Pith. sign in

REVIEW 4 major objections 4 minor 22 references

Context-Aware Concept Distillation for Trustworthy Flood Prediction

T0 review · 4 major / 4 minor · reviewed 2026-07-31 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper proposes Context-Aware Concept Distillation (CACD), a framework that compresses a black-box LSTM flood forecaster into six human-interpretable hydrological concepts and a context-modulated linear decoder, claiming this interpretab

desk verdict Useful open-sourced concept-distillation method for hydrology, but the headline fidelity gap rests on a single run and an under-specified SHAP clustering split. read the letter →

arxiv 2607.23237 v1 pith:V6CAPGHB submitted 2026-07-25 cs.LG cs.AI

Eli Levinkopf , Efrat Morin , Claudia V. Goldman This is my paper · ORCID
classification cs.LGcs.AI
keywords conceptdistillationfloodforecastinginterpretablemachinelearningSHAPclusteringresidualhypernetworkLSTMsurrogatehydrologicalconceptsexplainableAI
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to show that the reasoning of a high-dimensional LSTM flood forecaster can be compressed into six human-interpretable hydrological concepts—discovered automatically from SHAP attributions—and that a surrogate model which predicts streamflow as a linear combination of these concepts, with coefficients modulated by static basin attributes, reconstructs the teacher LSTM's predictions on unseen future years better than a generic nonlinear decoder (Median NSE 0.70 vs 0.60). The framework is called Context-Aware Concept Distillation (CACD). The claim is significant because it suggests that the structural bias of separating temporal forcing from spatial sensitivity acts as a regularizer, and that verifiable, concept-level explanations can be operationally useful rather than merely descriptive. A sympathetic reader would care because if true, flood warning systems could audit AI predictions against a small vocabulary of physical mechanisms before alerts.

What carries the argument

The key machinery is (1) an unsupervised pipeline that clusters SHAP attribution matrices into six concepts using log-temporal windows and robust statistics (net influence, magnitude, soft peak lag, concentration), and (2) a Context-Aware Decoder built from a Residual Hypernetwork: a static-attribute MLP produces a latent basin embedding z, and two linear heads generate residual corrections ΔW and Δb to a global base weight W_base and bias b_base, so the final prediction ŷ = W(x_s)·p + b(x_s) remains linear in the concept probabilities p while being nonlinear in the static context x_s. This enforces a strict separation of temporal forcing from spatial sensitivity and keeps the concept-to-flo

What would settle it

Re-run the pipeline with the unsupervised clustering (Section 3.1) restricted strictly to SHAP values computed on 2001–2014 training data, then evaluate the surrogate on 2018–2020; if the Median NSE falls to or below the Deep MLP baseline (0.60), the claim that the structural bias itself drives generalization would be refuted. Also, recompute basin-level NSE excluding the low-variance catchments (log10 σ² < −1) and check whether the remaining basins still show a median advantage.

Watch

Extended reading notes

Core claim

The central claim is that an interpretable surrogate—trained to imitate an LSTM teacher using six unsupervised-discovered concepts (the 'Hydrological Language') and a Residual Hypernetwork that generates basin-specific weights from 51 static attributes—achieves higher fidelity on unseen test years than a non-interpretable MLP decoder (Median NSE 0.70 vs 0.60 on 5,203 basins). The paper argues this shows the teacher's reasoning is intrinsically low-dimensional and context-dependent, and that enforcing the structure 'streamflow = basin sensitivity × dynamic forcing' prevents overfitting to training correlations.

Load-bearing premise

The core premise is that the SHAP-based concept vocabulary is discovered from the training period only and remains stable enough to generalize to future data; the paper does not explicitly state which time split is used for the unsupervised clustering in Section 3.1, so if test-period SHAP values leaked into the clustering, the reported test fidelity would be inflated by information leakage.

Editorial extensions

If this is right

  • If the framework is correct, emergency authorities could audit an LSTM flood forecast by reading which of six concepts drove it and how basin attributes modulate that concept's impact, without losing accuracy relative to a black-box decoder.
  • The structural bias of separating global hydrological laws from local adaptations appears to generalize better than unconstrained nonlinear decoders on unseen future periods.
  • The discovered concept vocabulary is sufficient to reconstruct streamflow predictions, providing a shared 'language' between the model and domain experts that can be validated against physical intuition.
  • The success depends on high-fidelity feature engineering (log-temporal windows and morphology statistics), since simple summation collapses success rate from 50.3% to 10.9%.
  • Low-fidelity basins are concentrated in low-variance catchments with small absolute errors, suggesting NSE-based evaluation penalizes quiescent rivers independently of model skill.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the conceptual vocabulary is stable across time, the same six concepts could serve as a universal 'hydrological language' for auditing other deep rainfall-runoff models, not just the LSTM used here.
  • The residual hypernetwork idea might transfer to other environmental prediction tasks where a global law is modulated by local context (e.g., water quality, crop yield), provided the teacher's explanations can be clustered into concepts.
  • A direct test of the paper's causal claim would be to perturb concept probabilities and check whether streamflow predictions change in the direction hydrologists expect; the paper does not report such counterfactual tests.
  • The paper leaves the data split for concept discovery unspecified; verifying that clustering uses only training-period SHAP values is a prerequisite for trusting the test-fidelity numbers.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper presents Context-Aware Concept Distillation (CACD), a pipeline for distilling a black-box LSTM streamflow forecaster into an interpretable surrogate. Attribution matrices from SHAP are aggregated with log-temporal windows and robust statistics, then hierarchically clustered into K=6 concepts ('Hydrological Language'). A Concept Encoder maps the LSTM hidden state to concept probabilities, and a Residual Hypernetwork generates basin-specific linear weights from static attributes so the final prediction remains linear in concepts. The method is evaluated on 5,203 Caravan basins with a temporal split (train 2001-2014, validation 2015-2017, test 2018-2020) by measuring fidelity to the teacher LSTM's predictions. The main claim is that the context-aware surrogate achieves median NSE 0.70 on test data, compared with 0.60 for a Deep MLP decoder, and that this demonstrates six concepts are sufficient to reconstruct the teacher's reasoning.

Significance. If the out-of-sample claim holds, the paper makes a useful contribution: it turns local SHAP attributions into a compact, semantically inspectable vocabulary, imposes a physically motivated separation between temporal concepts and static basin context, and evaluates at global scale. The open-sourced code, pretrained models, and extracted concept dataset are concrete reproducibility assets. The 'six concepts suffice' result is a falsifiable compression claim rather than a purely descriptive XAI exercise. That said, the headline comparison currently rests on an ambiguous data-split statement and on point estimates without uncertainty quantification; the abstract and conclusion also overstate what fidelity to the teacher means. These issues are fixable but must be addressed before the central claim can be accepted.

major comments (4)
  1. [§3.1 / §4.3] Potential information leakage in concept discovery. The text describes clustering of SHAP feature vectors v in §3.1 but never states that the SHAP matrices are computed only on the training split (2001-2014). §4.3 later defines the extreme-event set D as the top 10% of days 'from the training period,' but that sentence comes after the clustering methodology and does not explicitly bind the SHAP computation. If SHAP values from 2018-2020 are used to form clusters or to select K, then the Concept Encoder labels encode test-period information and the Table 1 comparison (0.70 vs 0.60) is not a clean out-of-sample result. Please state explicitly which split is used for every SHAP computation, and confirm that cluster assignment, silhouette optimization, and K selection are train-only. A short code trace would settle this.
  2. [Table 1 / §5.1] The central claim 'significantly outperforms' is unsupported by any uncertainty quantification. Only point estimates are reported for Global NSE, Median NSE, and Success Rate. With 5,203 basins, a paired per-basin comparison is straightforward: report bootstrap confidence intervals for the median difference and a paired test (e.g., Wilcoxon signed-rank) for the Deep MLP vs. Context-Aware comparison. Also report variance across random seeds or initializations of both the decoder and the teacher, since the distillation pipeline involves stochastic training.
  3. [Abstract / Conclusion / §4.4] The metrics in §4.4 explicitly use the LSTM's predictions y_LSTM as the target, not observed streamflow. Thus 'Median NSE 0.70' is a fidelity-to-teacher score, and the Conclusion's phrase 'accurate streamflow predictions can be reconstructed using only these concepts' overstates the finding: a surrogate that perfectly reproduced the teacher would still have the teacher's 0.60 NSE against observations. Please replace 'accurate streamflow predictions' with 'faithful reconstructions of the teacher's predictions' or an equivalent qualification throughout the abstract and conclusion. This is a wording issue, but it is load-bearing for the paper's scientific claim.
  4. [§5.1 (Variance Paradox)] The claim that low-fidelity basins are artifacts of NSE in low-variance regimes is supported only by two median RMSE values (0.14 vs 0.51) without distributions, sample sizes, or a statistical test. The threshold log10(sigma^2)<-1 and the comparison group ('high-fidelity basins') should be defined before the analysis. Please provide confidence intervals or an effect size for the RMSE difference and, if possible, a scatter plot of NSE vs. variance with the RMSE comparison shown. This analysis is secondary, but as written it does not justify the conclusion that these are not predictive failures.
minor comments (4)
  1. [§3.1] The notation 'F represents the set of dynamic and static input features' is ambiguous; if F is the number of features, it should be used consistently in the dimension of the attribution matrix.
  2. [Figure 4] The CDF figure lacks explicit axis labels and a legend in the caption; please add them so the stochastic-dominance claim is directly readable from the figure.
  3. [§4.5 / Table 2] The 'Simple Summation' baseline is described as using 'three broad windows,' but the number of windows and their boundaries are not given. State them for reproducibility.
  4. [General] Several equations (e.g., Eq. 1-2) define statistics per feature f and window W, but the notation in the text sometimes omits the f,W index. A notation table would help.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central fidelity claim is a controlled empirical comparison of decoders sharing a fixed concept encoder, not a definitional reduction.

full rationale

The paper's main result is an empirical comparison between two decoder architectures (Context-Aware Residual Hypernetwork vs. Deep MLP) that use the same frozen concept encoder and are evaluated on the same out-of-sample teacher targets. The concept vocabulary is discovered by unsupervised clustering of SHAP attributions from the frozen teacher; this is an interpretability prior, not a fitted target. The decoder is trained to minimize MSE to the teacher's predictions, and fidelity is measured on the 2018–2020 test period. No equation defines the predicted quantity in terms of the fitted parameters by construction: ŷ = W(x_s)·p + b(x_s) is a supervised regression with p fixed from Stage 1 and W,b learned from training data. The Deep MLP baseline is a controlled ablation of architectural inductive bias, not a renamed fit. The only caveat is that §3.1 does not explicitly state whether the SHAP clustering is restricted to the training-period samples defined in §4.3; if clustering used test-period SHAP values, the 'unseen future data' claim could be inflated. This is a data-leakage/correctness concern, not a circularity of derivation, and it is checkable from the open-source code. Self-citations (e.g., Morin et al. 2009; Nevo et al. 2022) are domain background, not load-bearing. Therefore no circular step is identified.

Assumptions & free parameters 5 free parameters · 5 assumptions · 2 invented entities

The framework rests on a large number of hand-chosen hyperparameters (concept count, window bins, statistics, network sizes) and on the assumption that SHAP clusters are semantically valid. No formal sensitivity analysis or external validation is provided.

free parameters (5)
  • Number of concepts (K_super=2, K_sub=3, K=6) = 6 total
    Chosen via PCA and Silhouette optimization; not independently motivated for all basins.
  • Log-temporal window boundaries = [0-2],[3-7],[8-14],[15-30],[31-60],[61-180],[181-364]
    Hand-designed to approximate hydrologic memory decay; no sensitivity analysis.
  • Robust statistics (I, M, soft peak lag, concentration) = 4 statistics per window
    Selected by the authors; no comparison to alternative aggregations except simple sum.
  • Alpha and beta scaling in hypernetwork = learnable scalars
    Learned during training; their values and impact are not reported.
  • Extreme event threshold = top 10% of days by streamflow
    Arbitrary cutoff for defining the distillation dataset.
assumptions (5)
  • domain assumption SHAP values faithfully represent the LSTM's reasoning
    The whole concept discovery pipeline relies on SHAP being a valid attribution. SHAP is an approximation and can be inconsistent for recurrent models.
  • ad hoc to paper Clustering SHAP statistics yields stable, meaningful semantic concepts
    No external validation (e.g., against known hydrology) beyond the qualitative label check by a co-author.
  • domain assumption The teacher LSTM is a sufficiently accurate flood model (Median NSE 0.60)
    The surrogate is only as good as the teacher; the paper does not question teacher biases.
  • domain assumption The concept encoder can learn cluster labels from hidden states with high accuracy
    The paper does not report classification accuracy or confidence for the concept encoder; a poor encoder would invalidate the interpretation.
  • domain assumption The temporal split (2001-2014/2015-2017/2018-2020) is representative for generalization
    Assumes climate and basin behavior in the test period are within the domain of the training period; no stationary adjustment.
invented entities (2)
  • The six-concept 'Hydrological Language' (C1-C6)
    purpose: Interpretable vocabulary to represent the LSTM's flood-driving mechanisms
    The concepts are derived from the same teacher model they explain; labels are candidate and expert-verified only qualitatively. No external data or quantitative validation is provided.
  • Basin context embedding z
    purpose: Low-dimensional signature of basin physical identity used to modulate concept weights
    A latent vector learned inside the hypernetwork; no independent physical interpretation is claimed beyond being a learned representation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Context-Aware Concept Distillation for Trustworthy Flood Prediction." pith.science (2026). https://pith.science/paper/V6CAPGHB

@misc{pith2026260723237,
  author       = {Pith},
  title        = {Pith review of: Context-Aware Concept Distillation for Trustworthy Flood Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V6CAPGHB}},
  note         = {Machine review of arXiv:2607.23237}
}
read the original abstract

Effective flood risk management relies on accurate forecasting, yet the "black box" nature of stateof-the-art Deep Learning models creates a barrier to trust and accountability in high-stakes public safety decisions. While existing Explainable AI (XAI) methods offer local attributions, they fail to provide the verifiable, operationally meaningful causal narratives required by disaster response authorities. To address this societal challenge, we propose Context-Aware Concept Distillation (CACD), a framework developed in collaboration with domain experts to distill opaque LSTMs into interpretable, hydrology-aware surrogate models. We introduce an unsupervised pipeline to discover a "Hydrological Language" and a Residual Hypernetwork that dynamically modulates these concepts based on static basin characteristics. Evaluated on 5,203 basins globally, our model achieves high fidelity (Median NSE 0.70), significantly outperforming black-box baselines (e.g., Multi Layer Perceptrons) on unseen future data. By demonstrating that human-interpretable concepts are sufficient to reconstruct flood dynamics, this work balances AI accuracy with the transparency required for responsible environmental decision-making.

Figures

Figures reproduced from arXiv: 2607.23237 by the authors.

Figure 1
Figure 1. Context-Aware Concept Distillation Architecture. The inputs are split into dynamic forcings (Xd) and static attributes (Xs). The Stage 1 Encoder distills the LSTM hidden state (h) into interpretable concept probabilities (p). The Stage 2 Context-Aware Decoder (Dψ) uses the static context (Xs) to modulate the mapping from concepts (p) to streamflow (yˆ), allowing for basin-specific adaptations. Stage 1: Distillation … view at source ↗
Figure 2
Figure 2. Zoom-in of the Stage 2 Context-Aware Decoder. The Residual Hypernetwork processes static attributes (Xs) to generate a compact latent context embedding (z). This embedding drives two linear heads to produce residual corrections (∆W, ∆b), which shift the global hydrological laws (Wbase, bbase) to create basin-specific weights (W(xs), b(xs)). zi . We parameterize this as a 2-layer MLP (51 → 64 → 64) with ReLU activati… view at source ↗
Figure 3
Figure 3. The Discovered Hydrological Language. SHAP summary plots for the K = 6 concept clusters. These distinct patterns serve as the ground-truth targets for the Concept Encoder. The model successfully disentangles distinct physical mechanisms. Top Row (Memory￾Driven): Concepts like C2 (Delayed Snowmelt) are driven by temperature and snow depth with long temporal lags. Bottom Row (Event￾Driven): Concepts like C6 (Convectiv… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Cumulative Distribution of NSE Fidelity Scores. Com￾parison of per-basin performance across 5,203 basins. The Context￾Aware model (Blue) consistently dominates the black-box MLP (Red). Furthermore, the gap between Blue and Orange (Simple Sum) visually demonstrates the …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 1 linked inside Pith

  1. [1]

    Bramm, Pavel V

    [Brammet al., 2025 ] Andrei M. Bramm, Pavel V . Matrenin, and Alexandra I. Khalyasmaa. A review of xai methods applications in forecasting runoff and water level hydro- logical tasks.Mathematics, 13(17):2830,

    2025

  2. [5]

    Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (tcav)

    [Kimet al., 2018 ] Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, and Rory Sayres. Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (tcav). InProceedings of the 35th International Conference on Machine Learning (ICML), pages 2668–2677. PMLR,

    2018

  3. [10]

    Lundberg and Su-In Lee

    [Lundberg and Lee, 2017] Scott M. Lundberg and Su-In Lee. A unified approach to interpreting model predic- tions. InAdvances in Neural Information Processing Sys- tems (NeurIPS), volume 30, pages 4765–4774,

    2017

  4. [12]

    [Moriasiet al., 2007 ] Daniel N Moriasi, Jeffrey G Arnold, Michael W Van Liew, Ronald L Bingner, Robert Daren Harmel, and Tamie L. Veith. Model evaluation guide- lines for systematic quantification of accuracy in water- shed simulations.Transactions of the ASABE, 50(3):885– 900,

    2007

  5. [15]

    Flood forecasting with machine learning models in an operational framework.Hydrology and Earth System Sciences, 26(15):4013–4032,

    [Nevoet al., 2022 ] Sella Nevo, Efrat Morin, Adi Gerzi Rosenthal, Asher Metzger, Chen Barshai, Dana Weitzner, Dafi V oloshin, Frederik Kratzert, Gal Elidan, Gideon Dror, Gregory Begelman, Grey Nearing, Guy Shalev, Hila Noga, Ira Shavitt, Liora Yuklea, Moriah Royz, Niv Giladi, Nofar Peled Levi, Ofir Reich, Oren Gilon, Ronnie Maor, Shahar Timnat, Tal Shecht...

    2022

  6. [16]

    ”why should i trust you?”: Explain- ing the predictions of any classifier

    [Ribeiroet al., 2016 ] Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. ”why should i trust you?”: Explain- ing the predictions of any classifier. InProceedings of the 22nd ACM SIGKDD International Conference on Knowl- edge Discovery and Data Mining (KDD), pages 1135– 1144,

    2016

  7. [18]

    Explainable artificial intelligence (xai) on timeseries data: A survey

    [Rojatet al., 2021 ] Thomas Rojat, Rapha ¨el Puget, David Filliat, Javier Del Ser, Rodolphe Gelin, and Na- talia D ´ıaz-Rodr´ıguez. Explainable artificial intelligence (xai) on timeseries data: A survey. arXiv preprint arXiv:2104.00950,

    arXiv 2021

  8. [20]

    Deep inside convolutional networks: Visualising image classification models and saliency maps

    [Simonyanet al., 2014 ] Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. Deep inside convolutional networks: Visualising image classification models and saliency maps. InProceedings of the International Con- ference on Learning Representations (ICLR) Workshop,

    2014

Show all 22 references
  1. [22]

    From black box to insight: Explainable ai for extreme event pre- paredness

    [Vuet al., 2025 ] Kiana Vu, ˙Ismet Selc ¸uk¨Ozer, Phung Lai, Zheng Wu, Thilanka Munasinghe, and Jennifer Wei. From black box to insight: Explainable ai for extreme event pre- paredness. arXiv preprint arXiv:2511.13712, 2025

    2025

  2. [2007]

    Towards flash-flood prediction in the dry dead sea region utilizing radar rainfall information

    [Morinet al., 2009 ] Efrat Morin, Yael Jacoby, Shilo Navon, and Erez Bet-Halachmi. Towards flash-flood prediction in the dry dead sea region utilizing radar rainfall information. Advances in Water Resources, 32(7):1066–1076,

    2009

  3. [2009]

    Global prediction of extreme floods in ungauged watersheds.Nature, 627:559–563,

    [Nearinget al., 2024 ] Grey Nearing, Deborah Cohen, Vusumuzi Dube, Martin Gauch, Oren Gilon, Shaun Harri- gan, Avinatan Hassidim, Daniel Klotz, Frederik Kratzert, Asher Metzger, Sella Nevo, Florian Pappenberger, Christel Prudhomme, Guy Shalev, Shlomo Shenzis, Tadele Yednkachw ...

    2024

  4. [2014]

    Transparency in ai for emer- gency management: building trust and accountability.AI and Ethics, 5:3967–3980, Mar

    [Visave, 2025] Jaideep Visave. Transparency in ai for emer- gency management: building trust and accountability.AI and Ethics, 5:3967–3980, Mar

    2025

  5. [2016]

    Controls of flash flood peak discharge in mediterranean basins and the special role of runoff-contributing areas.Journal of Hydrology, 565:846– 860,

    [Rinatet al., 2018 ] Yair Rinat, Francesco Marra, Davide Zoccatelli, and Efrat Morin. Controls of flash flood peak discharge in mediterranean basins and the special role of runoff-contributing areas.Journal of Hydrology, 565:846– 860,

    2018

  6. [2017]

    [Mamalakiset al., 2022 ] Antonios Mamalakis, Imme Ebert- Uphoff, and Elizabeth A. Barnes. Explainable artificial in- telligence in meteorology and climate science: Model fine- tuning, calibrating trust and learning new science. In An- dreas Holzinger, Randy Goebel, Ruth Fong, ...

    2022

  7. [2018]

    Concept bottleneck models

    [Kohet al., 2020 ] Pang Wei Koh, Thao Nguyen, Yew Siang Tang, Stephen Mussmann, Emma Pierson, Been Kim, and Percy Liang. Concept bottleneck models. InProceedings of the 37th International Conference on Machine Learning (ICML), pages 5338–5348. PMLR,

    2020

  8. [2019]

    Caravan: A global community dataset for large-sample hydrology.Sci- entific Data, 10(1):61,

    [Kratzertet al., 2023 ] Frederik Kratzert, Grey Nearing, Nans Addor, Tyler Erickson, Martin Gauch, Oren Gilon, Lukas Gudmundsson, Avinatan Hassidim, Daniel Klotz, Sella Nevo, Guy Shalev, and Yossi Matias. Caravan: A global community dataset for large-sample hydrology.Sci- enti...

    2023

  9. [2020]

    [Kratzertet al., 2019 ] Frederik Kratzert, Daniel Klotz, Guy Shalev, G ¨unter Klambauer, Sepp Hochreiter, and Grey Nearing. Towards learning universal, regional, and lo- cal hydrological behaviors via machine learning applied to large-sample datasets.Hydrology and Earth System...

    2019

  10. [2021]

    Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra

    [Selvarajuet al., 2017 ] Ramprasaath R. Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. Grad-cam: Visual explana- tions from deep networks via gradient-based localization. InProceedings of the IEEE International Conference on Comp...

    2017

  11. [2022]

    River flooding mechanisms and their changes in europe revealed by explainable machine learn- ing.Hydrology and Earth System Sciences, 26(24):6339– 6359,

    [Jianget al., 2022 ] Shijie Jiang, Emanuele Bevacqua, and Jakob Zscheischler. River flooding mechanisms and their changes in europe revealed by explainable machine learn- ing.Hydrology and Earth System Sciences, 26(24):6339– 6359,

    2022

  12. [2023]

    Probing the limit of hydrologic predictability with the transformer network.Journal of Hydrology, 637:131389,

    [Liuet al., 2024 ] Jiangtao Liu, Yuchen Bian, Kathryn Law- son, and Chaopeng Shen. Probing the limit of hydrologic predictability with the transformer network.Journal of Hydrology, 637:131389,

    2024

  13. [2024]

    Frame, Frederik Kratzert, Daniel Klotz, Martin Gauch, Guy Shalev, Oren Gilon, Lorne M

    [Frameet al., 2022 ] Jonathan M. Frame, Frederik Kratzert, Daniel Klotz, Martin Gauch, Guy Shalev, Oren Gilon, Lorne M. Qualls, Hoshin V . Gupta, and Grey S. Near- ing. Deep learning rainfall–runoff predictions of ex- treme events.Hydrology and Earth System Sciences, 26(13):3377–3392,

    2022

  14. [2025]

    Matin, Ghassan Beydoun, M

    [Dikshitet al., 2024 ] Abhirup Dikshit, Biswajeet Pradhan, Sahar S. Matin, Ghassan Beydoun, M. Santosh, Hyuck-Jin Park, and Khairul Nizam Abdul Maulud. Artificial intel- ligence: A new era for spatial modelling and interpreting climate-induced hazard assessment.Geoscience Fron...

    2024

Pith tools

Reviewed July 31, 2026 · model on record in the stance chip above.