A Novel Non-Stationary Bias-Correction Framework for Hurricane Risk Assessment under Climate Change
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.
Kai: We're looking at the paper: "A Novel Non-Stationary Bias-Correction Framework for Hurricane Risk Assessment under Climate Change".
Mira: Here's the gist of the paper. A novel dynamic bias correction framework is presented to enhance hurricane track simulations under climate change by learning a mapping between current and future Global Climate Model outputs <ref:2505.00832#pg2> This approach addresses the non-stationarity of climate biases, providing more accurate and relevant adjustments for future scenarios <ref:2505.00832#pg2>.
Kai: Framework Overview The proposed methodology introduces a novel machine learning approach to address the non-stationarity of climate biases by learning a mapping between current and future Global Climate Model outputs <ref:2505.00832#pg2> This method differs from traditional bias correction by focusing on predicting future biases, making it more suitable for long-term climate projections under climate change <ref:2505.00832#pg2>.
Mira: The framework involves two main stages to address the non-stationary bias issue and improve extreme event predictions <ref:2505.00832#pg2>.
Kai: Data Processing Steps The data-driven bias-correction framework employs a two-step process involving dimensionality reduction and surrogate modeling <ref:2505.00832#pg2> 1.
Paper summary: Kai: To summarize this paper, they are presenting a novel dynamic bias correction framework aimed at improving hurricane track simulations when dealing with climate change projections. The central thesis is that conventional methods fail because Global Climate Models have biases that are not constant; they evolve as the climate changes over time.
Mira: They claim their approach addresses this non-stationarity by using machine learning to learn a mapping between current and future Global Climate Model outputs. This allows for adaptive corrections that account for the changing climate conditions, rather than just applying a fixed adjustment based on past data.
Lev: So, the main claim is that by predicting these evolving GCM biases dynamically, they can produce more realistic spatial distributions of environmental parameters under future scenarios. It’s about capturing how the system itself is shifting.
Kai: They achieve this by combining dimensionality reduction—using singular value decomposition on historical and projected data to extract modes—with a data-driven surrogate modeling technique to map those historical patterns onto future ones. This is their core methodology for getting those realistic distributions.
Mira: Essentially, they use these steps—reduction, projection, and mapping—to produce a tool that can adapt to changing climate conditions by learning the underlying system dynamics from the GCM outputs themselves. They are creating a method that accounts for model simplifications and other biases inherent in the models themselves.
Lev: What this means for the world is that instead of using static estimates, we get corrections that evolve alongside climate change projections, which should give us a more accurate picture of hurricane intensity and track under future scenarios.
Conclusion: Kai: So looking at the title, "A Novel Non-Stationary Bias-Correction Framework for Hurricane Risk Assessment under Climate Change," it really sums up what they did: they developed a new way to correct errors that change over time specifically for assessing hurricane risk under climate change.
Mira: The authors are essentially saying that because the biases in current models are dynamic, we need a framework that can dynamically adjust itself to those changes. It’s about moving beyond simple static corrections to something more responsive.
Lev: The implication for us is that if we want reliable risk assessments for coastal communities facing climate change, we can't just trust the standard GCM outputs anymore; we need this kind of adaptive tool.
Kai: It underscores the need for proactive planning because these results show that hurricane risks aren't just increasing linearly; they are becoming more intense in ways that demand a framework that evolves with those changes.
Mira: So, in simple terms, they’ve shown how to build a system that learns the climate's changing biases so we can better prepare for hurricanes and the heavy rainfall events they’re predicting.
Lev: It highlights the necessity of adaptive bias correction techniques when dealing with complex systems like climate models, telling us that future planning needs to be flexible and constantly updated.
Reda Snaiki, Teng Wu
Department of Construction Engineering, École de Technologie Supérieure, Université du Québec · Department of Civil, Structural and Environmental Engineering, University at Buffalo
physics.ao-ph
Submitted: 2025-05-01
Updated: 2026-11-16
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
Importance score: 72/100
The gist: A novel dynamic bias correction framework is presented to enhance hurricane track simulations under climate change by learning a mapping between current and future Global Climate Model outputs…
Key concepts
- Non-Stationarity of Climate Biases
- Climate model biases are not constant over time; they change as the climate shifts due to climate change. This concept means that a bias correction method needs to adapt because the errors in predicting future weather patterns are constantly evolving, requiring a more flexible approach than traditional methods.
- Dimensionality Reduction using POD
- This technique simplifies complex climate data by using Singular Value Decomposition (SVD) to find the most important patterns, or 'modes,' in the historical data. By keeping only the modes that capture most of the variance, researchers can reduce noise and focus on the core physical drivers of climate variability.
- Surrogate Model Mapping
- A surrogate model is a simplified machine learning tool trained to predict how patterns from past climate data (historical POD coefficients) will translate into future patterns (future POD coefficients). This mapping allows the framework to project how current biases will evolve into future biases, enabling better long-term projections.
Terminology
Summary
A novel dynamic bias correction framework is presented to enhance hurricane track simulations under climate change by learning a mapping between current and future Global Climate Model outputs <ref:2505. p. 3 Therefore,
advanced techniques should be explored that can explicitly learn how biases might change under
different climate scenarios. This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. This proposed approach differs from traditional bias correction by shifting the
focus from reducing historical biases to predicting future biases, making it more suitable for longterm climate projections under climate change. By explicitly incorporating non-stationarity, this
method can provide more accurate projections compared to conventional techniques.Where the paper says it
Where the paper says it
Framework Overview
The proposed methodology introduces a novel machine learning approach to address the non-stationarity of climate biases by learning a mapping between current and future Global Climate Model outputs <ref:2505.00832#pg2> This method differs from traditional bias correction by focusing on predicting future biases, making it more suitable for long-term climate projections under climate change <ref:2505. p. 3 This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. This approach
differs from standard bias correction by focusing on predicting future biases, making it more
suitable for long-term climate projections. This approach
differs from standard bias correction by focusing on predicting future biases, making it more
suitable for long-term climate projections. The proposed machine learning technique involves two
main stages to address the non-stationary bias issue and improve extreme event predictions.Where the paper says it
Where the paper says it
Data Processing Steps
The data-driven bias-correction framework employs a two-step process involving dimensionality reduction and surrogate modeling <ref:2505.00832#pg2>
-
Dimensionality Reduction using POD involves computing Singular Value Decomposition (SVD) on the historical and projected data from the GCM model to extract the POD modes <ref:2505. p. 7
00832#pg2>.Where the paper says it
, SSP585) Step 1: Dimensionality Reduction using POD 1. Compute Singular Value Decomposition (SVD): Perform SVD on the historical and projected data from the GCM model to extract the POD modes.
-
Projection onto Reduced Subspace requires selecting the first K POD modes that capture at least 95% of the total variance in the data <ref:2505. p. 7
00832#pg2>.Where the paper says it
𝐺𝐶𝑀𝐻𝑖𝑠𝑡𝑜𝑟𝑖𝑐𝑎𝑙 = 𝑈1𝑆1𝑉1 𝑇 𝐺𝐶𝑀𝐹𝑢𝑡𝑢𝑟𝑒 = 𝑈2𝑆2𝑉2 𝑇 2. POD Mode Selection: Select the first 𝐾 POD modes from 𝑈1 and 𝑈2 that capture at least 95% of the total variance in the data.
-
A surrogate model is then trained to map historical POD coefficients to future POD coefficients, denoted as ftheta: a H → a P <ref:2505. p. 8
00832#pg2>.Where the paper says it
The dynamics of the reduced system are predicted using a surrogate model. While various surrogate models can be employed, this study utilizes a LSTM network, which maps the historical POD coefficients (𝑎 𝐻 as input) to the future POD coefficients (𝑎 𝑃 as output).
Simulation and Results
The proposed bias-correction methodology was applied to six selected locations, simulating historical and future climate conditions across three periods: 1979–2014, 2024–2059, and 2060–2095 <ref:2505.00832#pg2>. The simulation results demonstrate significant differences in wind speed and rain rates return periods between historical and future scenarios, particularly for longer return periods <ref:2505. p. 0 Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Model Performance Comparison
The analysis of percentage differences between the simulated scenarios and GFDL-CM4 data highlights that the data-driven bias-corrected framework generally underestimates storm intensity (wind speed) compared to GFDL-CM4 for the 2024–2060 period <ref:2505.00832#pg2>. However, this trend is reversed in certain locations, such as Atlantic City and Halifax, where the data-driven framework consistently predicts stronger storms <ref:2505. p. 9 69°),
Atlantic City (-74. 38°) and Halifax (-63.Where the paper says it
Conclusion
The study introduces a novel data-driven bias correction framework to enhance the accuracy of hurricane track simulations under future climate scenarios <ref:2505. p. 3 Therefore,
advanced techniques should be explored that can explicitly learn how biases might change under
different climate scenarios. This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. Therefore,
advanced techniques should be explored that can explicitly learn how biases might change under
different climate scenarios. This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. Therefore,
advanced techniques should be explored that can explicitly learn how biases might change under
different climate scenarios. This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. Therefore,
advanced techniques should be explored that can explicitly learn how biases might change under
different climate scenarios. This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. Therefore,
advanced techniques should be explored that can explicitly learn how biases might change under
different climate scenarios. This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. Therefore,
advanced techniques should be explored that can explicitly learn how biases might change under
different climate scenarios. This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. Therefore,
advanced techniques should be explored that can explicitly learn how biases might change under
different climate scenarios. This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. Therefore,
advanced techniques should be explored that can explicitly learn how biases might change under
different climate scenarios. This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. Therefore,
advanced techniques should be explored that can explicitly learn how biases might change under
different climate scenarios. This study proposes a novel, data-driven bias-correction approach that enhances hurricane track
data generation by learning a mapping between current and future GCM outputs. Our approach employs machine learning to predict evolving GCM biases, allowing
dynamic corrections that account for changing climate conditions. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. A key component of these methods is the creation of a large database of synthetic storms,
encompassing their entire life cycle from genesis to dissipation. By leveraging information from a
broader geographic area, this approach ensures sufficient data for estimating annual probabilities
of low-frequency, high-impact events in hurricane-prone regions. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 1 Hurricane track methodology
The downscaling method leverages information from a broader geographic region to populate the
target area with statistically robust data, enabling the estimation of the annual probability of lowfrequency, high-impact hurricane events. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 1 Hurricane track methodology
The downscaling method leverages information from a broader geographic region to populate the
target area with statistically robust data, enabling the estimation of the annual probability of lowfrequency, high-impact hurricane events. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 1 Hurricane track methodology
The downscaling method leverages information from a broader geographic region to populate the
target area with statistically robust data, enabling the estimation of the annual probability of lowfrequency, high-impact hurricane events. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 1 Hurricane track methodology
The downscaling method leverages information from a broader geographic region to populate the
target area with statistically robust data, enabling the estimation of the annual probability of lowfrequency, high-impact hurricane events. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 1 Hurricane track methodology
The downscaling method leverages information from a broader geographic region to populate the
target area with statistically robust data, enabling the estimation of the annual probability of lowfrequency, high-impact hurricane events. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 1 Hurricane track methodology
The downscaling method leverages information from a broader geographic region to populate the
target area with statistically robust data, enabling the estimation of the annual probability of lowfrequency, high-impact hurricane events. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 1 Hurricane track methodology
The downscaling method leverages information from a broader geographic region to populate the
target area with statistically robust data, enabling the estimation of the annual probability of lowfrequency, high-impact hurricane events. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 1 Hurricane track methodology
The downscaling method leverages information from a broader geographic region to populate the
target area with statistically robust data, enabling the estimation of the annual probability of lowfrequency, high-impact hurricane events. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 1 Hurricane track methodology
The downscaling method leverages information from a broader geographic region to populate the
target area with statistically robust data, enabling the estimation of the annual probability of lowfrequency, high-impact hurricane events. 5
Once the mapping function is identified, it is applied to ERA5 reanalysis data to produce a more
realistic spatial distribution of environmental parameters under future climate scenarios. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 5
Once the mapping function is identified, it is applied to ERA5 reanalysis data to produce a more
realistic spatial distribution of environmental parameters under future climate scenarios. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 5
Once the mapping function is identified, it is applied to ERA5 reanalysis data to produce a more
realistic spatial distribution of environmental parameters under future climate scenarios. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 5
Once the mapping function is identified, it is applied to ERA5 reanalysis data to produce a more
realistic spatial distribution of environmental parameters under future climate scenarios. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 5
Once the mapping function is identified, it is applied to ERA5 reanalysis data to produce a more
realistic spatial distribution of environmental parameters under future climate scenarios. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 5
Once the mapping function is identified, it is applied to ERA5 reanalysis data to produce a more
realistic spatial distribution of environmental parameters under future climate scenarios. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 5
Once the mapping function is identified, it is applied to ERA5 reanalysis data to produce a more
realistic spatial distribution of environmental parameters under future climate scenarios. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 5
Once the mapping function is identified, it is applied to ERA5 reanalysis data to produce a more
realistic spatial distribution of environmental parameters under future climate scenarios. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. 5
Once the mapping function is identified, it is applied to ERA5 reanalysis data to produce a more
realistic spatial distribution of environmental parameters under future climate scenarios. Ultimately, the goal is to evaluate the performance of the proposed bias-correction technique on
hurricane track generation and calculate return periods for wind speed and rain rate for North
American coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. By combining dimensionality
reduction with data-driven surrogate modeling, we capture the system's underlying dynamics to
produce realistic spatial distributions of environmental parameters under future scenarios. Using
the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall
across coastal cities. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. Our results reveal significant differences in projected risks with and without
dynamic bias correction, emphasizing the increased threat to critical infrastructure in hurricaneprone regions. This work highlights the necessity of adaptive techniques for accurately assessing
future climate impacts, offering a critical advancement in hurricane risk modeling and resilience
planning. This model outperforms traditional hurricane track models, which often
rely on simplified regression formulas and may struggle to capture the complex, non-linear
relationships within hurricane data. The genesis model employed in this study adopts a stochastic
approach, randomly seeding potential storm locations across both space and time (Emanuel et al.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Dynamic Bias Correction Framework Implementation: The improved AI system can dynamically correct biases in Global Climate Model outputs by employing a machine learning approach to predict evolving GCM biases, allowing for dynamic corrections that account for changing climate conditions. This moves beyond static bias correction by explicitly learning how biases might change under different climate scenarios.
-
Feature Engineering via Dimensionality Reduction: The system can perform dimensionality reduction using Singular Value Decomposition (SVD) to extract the most relevant features from the vast dimensionality of GCM output data, mitigating noise and redundancy before training the surrogate model. This step leads to a more efficient and robust subsequent modeling stage by capturing only the most significant variance in hurricane track data.
-
Surrogate Model Training for Future State Prediction: The AI can train a Long Short-Term Memory (LSTM) network to map historical POD coefficients to future POD coefficients, effectively predicting the reduced state dynamics. Once trained, this mapping function is applied to ERA5 reanalysis data to produce a more realistic spatial distribution of environmental parameters under future climate scenarios.
-
Scenario-Specific Risk Assessment: The system can generate synthetic hurricane datasets spanning 10,000 years for three distinct climate scenarios—Historical Baseline (1979-2014), Near-Future Scenario (2024-2059), and Far-Future Scenario (2060-2095)—allowing for the examination of both short-term and long-term climate effects. This enables the assessment of how wind speeds and rain rates return periods change across different future climate projections.
-
Quantifying Non-Stationary Risk Shifts: The system can quantify the percentage change in wind speeds and rain rates between historical and future scenarios, as shown in Table 1 and Table 2, to reveal trends such as
the far-future scenario forecasts a notable increase in wind speeds
orGalveston continues to show mostly negative differences,
providing localized insights into climate change impacts.
Abstract
Conventional hurricane track generation methods typically depend on biased outputs from Global Climate Models (GCMs), which undermines their accuracy in the context of climate change. We present a novel non-stationary bias correction framework, developed for and applied to the North Atlantic basin, that adaptively accounts for evolving biases in GCM outputs. The framework combines dimensionality reduction with data-driven surrogate modeling to learn the nonlinear evolution between historical and projected GCM climate states, allowing the bias correction to evolve with changing climate conditions. The learned relationship is subsequently applied to a reanalysis-based historical baseline to generate bias-corrected environmental fields under future climate scenarios. Using the empirical Weibull plotting approach, we calculate return periods for wind speed and rainfall across selected North American coastal cities. Our results reveal substantial differences in projected hurricane hazards with and without bias correction, highlighting the importance of accounting for evolving GCM biases in climate-risk assessment and resilience planning. This work highlights the necessity of adaptive techniques for accurately assessing future climate impacts, offering a critical advancement in hurricane risk modeling and resilience planning.
Related papers
- Anthropogenic Heat in Urban Climate Systems: Forcing, Sensitivity, and Feedback
- NORi: An ML-Augmented Ocean Boundary Layer Parameterization
- A Checklist to assess the energy and carbon impacts of ML/AI applications in Earth System Modeling
- Skillful Thunderstorm Forecasts Learned Directly from Observations
- Wind vector retrieval from miniaturized wave-enabled sea-surface drifters
- Stochastic coupling of climate variables and ice volume over the Late Pleistocene glacial cycles