Scholarlink

<p>Identification of representative vegetation patches using Earth Observation Satellite (EOS) imagery is a timely problem for land cover mapping, training classifications, and planning field campaigns. Still, the available methods are either costly or tedious and time-consuming. Here we investigated if an EOS-derived pattern of low local spectral variation may define structurally homogeneous vegetation patches across successional stages, from almost bare ground to dense forests. We chose 28 post-mining sites in southern Poland, jointly covering 1,758.8&#xa0;ha, with multiple successional stages occurring at relatively small distances. The Tasseled Cap Disturbance Index (TCDI) was obtained through free Sentinel-2 imagery; then, a 5-by-5-pixel (0.25&#xa0;ha) moving-window method was applied to derive maps of local TCDI standard deviation (TCDIsd). We compared the binary classifications, based on a range of TCDIsd threshold values: between the 10th and the 70th percentiles of a manually balanced sample, with both field-based classifications and with height-explicit airborne LiDAR-based assessments. We found that the share of LiDAR-homogeneous plots was higher in EOS-homogeneous patches, by up to 27%. Increasingly restricted TCDIsd thresholds identified vegetation of increasing successional stage, LiDAR-derived height and vertical complexity metrics (<i>p</i> &lt; 0.001), and a decreasing vertical heterogeneity (<i>p</i> &lt; 0.001). Moreover, the TCDIsd-based maps revealed a parabolic trajectory of spoil heap structural development, indicating that the density of homogeneous patches first increases and then decreases, peaking between 16 and 29% cover, for decreasing homogeneity. This pattern suggests ongoing human-induced disturbances and a deficiency of the late seral ecosystem services. Overall, the presented methods may readily provide ecologically meaningful, spectrally guided auxiliary data, informative of the site-level structural development of vegetation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Spectrally Guided Delineation of Structurally Homogeneous Vegetation Patches Using Sentinel-2 Imagery in Heterogeneous Post-Industrial Landscapes

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月20日

查看全文
  • Kamil Kędra,
  • Andrzej M. Jagodziński

摘要

Identification of representative vegetation patches using Earth Observation Satellite (EOS) imagery is a timely problem for land cover mapping, training classifications, and planning field campaigns. Still, the available methods are either costly or tedious and time-consuming. Here we investigated if an EOS-derived pattern of low local spectral variation may define structurally homogeneous vegetation patches across successional stages, from almost bare ground to dense forests. We chose 28 post-mining sites in southern Poland, jointly covering 1,758.8 ha, with multiple successional stages occurring at relatively small distances. The Tasseled Cap Disturbance Index (TCDI) was obtained through free Sentinel-2 imagery; then, a 5-by-5-pixel (0.25 ha) moving-window method was applied to derive maps of local TCDI standard deviation (TCDIsd). We compared the binary classifications, based on a range of TCDIsd threshold values: between the 10th and the 70th percentiles of a manually balanced sample, with both field-based classifications and with height-explicit airborne LiDAR-based assessments. We found that the share of LiDAR-homogeneous plots was higher in EOS-homogeneous patches, by up to 27%. Increasingly restricted TCDIsd thresholds identified vegetation of increasing successional stage, LiDAR-derived height and vertical complexity metrics (p < 0.001), and a decreasing vertical heterogeneity (p < 0.001). Moreover, the TCDIsd-based maps revealed a parabolic trajectory of spoil heap structural development, indicating that the density of homogeneous patches first increases and then decreases, peaking between 16 and 29% cover, for decreasing homogeneity. This pattern suggests ongoing human-induced disturbances and a deficiency of the late seral ecosystem services. Overall, the presented methods may readily provide ecologically meaningful, spectrally guided auxiliary data, informative of the site-level structural development of vegetation.