Shape selection in Landsat time series: a tool for monitoring forest dynamics

Abstract

We present a new methodology for fitting nonparametric shape-restricted regression splines to time series of Landsat imagery for the purpose of modeling, mapping, and monitoring annual forest disturbance dynamics over nearly three decades. For each pixel and spectral band or index of choice in temporal Landsat data, our method delivers a smoothed rendition of the trajectory constrained to behave in an ecologically sensible manner, reflecting one of seven possible 'shapes'. It also provides parameters summarizing the patterns of each change including year of onset, duration, magnitude, and pre- and postchange rates of growth or recovery. Through a case study featuring fire, harvest, and bark beetle outbreak, we illustrate how resultant fitted values and parameters can be fed into empirical models to map disturbance causal agent and tree canopy cover changes coincident with disturbance events through time. We provide our code in the r package ShapeSelectForest on the Comprehensive R Archival Network and describe our computational approaches for running the method over large geographic areas. We also discuss how this methodology is currently being used for forest disturbance and attribute mapping across the conterminous United States.

Phenology Modelling and Forest Disturbance Mapping with Sentinel-2 time series in Austria

“…RF or ANN), are needed. Furthermore, it can be worthwhile to categorise different 'shapes' of deviating courses based on an advanced threshold-based trajectory segmentation (Moisen et al 2016).…”

“…Previous research shows different methodical approaches to cope with these challenges. Most of the mainly LANDSAT-based TSAs are based on an image composite analysis (Nguyen et al 2018;Sebald et al 2019) or on some variant of a harmonic modelling approach (Zhu and Woodcock 2014;Moisen et al 2016;Pasquarella et al 2017;Nguyen et al 2018;Hermosilla et al 2019;Sebald et al 2019;Bullock et al 2020). Harmonic modelling approach robust but shows some limitations regarding the quality of temporal information (Jönsson et al 2018) and allow only little detail in reconstructing seasonal vegetation courses.…”

Analysis of Site-dependent Pinus halepensis Mill. Defoliation Caused by ‘Candidatus Phytoplasma pini’ through Shape Selection in Landsat Time Series

“…More recently, Landsat-based algorithms have emerged for the detection of the changes in the short-term anomalies of spectral trajectories related to the effects of defoliation caused by pests. These include the vegetation change tracker (VCT), vegetation continuous fields (VCF), LandTrendr, continuous change detection and classification (CCDC), multi-index integrated change algorithm (MIICA), a Fourier regression algorithm, a gradual ecosystem change algorithm and methods based on non-parametric statistical analysis [22]. Therefore, the relationship between the spectral trajectories of Landsat image time series and the temporal dynamics of pest disturbance can be applied to detect subtle and abrupt inter-annual forest changes [13,23].…”

Identifying Forest Impacted by Development in the Commonwealth of Virginia through the Use of Landsat and Known Change Indicators

“…While space precludes an exhaustive listing of multi-temporal change detection techniques, we did assess several likely possibilities, including shape selection, NLCD, Vegetation Change Tracker (VCT), EWMA-CD, LandTrendr, and relevant prior work by Sen et al [12,13,[19][20][21][22][23][24][25]. NLCD products are too infrequent to detect these subtle changes [12][13][14][22][23][24].…”

“…EWMA-CD may be too computationally and data intensive for many operational users at this juncture [25]. Shape selection probably holds the most promise, but given that forest disturbance and recovery differences among disturbance types often differs more in degree than in kind, further algorithm refinements may be necessary to separate low density development from other forest disturbances [19].…”