Leveraging Teleconnections with Physics-Informed Graph Attention Networks for Long-Range Extreme Rainfall Forecasting in Thailand

Jirawan Kamma; Kanoksri Sarinnapakorn; Kiattikun Chobtham; Kritanai Torsri; Prattana Deeprasertkul

arxiv: 2510.12328 · v5 · submitted 2025-10-14 · 💻 cs.LG

Leveraging Teleconnections with Physics-Informed Graph Attention Networks for Long-Range Extreme Rainfall Forecasting in Thailand

Kiattikun Chobtham , Kanoksri Sarinnapakorn , Kritanai Torsri , Prattana Deeprasertkul , Jirawan Kamma This is my paper

Pith reviewed 2026-05-18 07:36 UTC · model grok-4.3

classification 💻 cs.LG

keywords graph neural networksextreme rainfall forecastingphysics-informed machine learningteleconnectionsThailand rainfallgeneralized Pareto distributionlong-range forecastingspatiotemporal modeling

0 comments

The pith

Physics-informed graph attention networks with teleconnections and orographic physics improve long-range extreme rainfall forecasts at Thai gauge stations.

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

The paper develops a graph neural network model that treats rainfall gauge stations as nodes connected by edges whose initial features come from a simple physical model of orographic precipitation plus preprocessed climate teleconnection indices. Attention mechanisms in the graph layers highlight relevant connections, after which LSTM layers process the resulting embeddings to produce rainfall predictions. A spatial season-aware generalized Pareto distribution then maps these outputs to extreme value statistics. If the approach holds, it would allow more reliable multi-month ahead forecasts of heavy rainfall events that support water resource planning and flood preparedness in monsoon-affected regions. A sympathetic reader would care because accurate long-range extreme rainfall information directly informs agricultural, infrastructure, and disaster decisions.

Core claim

The authors establish that a Graph Attention Network combined with LSTM layers, using initial edge features derived from an orographic-precipitation physics formulation and climate teleconnection indices, followed by a Spatial Season-aware Generalized Pareto Distribution for extremes, produces rainfall predictions that outperform well-established baselines across most regions and remain competitive with the state of the art, while delivering practical gains over an operational forecasting system in extreme-event accuracy and high-resolution mapping for water management.

What carries the argument

The physics-informed Graph Attention Network (Attention-LSTM) that initializes edge features from orographic-precipitation physics and teleconnection indices, applies attention to capture spatiotemporal patterns, processes embeddings with LSTM layers, and integrates a Spatial Season-aware Generalized Pareto Distribution to handle extremes.

If this is right

The model outperforms well-established baselines across most regions, including areas prone to extremes.
It remains strongly competitive with the state of the art in rainfall prediction accuracy.
Compared with an operational forecasting system, it improves extreme-event prediction at gauge stations.
The method supports production of high-resolution maps useful for long-term water management decisions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The teleconnection-based explainability could be used to test which specific climate indices most strongly influence extremes in particular Thai sub-regions.
Adapting the same graph construction to other monsoon-influenced countries might yield similar gains if local orographic and teleconnection data are available.
Coupling the high-resolution output maps with existing hydrological routing models could produce improved flood extent forecasts as a direct next step.
Season-aware modeling of extremes might reduce over- or under-estimation during specific monsoon phases compared with non-seasonal extreme value methods.

Load-bearing premise

Deriving initial edge features from a simple orographic-precipitation physics formulation together with preprocessed climate teleconnection indices will capture the dominant spatiotemporal drivers of extreme rainfall sufficiently for the model to generalize reliably beyond the training gauges.

What would settle it

A test on independent gauge stations or future periods where the model shows no improvement over standard baselines or the operational forecasting system in predicting the occurrence or magnitude of extreme rainfall events would falsify the central claim.

Figures

Figures reproduced from arXiv: 2510.12328 by Jirawan Kamma, Kanoksri Sarinnapakorn, Kiattikun Chobtham, Kritanai Torsri, Prattana Deeprasertkul.

**Figure 1.** Figure 1: Simulation of the simple thermodynamic equation of Smith linear model 2.2 Recurrent Neural Networks (RNNs) and Graph Neural Networks (GNNs) Recent advancements in forecasting long-term climate variability have increasingly relied on ML and deep learning models. Recurrent Neural Networks (RNNs), LSTM, and Gated Recurrent Unit (GRU) were employed for sequence-to-sequence prediction (Hochreiter and Schmidhube… view at source ↗

**Figure 2.** Figure 2: The Attention-LSTM architecture 3.1 Dynamic Attention Coefficients Our novel approach introduces attention coefficients (weight matrices) of GATs into stateof-the-art LSTM. Static edges (E) of graph structures will be constructed using the Pearson correlation as the statistical method, referred to as teleconnection. This served as feature selection for data-preprocessing in a ML process. The weight matrix… view at source ↗

read the original abstract

Accurate rainfall forecasting, particularly for extreme events, remains a significant challenge in climatology and the Earth system. This paper presents novel physics-informed Graph Neural Networks (GNNs) combined with extreme-value analysis techniques to improve gauge-station rainfall predictions across Thailand. The model leverages a graph-structured representation of gauge stations to capture complex spatiotemporal patterns, and it offers explainability through teleconnections. We preprocess relevant climate indices that potentially influence regional rainfall. The proposed Graph Attention Network with Long Short-Term Memory (Attention-LSTM) applies the attention mechanism using initial edge features derived from simple orographic-precipitation physics formulation. The embeddings are subsequently processed by LSTM layers. To address extremes, we perform Peak-Over-Threshold (POT) mapping using the novel Spatial Season-aware Generalized Pareto Distribution (GPD) method, which overcomes limitations of traditional machine-learning models. Experiments demonstrate that our method outperforms well-established baselines across most regions, including areas prone to extremes, and remains strongly competitive with the state of the art. Compared with the operational forecasting system SEAS5, our real-world application improves extreme-event prediction and offers a practical enhancement to produce high-resolution maps that support decision-making in long-term water management.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper assembles orographic physics edges, teleconnection indices, attention-LSTM, and a spatial-seasonal GPD for Thai extreme rainfall forecasts, but the outperformance claims over SEAS5 rest on unshown numbers and missing ablations.

read the letter

The key takeaway is that this paper applies a physics-informed graph attention LSTM to Thai rainfall gauges, initializing edges with orographic precipitation physics and teleconnection indices, then uses a novel spatial-seasonal GPD for extreme value modeling. It claims practical improvements over SEAS5 for extreme events, but the abstract leaves the actual numbers and tests out of sight. What is new is the particular assembly for this setting: the orographic edge features, the attention on top of them, the LSTM processing, and the custom GPD that varies by space and season. The authors do a good job linking it to real-world needs like flood preparedness and high-resolution mapping for water management in Thailand. The explainability angle through teleconnections is a nice touch for domain users. The soft spots are around verification. The central performance claims lack the supporting scores, error bars, or cross-validation details that would let us judge if the gains are reliable. There is no mention of ablations that would show whether the physics edges add value over purely learned ones, or how sensitive the results are to the choice of teleconnection indices. The stress-test concern about generalization beyond training gauges is fair; without those checks the assumption that the initial features capture dominant drivers carries a lot of weight. This paper is for researchers and practitioners in climate informatics and hydrology who work on extreme rainfall in monsoon climates. A reader looking for examples of blending physics priors with GNNs for spatiotemporal extremes would get value from the methods. I think it deserves a serious referee. The application is relevant and the synthesis is honest, even if the results need more scrutiny to be fully convincing.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes a physics-informed Graph Attention Network with LSTM layers (Attention-LSTM) for long-range extreme rainfall forecasting across Thailand. Gauge stations are represented as a graph whose initial edge features are derived from a simple orographic-precipitation physics formulation together with preprocessed climate teleconnection indices. LSTM embeddings feed into a novel Spatial Season-aware Generalized Pareto Distribution (GPD) applied via Peak-Over-Threshold mapping to model extremes. The paper claims that the resulting model outperforms established baselines in most regions, remains competitive with the state of the art, and improves extreme-event prediction relative to the operational SEAS5 system while enabling high-resolution maps for water-management decisions.

Significance. If the performance and generalization claims are substantiated, the work would offer a concrete demonstration of how physics-derived edge features and teleconnection indices can be embedded in graph attention models to improve long-range extreme-value forecasting in monsoon-dominated regions. The season-aware GPD component directly addresses a known limitation of standard ML regressors on tail events, and the emphasis on explainability via teleconnections is a constructive addition. Practical utility for high-resolution mapping in Thailand would be a tangible outcome, provided the central assumptions about dominant drivers are shown to hold.

major comments (3)

[Abstract] Abstract: the central performance claim that the method 'outperforms well-established baselines across most regions' and 'improves extreme-event prediction' relative to SEAS5 is stated without any numerical scores, error bars, cross-validation protocol, or statistical tests. This absence prevents assessment of whether the data actually support the headline result.
[Experiments] Experiments section: no ablation is reported that isolates the contribution of the orographic-precipitation-derived edge features and teleconnection indices versus purely data-driven alternatives. Because the generalization claim rests on the assumption that these physics-informed features capture the dominant spatiotemporal drivers, the lack of such a controlled comparison is load-bearing for both the outperformance statement and the improvement over SEAS5.
[Methods] Methods section: the Spatial Season-aware GPD is presented as overcoming limitations of traditional ML models, yet the precise mechanism by which season-awareness is encoded (e.g., season-dependent shape/scale parameters) and whether those parameters are estimated independently of the target rainfall series is not fully specified, leaving open the possibility of circularity.

minor comments (2)

[Model Architecture] The description of how the initial edge features are computed from the orographic-precipitation formulation would benefit from an explicit equation or pseudocode block.
[Results] Figure captions for the high-resolution maps should include direct side-by-side comparison metrics with SEAS5 to facilitate reader evaluation.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed review. The comments highlight important areas for clarification and strengthening of the manuscript. We address each major comment below and indicate the revisions we will make.

read point-by-point responses

Referee: [Abstract] Abstract: the central performance claim that the method 'outperforms well-established baselines across most regions' and 'improves extreme-event prediction' relative to SEAS5 is stated without any numerical scores, error bars, cross-validation protocol, or statistical tests. This absence prevents assessment of whether the data actually support the headline result.

Authors: We agree that the abstract would benefit from quantitative support for the performance claims. In the revised manuscript, we will add specific metrics (e.g., RMSE or CRPS values with error bars), reference the cross-validation protocol, and note any statistical significance tests used to substantiate the outperformance relative to baselines and SEAS5. revision: yes
Referee: [Experiments] Experiments section: no ablation is reported that isolates the contribution of the orographic-precipitation-derived edge features and teleconnection indices versus purely data-driven alternatives. Because the generalization claim rests on the assumption that these physics-informed features capture the dominant spatiotemporal drivers, the lack of such a controlled comparison is load-bearing for both the outperformance statement and the improvement over SEAS5.

Authors: We recognize that an explicit ablation study is necessary to isolate the impact of the physics-informed components. The current manuscript compares the full model to baselines but does not include this controlled experiment. In the revision, we will add an ablation analysis that removes the orographic-precipitation edge features and teleconnection indices in turn, reporting the resulting performance changes to demonstrate their contribution. revision: yes
Referee: [Methods] Methods section: the Spatial Season-aware GPD is presented as overcoming limitations of traditional ML models, yet the precise mechanism by which season-awareness is encoded (e.g., season-dependent shape/scale parameters) and whether those parameters are estimated independently of the target rainfall series is not fully specified, leaving open the possibility of circularity.

Authors: We thank the referee for identifying this ambiguity. In the revised Methods section, we will explicitly describe the encoding of season-awareness (via season-specific shape and scale parameters in the GPD) and confirm that all parameters are fitted on historical data held out from the target forecast periods, thereby removing any risk of circularity with the evaluation series. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation remains self-contained

full rationale

The paper derives initial edge features from an orographic-precipitation physics formulation and preprocessed teleconnection indices that are independent of the target rainfall variable. These feed the Attention-LSTM, whose embeddings are then mapped via POT to a season-aware GPD for extremes. No quoted equation or step reduces a claimed prediction to a fitted parameter or self-defined quantity by construction. External comparisons to baselines and SEAS5 further anchor the claims outside the model's fitted weights. Absent any load-bearing self-citation chain or uniqueness theorem imported from the authors' prior work, the central forecasting pipeline does not collapse into its inputs.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 1 invented entities

The central performance claim rests on the effectiveness of physics-derived graph edges and a newly introduced season-aware GPD; both introduce domain assumptions and one invented modeling entity whose independent validation is not shown in the abstract.

free parameters (2)

Attention-LSTM weights and biases
Learned from training data to map graph embeddings to rainfall predictions.
Spatial season-aware GPD shape and scale parameters
Estimated per season and location from peak-over-threshold rainfall values.

axioms (2)

domain assumption Relevant climate teleconnections influence regional rainfall patterns across Thailand
Invoked when preprocessing climate indices as model inputs.
domain assumption A simple orographic-precipitation formulation provides useful initial edge features for the graph
Used to construct the attention mechanism's starting edge attributes.

invented entities (1)

Spatial Season-aware Generalized Pareto Distribution no independent evidence
purpose: To model the upper tail of rainfall extremes while respecting spatial and seasonal heterogeneity
Introduced to overcome limitations of standard machine-learning models on extremes; no external falsifiable evidence supplied in the abstract.

pith-pipeline@v0.9.0 · 5771 in / 1715 out tokens · 59609 ms · 2026-05-18T07:36:31.805023+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

23 extracted references · 23 canonical work pages

[1]

provided insights into the statistical analysis of climate variables or climate indices, by examining historical trends and precipitation patterns at both global and regional scales in many regions. In the northern peninsular Malaysia area, ENSO exerts a strong influence on the timing and intensity of the monsoon (Moten et al., 2014), affecting global rai...

work page 2014
[2]

was proposed to learn a spatiotemporal continuous-time process governed by atmospheric physics. Research recommendations emphasize that data-driven ML and traditional physical modeling can complement each other in predicting extreme events (Saha et al., 2024; Camps-Valls, 2023). However, the ML models remain limited to short-term predictions, and many mod...

work page 2024
[3]

can help elucidate the relative contributions of various climate drivers and their interactions in modulating climate (Runge et al., 2019; Du et al., 2024), but a causal model must be represented by a directed acyclic graph and time-series prediction remains particularly challenging (Ravuri et al., 2021). In this work, we cast the long-term rainfall predi...

work page 2019
[4]

(𝑘,l) = 𝐶!𝑖𝜎ℎ

2 Related Work 2.1 Orographic Precipitation As mentioned in the introduction, various studies suggest integrating physics with data-driven ML. The simple thermodynamic equation of Smith linear model (Smith, 2003), governed by the physics of orographic precipitation, provides the spectral formulation that captures the basic structure of rainfall fields. Th...

work page 2003
[5]

Recurrent Neural Networks (RNNs), LSTM, and Gated Recurrent Unit (GRU) were employed for sequence-to-sequence prediction (Hochreiter and Schmidhuber, 1997; Cho et al., 2014)

Simulation of the simple thermodynamic equation of Smith linear model 2.2 Recurrent Neural Networks (RNNs) and Graph Neural Networks (GNNs) Recent advancements in forecasting long-term climate variability have increasingly relied on ML and deep learning models. Recurrent Neural Networks (RNNs), LSTM, and Gated Recurrent Unit (GRU) were employed for sequen...

work page 1997
[6]

Recently, GNNs are a well-established neural network architecture designed for processing graph-structured data

with LSTM (Wavelet-CNN-LSTM and Wavelet-LSTM) showed superior performance in monthly gauge-rainfall predictions compared to traditional models. Recently, GNNs are a well-established neural network architecture designed for processing graph-structured data. GNNs can take input in the form of a directed or undirected graph describing the connectivity struct...

work page 2025
[7]

GCNs were formulated for undirected graphs, operating on the normalised graph Laplacian

and LSTMs and proposed GC-LSTM to predict future links, achieving outstanding performance and outperforming state-of-the-art methods. GCNs were formulated for undirected graphs, operating on the normalised graph Laplacian. Their convolution is implemented using a localised first-order Chebyshev polynomial of the Laplacian. Pareja et al. (2020) proposed Ev...

work page 2020
[8]

#$ !%&∗𝑌(𝑡−𝑖)+&𝛽!

The Attention-LSTM architecture 3.1 Dynamic Attention Coefficients Our novel approach introduces attention coefficients (weight matrices) of GATs into state-of-the-art LSTM. Static edges (E) of graph structures will be constructed using the Pearson correlation as the statistical method, referred to as teleconnection. This served as feature selection for d...

work page 1964
[9]

5 Dataset and Experimental Setup 5.1 Dataset and Graph Structure Modeling We use seven climate indices detailed in Table 2, processed from NOAA OISST (Huang et al

Workflow of the operational long-range rainfall forecasts based on physics-informed Attention-LSTM and Spatial Season-aware GPD extreme value mapping. 5 Dataset and Experimental Setup 5.1 Dataset and Graph Structure Modeling We use seven climate indices detailed in Table 2, processed from NOAA OISST (Huang et al. 2021), together with historical rainfall d...

work page 2021
[10]

The MJO can be characterized as an eastward-moving pulse of cloud and rainfall near the equator that typically recurs every 30 to 60 days

in the South China Sea and Gulf of Thailand Real-time Multivariate MJO series (RMM1) The major fluctuation in tropical weather occurs on weekly to monthly timescales. The MJO can be characterized as an eastward-moving pulse of cloud and rainfall near the equator that typically recurs every 30 to 60 days. It is calculated using multivariate EOF (MV-EOF) of...

work page 1981
[11]

(2024) for 1982–2024

Climate features used in the study Following the spectral method of orographic precipitation (Subsection 2.1), we simulate orographic rainfall using each TMD gauge station elevation and ERA5 monthly average U and V wind components at 200 hPa, following the study of Saha et al. (2024) for 1982–2024. These serve as the Attention-LSTM physics-informed initia...

work page 2024
[12]

The selection of climate indices is based on feature-selection criteria using Pearson correlation

Each climate feature influences rainfall at individual stations through spatial teleconnections and temporal relationships. The selection of climate indices is based on feature-selection criteria using Pearson correlation. We construct a teleconnection when the average absolute correlation is higher than 0.4. We incorporate domain knowledge into the GNN e...

work page 2016
[13]

of equation (13) evaluated with the Chi2 test. We consider a teleconnection present when the test p-value is below a significance threshold of 0.1, indicating that the lagged influence of the teleconnection improves each station-level rainfall predictability for each cluster in Thailand. Stations The clusters and High-Quality (HQ) stations Training, valid...

work page 1982
[14]

We then present statistics of Spatial Season-aware GPD mapping in Subsection 6.2 and compare predictive skill with related work in Subsection 6.3

Empirical Results Based on dataset and the experimental setup in Section 5, we report the graph-based model results in Subsection 6.1. We then present statistics of Spatial Season-aware GPD mapping in Subsection 6.2 and compare predictive skill with related work in Subsection 6.3. Finally, Subsection 6.4 presents operational forecasts evaluated against th...

work page 1982
[15]

(𝑘,l) and (b) twelve-cluster areas for HII rainfall stations The spectral transform of precipitation 𝑃

Reanalysis graph-based structure (a) the spectral transform of the precipitation 𝑃"(𝑘,l) and (b) twelve-cluster areas for HII rainfall stations The spectral transform of precipitation 𝑃"(𝑘,l) for each TMD rainfall station, simulated using ERA5 U and V wind compenents for 1982–2024, is shown in Figure 4a. A clear relationship between precipitation and high...

work page 1982
[16]

model using Accuracy for each cluster of TMD stations. 6.4 Real-world Applications compared with ECMWF Seasonal Forecasts In ECMWF operational seasonal forecasting system from SEAS5, probabilities of precipitation anomalies or the mean of precipitation over 51 ensembles summarise the expected climate over the coming months from a statistical view of poten...

work page 2025
[17]

Both predictions are initialized in March 2025 with a six-month lead time, covering March through August

provides the highest average Accuracy across clusters (see details in Table 5). Both predictions are initialized in March 2025 with a six-month lead time, covering March through August

work page 2025
[18]

For a preliminary assessment, we evaluated the 74 gauge stations as sample grid points

Because the inputs and outputs of ECMWF SEAS5 and our proposed model are different, a direct comparison is not strictly fair. For a preliminary assessment, we evaluated the 74 gauge stations as sample grid points. For monthly RMSE at these stations for March–August 2025, Attention-LSTM (Q =

work page 2025
[19]

(RMSE = 101.93 mm/month, average rainfall = 234.23 mm/month). 7 Conclusion and Future work This work presents an interpretable graph-based model for long-range rainfall forecasting at gauge stations across Thailand, along with the forecasting system for real-world application. The proposed Attention-LSTM is constructed using statistical feature selection ...

work page doi:10.3389/fclim.2022.1031226 2023
[20]

Camps-Valls, G., Gerhardus, A., Ninad, U., Varando, G., Martius, G., Balaguer-Ballester, E., Vinuesa, R., Diaz, E., Zanna, L., & Runge, J. (2023). Discovering causal relations and equations from data. Physics Reports, 1044, 1–68. https://doi.org/10.1016/j.physrep.2023.10.005 Chang, C.-H., Johnson, N. C., & Yoo, C. (2021). Evaluation of subseasonal impacts...

work page doi:10.1016/j.physrep.2023.10.005 2023
[21]

Saha, A., & Ravela, S. (2024). Statistical-physical adversarial learning from data and models for downscaling rainfall extremes. Journal of Advances in Modeling Earth Systems, 16, e2023MS003860. https://doi.org/10.1029/2023MS003860 Scarselli, F., Tsoi, A. C., & Hagenbuchner, M. (2018). The Vapnik–Chervonenkis dimension of graph and recursive neural networ...

work page doi:10.1029/2023ms003860 2024
[22]

https://doi.org/10.3390/app15168854 Smith, R. B. (2003). A linear upslope-time-delay model for orographic precipitation. Journal of Hydrology, 282(1–4), 2–9. https://doi.org/10.1016/S0022-1694(03)00248-8 Sun, Q., Zhang, X., Zwiers, F., Westra, S., & Alexander, L. V. (2021). A Global, Continental, and Regional Analysis of Changes in Extreme Precipitation. ...

work page doi:10.3390/app15168854 2003
[23]

https://doi.org/10.3390/w15162979 Waqas, M., Wannasingha Humphries, U., Chueasa, B., & Wangwongchai, A. (2025). Artificial intelligence and numerical weather prediction models: A technical survey. Natural Hazards Research, 5(2), 306–320. https://doi.org/10.1016/j.nhres.2024.11.004 Wu, T., Chen, F., & Wan, Y. (2018). Graph attention LSTM network: A new mod...

work page doi:10.3390/w15162979 2025

[1] [1]

provided insights into the statistical analysis of climate variables or climate indices, by examining historical trends and precipitation patterns at both global and regional scales in many regions. In the northern peninsular Malaysia area, ENSO exerts a strong influence on the timing and intensity of the monsoon (Moten et al., 2014), affecting global rai...

work page 2014

[2] [2]

was proposed to learn a spatiotemporal continuous-time process governed by atmospheric physics. Research recommendations emphasize that data-driven ML and traditional physical modeling can complement each other in predicting extreme events (Saha et al., 2024; Camps-Valls, 2023). However, the ML models remain limited to short-term predictions, and many mod...

work page 2024

[3] [3]

can help elucidate the relative contributions of various climate drivers and their interactions in modulating climate (Runge et al., 2019; Du et al., 2024), but a causal model must be represented by a directed acyclic graph and time-series prediction remains particularly challenging (Ravuri et al., 2021). In this work, we cast the long-term rainfall predi...

work page 2019

[4] [4]

(𝑘,l) = 𝐶!𝑖𝜎ℎ

2 Related Work 2.1 Orographic Precipitation As mentioned in the introduction, various studies suggest integrating physics with data-driven ML. The simple thermodynamic equation of Smith linear model (Smith, 2003), governed by the physics of orographic precipitation, provides the spectral formulation that captures the basic structure of rainfall fields. Th...

work page 2003

[5] [5]

Recurrent Neural Networks (RNNs), LSTM, and Gated Recurrent Unit (GRU) were employed for sequence-to-sequence prediction (Hochreiter and Schmidhuber, 1997; Cho et al., 2014)

Simulation of the simple thermodynamic equation of Smith linear model 2.2 Recurrent Neural Networks (RNNs) and Graph Neural Networks (GNNs) Recent advancements in forecasting long-term climate variability have increasingly relied on ML and deep learning models. Recurrent Neural Networks (RNNs), LSTM, and Gated Recurrent Unit (GRU) were employed for sequen...

work page 1997

[6] [6]

Recently, GNNs are a well-established neural network architecture designed for processing graph-structured data

with LSTM (Wavelet-CNN-LSTM and Wavelet-LSTM) showed superior performance in monthly gauge-rainfall predictions compared to traditional models. Recently, GNNs are a well-established neural network architecture designed for processing graph-structured data. GNNs can take input in the form of a directed or undirected graph describing the connectivity struct...

work page 2025

[7] [7]

GCNs were formulated for undirected graphs, operating on the normalised graph Laplacian

and LSTMs and proposed GC-LSTM to predict future links, achieving outstanding performance and outperforming state-of-the-art methods. GCNs were formulated for undirected graphs, operating on the normalised graph Laplacian. Their convolution is implemented using a localised first-order Chebyshev polynomial of the Laplacian. Pareja et al. (2020) proposed Ev...

work page 2020

[8] [8]

#$ !%&∗𝑌(𝑡−𝑖)+&𝛽!

The Attention-LSTM architecture 3.1 Dynamic Attention Coefficients Our novel approach introduces attention coefficients (weight matrices) of GATs into state-of-the-art LSTM. Static edges (E) of graph structures will be constructed using the Pearson correlation as the statistical method, referred to as teleconnection. This served as feature selection for d...

work page 1964

[9] [9]

5 Dataset and Experimental Setup 5.1 Dataset and Graph Structure Modeling We use seven climate indices detailed in Table 2, processed from NOAA OISST (Huang et al

Workflow of the operational long-range rainfall forecasts based on physics-informed Attention-LSTM and Spatial Season-aware GPD extreme value mapping. 5 Dataset and Experimental Setup 5.1 Dataset and Graph Structure Modeling We use seven climate indices detailed in Table 2, processed from NOAA OISST (Huang et al. 2021), together with historical rainfall d...

work page 2021

[10] [10]

The MJO can be characterized as an eastward-moving pulse of cloud and rainfall near the equator that typically recurs every 30 to 60 days

in the South China Sea and Gulf of Thailand Real-time Multivariate MJO series (RMM1) The major fluctuation in tropical weather occurs on weekly to monthly timescales. The MJO can be characterized as an eastward-moving pulse of cloud and rainfall near the equator that typically recurs every 30 to 60 days. It is calculated using multivariate EOF (MV-EOF) of...

work page 1981

[11] [11]

(2024) for 1982–2024

Climate features used in the study Following the spectral method of orographic precipitation (Subsection 2.1), we simulate orographic rainfall using each TMD gauge station elevation and ERA5 monthly average U and V wind components at 200 hPa, following the study of Saha et al. (2024) for 1982–2024. These serve as the Attention-LSTM physics-informed initia...

work page 2024

[12] [12]

The selection of climate indices is based on feature-selection criteria using Pearson correlation

Each climate feature influences rainfall at individual stations through spatial teleconnections and temporal relationships. The selection of climate indices is based on feature-selection criteria using Pearson correlation. We construct a teleconnection when the average absolute correlation is higher than 0.4. We incorporate domain knowledge into the GNN e...

work page 2016

[13] [13]

of equation (13) evaluated with the Chi2 test. We consider a teleconnection present when the test p-value is below a significance threshold of 0.1, indicating that the lagged influence of the teleconnection improves each station-level rainfall predictability for each cluster in Thailand. Stations The clusters and High-Quality (HQ) stations Training, valid...

work page 1982

[14] [14]

We then present statistics of Spatial Season-aware GPD mapping in Subsection 6.2 and compare predictive skill with related work in Subsection 6.3

Empirical Results Based on dataset and the experimental setup in Section 5, we report the graph-based model results in Subsection 6.1. We then present statistics of Spatial Season-aware GPD mapping in Subsection 6.2 and compare predictive skill with related work in Subsection 6.3. Finally, Subsection 6.4 presents operational forecasts evaluated against th...

work page 1982

[15] [15]

(𝑘,l) and (b) twelve-cluster areas for HII rainfall stations The spectral transform of precipitation 𝑃

Reanalysis graph-based structure (a) the spectral transform of the precipitation 𝑃"(𝑘,l) and (b) twelve-cluster areas for HII rainfall stations The spectral transform of precipitation 𝑃"(𝑘,l) for each TMD rainfall station, simulated using ERA5 U and V wind compenents for 1982–2024, is shown in Figure 4a. A clear relationship between precipitation and high...

work page 1982

[16] [16]

model using Accuracy for each cluster of TMD stations. 6.4 Real-world Applications compared with ECMWF Seasonal Forecasts In ECMWF operational seasonal forecasting system from SEAS5, probabilities of precipitation anomalies or the mean of precipitation over 51 ensembles summarise the expected climate over the coming months from a statistical view of poten...

work page 2025

[17] [17]

Both predictions are initialized in March 2025 with a six-month lead time, covering March through August

provides the highest average Accuracy across clusters (see details in Table 5). Both predictions are initialized in March 2025 with a six-month lead time, covering March through August

work page 2025

[18] [18]

For a preliminary assessment, we evaluated the 74 gauge stations as sample grid points

Because the inputs and outputs of ECMWF SEAS5 and our proposed model are different, a direct comparison is not strictly fair. For a preliminary assessment, we evaluated the 74 gauge stations as sample grid points. For monthly RMSE at these stations for March–August 2025, Attention-LSTM (Q =

work page 2025

[19] [19]

(RMSE = 101.93 mm/month, average rainfall = 234.23 mm/month). 7 Conclusion and Future work This work presents an interpretable graph-based model for long-range rainfall forecasting at gauge stations across Thailand, along with the forecasting system for real-world application. The proposed Attention-LSTM is constructed using statistical feature selection ...

work page doi:10.3389/fclim.2022.1031226 2023

[20] [20]

Camps-Valls, G., Gerhardus, A., Ninad, U., Varando, G., Martius, G., Balaguer-Ballester, E., Vinuesa, R., Diaz, E., Zanna, L., & Runge, J. (2023). Discovering causal relations and equations from data. Physics Reports, 1044, 1–68. https://doi.org/10.1016/j.physrep.2023.10.005 Chang, C.-H., Johnson, N. C., & Yoo, C. (2021). Evaluation of subseasonal impacts...

work page doi:10.1016/j.physrep.2023.10.005 2023

[21] [21]

Saha, A., & Ravela, S. (2024). Statistical-physical adversarial learning from data and models for downscaling rainfall extremes. Journal of Advances in Modeling Earth Systems, 16, e2023MS003860. https://doi.org/10.1029/2023MS003860 Scarselli, F., Tsoi, A. C., & Hagenbuchner, M. (2018). The Vapnik–Chervonenkis dimension of graph and recursive neural networ...

work page doi:10.1029/2023ms003860 2024

[22] [22]

https://doi.org/10.3390/app15168854 Smith, R. B. (2003). A linear upslope-time-delay model for orographic precipitation. Journal of Hydrology, 282(1–4), 2–9. https://doi.org/10.1016/S0022-1694(03)00248-8 Sun, Q., Zhang, X., Zwiers, F., Westra, S., & Alexander, L. V. (2021). A Global, Continental, and Regional Analysis of Changes in Extreme Precipitation. ...

work page doi:10.3390/app15168854 2003

[23] [23]

https://doi.org/10.3390/w15162979 Waqas, M., Wannasingha Humphries, U., Chueasa, B., & Wangwongchai, A. (2025). Artificial intelligence and numerical weather prediction models: A technical survey. Natural Hazards Research, 5(2), 306–320. https://doi.org/10.1016/j.nhres.2024.11.004 Wu, T., Chen, F., & Wan, Y. (2018). Graph attention LSTM network: A new mod...

work page doi:10.3390/w15162979 2025