ee-satellites 0.0.7

This library provides an easy-to-use, comprehensive, and flexible way to work with satellites data (for now: S1_GRD, S2_SR and LANDSAT8_T1_TOA). Its key advantages include: a well-documented API, support for the mainly used satellites, open-source code, and regular updates.
In addition to the already mentioned advantages, the implemented code relies on Google Earth Engine (GEE) APIs (used to access satellites data and perform some tasks such as cloud masking, image compositing, and time series selection) and exploits multi-threading (designed to work efficiently, by making a lot of parallel I/O requests to GEE).

These advantages make it an excellent tool for anyone working with satellites data, since it allows to generate datasets that can be both easily used for data analysis and efficiently integrated with well-known ML libraries, to deploy models.

How to install it?

$ pip install ee-satellites

This library is composed of lot of functions, the main ones are here described.

get_features() function

It allows to get from an input pandas DataFrame composed of fields information, an output DataFrame that contains for each time a selected satellite (sentinel-1, sentinel-2 or landsat-8) passed over the specified fields, within a given time period, all the mean values of some of the most used indexes (optical, radar or thermal).

The filters_params parameter is a list (default: None) containing the values of the Earth Engine filters to be used for extracting Image Collections, with the specified sentinel satellites. For Sentinel-2 the first parameter in the list is the value of the CLOUDY_PIXEL_PERCENTAGE filter (LESS OR EQUAL TO - values in range [0, 100]), whereas for Sentinel-1 the first parameter in the list is the value of the orbitProperties_pass filter (ASCENDING or DESCENDING).

The fields_threads parameter (default: 4) is the number of threads to dedicate to parallelization of GEE API requests over the fields level, the remaining ones instead are used to apply parallelization over dates level. The value of this parameter should be high (with respect to the overall number of threads exploitable) if you have a lot of crop fields but a little time-span to consider, whereas if you have fewer fields but a bigger time-span you should decrease this parameter. Finally, if you have a lot of fields with a lot of dates to process it is optimal to consider half of the overall number of threads available on the device used to perform the I/O requests.
A correct choice of this parameter can drastically reduce the features extraction time.

See the code for a better understanding.

Input DataFrame

The input DataFrame, lets suppose named fields_df, should be structured as follows (just columns position matters):

How to execute it?

First you have to initialize and authorize the Google Earth Engine APIs.

import ee, ee_satellites

ee.Authenticate()
ee.Initialize()

Then, supposing that you have already loaded the fields_df pandas DataFrame, you have just to run the following code. Change the satellite parameter value in case you want sentinel-1, or landasat-8, indexes extracted for the fields specified (within the time period selected - be careful that the given date format is consistent with ISO 8601 notation). See the above general description, instead, to comprehend what filters_params and fields_threads parameters are useful for.

df = ee_satellites.get_features(fields_df, '2022-01-01', '2022-12-31', satellite='sentinel-2', filters_params=['40'], fields_threads=3)

Output DataFrame

The output generated DataFrame, namely df, will be structured as follows:

What's new?

• version: 0.0.1:

  • This libary comes from the old one (named sentinel-satellites) where Landsat-8 indexes have been added

• versions: 0.0.2 to 0.0.7:

  • Improved descriptions
  • Added logo