TY - JOUR
T1 - Human or natural? Landscape context improves the attribution of forest disturbances mapped from Landsat in Central Europe
AU - Sebald, Julius
AU - Senf, Cornelius
AU - Seidl, Rupert
N1 - Publisher Copyright:
© 2021 The Author(s)
PY - 2021/9/1
Y1 - 2021/9/1
N2 - Disturbances have increased in Central Europe's forests, but whether changes in disturbance regimes are driven by natural or human causes remains unclear. Satellite-based remote sensing provides an important data source for quantifying forest disturbance change. Separating causes of forest disturbance is challenging, however, particularly in areas such as Central Europe where disturbance patches are small and disturbance agents interact strongly. Here we present a novel approach for the causal attribution of forest disturbance agents and illustrate its utility for 1.01 million disturbance patches mapped from Landsat data in Austria for the period 1986–2016. We gathered reference data on 2620 disturbance patches by conducting targeted field observations and structured interviews with 21 forest managers. We developed a novel indicator class characterizing the landscape context of a disturbance patch (i.e., the spatio-temporal autocorrelation of disturbance patches on the landscape), and combined it with other predictor variables describing the spectral signal, topography, and patch form of each disturbance patch. We used these predictors to identify the causal agents for disturbances mapped in Austria using Random Forest classification. Landscape context was the most important predictor of disturbance agent, improving model performance by up to 26 percentage points. Wind, bark beetles and timber harvesting were separated with an overall accuracy of 63%. Bark beetle patches were most difficult to identify correctly (producer's accuracy = 15%, user's accuracy = 30%), while regular timber harvesting was classified with highest certainty (producer's accuracy = 68%, user's accuracy = 82%). Harvesting dominates the disturbance regime of Austria's forests, with 70.5% of the disturbed area (76.7% of the disturbed patches) attributed to human causes and 29.5% (23.3%) to natural causes (wind: 23.0% [14.8%], bark beetles: 6.5% [8.5%]). Increases in disturbance since 1986 were driven by natural causes, with wind increasing by 408% and bark beetles increasing by 99% between the first and the second half of the observation period. Wind-disturbed patches were also considerably larger than those caused by bark beetles and harvesting (+102% and + 67%, respectively). Our novel approach to mapping causal agents of forest disturbance, applicable also to highly complex and interactive disturbance regimes, provides an important step towards a comprehensive monitoring and management of forest disturbances in a changing world.
AB - Disturbances have increased in Central Europe's forests, but whether changes in disturbance regimes are driven by natural or human causes remains unclear. Satellite-based remote sensing provides an important data source for quantifying forest disturbance change. Separating causes of forest disturbance is challenging, however, particularly in areas such as Central Europe where disturbance patches are small and disturbance agents interact strongly. Here we present a novel approach for the causal attribution of forest disturbance agents and illustrate its utility for 1.01 million disturbance patches mapped from Landsat data in Austria for the period 1986–2016. We gathered reference data on 2620 disturbance patches by conducting targeted field observations and structured interviews with 21 forest managers. We developed a novel indicator class characterizing the landscape context of a disturbance patch (i.e., the spatio-temporal autocorrelation of disturbance patches on the landscape), and combined it with other predictor variables describing the spectral signal, topography, and patch form of each disturbance patch. We used these predictors to identify the causal agents for disturbances mapped in Austria using Random Forest classification. Landscape context was the most important predictor of disturbance agent, improving model performance by up to 26 percentage points. Wind, bark beetles and timber harvesting were separated with an overall accuracy of 63%. Bark beetle patches were most difficult to identify correctly (producer's accuracy = 15%, user's accuracy = 30%), while regular timber harvesting was classified with highest certainty (producer's accuracy = 68%, user's accuracy = 82%). Harvesting dominates the disturbance regime of Austria's forests, with 70.5% of the disturbed area (76.7% of the disturbed patches) attributed to human causes and 29.5% (23.3%) to natural causes (wind: 23.0% [14.8%], bark beetles: 6.5% [8.5%]). Increases in disturbance since 1986 were driven by natural causes, with wind increasing by 408% and bark beetles increasing by 99% between the first and the second half of the observation period. Wind-disturbed patches were also considerably larger than those caused by bark beetles and harvesting (+102% and + 67%, respectively). Our novel approach to mapping causal agents of forest disturbance, applicable also to highly complex and interactive disturbance regimes, provides an important step towards a comprehensive monitoring and management of forest disturbances in a changing world.
UR - http://www.scopus.com/inward/record.url?scp=85106748748&partnerID=8YFLogxK
U2 - 10.1016/j.rse.2021.112502
DO - 10.1016/j.rse.2021.112502
M3 - Article
AN - SCOPUS:85106748748
SN - 0034-4257
VL - 262
JO - Remote Sensing of Environment
JF - Remote Sensing of Environment
M1 - 112502
ER -