks of the various approaches for calling the OpenET NDVI ET model are provided in the “examples” folder.

• Computing daily ET for a single Landsat image

• Computing a collection of “overpass” ET images

• Computing a collection of interpolated monthly ET images

Installation

The python OpenET NDVI based ET module can be installed via pip:

pip install openet-ndvi

Dependencies

• earthengine-api

• openet-core

OpenET Namespace Package

Each OpenET model is stored in the “openet” folder (namespace). The model can then be imported as a “dot” submodule of the main openet module.

import openet.ndvi as ndvi_et

Development and Testing

Please see the CONTRIBUTING.rst.

References