Scholarlink

<p>River deltas support human settlements, sustain ecosystems, and form hydrocarbon reservoirs in ancient sedimentary basins. Their morphologies are shaped by a complex interplay of environmental factors, posing challenges for predicting their evolution. Here we show that characteristics of delta shoreline and distributary channels are antagonistically influenced by rivers and waves. We analyzed morphologies of 1344 global deltas using a convolutional autoencoder, an unsupervised machine learning model, to encode their characteristics into 70-dimensional feature vectors serving as quantitative morphological metrics. The <i>X</i>-means clustering of these metrics revealed eight distinct delta morphotypes, adding a complementary perspective on their conventional classification schemes. We then performed multiple regression analyses to predict the relative sediment fluxes from rivers, tides, and waves from the morphological metrics. The results exhibited that shoreline protuberance and well-developed distributary channels are promoted by stronger fluvial influences and inhibited by stronger wave influences. In contrast, tidal influences were less clearly associated with these morphological features. This data-driven framework contributes to a better understanding of how specific delta morphologies respond to both natural variability and anthropogenic disturbances, offering additional avenues for sustainable management and future research.</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

Identifying controlling factors of delta morphology using a convolutional autoencoder

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月31日

查看全文
  • Ryusei Sato,
  • Hajime Naruse

摘要

River deltas support human settlements, sustain ecosystems, and form hydrocarbon reservoirs in ancient sedimentary basins. Their morphologies are shaped by a complex interplay of environmental factors, posing challenges for predicting their evolution. Here we show that characteristics of delta shoreline and distributary channels are antagonistically influenced by rivers and waves. We analyzed morphologies of 1344 global deltas using a convolutional autoencoder, an unsupervised machine learning model, to encode their characteristics into 70-dimensional feature vectors serving as quantitative morphological metrics. The X-means clustering of these metrics revealed eight distinct delta morphotypes, adding a complementary perspective on their conventional classification schemes. We then performed multiple regression analyses to predict the relative sediment fluxes from rivers, tides, and waves from the morphological metrics. The results exhibited that shoreline protuberance and well-developed distributary channels are promoted by stronger fluvial influences and inhibited by stronger wave influences. In contrast, tidal influences were less clearly associated with these morphological features. This data-driven framework contributes to a better understanding of how specific delta morphologies respond to both natural variability and anthropogenic disturbances, offering additional avenues for sustainable management and future research.