Content uploaded by Andrew Haywood
Author content
All content in this area was uploaded by Andrew Haywood on Jul 17, 2016
Content may be subject to copyright.
Forest Disturbance Mapping
We followed a two-phase classification approach based on Senf et al. (2015) to
map spatial and temporal patterns of natural and anthropogenic disturbance: First,
we classified the Landsat time series disturbance and recovery metrics into three
classes: 1) acute high severity disturbances; 2) low severity disturbances; and 3)
undisturbed areas. Second, we assigned all pixels identified within the acute high
severity in the first classification phase a likelihood of being disturbed by either
wildfire or clear-fell timber harvesting; we assigned all pixels identified within the
low severity disturbance a likelihood of being disturbed by planned burning,
selective timber harvesting, insects or drought.
Random Forests Model 1
Random Forest Model 2a
Random Forests Model 2b
This study presents an examination on the capacity of readily available open-
source tools to model spectral-temporal pixel time-series from annual Landsat
time series to map fire and timber disturbance dynamics within wet sclerophyll
forests in south-east Australia.
Study Area
Conclusions
Reference data:
Time series analysis:
We carried out the time series analysis using a number of standard R packages.
The time series analysis was conducted to:
1. extract spectral time series for each pixel using the raster package (Hijmans
and van Etten, 2012);
2. fit structural breakpoints to the time series using the bfast package (Verbesselt
et al., 2010); and
3. extract the following metrics from the pixel time series:
•For pixels without a breakpoint detected, the slope and intercept of the
linear trend of the time series was extracted
•For pixels with a breakpoint detected, the magnitude of the breakpoint was
calculated, the date of the breakpoint, the slope and intercept for the line
segments before and after the breakpoint were extracted
Introduction
Results
MAPPING DISTURBANCE DYNAMICS IN WET SCLEROPHYLL
FORESTS USING TIME SERIES LANDSAT
Andrew Haywood12, Jan Verbesselt 3, Patrick Baker 2
1EU REDD Facility, European Forest Institute, Malaysia 2Laboratory of Geo-information Science and Remote Sensing, Wageningen, Netherlands
3Royal Melbourne Institute of Technology, Melbourne, Australia
Methods
In this study we characterised acute high severity and low severity disturbance in
South-East Australia, using a well-established Landsat-based time series
technique.
From our results, we conclude that Landsat can be utilised to reliably distinguish
between acute severity disturbance agents (clear-felling and wildfire) in our study
region, using specific spectral time-series features. However, more research is
needed in distinguishing between the low severity disturbance agents (planned
burning and selective logging).
The resulting maps and estimates offer a combined and detailed picture of
disturbance dynamics in our study region through quantifying both the temporal
and spatial dynamics. These otherwise unavailable spatially explicit and quality
assured maps can help inform science and management needs.
Bibliography:
1. Hijmans, R.., van Etten, J., 2012. raster: Geographic Data Analysis and Modeling [WWW
Document]. R Packag. version 2.3-33. URL URL http://CRAN.R-project.org/package=raster.
(accessed 11.14.15).
2. Verbesselt, J., Hyndman, R., Newnham, G., Culvenor, D., 2010. Detecting trend and
seasonal changes in satellite image time series. Remote Sens. Environ. 114, 106–115.
doi:10.1016/j.rse.2009.08.014
3. Senf, C., Pflugmacher, D., Wulder, M.A., Hostert, P., 2015. Characterizing spectral–temporal
patterns of defoliator and bark beetle disturbances using Landsat time series. Remote Sens.
Environ. 170, 166–177. doi:10.1016/j.rse.2015.09.019
