The Yet Another TimeSeries Model (YATSM) algorithm is designed to monitor land surface phenomena, including land cover and land use change, using timeseries of remote sensing observations. The algorithm seeks to find distinct time periods, or time segments, within the timeseries by monitoring for disturbances. These time segments may be used to infer continuous periods of stable land cover, with breaks separating the segments representing ephemeral disturbances or permanent conversions in land cover or land use.
The "Yet Another..." part of the algorithm name is an acknowledgement of the influence a previously published timeseries algorithm - the Continuous Change Detection and Classification (CCDC) (Zhu and Woodcock, 2014) algorithm. While YATSM began as an extension from CCDC, it was never intended as a 1 to 1 port of CCDC and will continue to diverge in its own direction.
This algorithm is also influenced by other remote sensing algorithms which, like CCDC, are based in theory on tests for structural change from econometrics literature (Chow, 1960; Andrews, 1993; Chu et al, 1996; Zeileis, 2005). These other remote sensing algorithms include Break detection For Additive Season and Trend (BFAST) (Verbesselt et al, 2012) and LandTrendr (Kennedy et al, 2010). By combining ideas from CCDC, BFAST, and LandTrendr, this "Yet Another..." algorithm hopes to overcome weaknesses present in any single approach.
Please consider citing as:
Christopher E. Holden. (2015). Yet Another Time Series Model (YATSM). Zenodo. 10.5281/zenodo.17129
Documentation is available here.
Contributions to the documentation, especially for the user guide, is more than welcomed. The documentation for this project is built using Sphinx using the ReadTheDocs theme. See the docs/ folder for documentation source material.
The simplest way of using YATSM would be the pixel-by-pixel command line interface - run_yatsm.py.
We'll use the example Landsat stack from Chiapas, Mexico for this demonstration:
> run_yatsm.py --consecutive=5 --threshold=3 --min_obs=16 \
... --freq=1 --min_rmse 100 --test_indices "2 4 5" --screening RLM \
... --plot_index=2 --plot_style xkcd \
... ../landsat_stack/p022r049/images/ 133 106
Produces:
Python (2.7.x tested)
GDAL (1.10.0 tested)
Listed below are the Python library requirements for running YATSM. The version numbers listed are the versions I've used for development, but I suspect the versions are flexible.
cython >= 0.20.1
numpy >= 1.8.1
pandas >= 0.13.1
statsmodels >= 0.5.0
glmnet = 1.1-5 (see: https://github.com/dwf/glmnet-python)
scikit-learn >= 0.15.1
ggplot >= 0.5.8
docopt >= 0.6.1
patsy >= 0.3.0
These are listed in the requirements.txt file, which is usable by PIP as follows:
pip install -r requirements.txt
The R statistical software environment and the rpy2 Python to R interface are required for calculation of phenology metrics. You can find this requirement in requirements-pheno.txt.