Individual tree detection and classification from RGB satellite imagery with applications to wildfire fuel mapping and exposure assessments

L. Bennett; Z. Yu; R. Wasowski; S. Selland; S. Otway; J. Boisvert

doi:10.1071/WF24008

RESEARCH ARTICLE (Open Access)

Next Contents Vol 33(8)

Individual tree detection and classification from RGB satellite imagery with applications to wildfire fuel mapping and exposure assessments

L. Bennett

^A , Z. Yu ^A , R. Wasowski ^A , S. Selland ^A , S. Otway ^B and J. Boisvert ^A ^*

+ Author Affiliations

- Author Affiliations

^A University of Alberta, Edmonton, Alberta, Canada.

^B BarSO Enterprises Ltd., Wildwood, Alberta, Canada.

^* Correspondence to: jbb@ualberta.ca

International Journal of Wildland Fire 33, WF24008 https://doi.org/10.1071/WF24008

Submitted: 9 May 2023 Accepted: 2 July 2024 Published: 19 July 2024

© 2024 The Author(s) (or their employer(s)). Published by CSIRO Publishing on behalf of IAWF. This is an open access article distributed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC)

Abstract

Background

Wildfire fuels are commonly mapped via manual interpretation of aerial photos. Alternatively, RGB satellite imagery offers data across large spatial extents. A method of individual tree detection and classification is developed with implications to fuel mapping and community wildfire exposure assessments.

Methods

Convolutional neural networks are trained using a novel generational training process to detect trees in 0.50 m/px RGB imagery collected in Rocky Mountain and Boreal natural regions in Alberta, Canada by Pleiades-1 and WorldView-2 satellites. The workflow classifies detected trees as ‘green-in-winter’/‘brown-in-winter’, a proxy for coniferous/deciduous, respectively.

Key results

A k-fold testing procedure compares algorithm detections to manual tree identification densities reaching an R² of 0.82. The generational training process increased achieved R² by 0.23. To assess classification accuracy, satellite detections are compared to manual annotations of 2 cm/px drone imagery resulting in average F1 scores of 0.85 and 0.82 for coniferous and deciduous trees respectively. The use of model outputs in tree density mapping and community-scale wildfire exposure assessments is demonstrated.

Conclusion & Implications

The proposed workflow automates fine-scale overstorey tree mapping anywhere seasonal (winter and summer) 0.50 m/px RGB satellite imagery exists. Further development could enable the extraction of additional properties to inform a more complete fuel map.

Keywords: communities, community protection, machine learning, planning, remote sensing, satellite imagery, trees, wildfire, wildfire fuels, wildfire preplanning, wildland-urban interface.

References

Abozeid A, Alanazi R, Elhadad A, Taloba AI, Abd El-Aziz RM (2022) A large-scale dataset and deep learning model for detecting and counting olive trees in satellite imagery. Computational Intelligence and Neuroscience 2022, 1549842.
| Crossref | Google Scholar |

Agisoft (2023) Agisoft Metashape Professional. Available at https://www.agisoft.com/features/professional-edition/

Ahmed MR, Hassan QK (2023) Occurrence, area burned, and seasonality trends of forest fires in the Natural Subregions of Alberta over 1959–2021. Fire 6(3), 96.
| Crossref | Google Scholar |

Airbus Defense and Space Intelligence (2012) Pléiades. Airbus. (Taufkirchen: Germany). Available at https://intelligence.airbus.com/imagery/our-optical-and-radar-satellite-imagery/pleiades/

Alberta Sustainable Resource Development (2007) ‘R11 Forest Management Plan.’ (Rocky Mountain House: Alberta, Canada)

Alberta Wildfire (2020) ‘Wildland Urban Interface Triage Assessment Form.’ (Edmonton, AB, Canada) Available at https://open.alberta.ca/dataset/58d44774-be40-411b-96a9-0f4b175c902a/resource/53f6a266-739a-461e-b61a-3acf673f7fc8/download/ma-alberta-wildland-interface-fire-structure-protection-guidelines-2020.pdf

Badreldin N, Prieto B, Fisher R (2021) Mapping grasslands in mixed grassland ecoregion of Saskatchewan using big remote sensing data and machine learning. Remote Sensing 13, 4972.
| Crossref | Google Scholar |

Beaudoin A, Bernier PY, Guindon L, Villemaire P, Guo XJ, Stinson G, Bergeron T, Magnussen S, Hall RJ (2014) Mapping attributes of Canada’s forests at moderate resolution through kNN and MODIS imagery. Canadian Journal of Forest Research 44, 521-532.
| Crossref | Google Scholar |

Bennett L, Wilson B, Selland S, Qian L, Wood M, Zhao H, Boisvert J (2022) Image to attribute model for trees (ITAM-T): individual tree detection and classification in Alberta boreal forest for wildland fire fuel characterization. International Journal of Remote Sensing 43, 1848-1880.
| Crossref | Google Scholar |

Beverly JL, Bothwell P, Conner JCR, Herd EPK (2010) Assessing the exposure of the built environment to potential ignition sources generated from vegetative fuel. International Journal of Wildland Fire 19, 299-313.
| Crossref | Google Scholar |

Beverly JL, Braid G, Chapman L, Kelm S, Pozniak W, Stewart L, Johnston K (2018) ‘FireSmart Wildfire Exposure Assessment.’ (Government of Alberta: Edmonton, AB)

Beverly JL, McLoughlin N, Chapman E (2021) A simple metric of landscape fire exposure. Landscape Ecology 36, 785-801.
| Crossref | Google Scholar |

Brandt JP, Flannigan MD, Maynard DG, Thompson ID, Volney WJA (2013) An introduction to Canada’s boreal zone: ecosystem processes, health, sustainability, and environmental issues. Environmental Reviews 21, 207-226.
| Crossref | Google Scholar |

Canadian Forest Service Fire Danger Group (2021) An overview of the next generation of the Canadian Forest Fire Danger Rating System. Information Report GLC-X-26. Available at https://ostrnrcan-dostrncan.canada.ca/handle/1845/245411

Countryman CM (1972) ‘The fire environment concept.’ (USDA Forest Service, Pacific Southwest Forest and Range Experiment Station: Berkeley, CA)

Deur M, Gašparović M, Balenović I (2021) An evaluation of pixel- and object-based tree species classification in mixed deciduous forests using pansharpened very high spatial resolution satellite imagery. Remote Sensing 13, 1868.
| Crossref | Google Scholar |

DigitalGlobe (2009) WorldView-2. Available at https://www.spaceimagingme.com/downloads/sensors/datasheets/WorldView2-DS-WV2-Web.pdf

Forest Stewardship and Trade Branch (2022) ‘Alberta Vegetation Inventory Standards, Version 2.1.5.’ (Alberta Agriculture, Forestry and Rural Economic Development)

Forestry Canada Fire Danger Group (1992) ‘Development and Structure of the Canadian Forest Fire Behaviour Prediction System.’ (Forestry Canada: Ottawa, ON)

Frederick K (2012) ‘Revised Process to Convert Alberta Vegetation Inventory (AVI) Data to Canadian Forest Fire Behaviour Prediction System (FBP) Fuel Types.’ (Alberta Sustainable Resource Development Forest Protection Branch)

Freudenberg M, Magdon P, Nölke N (2022) Individual tree crown delineation in high-resolution remote sensing images based on U-Net. Neural Computing and Applications 34, 22197-22207.
| Crossref | Google Scholar |

Gomes MF, Maillard P, Deng H (2018) Individual tree crown detection in sub-meter satellite imagery using Marked Point Processes and a geometrical-optical model. Remote Sensing of Environment 211, 184-195.
| Crossref | Google Scholar |

Government of Alberta (2013) ‘FireSmart Guidebook for Community Protection: A Guidebook for Wildland/Urban Interface Communities.’ (Edmonton, AB). Available at https://wildfire.alberta.ca/firesmart/documents/FireSmart-GuideCommunityProtection-Nov2013.pdf

Hirsch KG (1996) ‘Canadian forest fire behavior prediction (FBP) system: user’s guide.’ (Natural Resources Canada, Canadian Forest Service, Northwest Region, Northern Forestry Centre: Edmonton, AB)

Hoff DL, Will RE, Zou CB, Weir JR, Gregory MS, Lillie ND (2018) Estimating increased fuel loading within the Cross Timbers forest matrix of Oklahoma, USA due to an encroaching conifer, Juniperus virginiana, using leaf-off satellite imagery. Forest Ecology and Management 409, 215-224.
| Crossref | Google Scholar |

Jocher G, Chaurasia A, Qiu J (2023) YOLO by Ultralytics. Available at https://github.com/ultralytics/ultralytics

Johnston LM, Flannigan MD (2018) Mapping Canadian wildland fire interface areas. International Journal of Wildland Fire 27, 1-14.
| Crossref | Google Scholar |

Kattenborn T, Leitloff J, Schiefer F, Hinz S (2021) Review on Convolutional Neural Networks (CNN) in vegetation remote sensing. ISPRS Journal of Photogrammetry and Remote Sensing 173, 24-49.
| Crossref | Google Scholar |

Linn RR, Goodrick SL, Brambilla S, Brown MJ, Middleton RS, O’Brien JJ, Hiers JK (2020) QUIC-fire: a fast-running simulation tool for prescribed fire planning. Environmental Modelling & Software 125, 104616.
| Crossref | Google Scholar |

Mubin NA, Nadarajoo E, Shafri HZM, Hamedianfar A (2019) Young and mature oil palm tree detection and counting using convolutional neural network deep learning method. International Journal of Remote Sensing 40, 7500-7515.
| Crossref | Google Scholar |

Natesan S, Armenakis C, Vepakomma U (2020) Individual tree species identification using Dense Convolutional Network (DenseNet) on multitemporal RGB images from UAV. Journal of Unmanned Vehicle Systems 8, 310-333.
| Crossref | Google Scholar |

Natural Resources Canada (2019) ‘Canadian Forest FBP Fuel Types (CanFG).’ (Canadian Forestry Service, Northern Forestry Centre: Edmonton, AB) Available at https://cwfis.cfs.nrcan.gc.ca/downloads/fuels/development/

Natural Regions Committee (2006) Natural Regions and Subregions of Alberta. Compiled by D. J. Downing and W. W. Pettapiece. Government of Alberta. Pub. No. T/852.

Partners in Protection (2003) ‘FireSmart: Protecting Your Community from Wildfire.’ (Natural Resources Canada: Edmonton AB)

Redmon J, Divvala S, Girshick R, Farhadi A (2015) You Only Look Once: Unified, Real-Time Object Detection. 10.48550/arXiv.1506.02640

Sanchez-Azofeifa GA, Chong M, Sinkwich J, Mamet S (2004) Alberta Ground Cover Characterization (AGCC) Training and Procedures Manual. Earth Observation Systems Laboratory, Department of Earth and Atmospheric Sciences, University of Alberta.

Sapkota BB, Liang L (2018) A multistep approach to classify full canopy and leafless trees in bottomland hardwoods using very high-resolution imagery. Journal of Sustainable Forestry 37, 339-356.
| Crossref | Google Scholar |

Stocks BJ, Lynham TJ, Lawson BD, Alexander ME, Wagner CE Van, McAlpine RS, Dubé DE (1989) The Canadian Forest Fire Danger Rating System: an overview. The Forestry Chronicle 65, 450-457.
| Crossref | Google Scholar |

Surový P, Kuželka K (2019) Acquisition of forest attributes for decision support at the forest enterprise level using remote-sensing techniques—A review. Forests 10, 273.
| Crossref | Google Scholar |

Terven J, Cordova-Esparza D (2023) A Comprehensive Review of YOLO: From YOLOv1 and Beyond. Available at http://arxiv.org/abs/2304.00501.

Tymstra C, Bryce RW, Wotton BM, Taylor SW, Armitage OB (2010) Development and structure of Prometheus: the Canadian wildland fire growth simulation model. Information Report NOR-X-417. (Natural Resources Canada, Canadian Forest Service, Northern Forestry Centre: Edmonton, AB)

Van Wagner CE (1977) Conditions for the start and spread of crown fire. Canadian Journal of Forest Research 7, 23-34.
| Crossref | Google Scholar |

Van Wagner CE (1983). Fire behaviour in northern conifer forests and shrublands. In ‘Role Fire North. Circumpolar Ecosyst’. (Eds R Wein, D MacLean) pp. 65–80. (John Wiley and Sons: New York, NY, USA)

Varin M, Chalghaf B, Joanisse G (2020) Object-based approach using very high spatial resolution 16-Band WorldView-3 and LiDAR data for tree species classification in a broadleaf forest in Quebec, Canada. Remote Sensing 12, 3092.
| Crossref | Google Scholar |

Vermote EF, Skakun S, Becker-Reshef I, Saito K (2020) Remote sensing of coconut trees in Tonga using very high spatial resolution WorldView-3 data. Remote Sensing 12, 3113.
| Crossref | Google Scholar |

Westhaver A (2017) ‘Why some homes survived: Learning from the Fort McMurray wildland/urban interface fire disaster.’ (Institute for Catastrophic Loss Reduction: Toronto, Canada)

Yao L, Liu T, Qin J, Lu N, Zhou C (2021) Tree counting with high spatial-resolution satellite imagery based on deep neural networks. Ecological Indicators 125, 107591.
| Crossref | Google Scholar |