跳到正文
arXiv cs.AI· Coen van den Elsen, Laurens Geffert·· 56 分钟前AI 评分36

DARWIN:面向海洋智能的地理空间基础模型

DARWIN: a Geospatial Foundation Model for Marine Intelligence

AI 导读

研究者推出面向海洋智能的地理空间基础模型 DARWIN,通过缺失感知编码、扩展时空掩码和傅里叶式位置嵌入适配海洋数据特性,并基于 26 年 Copernicus 全球再分析数据与 GEBCO 水深数据预训练。在叶绿素-a 预测任务上,其 2022 年每月相对误差均为最低,仅用 10% 监督数据即超过用全部数据训练的随机初始化编码器,可支撑异常检测、物种分布建模和环境监测等下游应用。

正文 · 原文

View PDF HTML (experimental)

Abstract:We present DARWIN, a geospatial foundation model for marine intelligence. Recent Earth-observation foundation models have concentrated on terrestrial domains. The marine realm poses distinct challenges: missingness, multi-modal physical and biogeochemical variables, sparse labels, and dynamics that are transient, multi-scale, and non stationary over time. DARWIN adapts a masked-autoencoder backbone to these properties through missingness-aware encoding, expanded spatial and temporal masking, and Fourier-style positional embeddings, and is pretrained on 26 years of global Copernicus reanalysis together with GEBCO bathymetry.
On predicting chlorophyll-a, the pretrained backbone attains the lowest monthly relative error in every month of 2022, outperforming both no-encoder baselines and an identically architected random-init encoder. At 10% of the supervised training data, DARWIN already surpasses the random-init model trained on 100% of the data. Embedding analysis and qualitative reconstructions show that DARWIN captures coherent oceanographic structure rather than memorising input statistics, providing a scalable backbone for downstream marine applications such as anomaly detection, species distribution modelling, and environmental monitoring.
Subjects: Artificial Intelligence (cs.AI); Atmospheric and Oceanic Physics (physics.ao-ph)
Cite as: arXiv:2611.22496 [cs.AI]
  (or arXiv:2611.22496v1 [cs.AI] for this version)
  https://doi.org/10.48550/arXiv.2611.22496

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Jan Laurens Geffert [view email]
[v1] Tue, 17 Nov 2026 09:51:26 UTC (28,040 KB)

来源:arXiv cs.AI · arxiv.org