大气与海洋物理
查看 最近的 文章
显示 2025年09月29日， 星期一 新的列表
- [1] arXiv:2509.21329 [中文pdf, pdf, html, 其他]
-
标题： 挪威海岸的海洋水动力分析降尺度——NORA-SARAH开放框架标题： Down-scale marine hydrodynamic analysis at the Norwegian coast -- the NORA-SARAH open framework主题： 大气与海洋物理 (physics.ao-ph) ; 流体动力学 (physics.flu-dyn)
在近海波浪研究中，通常假设高斯过程和均匀的波浪场。 然而，当波浪接近海岸线时，复杂的海岸地形和水深变化会引起折射、绕射、反射和破碎等变化，导致非线性增强和特定地点的波浪特征。 这种复杂性需要详细的特定地点研究，以用于海岸基础设施设计和蓝色经济规划。 本研究提出了一种降尺度程序，用于分析从远海气象海洋条件到结构物的波浪-结构相互作用。 开源的NORA3和NORA10EI历史数据库定义了远海海况，这些数据通过相位平均波浪模型SWAN传播到近岸区域。 输出结果用于相位解析模拟，使用完全非线性势流求解器REEF3D::FNPF，结合任意欧拉-拉格朗日（ALE）方法，通过莫里森公式计算波浪力并筛选极端事件。 进一步使用完全粘性纳维-斯托克斯求解器REEF3D::CFD来研究极端波浪荷载。 势流求解器和粘性求解器之间的一向水动力耦合（HDC）确保了信息的准确传递。 所提出的NORA-SARAH框架将NORA数据库与SWAN、REEF3D、ALE和HDC相结合，为复杂的海岸环境提供了一种稳健的方法。 一项在挪威南部的案例研究表明，该方法优于传统的显著波高（Hs）为基础或相位平均建模方法，突显了这种降尺度方法的必要性。
Offshore wave studies often assume Gaussian processes and homogeneous wave fields. However, as waves approach the shoreline, complex coastal topo-bathymetry induces transformations such as shoaling, refraction, diffraction, reflection, and breaking, leading to increased nonlinearity and site-specific wave characteristics. This complexity necessitates detailed site-specific studies for coastal infrastructure design and blue economy planning. This work presents a downscaling procedure for analyzing wave-structure interactions from offshore metocean conditions. The open-access NORA3 and NORA10EI hindcast databases define offshore sea states, which are propagated to nearshore regions using the phase-averaged wave model SWAN. The outputs inform phase-resolving simulations with the fully nonlinear potential flow solver REEF3D::FNPF, incorporating an Arbitrary Eulerian-Lagrangian (ALE) method to compute wave forces via Morisons formulation and to screen for extreme events. Extreme wave loads are further examined using the fully viscous Navier-Stokes solver REEF3D::CFD. A one-way hydrodynamic coupling (HDC) between the potential flow and viscous solvers ensures accurate information transfer. The proposed NORA-SARAH framework, integrating NORA databases with SWAN, REEF3D, ALE, and HDC, offers a robust approach for complex coastal environments. A case study in Southern Norway demonstrates its advantages over traditional significant wave height (Hs)-based or phase-averaged modeling practices, highlighting the necessity of this downscaling method.
- [2] arXiv:2509.21333 [中文pdf, pdf, html, 其他]
-
标题： 从旋转的海冰浮冰到边缘冰区海洋涡旋能量谱标题： From spinning sea ice floes to ocean enstrophy spectra in the Marginal Ice ZoneMinki Kim, Georgy E. Manucharyan, Michelle H. DiBenedetto, Ellen M. Buckley, Daniel M. Watkins, Monica M. Wilhelmus主题： 大气与海洋物理 (physics.ao-ph)
量化北极海洋中的动能（KE）和涡旋度传递、混合和耗散是理解极地海洋动力学的关键，这些是全球气候系统的重要组成部分。 然而，在冰覆盖区域，有限的涡旋解析观测挑战了在不同尺度上表征动能和涡旋度传递。 在此，我们使用卫星获得的海冰浮冰旋转速率来推断边缘冰区的表面海洋涡旋度谱。 采用一种粗粒化方法，我们将每个浮冰视为局部空间滤波器。 该方法通过理想化的海冰-海洋模拟进行验证，并应用于贝弗特环流中的浮冰观测。 我们的结果揭示了低海冰浓度下的陡峭谱斜率，表明春季到夏季过渡期间中尺度活动增强。 高分辨率模拟支持这些发现，但高估了涡旋度，突显出需要更密集的观测阵列。 我们的二维谱估计是首次此类研究，提供了一种可扩展的方法来绘制北极海洋特征。
Quantifying kinetic energy (KE) and enstrophy transfer, mixing, and dissipation in the Arctic Ocean is key to understanding polar ocean dynamics, which are critical components of the global climate system. However, in ice-covered regions, limited eddy-resolving observations challenge characterizing KE and enstrophy transfer across scales. Here, we use satellite-derived sea ice floe rotation rates to infer the surface ocean enstrophy spectra in the marginal ice zone. Employing a coarse-graining approach, we treat each floe as a local spatial filter. The method is validated with idealized sea ice-ocean simulations and applied to floe observations in the Beaufort Gyre. Our results reveal steepened spectral slopes at low sea ice concentrations, indicating enhanced mesoscale activity during the spring-to-summer transition. High-resolution simulations support these findings but overestimate enstrophy, highlighting a denser array of observations. Our two-dimensional spectral estimates are the first of their kind, providing a scalable approach for mapping Arctic Ocean characteristics.
- [3] arXiv:2509.21334 [中文pdf, pdf, html, 其他]
-
标题： 气候变化：时间与频率方面标题： Climate change: across time and frequencies主题： 大气与海洋物理 (physics.ao-ph) ; 计量经济学 (econ.EM)
我们使用连续小波工具来表征气候变化在时间和频率上的动态特性。 这种方法使我们能够捕捉全球平均温度异常与气候强迫之间关系的变化模式。 使用1850年至2022年的历史数据，我们发现温室气体，尤其是CO$_2$，在控制自然强迫影响后，在温度的非常低频趋势行为中起到了重要作用。 在较短的频率下，强迫对温度的影响会开启和关闭，这很可能是由于地球气候系统中的复杂反馈机制。
We use continuous wavelet tools to characterize the dynamics of climate change across time and frequencies. This approach allows us to capture the changing patterns in the relationship between global mean temperature anomalies and climate forcings. Using historical data from 1850 to 2022, we find that greenhouse gases, and CO$_2$ in particular, play a significant role in driving the very low frequency trending behaviour in temperatures, even after controlling for the effects of natural forcings. At shorter frequencies, the effect of forcings on temperatures switches on and off, most likely because of complex feedback mechanisms in Earth's climate system.
- [4] arXiv:2509.21349 [中文pdf, pdf, 其他]
-
标题： 使用非迭代时空变换器模型的准确台风强度预测标题： Accurate typhoon intensity forecasts using a non-iterative spatiotemporal transformer model评论： 41页，正文中有5张图表，附录中有6张图表。 提交至《npj Climate and Atmospheric Science》主题： 大气与海洋物理 (physics.ao-ph) ; 机器学习 (cs.LG)
准确预测热带气旋（TC）强度——特别是在快速增强和快速减弱期间——仍然是业务气象学的挑战，对灾害准备和基础设施韧性具有高风险影响。 机器学习的最新进展在TC预测方面取得了显著进展；然而，大多数现有系统提供的预测在极端情况下迅速退化，并且缺乏长期一致性。 在这里，我们介绍了TIFNet，这是一种基于Transformer的预测模型，通过将高分辨率全球预测与历史演变融合机制相结合，生成非迭代的5天强度轨迹。 TIFNet在再分析数据上进行训练，并使用业务数据进行微调，在所有预测时间范围内始终优于业务数值模型，为弱、强和超级台风类别提供了稳健的改进。 在长期被视为最难预测的快速强度变化情况下，TIFNet相对于当前业务基准将预测误差降低了29-43%。 这些结果代表了基于人工智能的TC强度预测的重大进展，尤其是在传统模型持续表现不佳的极端条件下。
Accurate forecasting of tropical cyclone (TC) intensity - particularly during periods of rapid intensification and rapid weakening - remains a challenge for operational meteorology, with high-stakes implications for disaster preparedness and infrastructure resilience. Recent advances in machine learning have yielded notable progress in TC prediction; however, most existing systems provide forecasts that degrade rapidly in extreme regimes and lack long-range consistency. Here we introduce TIFNet, a transformer-based forecasting model that generates non-iterative, 5-day intensity trajectories by integrating high-resolution global forecasts with a historical-evolution fusion mechanism. Trained on reanalysis data and fine-tuned with operational data, TIFNet consistently outperforms operational numerical models across all forecast horizons, delivering robust improvements across weak, strong, and super typhoon categories. In rapid intensity change regimes - long regarded as the most difficult to forecast - TIFNet reduces forecast error by 29-43% relative to current operational baselines. These results represent a substantial advance in artificial-intelligence-based TC intensity forecasting, especially under extreme conditions where traditional models consistently underperform.
- [5] arXiv:2509.21844 [中文pdf, pdf, html, 其他]
-
标题： 生成式人工智能-大型集合项目下放前所未有的未来干旱标题： Generative AI-Downscaling of Large Ensembles Project Unprecedented Future DroughtsHamish Lewis, Neelesh Rampal, Peter B. Gibson, Luke J. Harrington, Chiara M. Holgate, Anna Ukkola, Nicola M. Maher评论： 61页，37图主题： 大气与海洋物理 (physics.ao-ph)
理解干旱在未来可能如何变化对于预见和减轻其不利影响至关重要。 然而，可靠的气候预测需要大量的高分辨率气候模拟，特别是对于评估极端事件而言。 在此，我们使用了一个新的数据集，即多个全球气候模型（GCMs）的大集合，通过生成式人工智能降尺度至12公里，以量化新西兰未来气象干旱的风险。 这些集合包括20个GCMs，其中包括两个单一模型初始条件大集合。 该人工智能被训练以模仿用于动态降尺度的基于物理的区域气候模型（RCM），并在降水和干旱指标方面提供了与RCM相当的价值。 在所有集合中都发现了降水变异性显著增加，同时平均降水的变化高度不确定。 未来预测显示，全国大部分地区干旱将变得更加严重，然而，内部变率和模型不确定性掩盖了全国大部分地区干旱持续时间和频率的未来变化。 使用较少数量的动态降尺度模拟会低估这种不确定性。 我们发现证据表明，在当前气候下，整个国家可能会出现比小集合中发现的干旱长两倍的极端干旱，这突显了长时间降尺度模拟在采样罕见事件方面的价值。 在高排放情景SSP3-7.0下，许多地区的极端长期干旱长度增加，导致一些地区出现约30个月长的事件。
Understanding how droughts may change in the future is essential for anticipating and mitigating their adverse impacts. However, robust climate projections require large amounts of high-resolution climate simulations, particularly for assessing extreme events. Here, we use a novel dataset, multiple large-ensembles of Global Climate Models (GCMs), downscaled to 12km using generative AI, to quantify the future risk of meteorological drought across New Zealand. The ensembles consists of 20 GCMs, including two single-model initial condition large ensembles. The AI is trained to emulate a physics-based regional climate model (RCM) used in dynamically downscaling, and adds a similar amount of value as the RCM across precipitation and drought metrics. Marked increases in precipitation variability are found across all ensembles, alongside highly uncertain changes in mean precipitation. Future projections show droughts will become more intense across the majority of the country, however, internal variability and model uncertainty obscure future changes in drought durations and frequency across large portions of the country. This uncertainty is understated using a smaller number of dynamically-downscaled simulations. We find evidence that extreme droughts up to twice as long as those found in smaller ensembles, could occur across the entirety of the country in the current climate, highlighting the value of long-duration downscaled simulations to sample rare events. These extremely long droughts increase in length in many locations under a high emissions SSP3-7.0 scenario giving rise to events around 30 months long in some locations.
- [6] arXiv:2509.21939 [中文pdf, pdf, html, 其他]
-
标题： 非经典重力波动力学对中间大气平均流和太阳潮汐的影响标题： Impact of Non-Classical Gravity-Wave Dynamics on Middle-Atmosphere Mean Flow and Solar Tides评论： 提交至《地球物理研究杂志》主题： 大气与海洋物理 (physics.ao-ph)
传统重力波（GW）参数化忽略了GW动力学的三个方面。 它们不使用动量和熵通量，而是使用Eliassen-Palm通量，从而忽略了分辨流可能不在地转和静力平衡中的可能性。 通过在每个时间步确定如果垂直重力波传播是瞬时的，将产生的重力波通量的平衡剖面，它们忽略了重力波场和分辨流的瞬变性。 此外，它们也不考虑横向重力波传播和水平重力波通量。 由于预报性重力波模型MS-GWaM不需要做出这些假设，它已被用于全球天气和气候代码ICON中，以研究它们对月平均纬向平均流和太阳潮汐模拟结果的影响。 发现这三个方面都会显著影响模拟结果。 其中，瞬变性和横向传播的影响最强。 中间层和低热层的平均环流在所有纬度以及平流层高海拔地区都受到影响。 这与直接修改的重力波强迫一起，也导致了太阳潮汐的迁移和非迁移成分的显著差异。 如果同时考虑这两个方面，与从卫星数据中获取的潮汐比较最为有利。 这表明，在参数化中应相应地对重力波动力学进行广义处理，作为允许重力波模拟的高效替代方案。
Conventional gravity-wave (GW) parameterizations neglect three aspects of GW dynamics. Instead of momentum and entropy fluxes they use Eliassen-Palm fluxes, thereby neglecting the possibility that resolved flow are not in geostrophic and hydrostatic balance. They neglect the transience of the GW field and of the resolved flow, by determining at every time step equilibrium profiles of GW fluxes that would result if the vertical GW propagation were instantaneous. Moreover, they also do not take into account lateral GW propagation and horizontal GW fluxes. Because the prognostic GW model MS-GWaM does not need to make these assumptions, it has been used in the global weather and climate code ICON to investigate their consequences for the simulation of monthly mean zonal mean flows and of solar tides. All three aspects are found to influence the simulation results significantly. Among those, transience and lateral propagation have the strongest impact. The mean circulation in the mesosphere and lower thermosphere is affected at all latitudes and in the stratosphere at high altitudes as well. This together with the directly modified GW forcing leads also to significant differences in the migrating and nonmigrating components of solar tides. Comparisons with tides retrieved from satellite data are most favorable if both aspects are taken into account. This argues for a correspondingly generalized treatment of GW dynamics in their parameterization, as an efficient alternative to GW permitting simulations.
- [7] arXiv:2509.21957 [中文pdf, pdf, html, 其他]
-
标题： 大堡礁数字孪生研究：极端模型分辨率对潮汐模拟的影响标题： Toward a digital twin of the Great Barrier Reef: impact of extreme model resolution on tidal simulations主题： 大气与海洋物理 (physics.ao-ph)
珊瑚礁是拓扑结构复杂的环境，在小空间尺度上变化很大。 高分辨率数据（米级）用于研究这些环境的可用性迅速增加，以至于许多研究人员正在积极创建这些环境的“数字孪生体”，以帮助保护和管理。 然而，就像任何模型一样，数字孪生体的有用性仅取决于用于创建它的数据。 以前对珊瑚礁的数值模拟工作是在从几十米到几千米的多种分辨率下进行的，但迄今为止，还没有关于米级极端模型分辨率影响的全面研究。 在这里，我们使用网格尺度从20,000米到5米的高分辨率、多尺度模型来模拟大堡礁的Capricorn Bunker区域，并将其与网格尺度最小为250米和50米的模型进行比较。结果表明，可观测的物理过程在极高的分辨率下得到最佳模拟，尽管中间分辨率模型也表现良好。 低分辨率模型虽然使用的分辨率与一些先前的研究相当，但不足以捕捉局部尺度的过程。 数值模型在创建沿海海域的数字孪生体中起着至关重要的作用，因为它们包含了生物物理和化学过程的数学表示，但目前的分辨率比可用于构建数字孪生体的卫星和水深数据要粗。 弥合这一分辨率差距仍然是一个挑战。
Coral reefs are topologically complex environments with a large variation over small spatial-scales. The availability of high resolution data (metre-scale) to study these environments has increased rapidly such that many researchers are actively engaged in creating a `digital twin' of these environments to aid protection and management. However, as with any model, a digital twin will only be as useful as the data used to create it. Previous numerical modelling work on coral reefs has been carried out at a range of resolutions from 10s to 1000s of metres, but to date there has been no comprehensive study on the impact of extreme model resolution at metre-scale. Here, we simulate the Capricorn Bunker region of the GBR in a high resolution, multi-scale model using grid scales of 20,000 m to 5 m and compare that to the models with minimum grid scales of 250 m and 50 m. It is shown that the observable physical processes are best simulated at extremely high resolutions, though the intermediate resolution model performs well also. The low resolution model, whilst using a resolution comparable to a number of previous studies, does not sufficiently capture local-scale processes. Numerical models play a vital role in creating a digital twin of coastal seas as they contain the mathematical representation of the biophysical and chemical processes present but are currently at a coarser resolution than satellite and bathymetric data on which digital twins could be based. Bridging this resolution gap remains a challenge.
- [8] arXiv:2509.22359 [中文pdf, pdf, html, 其他]
-
标题： 利用昨日气候预测未来：人工智能天气和气候模型中的温度偏差标题： Forecasting the Future with Yesterday's Climate: Temperature Bias in AI Weather and Climate Models评论： 13页，5图主题： 大气与海洋物理 (physics.ao-ph) ; 人工智能 (cs.AI)
基于人工智能的气候和天气模型迅速获得了流行，提供了速度更快且技能可以与甚至超过传统动力模型的预测。 尽管取得了成功，这些模型面临一个关键挑战：仅使用历史数据进行训练时预测未来气候。 在本研究中，我们通过分析AI天气和气候模型中的北半球冬季陆地温度偏差来研究这个问题。 我们检查了两个天气模型，FourCastNet V2 Small（FourCastNet）和Pangu Weather（Pangu），评估它们对2020-2025年的预测以及Ai2 Climate Emulator版本2（ACE2）对1996-2010年的预测。 这些时间段超出了各自模型的训练集，并且比它们大部分训练数据要新得多，使我们能够评估模型在新的即更现代条件下的泛化能力。 我们发现所有三个模型都产生了冷偏差的平均温度，类似于它们所预测时期之前15-20年的气候。 在某些地区，如美国东部，预测类似于比所预测时期早20-30年的气候。 进一步分析显示，FourCastNet和Pangu的冷偏差在最热的预测温度中最强，表明它们对现代极端高温事件的训练接触有限。 相比之下，ACE2的偏差分布更为均匀，但在气候变化最显著的地区、季节和温度分布部分最大。 这些发现强调了仅使用历史数据训练人工智能模型的挑战，并突出了在将它们应用于未来气候预测时需要考虑此类偏差的必要性。
AI-based climate and weather models have rapidly gained popularity, providing faster forecasts with skill that can match or even surpass that of traditional dynamical models. Despite this success, these models face a key challenge: predicting future climates while being trained only with historical data. In this study, we investigate this issue by analyzing boreal winter land temperature biases in AI weather and climate models. We examine two weather models, FourCastNet V2 Small (FourCastNet) and Pangu Weather (Pangu), evaluating their predictions for 2020-2025 and Ai2 Climate Emulator version 2 (ACE2) for 1996-2010. These time periods lie outside of the respective models' training sets and are significantly more recent than the bulk of their training data, allowing us to assess how well the models generalize to new, i.e. more modern, conditions. We find that all three models produce cold-biased mean temperatures, resembling climates from 15-20 years earlier than the period they are predicting. In some regions, like the Eastern U.S., the predictions resemble climates from as much as 20-30 years earlier. Further analysis shows that FourCastNet's and Pangu's cold bias is strongest in the hottest predicted temperatures, indicating limited training exposure to modern extreme heat events. In contrast, ACE2's bias is more evenly distributed but largest in regions, seasons, and parts of the temperature distribution where climate change has been most pronounced. These findings underscore the challenge of training AI models exclusively on historical data and highlight the need to account for such biases when applying them to future climate prediction.
新提交 (展示 8 之 8 条目 )
- [9] arXiv:2509.21378 (替换) [中文pdf, pdf, 其他]
-
标题： 北极卤跃层的不稳定性标题： Instability of the halocline at the North Pole主题： 地球物理 (physics.geo-ph) ; 偏微分方程分析 (math.AP) ; 大气与海洋物理 (physics.ao-ph)
在本文中，我们研究了近惯性Pollard波的稳定性，作为北极洋区域以北极为中心的盐跃层的模型，该模型来源于Puntini (2025a)。 采用短波长不稳定性方法，此类流动的稳定性简化为沿流体轨迹研究常微分方程组的稳定性，得出的结果是，当近惯性Pollard波的陡度超过特定阈值时，这些波是线性不稳定的。 该模型的显式色散关系使得可以很容易地计算出该阈值，已知水柱的物理特性。
In this paper we address the issue of stability for the near-inertial Pollard waves, as a model for the halocline in the region of the Arctic Ocean centered around the North Pole, derived in Puntini (2025a). Adopting the short-wavelength instability approach, the stability of such flows reduces to study the stability of a system of ODEs along fluid trajectories, leading to the result that, when the steepness of the near-inertial Pollard waves exceeds a specific threshold, those waves are linearly unstable. The explicit dispersion relation of the model allows to easily compute such threshold, knowing the physical properties of the water column.
- [10] arXiv:2509.21440 (替换) [中文pdf, pdf, 其他]
-
标题： 碳负通勤：整合城市设计、行为和技术的气候正向交通标题： Carbon-Negative Commuting: Integrating Urban Design, Behavior, and Technology for Climate-Positive Mobility主题： 物理与社会 (physics.soc-ph) ; 大气与海洋物理 (physics.ao-ph)
通勤对城市温室气体排放有显著贡献，是气候减缓努力的关键重点。 本文通过分析城市形态、社会经济属性和个体行为的影响，探讨了与通勤相关的二氧化碳排放的多方面特性。 它回顾了用于理解和减少排放的分析方法，包括结构方程建模、多目标优化和基于代理的模拟。 在这些见解的基础上，本文提出了一个碳负通勤的概念框架，该框架结合了空间规划、行为干预、技术创新和碳抵消策略。 来自不同全球背景的案例研究展示了实施这些干预措施的可行性和挑战。 讨论强调了关键权衡、公平性考虑和治理障碍，同时识别了如改善公共健康和城市韧性等协同效益。 本文最后强调了需要跨学科研究和适应性政策制定，以实现碳负通勤，并使城市交通系统与全球脱碳目标保持一致。
Commuting contributes substantially to urban greenhouse gas emissions and represents a critical focus for climate mitigation efforts. This paper explores the multifaceted nature of commuting-related carbon dioxide emissions by analyzing the influence of urban form, socio-economic attributes, and individual behaviors. It reviews analytical approaches including structural equation modeling, multi-objective optimization, and agent-based simulations that have been employed to understand and mitigate emissions. Building on these insights, the paper develops a conceptual framework for carbon-negative commuting that integrates spatial planning, behavioral interventions, technological innovations, and carbon offsetting strategies. Case studies from diverse global contexts illustrate both the feasibility and challenges of implementing these interventions. The discussion highlights key trade-offs, equity considerations, and governance barriers while identifying co-benefits such as improved public health and urban resilience. The paper concludes by emphasizing the need for interdisciplinary research and adaptive policymaking to operationalize carbon-negative commuting and align urban mobility systems with global decarbonization goals.
- [11] arXiv:2509.21477 (替换) [中文pdf, pdf, html, 其他]
-
标题： VISION：从不完整的观测中提示海洋垂直速度重建标题： VISION: Prompting Ocean Vertical Velocity Reconstruction from Incomplete Observations主题： 机器学习 (cs.LG) ; 计算机视觉与模式识别 (cs.CV) ; 大气与海洋物理 (physics.ao-ph)
从不完整的表层观测中重建海底海洋动力学，如垂直速度场，是地球科学中的一个关键挑战，该领域长期以来因缺乏标准化、可分析的基准而受到阻碍。 为系统解决这一问题并推动研究，我们首先构建并发布了KD48，这是一个高分辨率的海洋动力学基准，源自拍字节级的模拟，并通过专家驱动的去噪进行整理。 基于这个基准，我们引入了VISION，这是一种基于动态提示的新重建范式，旨在解决现实观测中数据缺失的核心问题。 VISION的核心在于其能够从任何可用的观测子集实时生成视觉提示，该提示同时编码数据可用性和海洋的物理状态。 更重要的是，我们设计了一个状态条件提示模块，可以将此提示高效注入一个通用主干网络中，该网络配备了几何和尺度感知的操作器，以引导其自适应调整计算策略。 这种机制使VISION能够精确处理不同输入组合带来的挑战。 在KD48基准上的大量实验表明，VISION不仅显著优于最先进的模型，而且在极端数据缺失场景下也表现出强大的泛化能力。 通过提供高质量的基准和稳健的模型，我们的工作为数据不确定性下的海洋科学研究建立了坚实的基础设施。 我们的代码可在以下链接获取：https://github.com/YuanGao-YG/VISION.
Reconstructing subsurface ocean dynamics, such as vertical velocity fields, from incomplete surface observations poses a critical challenge in Earth science, a field long hampered by the lack of standardized, analysis-ready benchmarks. To systematically address this issue and catalyze research, we first build and release KD48, a high-resolution ocean dynamics benchmark derived from petascale simulations and curated with expert-driven denoising. Building on this benchmark, we introduce VISION, a novel reconstruction paradigm based on Dynamic Prompting designed to tackle the core problem of missing data in real-world observations. The essence of VISION lies in its ability to generate a visual prompt on-the-fly from any available subset of observations, which encodes both data availability and the ocean's physical state. More importantly, we design a State-conditioned Prompting module that efficiently injects this prompt into a universal backbone, endowed with geometry- and scale-aware operators, to guide its adaptive adjustment of computational strategies. This mechanism enables VISION to precisely handle the challenges posed by varying input combinations. Extensive experiments on the KD48 benchmark demonstrate that VISION not only substantially outperforms state-of-the-art models but also exhibits strong generalization under extreme data missing scenarios. By providing a high-quality benchmark and a robust model, our work establishes a solid infrastructure for ocean science research under data uncertainty. Our codes are available at: https://github.com/YuanGao-YG/VISION.