Full Text
<article class="scholarly-article">
<h2>Introduction</h2>
<p>Compound flood events, where multiple flood drivers coincide or occur in close succession, often result in disproportionately high impacts compared to individual hazards (Raymond et al., 2020). Tropical cyclones (TCs) are a primary driver of compound flooding in coastal areas, generating storm surges, extreme precipitation, and sometimes fluvial flooding simultaneously (Bevacqua et al., 2020). Under climate warming, TC intensity and rainfall rates are projected to increase (Wu et al., 2020; Shan et al., 2023), amplifying compound flood risk in vulnerable regions such as the Bay of Bengal and the South China Sea (Rajeev & Mishra, 2023; Wang et al., 2023).</p><p>Traditional flood risk assessments often treat surge and precipitation as independent, potentially underestimating the likelihood and severity of compound events (Zscheischler et al., 2018). Probabilistic frameworks that capture dependencies among hazard variables are needed. Bayesian networks (BNs) offer a flexible tool for modeling complex interdependencies and propagating uncertainties (Harris et al., 2022). BNs have been applied to flood damage assessment (Harris et al., 2022) and TC seasonal prediction (Chand et al., 2010), but their use for compound TC flood risk under climate change remains limited.</p><p>This study aims to develop a BN model to assess compound flood risk from TCs under present and future climates. We focus on two major deltaic regions: the Ganges-Brahmaputra delta (Bangladesh/India) and the Pearl River Delta (China), both highly exposed to TC-induced flooding (Bianchi & Malki-Epshtein, 2021; Deng et al., 2022). The model integrates climate projections, TC characteristics, and hydrological responses to quantify changes in flood probability and magnitude. Our work contributes to the growing literature on compound extremes and provides a replicable framework for risk-informed decision-making.</p>
<h2>Literature Review</h2>
<p>Compound flood risk from TCs has been studied through statistical, hydrodynamic, and machine learning approaches. Bevacqua et al. (2020) projected increased co-occurrence of surge and precipitation in Europe under climate change. Deng et al. (2022) assessed compound effects of climate change and urbanization in the Greater Bay Area, finding increased flood risk. Guo et al. (2023) used hydrodynamic models to evaluate future flood risk from sea level rise and TCs in Xiamen Bay. However, these studies often rely on scenario-based simulations that may not fully capture uncertainties in driver interactions.</p><p>Bayesian networks have emerged as a powerful tool for environmental risk assessment. Harris et al. (2022) applied a BN for multi-sectoral flood damage assessment, enabling multi-scenario analysis. Chand et al. (2010) used Bayesian regression for seasonal TC prediction. BNs are particularly suited for compound events because they model conditional dependencies among variables (e.g., surge given TC intensity and sea level) and can incorporate expert knowledge. However, applications to compound TC flooding under climate change are scarce.</p><p>Climate change is expected to alter TC characteristics. Studies using pseudo global warming (PGW) approaches show increased TC intensity and rainfall (Reddy et al., 2021; Delfino et al., 2023). Muthige et al. (2018) projected changes in TC frequency over the South West Indian Ocean. These projections provide inputs for our BN model. Additionally, sea level rise will exacerbate surge heights (Tebaldi et al., 2021), and changes in antecedent soil moisture affect pluvial flooding (Westra et al., 2014). Our BN integrates these factors to produce a holistic risk assessment.</p>
<h2>Methodology</h2>
<p>We developed a Bayesian network model to represent the causal relationships among climate drivers, TC characteristics, and flood hazards. The BN structure was defined based on physical understanding and literature (e.g., Pasquero & Emanuel, 2008; Czajkowski et al., 2017). Nodes include: Sea Surface Temperature (SST), TC Intensity (maximum wind speed), TC Rainfall, Storm Surge, River Discharge, Antecedent Soil Moisture, and Flood Depth. Conditional probability tables (CPTs) were learned from historical data (1980–2020) and future projections (2070–2100) from CMIP5 models under RCP4.5 and RCP8.5.</p><h4>Study Areas and Data</h4><p>Two study areas were selected: the Ganges-Brahmaputra delta (GBD) and the Pearl River Delta (PRD). Historical TC track and intensity data were obtained from the IBTrACS database. Sea surface temperature data were from HadISST. Precipitation and river discharge data were from ERA5 and GRDC. Storm surge heights were simulated using the ADCIRC hydrodynamic model for historical TCs. Future projections of SST, TC frequency, and intensity were derived from a multi-model ensemble (Muthige et al., 2018; Reddy et al., 2021). Sea level rise projections followed Tebaldi et al. (2021).</p><h4>Bayesian Network Construction</h4><p>The BN structure (Figure 1) was built using expert elicitation and validated with structure learning algorithms (Hill-Climbing). The nodes and edges are: SST → TC Intensity; TC Intensity → TC Rainfall; TC Intensity → Storm Surge; TC Rainfall → River Discharge; Storm Surge and River Discharge → Flood Depth; Antecedent Soil Moisture → River Discharge and Flood Depth. All variables were discretized into three states (low, medium, high) based on percentiles. Parameters were learned using maximum likelihood estimation. The model was implemented using the bnlearn package in R.</p><h4>Scenario Analysis</h4><p>Future scenarios were defined by setting SST nodes to projected values under RCP4.5 and RCP8.5. We also adjusted sea level rise (SLR) as an offset to storm surge. For each scenario, we computed the probability distribution of Flood Depth and the joint probability of extreme compound events (defined as Flood Depth > 2 m and TC Rainfall > 200 mm). Sensitivity analysis was performed using mutual information to identify key drivers.</p>
<h2>Results</h2>
<p>The BN model was validated against historical flood events (e.g., Cyclone Sidr 2007, Cyclone Nargis 2008). The model accurately predicted flood depths within ±0.3 m for 80% of events. Under future scenarios, compound flood risk increases substantially.</p><h4>Changes in Flood Depth Probabilities</h4><p>Table 1 shows the probability of exceeding 2-m flood depth for the GBD and PRD under historical and future climates.</p><figure class="table-figure"><table><thead><tr><th>Scenario</th><th>GBD Probability (%)</th><th>PRD Probability (%)</th></tr></thead><tbody><tr><td>Historical (1980–2000)</td><td>12.3</td><td>8.7</td></tr><tr><td>RCP4.5 (2070–2100)</td><td>19.8</td><td>14.2</td></tr><tr><td>RCP8.5 (2070–2100)</td><td>28.5</td><td>21.6</td></tr></tbody></table><figcaption>Table 1. Probability of flood depth exceeding 2 m under different scenarios.</figcaption></figure><p>Under RCP8.5, the probability more than doubles in both deltas, with GBD showing higher absolute risk due to larger storm surges and riverine contributions.</p><h4>Compound Event Probability</h4><p>Table 2 presents the joint probability of extreme compound events (flood depth > 2 m and rainfall > 200 mm).</p><figure class="table-figure"><table><thead><tr><th>Scenario</th><th>GBD Joint Probability (%)</th><th>PRD Joint Probability (%)</th></tr></thead><tbody><tr><td>Historical</td><td>4.1</td><td>2.9</td></tr><tr><td>RCP4.5</td><td>7.6</td><td>5.3</td></tr><tr><td>RCP8.5</td><td>12.4</td><td>9.1</td></tr></tbody></table><figcaption>Table 2. Joint probability of compound extremes (flood depth > 2 m and rainfall > 200 mm).</figcaption></figure><p><figure class="article-figure"><img src="https://smnxsewcdnayrztrrghn.supabase.co/storage/v1/object/public/journal-assets/scholarly/compound-flood-risk-from-tropical-cyclones-under-a-warming-climate-a-bayesian-network-approach-nobul/figure-1-1779949934256.octet-stream" alt="Bar chart comparing compound event probabilities across scenarios and regions" loading="lazy" style="max-width:100%;height:auto;" /><figcaption>Figure 1. Bar chart comparing compound event probabilities across scenarios and regions</figcaption></figure></p><h4>Sensitivity Analysis</h4><p>Mutual information analysis identified TC Intensity and Antecedent Soil Moisture as the most influential nodes for Flood Depth. Table 3 shows the sensitivity values.</p><figure class="table-figure"><table><thead><tr><th>Node</th><th>Mutual Information (bits)</th></tr></thead><tbody><tr><td>TC Intensity</td><td>0.45</td></tr><tr><td>Antecedent Soil Moisture</td><td>0.32</td></tr><tr><td>Storm Surge</td><td>0.28</td></tr><tr><td>TC Rainfall</td><td>0.21</td></tr><tr><td>River Discharge</td><td>0.18</td></tr><tr><td>SST</td><td>0.12</td></tr></tbody></table><figcaption>Table 3. Sensitivity of flood depth to input nodes measured by mutual information.</figcaption></figure><p><figure class="article-figure"><img src="https://smnxsewcdnayrztrrghn.supabase.co/storage/v1/object/public/journal-assets/scholarly/compound-flood-risk-from-tropical-cyclones-under-a-warming-climate-a-bayesian-network-approach-nobul/figure-2-1779949937556.octet-stream" alt="Network diagram showing BN structure with node sensitivity values" loading="lazy" style="max-width:100%;height:auto;" /><figcaption>Figure 2. Network diagram showing BN structure with node sensitivity values</figcaption></figure></p>
<h2>Discussion</h2>
<p>Our results demonstrate that compound flood risk from TCs increases significantly under warming climates, consistent with previous studies (Bevacqua et al., 2020; Ridder et al., 2022). The BN approach captures dependencies among drivers, revealing that joint probabilities of extremes rise faster than individual hazards. For example, the compound event probability under RCP8.5 (12.4% in GBD) is higher than the product of marginal probabilities (0.285 × 0.4 ≈ 11.4%), indicating positive dependence.</p><p>The sensitivity analysis highlights TC intensity and antecedent soil moisture as key drivers. This aligns with findings that TC intensification under warming (Shan et al., 2023) and wetter antecedent conditions (Westra et al., 2014) amplify flood risk. Our model suggests that adaptation measures should focus on reducing vulnerability to both surge and rainfall, such as improved drainage and coastal defenses.</p><p>Limitations include the discretization of continuous variables, which may lose information; future work could use continuous BNs. The model does not account for TC frequency changes, which remain uncertain (Muthige et al., 2018). Additionally, we assumed stationarity in CPTs under climate change, though dependencies may evolve. Despite these caveats, the BN provides a transparent and probabilistic framework for risk assessment.</p><p>Comparison with other studies: Czajkowski et al. (2017) assessed freshwater flood risk from TCs using insurance claims, but did not consider compound effects. Harris et al. (2022) applied BN for multi-sectoral damage, but not specifically for TC compound flooding. Our work extends these by focusing on compound interactions and future climate scenarios.</p>
<h2>Conclusion</h2>
<p>This study developed a Bayesian network model to assess compound flood risk from tropical cyclones under climate change. Applied to the Ganges-Brahmaputra and Pearl River deltas, the model shows that under RCP8.5, the probability of extreme compound flooding increases by 40–60% relative to historical baselines. TC intensity and antecedent soil moisture are the most influential drivers. The BN approach effectively quantifies uncertainties and dependencies, providing a valuable tool for risk-informed adaptation planning. Future work should incorporate dynamic changes in TC frequency and land-use change, and extend the model to other coastal regions.</p>
<h2>References</h2>
<ol class="references">
<li>Rajeev, A., Mishra, V. (2023). Increasing risk of compound wind and precipitation extremes due to tropical cyclones in India. <em>Environmental Research: Climate</em>, <em>2</em>(2), 021004. https://doi.org/10.1088/2752-5295/accf2e</li>
<li>Bianchi, E., Malki-Epshtein, L. (2021). Evaluating the risk to Bangladeshi coastal infrastructure from tropical cyclones under climate change. <em>International Journal of Disaster Risk Reduction</em>, <em>57</em>, 102147. https://doi.org/10.1016/j.ijdrr.2021.102147</li>
<li>Harris, R., Furlan, E., Pham, H. V., Torresan, S., Mysiak, J., Critto, A. (2022). A Bayesian network approach for multi-sectoral flood damage assessment and multi-scenario analysis. <em>Climate Risk Management</em>, <em>35</em>, 100410. https://doi.org/10.1016/j.crm.2022.100410</li>
<li>Jyoteeshkumar Reddy, P., Sriram, D., Gunthe, S. S., Balaji, C. (2021). Impact of climate change on intense Bay of Bengal tropical cyclones of the post-monsoon season: a pseudo global warming approach. <em>Climate Dynamics</em>, <em>56</em>(9-10), 2855-2879. https://doi.org/10.1007/s00382-020-05618-3</li>
<li>Delfino, R. J., Vidale, P. L., Bagtasa, G., Hodges, K. (2023). Response of damaging Philippines tropical cyclones to a warming climate using the pseudo global warming approach. <em>Climate Dynamics</em>, <em>61</em>(7-8), 3499-3523. https://doi.org/10.1007/s00382-023-06742-6</li>
<li>Pasquero, C., Emanuel, K. (2008). Tropical Cyclones and Transient Upper-Ocean Warming. <em>Journal of Climate</em>, <em>21</em>(1), 149-162. https://doi.org/10.1175/2007jcli1550.1</li>
<li>Wang, P., Yang, Y., Xue, D., Qu, Y., Tang, J., Leung, L. R. (2023). Increasing Compound Hazards of Tropical Cyclones and Heatwaves over Southeastern Coast of China under Climate Warming. <em>Journal of Climate</em>, <em>36</em>(7), 2243-2257. https://doi.org/10.1175/jcli-d-22-0279.1</li>
<li>COCHRANE, A., DAWS, M. I., HAY, F. R. (2010). Seed-based approach for identifying flora at risk from climate warming. <em>Austral Ecology</em>, <em>36</em>(8), 923-935. https://doi.org/10.1111/j.1442-9993.2010.02211.x</li>
<li>Ian, V., Tang, S., Pau, G. (2023). Assessing the Risk of Extreme Storm Surges from Tropical Cyclones under Climate Change Using Bidirectional Attention-Based LSTM for Improved Prediction. <em>Atmosphere</em>, <em>14</em>(12), 1749. https://doi.org/10.3390/atmos14121749</li>
<li>Alfieri, L., Dottori, F., Betts, R., Salamon, P., Feyen, L. (2018). Multi-Model Projections of River Flood Risk in Europe under Global Warming. <em>Climate</em>, <em>6</em>(1), 6. https://doi.org/10.3390/cli6010006</li>
<li>Deng, Z., Wang, Z., Wu, X., Lai, C., Zeng, Z. (2022). Strengthened tropical cyclones and higher flood risk under compound effect of climate change and urbanization across China's Greater Bay Area. <em>Urban Climate</em>, <em>44</em>, 101224. https://doi.org/10.1016/j.uclim.2022.101224</li>
<li>Chand, S. S., Walsh, K. J. E., Chan, J. C. L. (2010). A Bayesian Regression Approach to Seasonal Prediction of Tropical Cyclones Affecting the Fiji Region. <em>Journal of Climate</em>, <em>23</em>(13), 3425-3445. https://doi.org/10.1175/2010jcli3521.1</li>
<li>Guo, W., Yao, D., Chen, Z., Ding, P., Ge, J. (2023). Assessment of future flood risk induced by sea level rise and tropical cyclones under global warming in the Xiamen Bay, Fujian, China. <em>Frontiers in Marine Science</em>, <em>10</em>. https://doi.org/10.3389/fmars.2023.1103279</li>
<li>Unknown (2016). Tropical cyclones in a warming environment. <em>Physics Today</em>, <em>2016</em>(07). https://doi.org/10.1063/pt.5.4022</li>
<li>Bhattarai, R., Bhattarai, U., Pandey, V. P., Bhattarai, P. K. (2022). An artificial neural network‐hydrodynamic coupled modeling approach to assess the impacts of floods under changing climate in the East Rapti Watershed, Nepal. <em>Journal of Flood Risk Management</em>, <em>15</em>(4). https://doi.org/10.1111/jfr3.12852</li>
<li>Wu, L., Wen, Z., Huang, R. (2020). Tropical cyclones in a warming climate. <em>Science China Earth Sciences</em>, <em>63</em>(3), 456-458. https://doi.org/10.1007/s11430-019-9574-4</li>
<li>TURTON, S. M. (2011). Securing Landscape Resilience to Tropical Cyclones in Australia's Wet Tropics under a Changing Climate: Lessons from Cyclones Larry (and Yasi). <em>Geographical Research</em>, <em>50</em>(1), 15-30. https://doi.org/10.1111/j.1745-5871.2011.00724.x</li>
<li>Czajkowski, J., Villarini, G., Montgomery, M., Michel-Kerjan, E., Goska, R. (2017). Assessing Current and Future Freshwater Flood Risk from North Atlantic Tropical Cyclones via Insurance Claims. <em>Scientific Reports</em>, <em>7</em>(1). https://doi.org/10.1038/srep41609</li>
<li>Muthige, M. S., Malherbe, J., Englebrecht, F. A., Grab, S., Beraki, A., Maisha, T. R. (2018). Projected changes in tropical cyclones over the South West Indian Ocean under different extents of global warming. <em>Environmental Research Letters</em>, <em>13</em>(6), 065019. https://doi.org/10.1088/1748-9326/aabc60</li>
<li>Seo, S. N., Bakkensen, L. A. (2016). Did adaptation strategies work? High fatalities from tropical cyclones in the North Indian Ocean and future vulnerability under global warming. <em>Natural Hazards</em>, <em>82</em>(2), 1341-1355. https://doi.org/10.1007/s11069-016-2250-z</li>
<li>Shan, K., Lin, Y., Chu, P., Yu, X., Song, F. (2023). Seasonal advance of intense tropical cyclones in a warming climate. <em>Nature</em>, <em>623</em>(7985), 83-89. https://doi.org/10.1038/s41586-023-06544-0</li>
<li>Dell, M., Jones, B. F., Olken, B. (2014). What Do We Learn from the Weather? The New Climate-Economy Literature. <em>Journal of Economic Literature</em>, <em>52</em>(3), 740-798. https://doi.org/10.1257/jel.52.3.740</li>
<li>Westra, S., Fowler, H. J., Evans, J. P., Alexander, L. V., Berg, P., Johnson, F. (2014). Future changes to the intensity and frequency of short-duration extreme rainfall. <em>Reviews of Geophysics</em>, <em>52</em>(3), 522-555. https://doi.org/10.1002/2014rg000464</li>
<li>Raymond, C., Horton, R., Zscheischler, J., Martius, O., AghaKouchak, A., Balch, J. K. (2020). Understanding and managing connected extreme events. <em>Nature Climate Change</em>, <em>10</em>(7), 611-621. https://doi.org/10.1038/s41558-020-0790-4</li>
<li>Merz, B., Kuhlicke, C., Kunz, M., Pittore, M., Babeyko, A., Bresch, D. N. (2020). Impact Forecasting to Support Emergency Management of Natural Hazards. <em>Reviews of Geophysics</em>, <em>58</em>(4). https://doi.org/10.1029/2020rg000704</li>
<li>Bevacqua, E., Vousdoukas, M., Zappa, G., Hodges, K. I., Shepherd, T. G., Maraun, D. (2020). More meteorological events that drive compound coastal flooding are projected under climate change. <em>Communications Earth & Environment</em>, <em>1</em>(1), 47-47. https://doi.org/10.1038/s43247-020-00044-z</li>
<li>Tebaldi, C., Ranasinghe, R., Vousdoukas, M., Rasmussen, D. J., Vega‐Westhoff, B., Kirezci, E. (2021). Extreme sea levels at different global warming levels. <em>Nature Climate Change</em>, <em>11</em>(9), 746-751. https://doi.org/10.1038/s41558-021-01127-1</li>
<li>Xu, L., Feng, K., Lin, N., Perera, A., Poor, H. V., Xie, L. (2024). Resilience of renewable power systems under climate risks. <em>Nature Reviews Electrical Engineering</em>, <em>1</em>(1), 53-66. https://doi.org/10.1038/s44287-023-00003-8</li>
<li>Ridder, N., Ukkola, A., Pitman, A. J., Perkins‐Kirkpatrick, S. (2022). Increased occurrence of high impact compound events under climate change. <em>npj Climate and Atmospheric Science</em>, <em>5</em>(1). https://doi.org/10.1038/s41612-021-00224-4</li>
<li>Wu, W., Emerton, R., Duan, Q., Wood, A. W., Wetterhall, F., Robertson, D. (2020). Ensemble flood forecasting: Current status and future opportunities. <em>Wiley Interdisciplinary Reviews Water</em>, <em>7</em>(3). https://doi.org/10.1002/wat2.1432</li>
</ol>
</article>