Using Space-Time Features to Improve Detection of Forest Disturbances from Landsat Time Series

Current research on forest change monitoring using medium spatial resolution Landsat satellite data aims for accurate and timely detection of forest disturbances. However, producing forest disturbance maps that have both high spatial and temporal accuracy is still challenging because of the trade-off between spatial and temporal accuracy. Timely detection of forest disturbance is often accompanied by many false detections, and existing approaches for reducing false detections either compromise the temporal accuracy or amplify the omission error for forest disturbances. Here, we propose to use a set of space-time features to reduce false detections. We first detect potential forest disturbances in the Landsat time series based on two consecutive negative anomalies, and subsequently use space-time features to confirm forest disturbances. A probability threshold is used to discriminate false detections from forest disturbances. We demonstrated this approach in the UNESCO Kafa Biosphere Reserve located in the southwest of Ethiopia by detecting forest disturbances between 2014 and 2016. Our results show that false detections are reduced significantly without compromising temporal accuracy. The user’s accuracy was at least 26% higher than the user’s accuracies obtained when using only temporal information (e.g., two consecutive negative anomalies) to confirm forest disturbances. We found the space-time features related to change in spatio-temporal variability, and spatio-temporal association with non-forest areas, to be the main predictors for forest disturbance. The magnitude of change and two consecutive negative anomalies, which are widely used to distinguish real changes from false detections, were not the main predictors for forest disturbance. Overall, our findings indicate that using a set of space-time features to confirm forest disturbances increases the capacity to reject many false detections, without compromising the temporal accuracy.

[1]  Warren B. Cohen,et al.  Detecting trends in forest disturbance and recovery using yearly Landsat time series: 2. TimeSync — Tools for calibration and validation , 2010 .

[2]  Scott J. Goetz,et al.  National-scale estimation of gross forest aboveground carbon loss: a case study of the Democratic Republic of the Congo , 2013 .

[3]  Jan Verbesselt,et al.  Monitoring Deforestation at Sub-Annual Scales as Extreme Events in Landsat Data Cubes , 2016, Remote. Sens..

[4]  Jan Verbesselt,et al.  Using spatial context to improve early detection of deforestation from Landsat time series , 2016 .

[5]  M. Herold,et al.  Near real-time disturbance detection using satellite image time series , 2012 .

[6]  Giles M. Foody,et al.  Good practices for estimating area and assessing accuracy of land change , 2014 .

[7]  M. Herold,et al.  Improving near-real time deforestation monitoring in tropical dry forests by combining dense Sentinel-1 time series with Landsat and ALOS-2 PALSAR-2 , 2018 .

[8]  Jan Verbesselt,et al.  Characterizing Forest Change Using Community-Based Monitoring Data and Landsat Time Series , 2016, PloS one.

[9]  Christopher Justice,et al.  Towards a Generalized Approach for Correction of the BRDF Effect in MODIS Directional Reflectances , 2009, IEEE Transactions on Geoscience and Remote Sensing.

[10]  Markus Reichstein,et al.  Detection and attribution of large spatiotemporal extreme events in Earth observation data , 2013, Ecol. Informatics.

[11]  M. Uddstrom,et al.  On the use of high‐resolution satellite data to describe the spatial and temporal variability of sea surface temperatures in the New Zealand region , 1999 .

[12]  Dirk Pflugmacher,et al.  Cross-border forest disturbance and the role of natural rubber in mainland Southeast Asia using annual Landsat time series , 2015 .

[13]  Belinda A. Margono,et al.  Humid tropical forest disturbance alerts using Landsat data , 2016 .

[14]  Martin Herold,et al.  Design and Implementation of an Interactive Web-Based Near Real-Time Forest Monitoring System , 2016, PloS one.

[15]  Chris A. Glasbey,et al.  Spatio-temporal variability of solar energy across a region: a statistical modelling approach , 2001 .

[16]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[17]  Jan Verbesselt,et al.  A Bayesian Approach to Combine Landsat and ALOS PALSAR Time Series for Near Real-Time Deforestation Detection , 2015, Remote. Sens..

[18]  Brendan Fisher,et al.  African exception to drivers of deforestation , 2010 .

[19]  Zhe Zhu,et al.  Object-based cloud and cloud shadow detection in Landsat imagery , 2012 .

[20]  Eric A. Lehmann,et al.  SAR and optical remote sensing: Assessment of complementarity and interoperability in the context of a large-scale operational forest monitoring system , 2015 .

[21]  Belen Franch,et al.  Preliminary analysis of the performance of the Landsat 8 / OLI land surface 1 reflectance product 2 3 , 2016 .

[22]  Stephen V. Stehman,et al.  Sampling designs for accuracy assessment of land cover , 2009 .

[23]  M. Herold,et al.  Robust monitoring of small-scale forest disturbances in a tropical montane forest using Landsat time series , 2015 .

[24]  C. Tucker.  Red and photographic infrared linear combinations for monitoring vegetation , 1979 .

[25]  M. Herold,et al.  Monitoring forest cover loss using multiple data streams, a case study of a tropical dry forest in Bolivia , 2015 .

[26]  Eric A. Lehmann,et al.  Joint Processing of Landsat and ALOS-PALSAR Data for Forest Mapping and Monitoring , 2012, IEEE Transactions on Geoscience and Remote Sensing.

[27]  Martin Herold,et al.  Mobile Devices for Community-Based REDD+ Monitoring: A Case Study for Central Vietnam , 2012, Sensors.

[28]  C. Woodcock,et al.  Continuous change detection and classification of land cover using all available Landsat data , 2014 .

[29]  C. Woodcock,et al.  Continuous monitoring of forest disturbance using all available Landsat imagery , 2012 .

[30]  Martin Herold,et al.  Fuelwood Savings and Carbon Emission Reductions by the Use of Improved Cooking Stoves in an Afromontane Forest, Ethiopia , 2014 .

[31]  Robert E. Wolfe,et al.  A Landsat surface reflectance dataset for North America, 1990-2000 , 2006, IEEE Geoscience and Remote Sensing Letters.

[32]  B. Holben,et al.  A spatio‐temporal approach for global validation and analysis of MODIS aerosol products , 2002 .

[33]  J. A. Schell,et al.  Monitoring vegetation systems in the great plains with ERTS , 1973 .

[34]  Michael Schultz,et al.  Performance of vegetation indices from Landsat time series in deforestation monitoring , 2016, Int. J. Appl. Earth Obs. Geoinformation.

[35]  M. Herold,et al.  Fusing Landsat and SAR time series to detect deforestation in the tropics , 2015 .

[36]  Achim Zeileis,et al.  BMC Bioinformatics BioMed Central Methodology article Conditional variable importance for random forests , 2008 .

[37]  José dos Campos,et al.  Space-time dynamics of deforestation in Brazilian Amazônia , 2002 .

[38]  Stephen V. Stehman,et al.  Impact of sample size allocation when using stratified random sampling to estimate accuracy and area of land-cover change , 2012 .

[39]  Joanne C. White,et al.  Generation of dense time series synthetic Landsat data through data blending with MODIS using a spatial and temporal adaptive reflectance fusion model. , 2009 .

[40]  Zhe Zhu,et al.  Toward near real-time monitoring of forest disturbance by fusion of MODIS and Landsat data , 2013 .

[41]  S. Goward,et al.  Dynamics of national forests assessed using the Landsat record: Case studies in eastern United States , 2009 .

[42]  Arun Kumar Pratihast,et al.  Combining Satellite Data and Community-Based Observations for Forest Monitoring , 2014 .