Mapping dead trees in Rhode Island using semi-automated classification of summertime aerial imagery

Abstract

Mapping individual dead trees from aerial imagery is useful for assessing habitat quality, monitoring forest health, identifying risks to infrastructure, and guiding forest management strategies. Dead trees can be identified in summertime aerial imagery but computer-assisted methods are needed for efficient mapping across large areas. This study evaluated pixel- and object-based unsupervised classification and Deep Learning for mapping individual dead trees using summertime true-color aerial imagery. The study area was located in Rhode Island forests which had high rates of deciduous tree mortality due to a Spongy moth outbreak in 2015–2017. The unsupervised approaches included spectral and morphological filters to reduce commission error. The Deep Learning approach used a training/validation dataset with 22,504-point features representing dead trees. The pixel-based method had the best performance (F1 = 0.84), the object-based method had slightly lower performance (F1 = 0.79), and Deep Learning had the worst performance (F1 = 0.67). The pixel-based method had the added benefits of being automatable and not requiring training data whereas the object-based method could not be automated and Deep Learning required a large training dataset. The study showed that dead deciduous trees can be mapped from true-color aerial imagery and that an automated classification can yield high accuracy.

Keywords

Forest mortality / Dead trees / Aerial imagery / ISODATA / Unsupervised classification / Deep learning

References

[1] Abbas A, Minallh N, Ahmad N, Abid SAR, Khan MA. K-Means and ISODATA clustering algorithms for landcover classification using remote sensing. Sindh Univ Res J-SURJ (Sci Ser), 2016, 48: 315-318

[2] Bergmüller KO, Vanderwel MC. Predicting tree mortality using spectral indices derived from multispectral UAV imagery. Remote Sens, 2022, 14(9): 2195

[3] Briechle S, Krzystek P, Vosselman G. Silvi-net–a dual-CNN approach for combined classification of tree species and standing dead trees from remote sensing data. Int J Appl Earth Obs Geoinf, 2021, 98 102292

[4] Bucha T, Slávik M. Improved methods of classification of multispectral aerial photographs: evaluation of floodplain forests in the inundation area of the Danube. Folia for Pol Ser A for, 2013, 55(258-71

[5] Campbell JL, Harmon ME, Mitchell SR (2012) Can fuel-reduction treatments really increase forest carbon storage in the western US by reducing future fire emissions? Front Ecol Environ 10(2):83–90. https://doi.org/10.1890/110057

[6] Chiang CY, Barnes C, Angelov P, Jiang R. Deep learning-based automated forest health diagnosis from aerial images. IEEE Access, 2020, 8: 144064-144076

[7] Dumarevskaya L, Parent JR. Modeling spongy moth forest mortality in Rhode Island temperate deciduous forest. Forests, 2025, 16(1 93

[8] Enser R, Gregg D, Sparks C, August P, Jordan P, Coit J, Raithel C, Tefft B, Payton B, Brown C, LaBash C, Comings S, Ruddock K (2011) Rhode island ecological communities classification. Technical Report. Rhode Island Natural History Survey, Kingston, RI

[9] ESRI (2022) Segment mean shift. Retrieved from https://pro.arcgis.com/en/pro-app/tool-reference/spatial-analyst/segment-mean-shift.htm. Accessed July 2022

[10] Fröhlich A, Ciach M. Dead tree branches in urban forests and private gardens are key habitat components for woodpeckers in a city matrix. Landsc Urban Plan, 2020, 202 103869

[11] Goodwin NR, Coops NC, Wulder MA, Gillanders S, Schroeder TA, Nelson T. Estimation of insect infestation dynamics using a temporal sequence of Landsat data. Remote Sens Environ, 2008, 112(9): 3680-3689

[12] Guggenmoos S (2003) Effects of tree mortality on power line security. J Arboricult 29:181–196

[13] Guo QH, Kelly M, Gong P, Liu DS. An object-based classification approach in mapping tree mortality using high spatial resolution imagery. Gisci Remote Sens, 2007, 44(1): 24-47

[14] Hell M, Brandmeier M, Briechle S, Krzystek P. Classification of tree species and standing dead trees with lidar point clouds using two deep neural networks: PointCNN and 3dmFV-net. PFG, 2022, 90(2): 103-121

[15] Janisch JE, Harmon ME. Successional changes in live and dead wood carbon stores: implications for net ecosystem productivity. Tree Physiol, 2002, 22(2–3): 77-89

[16] Jiang XT, Wu ZY, Han SY, Yan H, Zhou B, Li JJ. A multi-scale approach to detecting standing dead trees in UAV RGB images based on improved faster R-CNN. PLoS ONE, 2023, 18(2 e0281084

[17] Kamińska A, Lisiewicz M, Stereńczak K, Kraszewski B, Sadkowski R (2018) Species-related single dead tree detection using multi-temporal ALS data and CIR imagery. Remote Sens Environ 219:31–43. https://doi.org/10.1016/j.rse.2018.10.005

[18] Lemenkova P. Evaluating land cover types from Landsat TM using SAGA GIS for vegetation mapping based on ISODATA and K-means clustering. Acta Agric Serb, 2021, 26(52): 159-165

[19] Lemenkova P. ISO cluster classifier by ArcGIS for unsupervised classification of the Landsat TM image of Reykjavík. Bull Nat Sci Res, 2021, 11(1): 29-37

[20] Li ZC, Yang RL, Cai WW, Xue YF, Hu YW, Li LJ. LLAM-MDCNet for detecting remote sensing images of dead tree clusters. Remote Sens, 2022, 14(15): 3684

[21] Long JA, Lawrence RL. Mapping percent tree mortality due to mountain pine beetle damage. For Sci, 2016, 62(4392-402

[22] Marcot BG, Ohmann JL, Mellen-McLean K, Waddell KL. Synthesis of regional wildlife and vegetation field studies to guide management of standing and down dead trees. For Sci, 2010, 56(4): 391-404

[23] Meddens AJH, Hicke JA. Spatial and temporal patterns of Landsat-based detection of tree mortality caused by a mountain pine beetle outbreak in Colorado, USA. For Ecol Manag, 2014, 322: 78-88

[24] Meng R, Gao RJ, Zhao F, Huang CQ, Sun R, Lv ZG, Huang ZH. Landsat-based monitoring of southern pine beetle infestation severity and severity change in a temperate mixed forest. Remote Sens Environ, 2022, 269 112847

[25] NCDC (1971) National climatic data center. Climates of the States, Volume 1. Distributed by the U.S. Govt. Print. Off., Washington. (Climatography of the United States). Retrieved from https://books.google.com/books?id=pfHFwgEACAAJ

[26] Nielsen MM, Heurich M, Malmberg B, Brun A. Automatic mapping of standing dead trees after an insect outbreak using the window independent context segmentation method. J for, 2014, 112(6): 564-571

[27] NOAA (2021) National Oceanographic and Atmospheric Administration. Coastal Change Analysis Program (C-CAP). High-Resolution Land Cover dataset for Rhode Island. Retrieved from https://coast.noaa.gov/digitalcoast/data/ on July 2022.

[28] Pasher J, King DJ (2009) Mapping dead wood distribution in a temperate hardwood forest using high resolution airborne imagery. Forest Ecol Manag 258:1536–1548. https://doi.org/10.1016/j.foreco.2009.07.009

[29] Poulos HM (2010) Spatially explicit mapping of hurricane risk in New England, USA using ArcGIS. Nat Hazard 54(3):1015–1023. https://doi.org/10.1007/s11069-010-9502-0

[30] Radu S (2006) The ecological role of deadwood in natural Forests. In Gafta, D., & Akeroyd, J. (Eds.), Nature Conservation. Environmental Science and Engineering. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-47229-2_16.

[31] Sani-Mohammed A, Yao W, Heurich M. Instance segmentation of standing dead trees in dense forest from aerial imagery using deep learning. ISPRS Open J Photogramm Remote Sens, 2022, 6 100024

[32] Tao H, Li CJ, Zhao D, Deng SQ, Hu HT, Xu XL, Jing WB. Deep learning-based dead pine tree detection from unmanned aerial vehicle images. Int J Remote Sens, 2020, 41(21): 8238-8255

[33] Wang XW, Zhao QZ, Jiang P, Zheng YC, Yuan L, Yuan PL. LDS-YOLO: a lightweight small object detection method for dead trees from shelter forest. Comput Electron Agric, 2022, 198 107035

[34] Weinstein BG, Marconi S, Bohlman S, Zare A, White E. Individual tree-crown detection in RGB imagery using semi-supervised deep learning neural networks. Remote Sens, 2019, 11(11): 1309

[35] Westveld M. Natural forest vegetation zones of New England. J for, 1956, 54(5332-338

[36] Zhan ZY, Yu LF, Li Z, Ren LL, Gao BT, Wang LX, Luo YQ. Combining GF-2 and sentinel-2 images to detect tree mortality caused by red turpentine beetle during the early outbreak stage in North China. Forests, 2020, 11(2 172

RIGHTS & PERMISSIONS

The Author(s)