Jorge A. Duarte
Jorge A. Duarte has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
- Pierre-Emmanuel Kirstetter
Advanced Radar Research Center University of Oklahoma Norman OK USA
4 shared publications - Devon Woods
Univ Oklahoma, Sch Civil Engn & Environm Sci, Norman, OK 73019 USA
3 shared publications - Humberto Vergara
Univ Iowa, Coll Engn, Iowa City, IA USA
3 shared publications - Jonathan J. Gourley
NOAA/OAR National Severe Storms Laboratory, Norman, OK, 73072, USA
3 shared publications Charles D. Nicholson
School of Industrial & Systems Engineering, University of Oklahoma, Norman, Oklahoma
2 shared publications- Jeffrey B. Basara
University of Massachusetts Lowell, Unknown, Lowell, MA, United States [US
2 shared publications
Latest Publications
- Open Access
- 11 September 2026
- Published by Elsevier in International Journal of Disaster Risk Reduction
- Vol. 145
Equity-aware interpretable policy rules for personalized hurricane risk communication
- 11 September 2026
- Published by Elsevier in International Journal of Disaster Risk Reduction
- Vol. 145
Open AccessEquity-aware interpretable policy rules for personalized hurricane risk communication
- 1 April 2026
- Published by American Meteorological Society in Artificial Intelligence for the Earth Systems
- Vol. 5 (2)
Large Language Model–Based Classification of Flash Flood Impacts across the United States
In late 2019, the National Weather Service (NWS) transitioned its warning paradigm into an impact-based warning (IBW) format. This new format applies to flash flood warnings and intends to provide detailed information about the hazard, its source, an impact narrative, and a flash flood damage threat tag. These damage threat tags and additional warning information aim to elicit different calls to action for the emergency management community and the public (e.g., Wireless Emergency Alerts). With the advent of highly performant, widely available, and affordable access to pretrained large language models (LLMs) like ChatGPT, in conjunction with the flash flood severity index (a systematically conceived framework for assessing and therefore classifying flash flood severity from textual flash flood reports), the present work explores a systematic LLM-based method for classifying flash flood report narratives into damage threat tags. Specifically, this work showcases the classification of flash flood reports into flash flood severity index (FFSI) impact categories using ChatGPT and prompts engineered to incorporate textualized forms of FFSI impact definitions based on the published literature. Report classifications were verified and contrasted with Flooded Locations and Simulated Hydrographs (FLASH) product outputs for each reported flood event within a given spatiotemporal 3D window. This approach was first tested on a reduced dataset of 663 expert-classified reports which enabled us to compare and contrast the method’s performance. Subsequently, it was used to classify over 22 000 historical local storm reports (LSRs) between May 2018 and June 2022. This unprecedented dataset is the cornerstone that has enabled the research and development of new experimental FLASH products, which look to inform forecasters of the anticipated impacts for any given flash flood forecast, in line with the NWS’s recently implemented IBW framework.
- 1 April 2026
- Published by American Meteorological Society in Artificial Intelligence for the Earth Systems
- Vol. 5 (2)
Large Language Model–Based Classification of Flash Flood Impacts across the United States
In late 2019, the National Weather Service (NWS) transitioned its warning paradigm into an impact-based warning (IBW) format. This new format applies to flash flood warnings and intends to provide detailed information about the hazard, its source, an impact narrative, and a flash flood damage threat tag. These damage threat tags and additional warning information aim to elicit different calls to action for the emergency management community and the public (e.g., Wireless Emergency Alerts). With the advent of highly performant, widely available, and affordable access to pretrained large language models (LLMs) like ChatGPT, in conjunction with the flash flood severity index (a systematically conceived framework for assessing and therefore classifying flash flood severity from textual flash flood reports), the present work explores a systematic LLM-based method for classifying flash flood report narratives into damage threat tags. Specifically, this work showcases the classification of flash flood reports into flash flood severity index (FFSI) impact categories using ChatGPT and prompts engineered to incorporate textualized forms of FFSI impact definitions based on the published literature. Report classifications were verified and contrasted with Flooded Locations and Simulated Hydrographs (FLASH) product outputs for each reported flood event within a given spatiotemporal 3D window. This approach was first tested on a reduced dataset of 663 expert-classified reports which enabled us to compare and contrast the method’s performance. Subsequently, it was used to classify over 22 000 historical local storm reports (LSRs) between May 2018 and June 2022. This unprecedented dataset is the cornerstone that has enabled the research and development of new experimental FLASH products, which look to inform forecasters of the anticipated impacts for any given flash flood forecast, in line with the NWS’s recently implemented IBW framework.
- Open Access
- 18 March 2025
- Published by Copernicus GmbH
StreamScope: Fixed-mount laser scanning instrumentation for remote stream gauging in shallow rivers with frequently changing geomorphologies
A novel approach has been developed to facilitate streaming gauging in shallow rivers with frequently changing geomorphologies. This effort, developed in partnership with the United States Geological Survey (USGS), aims to ultimately generate more accurate, real-time discharge estimates at previously ungauged locations in remote areas.This presentation will introduce the StreamScope, designed for automated bathymetry retrievals. It is a low-power laser scanning instrument that uses a Class II 620 nm laser and an ultrasonic sensor to remotely measure cross-sectional geometry and generate real-time stage-area ratings. By utilizing onboard automation and clustering algorithms, StreamScope can accurately measure channel bathymetry using multiple angles, determine stream width and stage, providing data to enable real-time discharge estimates from noncontact sensors. Results from laboratory experiments will be presented to evaluate the laser’s efficacy under different solar radiation conditions, varying turbidity levels at different water depths, different bottom substrates, and at varying heights above the water’s surface. Notably, the experiments revealed that the laser's distance retrievals may serve as a proxy for turbidity, offering a potential new method for assessing water clarity in real-time.
- 18 March 2025
- Published by Copernicus GmbH
Open AccessStreamScope: Fixed-mount laser scanning instrumentation for remote stream gauging in shallow rivers with frequently changing geomorphologies
A novel approach has been developed to facilitate streaming gauging in shallow rivers with frequently changing geomorphologies. This effort, developed in partnership with the United States Geological Survey (USGS), aims to ultimately generate more accurate, real-time discharge estimates at previously ungauged locations in remote areas.This presentation will introduce the StreamScope, designed for automated bathymetry retrievals. It is a low-power laser scanning instrument that uses a Class II 620 nm laser and an ultrasonic sensor to remotely measure cross-sectional geometry and generate real-time stage-area ratings. By utilizing onboard automation and clustering algorithms, StreamScope can accurately measure channel bathymetry using multiple angles, determine stream width and stage, providing data to enable real-time discharge estimates from noncontact sensors. Results from laboratory experiments will be presented to evaluate the laser’s efficacy under different solar radiation conditions, varying turbidity levels at different water depths, different bottom substrates, and at varying heights above the water’s surface. Notably, the experiments revealed that the laser's distance retrievals may serve as a proxy for turbidity, offering a potential new method for assessing water clarity in real-time.
- Open Access
- 18 March 2025
- Published by Copernicus GmbH
Towards a Road Weather Impacts Prediction System: Phase 1 - Flash Flood-Related Impacts
This work presents a prototype of a rapidly-updating, impact-based decision-support system (IDSS) for forecasting weather impacts on roadways. This effort, developed in partnership with the Missouri Department of Transportation (MODOT), aims to aid in effectively responding to road-weather hazards and impacts. While the main focus of this project's initial stage is on road impacts from flash flooding, the framework will be developed and later extended for additional road hazards such as winter weather. The presented work will showcase the foundation for developing the weather impact prediction capabilities for the aforementioned IDSS. Based on a comprehensive data set of road closures due to flooding in the state of Missouri between 2017 and 2023, current efforts have produced an event-based data set, which incorporates diverse data sources such as radar-based precipitation estimates and accumulations and hydrologic model outputs. A data-driven approach is used to define event scales (spatial and temporal), for matching road closure reports to model and product outputs. A pilot study demonstrating the feasibility of implementing machine learning-based approaches is being developed, which will provide state-wide probabilistic flash flood impacts on road networks. Results on data set preparation, exploration of variables comprising the road closure database, and model implementation will be shown, with an emphasis towards prospective multi-hazard, real-time implementation of a scalable road weather impacts prediction system (RWIPS).
- 18 March 2025
- Published by Copernicus GmbH
Open AccessTowards a Road Weather Impacts Prediction System: Phase 1 - Flash Flood-Related Impacts
This work presents a prototype of a rapidly-updating, impact-based decision-support system (IDSS) for forecasting weather impacts on roadways. This effort, developed in partnership with the Missouri Department of Transportation (MODOT), aims to aid in effectively responding to road-weather hazards and impacts. While the main focus of this project's initial stage is on road impacts from flash flooding, the framework will be developed and later extended for additional road hazards such as winter weather. The presented work will showcase the foundation for developing the weather impact prediction capabilities for the aforementioned IDSS. Based on a comprehensive data set of road closures due to flooding in the state of Missouri between 2017 and 2023, current efforts have produced an event-based data set, which incorporates diverse data sources such as radar-based precipitation estimates and accumulations and hydrologic model outputs. A data-driven approach is used to define event scales (spatial and temporal), for matching road closure reports to model and product outputs. A pilot study demonstrating the feasibility of implementing machine learning-based approaches is being developed, which will provide state-wide probabilistic flash flood impacts on road networks. Results on data set preparation, exploration of variables comprising the road closure database, and model implementation will be shown, with an emphasis towards prospective multi-hazard, real-time implementation of a scalable road weather impacts prediction system (RWIPS).
- 26 March 2024
- Published by Elsevier in Journal of Hydrology
- Vol. 635
Hydrologic Evaluation of the Global Precipitation Measurement Mission over the U.S.: Effect of Spatial and Temporal Scales
- 26 March 2024
- Published by Elsevier in Journal of Hydrology
- Vol. 635
Hydrologic Evaluation of the Global Precipitation Measurement Mission over the U.S.: Effect of Spatial and Temporal Scales
- 26 September 2023
- Published by Elsevier in Journal of Hydrology
- Vol. 626
Hydrologic Evaluation of the Global Precipitation Measurement Mission over the U.S.: Error Budget Analysis
- 26 September 2023
- Published by Elsevier in Journal of Hydrology
- Vol. 626
Hydrologic Evaluation of the Global Precipitation Measurement Mission over the U.S.: Error Budget Analysis
- 13 January 2023
- Published by Elsevier in Journal of Hydrology
- Vol. 617
Hydrologic evaluation of the global precipitation measurement mission over the U.S.: Flood peak discharge and duration
- 13 January 2023
- Published by Elsevier in Journal of Hydrology
- Vol. 617
Hydrologic evaluation of the global precipitation measurement mission over the U.S.: Flood peak discharge and duration
- Open Access
- 30 October 2021
- Published by Springer Nature in Optimization Letters
- Vol. 16 (3) , 789-819
Wildfire burn scar encapsulation
Wildfires burn annually across the United States (US), which threaten those in close proximity to them. Due to drastic alterations of soil properties and to the land surfaces by these fires, risks of flash floods, debris flows, and severe erosion increases for these areas, which can have catastrophic consequences for biota, people and property. Computational tools, such as the WildfireRain algorithm, have been designed and implemented to assess the potential occurrence of debris flows over burn scars. However, in order to efficiently operate these tools, they require independent, non-overlapping buffers around burned areas to be defined, which is not a trivial task. In this paper we consider the problem of efficiently subsetting the conterminous US (CONUS) domain into optimal subdomains around burn scars, aiming to enable domain-wide WildfireRain product outputs to be used for operations by the National Weather Service (NWS). To achieve this, we define the Object Encapsulation Problem, where burn scars are represented by single-cell objects in a gridded domain, and circular buffers must be constructed around them. We propose a Linear Programming (LP) model that solves this problem efficiently. Optimal results produced using this model are presented for both a simplified synthetic data set, as well as for a subset of burn scars produced by severe wildfires in 2012 over the CONUS.
- 30 October 2021
- Published by Springer Nature in Optimization Letters
- Vol. 16 (3) , 789-819
Open AccessWildfire burn scar encapsulation
Wildfires burn annually across the United States (US), which threaten those in close proximity to them. Due to drastic alterations of soil properties and to the land surfaces by these fires, risks of flash floods, debris flows, and severe erosion increases for these areas, which can have catastrophic consequences for biota, people and property. Computational tools, such as the WildfireRain algorithm, have been designed and implemented to assess the potential occurrence of debris flows over burn scars. However, in order to efficiently operate these tools, they require independent, non-overlapping buffers around burned areas to be defined, which is not a trivial task. In this paper we consider the problem of efficiently subsetting the conterminous US (CONUS) domain into optimal subdomains around burn scars, aiming to enable domain-wide WildfireRain product outputs to be used for operations by the National Weather Service (NWS). To achieve this, we define the Object Encapsulation Problem, where burn scars are represented by single-cell objects in a gridded domain, and circular buffers must be constructed around them. We propose a Linear Programming (LP) model that solves this problem efficiently. Optimal results produced using this model are presented for both a simplified synthetic data set, as well as for a subset of burn scars produced by severe wildfires in 2012 over the CONUS.
- Open Access
- 23 March 2020
- Published by Copernicus GmbH
Predicting flood responses from spatial rainfall variability and basin morphology through machine learning
Predicting flash floods at short time scales as well as their impacts is of vital interest to forecasters, emergency managers and community members alike. Particularly, characteristics such as location, timing, and duration are crucial for decision-making processes for the protection of lives, property and infrastructure. Even though these characteristics are primarily driven by the causative rainfall and basin geomorphology, untangling the complex interactions between precipitation and hydrological processes becomes challenging due to the lack of observational datasets which capture diverse conditions.This work follows upon previous efforts on incorporating spatial rainfall moments as viable predictors for flash flood event characteristics such as lag time and the exceedance of flood stage thresholds at gauged locations over the Conterminous United States (CONUS). These variables were modeled by applying various supervised machine learning techniques over a database of flood events. The data included morphological, climatological, streamflow and precipitation data from over 21,000 flood-producing rainfall events – that occurred over 900+ different basins throughout the CONUS between 2002-2011. This dataset included basin parameters and indices derived from radar-based precipitation, which represented sub-basin scale rainfall spatial variability for each storm event. Both classification and regression models were constructed, and variable importance analysis was performed in order to determine the relevant factors reflecting hydrometeorological processes. In this iteration, a closer look at model performance consistency and variable selection aims to further explore rainfall moments’ explanatory power of flood characteristics.
- 23 March 2020
- Published by Copernicus GmbH
Open AccessPredicting flood responses from spatial rainfall variability and basin morphology through machine learning
Predicting flash floods at short time scales as well as their impacts is of vital interest to forecasters, emergency managers and community members alike. Particularly, characteristics such as location, timing, and duration are crucial for decision-making processes for the protection of lives, property and infrastructure. Even though these characteristics are primarily driven by the causative rainfall and basin geomorphology, untangling the complex interactions between precipitation and hydrological processes becomes challenging due to the lack of observational datasets which capture diverse conditions.This work follows upon previous efforts on incorporating spatial rainfall moments as viable predictors for flash flood event characteristics such as lag time and the exceedance of flood stage thresholds at gauged locations over the Conterminous United States (CONUS). These variables were modeled by applying various supervised machine learning techniques over a database of flood events. The data included morphological, climatological, streamflow and precipitation data from over 21,000 flood-producing rainfall events – that occurred over 900+ different basins throughout the CONUS between 2002-2011. This dataset included basin parameters and indices derived from radar-based precipitation, which represented sub-basin scale rainfall spatial variability for each storm event. Both classification and regression models were constructed, and variable importance analysis was performed in order to determine the relevant factors reflecting hydrometeorological processes. In this iteration, a closer look at model performance consistency and variable selection aims to further explore rainfall moments’ explanatory power of flood characteristics.
- 29 August 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Latin America Transactions
- Vol. 14 (6) , 2783-2788
Analysis of Data Mining Techniques for Constructing a Predictive Model for Academic Performance
This paper presents and analyzes the experience of applying certain data mining methods and techniques on 932 Systems Engineering students' data, from El Bosque University in Bogotá, Colombia; effort which has been pursued in order to construct a predictive model for students' academic performance. Previous works were reviewed, related with predictive model construction within academic environments using decision trees, artificial neural networks and other classification techniques. As an iterative discovery and learning process, the experience is analyzed according to the results obtained in each of the process' iterations. Each obtained result is evaluated regarding the results that are expected, the data's input and output characterization, what theory dictates and the pertinence of the model obtained in terms of prediction accuracy. Said pertinence is evaluated taking into account particular details about the population studied, and the specific needs manifested by the institution, such as the accompaniment of students along their learning process, and the taking of timely decisions in order to prevent academic risk and desertion. Lastly, some recommendations and thoughts are laid out for the future development of this work, and for other researchers working on similar studies.
- 29 August 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Latin America Transactions
- Vol. 14 (6) , 2783-2788
Analysis of Data Mining Techniques for Constructing a Predictive Model for Academic Performance
This paper presents and analyzes the experience of applying certain data mining methods and techniques on 932 Systems Engineering students' data, from El Bosque University in Bogotá, Colombia; effort which has been pursued in order to construct a predictive model for students' academic performance. Previous works were reviewed, related with predictive model construction within academic environments using decision trees, artificial neural networks and other classification techniques. As an iterative discovery and learning process, the experience is analyzed according to the results obtained in each of the process' iterations. Each obtained result is evaluated regarding the results that are expected, the data's input and output characterization, what theory dictates and the pertinence of the model obtained in terms of prediction accuracy. Said pertinence is evaluated taking into account particular details about the population studied, and the specific needs manifested by the institution, such as the accompaniment of students along their learning process, and the taking of timely decisions in order to prevent academic risk and desertion. Lastly, some recommendations and thoughts are laid out for the future development of this work, and for other researchers working on similar studies.