Machine learning in calibrating tropical cyclone intensity forecast of <scp>ECMWF EPS</scp>

Chan, Ming Hei Kenneth; Wong, Wai Kin; Au‐Yeung, Kin Chung

doi:10.1002/met.2041

Cited by 14 publications

(4 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We note that there might exist biases between these historical tracks and the ECMWF forecast tracks [45,46] (c.f., Section 4.2) used for the impact forecast. Currently no bias correction is made, but this could be addressed in future iterations of the model.…”

Section: Calibration Of Impact Functions For Human Displacementmentioning

confidence: 99%

Impact-Based Forecasting of Tropical Cyclone-Related Human Displacement to Support Anticipatory Action

Kam,

Ciccone,

Kropf

et al. 2023

Preprint

View full text Add to dashboard Cite

Tropical cyclones (TCs) displace millions every year. While TCs pose hardships and threaten lives, their negative impacts can be reduced by anticipatory like evacuation and humanitarian aid coordination. In addition to weather forecasts, impact forecast enables effective response by providing richer information on the numbers and locations of people at risk of displacement. We introduce a fully open-source implementation of a globally consistent and regionally calibrated TC-related displacement forecast at low computational costs, combining meteorological forecast with population exposure and respective vulnerability. We present a case study of TC Yasa which hit Fiji in December 2020. We emphasise the importance of considering the uncertainties associated with hazard, exposure, and vulnerability in a global uncertainty analysis, which reveals the full range of possible outcomes. Additionally, we perform a sensitivity analysis on all recorded TC displacement events from 2017 to 2020 to understand how the forecast outcomes depend on these uncertain inputs. Our findings suggest that for longer forecast lead times, decision-making should focus more on meteorological uncertainty, while greater emphasis should be placed on the vulnerability of the local community shortly before TC landfall. Our open-source codes and implementations are readily transferable to other users, hazards, and impact types.

show abstract

Section: Calibration Of Impact Functions For Human Displacementmentioning

confidence: 99%

Impact-Based Forecasting of Tropical Cyclone-Related Human Displacement to Support Anticipatory Action

Kam,

Ciccone,

Kropf

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…The MLP model was based on deep-learning. Utilizing a decision-tree-based machine learning algorithm "XGBoost ", an increased TC intensity forecast skills have been achieved for ECMWF ensemble prediction system model outputs (Chan et al, 2021). Despite their usefulness in intensity prediction of TCs, the machine learning technique provides short lead-time predictions of the cyclones, due to which its utility is yet to be adopted by disaster managers for real-time operations (Chen et al, 2020).…”

Section: Techniques Utilizing Arti Cial Neural Network (Ann)mentioning

confidence: 99%

Systematic scientific strategies associated with mitigating the challenges in predicting track and intensity of cyclones across the world basins

Singh

2022

Preprint

View full text Add to dashboard Cite

The present study analyses the systematic scientific strategies associated with mitigating the challenges in predicting track and intensity of cyclones across the world basins. The relevant ‘51’ original publications on predicting track and intensity of cyclones across the world basins utilising numerical weather prediction (NWP) models, machine learning and satellite based approach were chosen using a focused orderly search and were used to develop the selection benchmarks for the systematic literature review. According to our systematic review, out of selected ‘51’ studies, a total of ‘19’(37.25%) and ‘12’ (23.52%) and ‘9’ (17.64%) studies were focused on Pacific (including south China sea) Atlantic and North Indian Ocean (NIO) basins respectively. Three studies were conducted over both the Pacific and Atlantic Basins. However, the remaining ‘8’ studies were not over any specific basin. After 2015, the widespread use of machine based algorithm and analysis began, with a focus on the improvement in intensity prediction of TCs. Satellite based approach was employed in nearly all of the studies, and data assimilation techniques were in nascent stage during 2006-2013. To create detailed cyclone prediction track and intensity of cyclones across the world basins, a balance among data assimilation systems, modelling systems, and machine learning to uncover predictability drivers is necessary. Globally, predicting track and intensity of cyclones should take into account future climate change scenarios in the world basins. In order to gain improvement in this direction, further research is needed to fill in the knowledge gaps that have been identified.

show abstract

“…Here, we aim to correct the intensity in the ERA5 reanalysis using the true intensity in IBTrACS as a reference. There are some related works to correct the intensity in operational forecasting [48][49][50]. Differently, we develop an adaptive learning approach based on deep neural networks and domain adaptation, which helps to solve the issues of data quality and weak model generalisability.…”

Section: Introductionmentioning

confidence: 99%

An Adaptive Learning Approach for Tropical Cyclone Intensity Correction

Chen,

Toumi,

Shi

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

Tropical cyclones (TCs) are dangerous weather events; accurate monitoring and forecasting can provide significant early warning to reduce loss of life and property. However, the study of tropical cyclone intensity remains challenging, both in terms of theory and forecasting. ERA5 reanalysis is a benchmark data set for tropical cyclone studies, yet the maximum wind speed error is very large (68 kts) and is still 19 kts after simple linear correction, even in the better sampled North Atlantic. Here, we develop an adaptive learning approach to correct the intensity in the ERA5 reanalysis, by optimising the inputs to overcome the problems caused by the poor data quality and updating the features to improve the generalisability of the deep learning-based model. Specifically, we use understanding of TC properties to increase the representativeness of the inputs so that the general features can be learned with deep neural networks in the sample space, and then use domain adaptation to update the general features from the known domain with historical storms to the specific features for the unknown domain of new storms. This approach can reduce the error to only 6 kts which is within the uncertainty of the best track data in the international best track archive for climate stewardship (IBTrACS) in the North Atlantic. The method may have wide applicability, such as when extending it to the correction of intensity estimation from satellite imagery and intensity prediction from dynamical models.

show abstract

Machine learning in calibrating tropical cyclone intensity forecast of ECMWF EPS

Cited by 14 publications

References 61 publications

Impact-Based Forecasting of Tropical Cyclone-Related Human Displacement to Support Anticipatory Action

Impact-Based Forecasting of Tropical Cyclone-Related Human Displacement to Support Anticipatory Action

Systematic scientific strategies associated with mitigating the challenges in predicting track and intensity of cyclones across the world basins

An Adaptive Learning Approach for Tropical Cyclone Intensity Correction

Contact Info

Product

Resources

About