Forest Change Detection Using Remote Sensing Time Series

Forest change detection with remote sensing time series has become a cornerstone for monitoring deforestation, degradation and recovery at regional to global scales. By exploiting decades of observations from instruments such as Landsat and Sentinel-2, researchers can track spectral signatures of vegetation over time, identify subtle transitions in canopy cover and quantify rates of disturbance and regrowth. Core methodologies include temporal segmentation, which partitions each pixel’s reflectance history into linear or harmonic segments, and index-based approaches that emphasise greenness signals. Advanced techniques combine spectral mixture analysis to resolve sub-pixel forest fractions, machine-learning classifiers to remove noise and cloud contamination, and cloud-computing platforms for rapid processing of petabyte-scale archives. These developments support policy evaluation, carbon accounting and biodiversity assessments by providing spatially explicit maps of forest change, enabling both retrospective analyses of long-term trends and near-real-time alerts for emerging losses. Global applications range from tracking Amazon fragmentation to assessing temperate reforestation efforts, underlining the practical value of time series approaches for sustainable forest management.

Research from Nature Portfolio

Recent studies have applied spectral-temporal segmentation to comprehensive Landsat archives to map forest disturbance in China over the past decades. One study employed an advanced segmentation algorithm across all available imagery to characterise disturbance rate, size, frequency and severity from 1986 to 2020, revealing that nearly 40 % of forests experienced disturbance and that national rates declined over time in response to policy interventions. Spatially explicit maps highlighted contrasting trends, with intensification of disturbances in the southeast and reduction in the northeast, providing a data layer for understanding drivers of forest dynamics across diverse biomes.

Forest Change Detection Using Remote Sensing Time Series publication trend

The graph below shows the total number of articles in forest change detection using remote sensing time series across all publications each year (not limited to Nature Index journals).

Technical terms

Spectral-temporal segmentation: Partitioning of a pixel’s spectral time series into continuous segments bounded by change points, allowing precise detection of disturbance onset, duration and recovery.

Spectral mixture analysis (SMA): Decomposition of mixed pixel reflectance into constituent endmember spectra (for example vegetation, soil and shadow) to estimate sub-pixel cover fractions.

Continuous Change Detection and Classification (CCDC): Algorithm that fits harmonic or linear models to time series data and flags temporal breakpoints indicating land cover change events.

LandTrendr algorithm: Temporal segmentation method that fits straight-line segments to multispectral time series, extracting the timing, magnitude and duration of forest change events.

Vegetation index: Mathematical combination of spectral bands (such as NDVI or EVI) designed to highlight vegetation greenness and health for change analysis.

References

1. A novel approach towards continuous monitoring of forest change dynamics in fragmented landscapes using time series Landsat imagery. International Journal of Applied Earth Observation and Geoinformation (2023).
2. Forest disturbance decreased in China from 1986 to 2020 despite regional variations. Communications Earth & Environment (2023).
3. Implementation of the LandTrendr Algorithm on Google Earth Engine. Remote Sensing (2018).