ee.Algorithms.TemporalSegmentation.C2c

Page Summary

outlined_flag

• The Composite 2 Change (C2C) algorithm segments a time series using a piecewise linear fit to minimize the number of segments while staying within a maximum root mean squared error (RMSE).

• For each band processed, the algorithm returns several output bands including change date, value, magnitude, duration, and rate, describing the characteristics of each fitted segment.

• The algorithm is in preview and its implementation details are subject to change.

An implementation of the Composite 2 Change (C2C) algorithm. This algorithm segments a time series using a piecewise linear fit with the minimum number of segments required to fit the data within the given maximum root mean squared error (RMSE). For each input band, the algorithm returns the following output bands:

• changeDate (Array[Double]): The dates at which changes are detected. The date format is determined by the dateFormat argument.
• value (Array[Double]): Value of the band at each changeDate.
• magnitude (Array[Double]): The difference between the values before and after a change date. The first magnitude is always NaN.
• duration (Array[Double]): The duration of the segment preceding the change date. The first duration is always NaN.
• rate (Array[Double]): Rate of change of the data preceding the change date. The first rate is always NaN.

If includePostMetrics is true the following variables are included per-band.

• postMagnitude (Array[Double]): The absolute difference between the value at the start of the next segment and the value at the change date. The last postMagnitude is always NaN.
• postDuration (Array[Double]): The duration of the segment following the change date. The last postDuration is always NaN.
• postRate (Array[Double]): The rate of change of the data following the change date. The last postRate is always NaN.

If includeRegrowth is true the following variables are included per-band.

• indexRegrowth (Array[Double]): The difference between the value at the change date and the value five data points after.
• recoveryIndicator (Array[Double]): The ratio of indexRegrowth to magnitude.
• regrowth60 (Array[Double]): Time difference between the change date and the data point where the series value is 60% of the pre-disturbance value.
• regrowth80 (Array[Double]): Time difference between the change date and the data point where the series value is 80% of the pre-disturbance value.
• regrowth100 (Array[Double]): Time difference between the change date and the data point where the series value is 100% of the pre-disturbance value.

See: Hermosilla et al. (2015) https://doi.org/10.1016/j.rse.2014.11.005 for further details on the original algorithm. Algorithm implementation can be found on GitHub: https://github.com/saveriofrancini/C2C-GEE Acknowledgements: FORWARDS and NextGenCarbon.

Citation: Txomin Hermosilla, Michael A. Wulder, Joanne C. White, Nicholas C. Coops, Daniel Coelho, Giovanni Ciatto, Noel Gorelick, and Saverio Francini. In preparation. Image compositing, time-series change detection and temporal metrics: Implementation of the Composite2Change (C2C) algorithm on Google Earth Engine.

This algorithm is in preview and is subject to change.

Usage	Returns
ee.Algorithms.TemporalSegmentation.C2c(collection, dateFormat , maxErrorList , spikesToleranceList , spikeRemovalMagnitudeList , maxError , maxSegments , infill , spikesTolerance , spikeRemovalMagnitude , includePostMetrics , includeRegrowth , interpolateRegrowth , useRelativeRegrowth , negativeMagnitudeOnly )	Image

Argument	Type	Details
collection	ImageCollection	Collection of images on which to run C2C.
dateFormat	Integer, default: 0	The time representation to use during fitting: 0 = jDays, 1 = fractional years, 2 = unix time in milliseconds. The start, end and break times for each temporal segment will be encoded this way.
maxErrorList	List, default: {}	List of maximum error (RMSE) values to be used for each band. If not provided, the maxError value will be used for all bands.
spikesToleranceList	List, default: {}	List of spike tolerance values to be used for each band. A value of 1 indicates no spike removal. If not provided, the spikesTolerance value will be used for all bands.
spikeRemovalMagnitudeList	List, default: {}	List of spike removal magnitude values to be used for each band. Spikes with a magnitude above this value are removed. If not provided, the spikeRemovalMagnitude value will be used for all bands.
maxError	Float, default: 0.075	Maximum allowed RMSE of the piecewise linear fit; controls segmentation sensitivity.
maxSegments	Integer, default: 6	Maximum number of segments permitted in the fitted tajectory.
infill	Boolean, default: true	Enables gap infilling within the time series to support stable fitting in the presence of missing values (i.e., values equal to 0).
spikesTolerance	Float, default: 0.85	Controls tolerance of spikes in the time series. Ranges from 0 to 1. A value of 1 indicates no spike removal, lower values are more aggressive.
spikeRemovalMagnitude	Float, default: 0.1	Spike removal magnitude threshold. Spikes with a magnitude (absolute difference from the average of neighbors) above this value are removed.
includePostMetrics	Boolean, default: true	Returns post-change descriptors (postMagnitude, postDuration, postRate).
includeRegrowth	Boolean, default: false	Returns recovery/regrowth metrics (indexRegrowth, recoveryIndicator, regrowth60/80/100).
interpolateRegrowth	Boolean, default: true	Linearly interpolate the time series using the detected changes before calculating the regrowth metrics.
useRelativeRegrowth	Boolean, default: false	Computes regrowth thresholds in relative terms to pre-disturbance conditions.
negativeMagnitudeOnly	Boolean, default: false	Retains only breakpoints associated with negative changes (directional filtering).