ee.Algorithms.TemporalSegmentation.LandTrendr

AI-generated Key Takeaways

outlined_flag

• LandTrendr is a temporal segmentation algorithm designed to detect trends in disturbance and recovery within yearly Landsat time-series data.

• It identifies breakpoints in spectral trajectories, using the first band of the image collection for initial detection and then fitting the breakpoints to all other bands.

• These breakpoints, representing changes in land cover, are fitted to a model assuming increasing values indicate disturbance and decreasing values signify recovery.

• The algorithm offers parameters for controlling spike dampening, segment recovery rates, model selection, and minimum data requirements to fine-tune the analysis.

• The output is an image containing the original and fitted values, segment vertices, and optionally fitted values for additional bands.

Landsat-based detection of Trends in Disturbance and Recovery: temporally segments a time-series of images by extracting the spectral trajectories of change over time. The first band of each image is used to find breakpoints, and those breakpoints are used to perform fitting on all subsequent bands. The breakpoints are returned as a 2-D matrix of 4 rows and as many columns as images. The first two rows are the original X and Y values. The third row contains the Y values fitted to the estimated segments, and the 4th row contains a 1 if the corresponding point was used as a segment vertex or 0 if not. Any additional fitted bands are appended as rows in the output. Breakpoint fitting assumes that increasing values represent disturbance and decreasing values represent recovery.

See: Kennedy, R.E., Yang, Z. and Cohen, W.B., 2010. Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr - Temporal segmentation algorithms. Remote Sensing of Environment, 114(12), pp.2897-2910.

Usage	Returns
ee.Algorithms.TemporalSegmentation.LandTrendr(timeSeries, maxSegments, spikeThreshold, vertexCountOvershoot, preventOneYearRecovery, recoveryThreshold, pvalThreshold, bestModelProportion, minObservationsNeeded)	Image

Argument	Type	Details
timeSeries	ImageCollection	Yearly time-series from which to extract breakpoints. The first band is usedto find breakpoints, and all subsequent bands are fitted using those breakpoints.
maxSegments	Integer	Maximum number of segments to be fitted on the time series.
spikeThreshold	Float, default: 0.9	Threshold for dampening the spikes (1.0 means no dampening).
vertexCountOvershoot	Integer, default: 3	The initial model can overshoot the maxSegments + 1 vertices by this amount. Later, it will be pruned down to maxSegments + 1.
preventOneYearRecovery	Boolean, default: false	Prevent segments that represent one year recoveries.
recoveryThreshold	Float, default: 0.25	If a segment has a recovery rate faster than 1/recoveryThreshold (in years), then the segment is disallowed.
pvalThreshold	Float, default: 0.1	If the p-value of the fitted model exceeds this threshold, then the current model is discarded and another one is fitted using the Levenberg-Marquardt optimizer.
bestModelProportion	Float, default: 0.75	Allows models with more vertices to be chosen if their p-value is no more than (2 - bestModelProportion) times the p-value of the best model.
minObservationsNeeded	Integer, default: 6	Min observations needed to perform output fitting.