Searcharxiv⌕ Search

arXiv · 2611.11563

HydroTuring: A physics probe benchmark testing mass, energy, and momentum conservation in hydrologic models

Abstract

Hydrologic models, including deep learning systems now deployed for operational flood forecasting, are evaluated primarily by how closely their simulated hydrographs match observed streamflow. However, high hydrograph skill does not guarantee physical consistency: in a synthetic calibration experiment, a conceptual model that loses 15% of its water input to an unreported sink still reaches a Nash-Sutcliffe efficiency of 0.93, and no shared benchmark tests whether models obey fundamental conservation laws. Here we introduce HydroTuring (v0.1.0), a black-box benchmark of 37 physics probes that test mass (24 probes), energy (7) and momentum (6) conservation in hydrologic models. Each probe runs a controlled experiment on forcing generated at run time from a recorded seed and applies binary pass or fail criteria validated against 18 physically exact and 59 deliberately broken reference models. Across 32 evaluated models (3 exact references, 7 AI-based models and 22 process-based models), only the 3 exact references pass every probe among the 28 models scored on 5 or more probes. Process-based and AI-based models pass similar shares of their scored probes (76% and 74%), and a pre-trained flood-forecasting LSTM fails 6 of the 11 runoff-only probes it can be scored on. Furthermore, 49% of all 1,184 model-probe pairs remain unscored because models omit the required state or flux variables, while the scored failures occur in perturbation responses and state bounds rather than long-term water-budget closure. HydroTuring provides a reproducible smoke test for model development, peer review and operational deployment under an open community protocol.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Zhi Li, Changming Li, Siavash Shams, Han Wang, Qingyi Yang, Yuanhang Liu, Taiqi Lian, Xin Lan, Shunan Zhou, Binlan Zhang, Siddik Barbhuiya, Songkun Yan, Bing Li, Jinlong Hu, Yaji Wang, Zhenjiang Wu, Mofan Zhang, Kaihao Long, Jingzhi Chen, Yuan-Heng Wang. 2026-11-08. HydroTuring: A physics probe benchmark testing mass, energy, and momentum conservation in hydrologic models. https://searcharxiv.com/abs/2611.11563

Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Searcharxiv by Indiciae: BibTeX · RIS.

KEEP EXPLORING

Related papers

PyOTIS: an open-source Python package for stream solute transport with transient storage

OTIS and OTIS-P provide transient-storage transport and parameter estimation, whereas E-OTIS adds coupled water-quality reactions in a separate MATLAB implementation. PyOTIS brings these workflows into an open-source Python package with legacy OTIS input support. Its contribution is software integration: a scriptable interface, structured simulation outputs, Jacobian-based parameter uncertainty, sensitivity analysis, and selectable transport schemes. A browser-based graphical user interface (GUI) provides example loading, custom parameter entry, breakthrough-curve and storage-zone plots, reference overlays, and calibration of supplied cases with confidence intervals. The package retains the OTIS Crank--Nicolson formulation and the E-OTIS reaction network; these are inherited model components rather than new process formulations. Evaluation distinguishes code comparisons, analytical convergence, and fits to observations. For the six OTIS manual examples, the comparisons distinguish model outputs from field observations and include agreement to seven significant figures for the steady-state iron profile. The nine-species E-OTIS comparison reports maximum absolute differences of order $10^{-14}$\,mg\,L$^{-1}$ against MATLAB. The conservative analytical benchmark shows approximately second-order convergence. For Clackamas River calibration, residual sums of squares differ from OTIS-P by 3.65--5.12\%, although individual parameter estimates differ by up to 39.1\%. These results support implementation consistency for the tested cases without establishing unique parameter estimates or validating every reaction pathway. Source code, example inputs, and verification scripts accompany the package.

physics.geo-ph↗

2.5D Transformer: An Efficient 3D Seismic Interpolation Method without Full 3D Training

Transformer has emerged as a powerful deep-learning technique for two-dimensional (2D) seismic data interpolation, owing to its global modeling ability. However, its core operation introduces heavy computational burden due to the quadratic complexity, hindering its further application to higher-dimensional data. To achieve Transformer-based three-dimensional (3D) seismic interpolation, we propose a 2.5-dimensional Transformer network (T-2.5D) that adopts a cross-dimensional transfer learning (TL) strategy, so as to adapt the 2D Transformer encoders to 3D seismic data. The proposed T-2.5D is mainly composed of 2D Transformer encoders and 3D seismic dimension adapters (SDAs). Each 3D SDA is placed before a Transformer encoder to learn spatial correlation information across seismic lines. The proposed cross-dimensional TL strategy comprises two stages: 2D pre-training and 3D fine-tuning. In the first stage, we optimize the 2D Transformer encoders using a large amount of 2D data patches. In the second stage, we freeze the 2D Transformer encoders and fine-tune the 3D SDAs using limited 3D data volumes. Extensive experiments on multiple datasets are conducted to assess the effectiveness and efficiency of T-2.5D. Experimental results demonstrate that the proposed method achieves comparable performance to that of full 3D Transformer at a significantly low cost.

physics.geo-ph↗

RevealLO: Global-Scale FWI Including Long-Period Major-Arc Surface Wave and Multi-Orbit Body Wave Data

Global full-waveform inversion (FWI) has matured significantly in recent years, but resolution in the southern hemisphere often remains limited due to sparse receiver coverage. To address this problem, we constructed RevealLO, a global-scale FWI model that updates its predecessor REVEAL by expanding the inversion time window from 1 hour to 3.5 hours. The extended time window incorporates both minor-arc ($R_1, G_1$) and major-arc ($R_2, G_2$) surface wave arrivals, as well as late-arriving multi-orbit body waves, thereby increasing azimuthal coverage beneath ocean basins and the southern hemisphere. We simulate 3-D viscoelastic wavefields using the spectral-element solver Salvus on global fourth-order cubed-sphere meshes, and update velocity structure by successively broadening the period band from 100 - 200 s to 60 - 200 s. Measurements and adjoint sources are based on time-frequency phase misfits, which are suitable for both large-amplitude surface and low-amplitude multiply-reflected body waves. For iterative optimization, we employ a mini-batch stochastic trust-region L-BFGS algorithm, followed by mesh reprojection onto the $33~s$ REVEAL baseline mesh and lowermost-mantle regularization. RevealLO preserves short-period whole-Earth body wave fits while improving azimuthal sampling and mitigating trade-offs between unmodeled azimuthal anisotropy and isotropic structure in the upper and mid mantle.

physics.geo-ph↗