SimpleSpatialGapfiller - python class for filling gaps in matrices based on machine learing techniques. Main purpose is to provide convenient and simple instruments for modeling geophysical parameters, derived with Earth Remote Sensing, under clouds. But it also could be used for any matrices.

Sarafanov, M.; Kazakov, E.; Nikitin, N.O.; Kalyuzhnaya, A.V. A Machine Learning Approach for Remote Sensing Data Gap-Filling with Open-Source Implementation: An Example Regarding Land Surface Temperature, Surface Albedo and NDVI. Remote Sens. 2020, 12, 3865.

'python>=3.7',
'gdal>=2.4',
'numpy',
'scikit-learn==0.21.3',
'pandas',
'scipy',
'netCDF4',
'pyproj' 

conda install -c conda-forge gdal

pip install git+https://github.com/Dreamlone/SSGP-toolbox

For now SSGP-toolbox is presented with:

• Gapfiller class
• Discretizator class
• Several preparators: for Sentinel 3 LST data; for MODIS LST products; for MODIS NDVI based on reflectance product.
• Algorithm for identifying cloud-shaded pixels in temperature field

By the way, you can prepare any data by yourself, it must be in binary numpy matrices format (.npy) and organized in several directories, as shown in docs.

All documentation and examples for now are described in Markdown files and Jupyter Notebooks:

• Theoretical basis
• Sentinel 3 LST preparation
• MODIS LST preparation
• Identifying cloud-shaded pixels

• Gapfiller class, how to organize data and how to use

• Discretizator class, how to organize data and how to use

If you want to compare the accuracy of your algorithm with ours, you can use the dataset we have prepared. You can find it in the "Comparison" folder. The dataset already contains the layers filled in by our model, as well as the "CRAN gapfill" and "gapfilling rasters" layers.

• Detailed description of the data

Feel free to contact us:

Mikhail Sarafanov (maintainer) | [email protected]

Eduard Kazakov | [email protected]