Revealing forest disturbances and post-disturbance recovery dynamics in Ethiopia using fractional vegetation cover time series

Abstract

Accurate monitoring of forest disturbance and recovery dynamics is essential for understanding ecosystem resilience. However, reliable monitoring remains challenging in heterogeneous tropical landscapes using satellite observations. In this study, we developed a deep learning–based spectral unmixing approach to estimate annual fractional vegetation cover (FVC) of trees, shrubs, herbaceous species, and bare ground across Ethiopia using dense Landsat and Sentinel-2A/B time series. The proposed attention-based bidirectional long short-term memory (Bi-LSTM) model effectively captured spectral–temporal patterns and achieved robust performance, with mean absolute errors (MAEs) below 14% across all vegetation components. Model accuracy varied distinctly among different land cover types as tree cover had the highest accuracy while bare ground showed lower performance. Based on the derived FVC, we constructed a Normalized Difference Fraction Index (NDFI) to characterize the relative dominance of vegetation cover components and applied time-series trend analysis to detect disturbance events and recovery trajectories. A major disturbance event occurred in 2003, affecting over 564 km² of forest and reducing tree cover by 34.2%. In the Harenna moist evergreen montane forest, the affected stands required an average of 8.4 ± 0.6 years to recover. The post-disturbance bare ground fraction was negatively associated with recovery probability, indicating its potential as an early-warning indicator of delayed forest recovery and regeneration. Our findings demonstrate that integrating deep learning–based spectral unmixing with dense satellite time series enables effective monitoring of forest dynamics. This framework provides a scalable solution for assessing disturbance and recovery processes and supports more timely and informed forest management in data-limited regions.

Keywords

Fractional vegetation cover dynamics / Spectral unmixing / Bi-LSTM / Forest disturbance recovery / Ethiopia

References

[1] Algarni M, Saeed F, Al-Hadhrami T, Ghabban F, Al-Sarem M. Deep learning-based approach for emotion recognition using electroencephalography (EEG) signals using bi-directional long short-term memory (Bi-LSTM). Sensors (Basel), 2022, 22(82976

[2] Anderson-Teixeira KJ, Miller AD, Mohan JE, Hudiburg TW, Duval BD, DeLucia EH. Altered dynamics of forest recovery under a changing climate. Glob Change Biol, 2013, 19(7): 2001-2021

[3] Asefa M, Cao M, He YY, Mekonnen E, Song XY, Yang J. Ethiopian vegetation types, climate and topography. Plant Divers, 2020, 42(4): 302-311

[4] Bueno IT, McDermid GJ, Silveira EMO, Hird JN, Domingos BI, Acerbi Júnior FW. Spatial agreement among vegetation disturbance maps in tropical domains using Landsat time series. Remote Sens, 2020, 12(18): 2948

[5] Buisson E, Le Stradic S, Silveira FAO, Durigan G, Overbeck GE, Fidelis A, Fernandes GW, Bond WJ, Hermann JM, Mahy G, Alvarado ST, Zaloumis NP, Veldman JW. Resilience and restoration of tropical and subtropical grasslands, savannas, and grassy woodlands. Biol Rev, 2019, 94(2): 590-609

[6] Cao S, Feng JL, Hu ZW, Li QQ, Wu GF. Improving estimation of urban land cover fractions with rigorous spatial endmember modeling. ISPRS J Photogramm Remote Sens, 2022, 189: 36-49

[7] Cardille JA, Perez E, Crowley MA, Wulder MA, White JC, Hermosilla T. Multi-sensor change detection for within-year capture and labelling of forest disturbance. Remote Sens Environ, 2022, 268: 112741

[8] Cole LES, Bhagwat SA, Willis KJ. Recovery and resilience of tropical forests after disturbance. Nat Commun, 2014, 5: 3906

[9] Crisóstomo de Castro Filho H, Abílio de Carvalho Júnior O, Ferreira de Carvalho OL, Pozzobon de Bem P, dos Santos de Moura R, Olino de Albuquerque A, Rosa Silva C, Guimarães Ferreira PH, Fontes Guimarães R, Trancoso Gomes RA. Rice crop detection using LSTM, Bi-LSTM, and machine learning models from sentinel-1 time series. Remote Sens, 2020, 12(16): 2655

[10] Cuevas-González M, Gerard F, Balzter H, Riaño D. Analysing forest recovery after wildfire disturbance in boreal Siberia using remotely sensed vegetation indices. Glob Change Biol, 2009, 15(3): 561-577

[11] De Keersmaecker W, Rodríguez-Sánchez P, Milencović M, Herold M, Reiche J, Verbesselt J. Evaluating recovery metrics derived from optical time series over tropical forest ecosystems. Remote Sens Environ, 2022, 274: 112991

[12] FAO (2012) Forest resources assessment working paper 180—terms and definitions

[13] Flores BM, Staal A, Jakovac CC, Hirota M, Holmgren M, Oliveira RS. Soil erosion as a resilience drain in disturbed tropical forests. Plant Soil, 2020, 450(1): 11-25

[14] Hansen MC, Potapov PV, Moore R, Hancher M, Turubanova SA, Tyukavina A, Thau D, Stehman SV, Goetz SJ, Loveland TR, Kommareddy A, Egorov A, Chini L, Justice CO, Townshend JRG. High-resolution global maps of 21st-century forest cover change. Science, 2013, 342(6160): 850-853

[15] Harkort L, Okujeni A, Amputu V, Mahler J, Nill L, Pflugmacher D, Röder A, Hostert P. Mapping fractional vegetation cover in Sub-Saharan rangelands using phenological feature spaces. Remote Sens Environ, 2025, 319: 114646

[16] Heim RJ, Rocha AV, Zemlianskii V, Barrett K, Bültmann H, Breen A, Frost GV, Hollingsworth TN, Jandt R, Kozlova M, Kurka A, Jorgenson MT, Landhäusser SM, Loranty MM, Miller EA, Narita K, Pravdolyubova E, Hölzel N, Schaepman-Strub G. Arctic tundra ecosystems under fire—alternative ecosystem states in a changing climate?. J Ecol, 2025, 113(5): 1042-1056

[17] Hishe H, Oosterlynck L, Giday K, De Keersmaecker W, Somers B, Muys B. A combination of climate, tree diversity and local human disturbance determine the stability of dry Afromontane forests. For Ecosyst, 2021, 8: 16

[18] Kowalski K, Okujeni A, Brell M, Hostert P. Quantifying drought effects in Central European grasslands through regression-based unmixing of intra-annual Sentinel-2 time series. Remote Sens Environ, 2022, 268: 112781

[19] LeCun Y, Bengio Y, Hinton G. Deep learning. Nature, 2015, 521(7553): 436-444

[20] Li KW, Xu EQ. Differentiating effects of salvage logging and recovery patterns on post-fire boreal forests in Northeast China using a modified forest disturbance index. Gisci Remote Sens, 2023, 60: 2188674

[21] Madebo AA (2017) Vegetation ecology and land use/land cover changes in selected Afromontane forests along Gibe–Omo watershed, Southwest Ethiopia. In: Addis Ababa University

[22] Mandl L, Viana-Soto A, Seidl R, Stritih A, Senf C. Unmixing-based forest recovery indicators for predicting long-term recovery success. Remote Sens Environ, 2024, 308: 114194

[23] Meli P, Ellison D, de Barros Ferraz SF, Filoso S, Brancalion PHS. On the unique value of forests for water: hydrologic impacts of forest disturbances, conversion, and restoration. Glob Change Biol, 2024, 30(2): e17162

[24] Morresi D, Vitali A, Urbinati C, Garbarino M. Forest spectral recovery and regeneration dynamics in stand-replacing wildfires of central Apennines derived from Landsat time series. Remote Sens, 2019, 11(3): 308

[25] Nill L, Grünberg I, Ullmann T, Gessner M, Boike J, Hostert P. Arctic shrub expansion revealed by Landsat-derived multitemporal vegetation cover fractions in the Western Canadian Arctic. Remote Sens Environ, 2022, 281: 113228

[26] Okujeni A, Kowalski K, Lewińska KE, Schneidereit S, Hostert P. Multidecadal grassland fractional cover time series retrieval for Germany from the Landsat and Sentinel-2 archives. Remote Sens Environ, 2024, 302: 113980

[27] Rocchini D, Salvatori N, Beierkuhnlein C, Chiarucci A, de Boissieu F, Förster M, Garzon-Lopez CX, Gillespie TW, Hauffe HC, He KS, Kleinschmit B, Lenoir J, Malavasi M, Moudrý V, Nagendra H, Payne D, Šímová P, Torresani M, Wegmann M, Féret JB. From local spectral species to global spectral communities: a benchmark for ecosystem diversity estimate by remote sensing. Ecol Inform, 2021, 61: 101195

[28] Rybar J, Bosela M, Marcis P, Ujházyová M, Polťák D, Hederová L, Ujházy K. Effects of tree canopy on herbaceous understorey throughout the developmental cycle of a temperate mountain primary forest. For Ecol Manag, 2023, 546: 121353

[29] Seidl R, Turner MG. Post-disturbance reorganization of forest ecosystems in a changing world. Proc Natl Acad Sci U S A, 2022, 119(28): e2202190119

[30] Senf C, Laštovička J, Okujeni A, Heurich M, van der Linden S. A generalized regression-based unmixing model for mapping forest cover fractions throughout three decades of Landsat data. Remote Sens Environ, 2020, 240: 111691

[31] Senf C, Seidl R. Post-disturbance canopy recovery and the resilience of Europe’s forests. Glob Ecol Biogeogr, 2022, 31(1): 25-36

[32] Somers B, Asner GP, Tits L, Coppin P. Endmember variability in spectral mixture analysis: a review. Remote Sens Environ, 2011, 115(7): 1603-1616

[33] Su XP, Wang MH, Huang ZQ, Fu SL, Chen HYH. Forest understorey vegetation: colonization and the availability and heterogeneity of resources. Forests, 2019, 10(11): 944

[34] Tang XJ, Bratley KH, Cho K, Bullock EL, Olofsson P, Woodcock CE. Near real-time monitoring of tropical forest disturbance by fusion of Landsat, Sentinel-2, and Sentinel-1 data. Remote Sens Environ, 2023, 294: 113626

[35] Tian HQ, Lu CQ, Ciais P, Michalak AM, Canadell JG, Saikawa E, Huntzinger DN, Gurney KR, Sitch S, Zhang BW, Yang J, Bousquet P, Bruhwiler L, Chen GS, Dlugokencky E, Friedlingstein P, Melillo J, Pan SF, Poulter B, Prinn R, Saunois M, Schwalm CR, Wofsy SC. The terrestrial biosphere as a net source of greenhouse gases to the atmosphere. Nature, 2016, 531(7593): 225-228

[36] Viana-Soto A, Okujeni A, Pflugmacher D, García M, Aguado I, Hostert P. Quantifying post-fire shifts in woody-vegetation cover composition in Mediterranean pine forests using Landsat time series and regression-based unmixing. Remote Sens Environ, 2022, 281: 113239

[37] von Oppen J, Assmann JJ, Bjorkman AD, Treier UA, Elberling B, Nabe-Nielsen J, Normand S. Cross-scale regulation of seasonal microclimate by vegetation and snow in the Arctic tundra. Glob Change Biol, 2022, 28(24): 7296-7312

[38] Wakshum S, Sebsebe D, Tamrat B. Ecology of soil seed banks: implications for conservation and restoration of natural vegetation: a review. Int J Biodivers Conserv, 2018, 10(10): 380-393

[39] Wen DW, Huang X, Bovolo F, Li JY, Ke XL, Zhang AL, Benediktsson JA. Change detection from very-high-spatial-resolution optical remote sensing images: methods, applications, and future directions. IEEE Geosci Remote Sens Mag, 2021, 9(4): 68-101

[40] White JC, Wulder MA, Hermosilla T, Coops NC, Hobart GW. A nationwide annual characterization of 25 years of forest disturbance and recovery for Canada using Landsat time series. Remote Sens Environ, 2017, 194: 303-321

[41] Wu QH, Lee CKF, Wang JA, Zhao YY, Song GQ, Maeda EE, Su YJ, Huete A, Hughes AC, Wu J. Improved assessment of post-fire recovery trajectory of forests in Amazon’s protected areas. Remote Sens Environ, 2025, 326: 114802

[42] Yang QC, Wang LH, Foody G, Huang JL, Yan X, Du Y, Ling F. Untangling the spatial distribution of Alpine vegetation fractional covers from Sentinel-2 time series images by deep learning regression-based unmixing. Int J Remote Sens, 2025, 46(18): 6915-6936

[43] Yang QC, Wang LH, Li XQ, Yan X, Huang JL, Du Y, Ling F. Multi-decadal spatiotemporal dynamics of Alpine plant functional types (PFTs) inferred from Landsat-derived fractional cover across the Yarlung Zangbo river basin, China. Big Earth Data, 2025, 9(4): 913-937

[44] Zegeye H, Teketay D, Kelbessa E. Diversity and regeneration status of woody species in Tara Gedam and Abebaye forests, northwestern Ethiopia. J for Res, 2011, 22(3): 315-328

[45] Zeng YL, Hao DL, Huete A, Dechant B, Berry J, Chen JM, Joiner J, Frankenberg C, Bond-Lamberty B, Ryu Y, Xiao JF, Asrar GR, Chen M. Optical vegetation indices for monitoring terrestrial ecosystems globally. Nat Rev Earth Environ, 2022, 3(7): 477-493

[46] Zhang X, Liu LY, Chen XD, Gao Y, Xie S, Mi J. GLC_FCS30: global land-cover product with fine classification system at 30 m using time-series Landsat imagery. Earth Syst Sci Data, 2021, 13(6): 2753-2776

[47] Zhang YH, Ling F, Wang X, Foody GM, Boyd DS, Li XD, Du Y, Atkinson PM. Tracking small-scale tropical forest disturbances: fusing the Landsat and Sentinel-2 data record. Remote Sens Environ, 2021, 261: 112470

[48] Zhang YH, Wang X, Li XD, Du Y, Atkinson PM. Monitoring monthly tropical humid forest disturbances with Planet NICFI images in Cameroon. Agric for Meteorol, 2023, 341: 109676

[49] Zhang YH, Wang X, Li XD, Zhao WQ, Zhong XY, Yu BJ, Du Y, Atkinson PM. Characterizing tropical evergreen forest disturbances and post-disturbance recovery using time-series landsat canopy openings. IEEE Trans Geosci Remote Sens, 2025, 63: 4420823

[50] Zhou YN, Wang SY, Wu TJ, Feng L, Wu W, Luo JC, Zhang X, Yan NN. For-backward LSTM-based missing data reconstruction for time-series Landsat images. Gisci Remote Sens, 2022, 59(1): 410-430

[51] Zhu Z, Wang SX, Woodcock CE. Improvement and expansion of the Fmask algorithm: cloud, cloud shadow, and snow detection for Landsats 4–7, 8, and Sentinel 2 images. Remote Sens Environ, 2015, 159: 269-277

Funding

Youth Project of Natural Science Foundation of Hubei Province, China(2025AFB393)

National Key Research and Development Program of China(2024YFE0214500)

CAS-ANSO Sustainable Development Research Project(CAS-ANSO-SDRP-2024-04)

Sino-Africa Joint Research Center, CAS, China(SAJC202527ZD02)

International Science, Technology Cooperation Project of Hubei Province, China(2024EHA035)

Rights & permissions

Northeast Forestry University