liu.seSearch for publications in DiVA
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • oxford
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
An Informed System Development Approach to Tropical Cyclone Track and Intensity Forecasting
Linköping University, Department of Computer and Information Science, Human-Centered systems. Linköping University, Faculty of Science & Engineering. (MDA)ORCID iD: 0000-0002-3997-1088
2016 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

Introduction: Tropical Cyclones (TCs) inflict considerable damage to life and property every year. A major problem is that residents often hesitate to follow evacuation orders when the early warning messages are perceived as inaccurate or uninformative. The root problem is that providing accurate early forecasts can be difficult, especially in countries with less economic and technical means.

Aim: The aim of the thesis is to investigate how cyclone early warning systems can be technically improved. This means, first, identifying problems associated with the current cyclone early warning systems, and second, investigating if biologically based Artificial Neural Networks (ANNs) are feasible to solve some of the identified problems.

Method: First, for evaluating the efficiency of cyclone early warning systems, Bangladesh was selected as study area, where a questionnaire survey and an in-depth interview were administered. Second, a review of currently operational TC track forecasting techniques was conducted to gain a better understanding of various techniques’ prediction performance, data requirements, and computational resource requirements. Third, a technique using biologically based ANNs was developed to produce TC track and intensity forecasts. Systematic testing was used to find optimal values for simulation parameters, such as feature-detector receptive field size, the mixture of unsupervised and supervised learning, and learning rate schedule. Five types of 2D data were used for training. The networks were tested on two types of novel data, to assess their generalization performance.

Results: A major problem that is identified in the thesis is that the meteorologists at the Bangladesh Meteorological Department are currently not capable of providing accurate TC forecasts. This is an important contributing factor to residents’ reluctance to evacuate. To address this issue, an ANN-based TC track and intensity forecasting technique was developed that can produce early and accurate forecasts, uses freely available satellite images, and does not require extensive computational resources to run. Bidirectional connections, combined supervised and unsupervised learning, and a deep hierarchical structure assists the parallel extraction of useful features from five types of 2D data. The trained networks were tested on two types of novel data: First, tests were performed with novel data covering the end of the lifecycle of trained cyclones; for these test data, the forecasts produced by the networks were correct in 91-100% of the cases. Second, the networks were tested with data of a novel TC; in this case, the networks performed with between 30% and 45% accuracy (for intensity forecasts).

Conclusions: The ANN technique developed in this thesis could, with further extensions and up-scaling, using additional types of input images of a greater number of TCs, improve the efficiency of cyclone early warning systems in countries with less economic and technical means. The thesis work also creates opportunities for further research, where biologically based ANNs can be employed for general-purpose weather forecasting, as well as for forecasting other severe weather phenomena, such as thunderstorms.

Place, publisher, year, edition, pages
Linköping, Sweden: Linköping University Electronic Press, 2016. , p. 160
Series
Linköping Studies in Science and Technology. Dissertations, ISSN 0345-7524 ; 1734
Keywords [en]
Biologically based artificial neural networks; bi-directionally connected networks; temporal sequence learning; accurate tropical cyclone forecasting; informative warning message;
National Category
Other Computer and Information Science
Identifiers
URN: urn:nbn:se:liu:diva-123198DOI: 10.3384/diss.diva-123198ISBN: 978-91-7685-854-7 (print)OAI: oai:DiVA.org:liu-123198DiVA, id: diva2:882030
Public defence
2016-02-17, Alan Turing, Hus E, Campus Valla, Linköping, 13:15 (English)
Opponent
Supervisors
Available from: 2016-01-13 Created: 2015-12-06 Last updated: 2020-06-29Bibliographically approved
List of papers
1. The current cyclone early warning system in Bangladesh: Providers' and receivers' views
Open this publication in new window or tab >>The current cyclone early warning system in Bangladesh: Providers' and receivers' views
2015 (English)In: International Journal of Disaster Risk Reduction, E-ISSN 2212-4209, Vol. 12, p. 285-299Article in journal (Refereed) Published
Abstract [en]

Bangladesh has experienced several catastrophic Tropical Cyclones (TCs) during the last decades. Despite the efforts of disaster management organizations, as well as the Bangladesh Meteorological Department (BMD), there were lapses in the residents’ evacuation behavior. To examine the processes of TC forecasting and warning at BMD and to understand the reasons for residents’ reluctance to evacuate after a cyclone warning, we conducted an individual in-depth interview among the meteorologists at BMD, as well as a questionnaire survey among the residents living in the coastal areas. The results reveal that the forecasts produced by BMD are not reliable for longer than 12-h. Therefore, longer-term warnings have to be based on gross estimates of TC intensity and motion, which renders the disseminated warning messages unreliable. Our results indicate that residents in the coastal areas studied, do not follow the evacuation orders due to mistrust of the warning messages—which can deter from early evacuation; and insufficient number of shelters and poor transportation possibilities—which discourages late evacuation. Suggestions made by the residents highlight the necessity of improved warning messages in the future. These findings indicate the need for improved forecasting, and more reliable and more informative warning messages for ensuring a timely evacuation response from residents.

Place, publisher, year, edition, pages
Elsevier, 2015
Keywords
Accurate tropical cyclone forecasting; Informative warning message; Warning message interpretation; Meteorologists’ perspective; Residents’ perspective; Principal component analysis
National Category
Social Sciences Interdisciplinary
Identifiers
urn:nbn:se:liu:diva-117927 (URN)10.1016/j.ijdrr.2015.02.004 (DOI)000357735000026 ()
Available from: 2015-05-18 Created: 2015-05-18 Last updated: 2019-07-05Bibliographically approved
2. Tropical cyclone track forecasting techniques: A review
Open this publication in new window or tab >>Tropical cyclone track forecasting techniques: A review
2012 (English)In: Atmospheric research, ISSN 0169-8095, E-ISSN 1873-2895, Vol. 104-105, p. 40-69Article, review/survey (Refereed) Published
Abstract [en]

Delivering accurate cyclone forecasts in time is of key importance when it comes to saving human lives and reducing economic loss. Difficulties arise because the geographical and climatological characteristics of the various cyclone formation basins are not similar, which entail that a single forecasting technique cannot yield reliable performance in all ocean basins. For this reason, global forecasting techniques need to be applied together with basin-specific techniques to increase the forecast accuracy. As cyclone track is governed by a range of factors variations in weather conditions, wind pressure, sea surface temperature, air temperature, ocean currents, and the earths rotational force-the coriolis force, it is a formidable task to combine these parameters and produce reliable and accurate forecasts. In recent years, the availability of suitable data has increased and more advanced forecasting techniques have been developed, in addition to old techniques having been modified. In particular, artificial neural network based techniques are now being considered at meteorological offices. This new technique uses freely available satellite images as input, can be run on standard PCs, and can produce forecasts with good accuracy. For these reasons, artificial neural network based techniques seem especially suited for developing countries which have limited capacity to forecast cyclones and where human casualties are the highest. © 2011 Elsevier B.V.

Place, publisher, year, edition, pages
Elsevier, 2012
Keywords
Artificial neural networks; Cyclone forecasting models; Cyclone forecasting techniques; Cyclone track forecasting; Hurricane; Typhoon
National Category
Engineering and Technology
Identifiers
urn:nbn:se:liu:diva-74115 (URN)10.1016/j.atmosres.2011.09.012 (DOI)
Available from: 2012-01-19 Created: 2012-01-19 Last updated: 2017-12-08Bibliographically approved
3. Tropical Cyclone Track Forecasting
Open this publication in new window or tab >>Tropical Cyclone Track Forecasting
2018 (English)In: Exploring Natural Hazards: A Case Study Approach / [ed] Darius Bartlett; Ramesh P. Singh, Taylor & Francis, 2018Chapter in book (Refereed)
Abstract [en]

Tropical cyclones are large-scale low-pressure systems that form over warm tropical and subtropical waters. These low-pressure systems are characterized by high-speed surface wind circulation, rotating spirals of thick clouds, heavy rain, and surges, the water masses sometimes reaching a height of 10 meters when they hit a coastline. Tropical cyclones are one of the most destructive meteorological disasters due to their high damaging power, both through strong winds and flooding. To minimize economic loss and to save human lives, meteorologists have developed a range of techniques for forecasting tropical cyclone track. The most common techniques utilize statistical and mathematical equations to integrate the movement pattern of historical tropical cyclones with the recently observed movement of the current tropical cyclone. Alternatively, forecasting techniques can focus on the forces responsible for tropical cyclone motion to produce a cyclone track forecast. Today, improved cyclone track forecasting techniques have enabled meteorological offices to warn residents in the affected areas before a tropical cyclone impact, and help to reduce the losses created by them.

Place, publisher, year, edition, pages
Taylor & Francis, 2018
Keywords
Tropical cyclone motion, track prediction techniques, track prediction accuracy
National Category
Meteorology and Atmospheric Sciences
Identifiers
urn:nbn:se:liu:diva-123197 (URN)10.1201/9781315166858 (DOI)9781315166858 (ISBN)
Available from: 2015-12-06 Created: 2015-12-06 Last updated: 2019-07-02
4. Local Feature Extraction—What Receptive Field Size Should Be Used?
Open this publication in new window or tab >>Local Feature Extraction—What Receptive Field Size Should Be Used?
2009 (English)In: Proceedings of International Conference on Image Processing, Computer Vision and Pattern Recognition, 2009Conference paper, Published paper (Refereed)
Abstract [en]

Biologically inspired hierarchical networks for image processing are based on parallel feature extraction across the image using feature detectors that have a limited Receptive Field (RF). It is, however, unclear how large these receptive fields should be. To study this, we ran systematic tests of various receptive field sizes using the same hierarchical network. After 40 epochs of training, we tested the network both by using similar but novel images of the same tropical cyclone that was used for training, and by using dissimilar images, depicting different cyclones. The results indicate that correct RF size is important for generalization in hierarchical networks, and that RF size should be chosen so that all RFs at least partially cover meaningful parts of the input image.

Keywords
pattern recognition, artificial neural networks, hierarchical networks
National Category
Computer Sciences
Identifiers
urn:nbn:se:liu:diva-55074 (URN)
Conference
International Conference on Image Processing, Computer Vision and Pattern Recognition
Available from: 2010-05-19 Created: 2010-04-28 Last updated: 2018-01-12Bibliographically approved
5. Bidirectional Hierarchical Neural Networks: Hebbian Learning Improves Generalization
Open this publication in new window or tab >>Bidirectional Hierarchical Neural Networks: Hebbian Learning Improves Generalization
2010 (English)In: Proceedings of the Fifth International Conference on Computer Vision Theory and Applications,  Volume 1, 2010, p. 105-111Conference paper, Published paper (Other academic)
Abstract [en]

Visual pattern recognition is a complex problem, and it has proven difficult to achieve satisfactorily instandard three-layer feed-forward artificial neural networks. For this reason, an increasing number ofresearchers are using networks whose architecture resembles the human visual system. These biologicallybasednetworks are bidirectionally connected, use receptive fields, and have a hierarchical structure, withthe input layer being the largest layer, and consecutive layers getting increasingly smaller. These networksare large and complex, and therefore run a risk of getting overfitted during learning, especially if smalltraining sets are used, and if the input patterns are noisy. Many data sets, such as, for example, handwrittencharacters, are intrinsically noisy. The problem of overfitting is aggravated by the tendency of error-drivenlearning in large networks to treat all variations in the noisy input as significant. However, there is one wayto balance off this tendency to overfit, and that is to use a mixture of learning algorithms. In this study, weran systematic tests on handwritten character recognition, where we compared generalization performanceusing a mixture of Hebbian learning and error-driven learning with generalization performance using pureerror-driven learning. Our results indicate that injecting even a small amount of Hebbian learning, 0.01 %,significantly improves the generalization performance of the network.

Keywords
generalization, image processing, bidirectional hierarchical neural networks, Hebbian learning, feature extraction, object recognition
National Category
Engineering and Technology
Identifiers
urn:nbn:se:liu:diva-77026 (URN)10.5220/0002835501050111 (DOI)978-989-674-028-3 (ISBN)
Conference
Fifth International Conference on Computer Vision Theory and Applications (VISAPP'10), May 17-21, 2010, Angers, France
Available from: 2012-08-28 Created: 2012-05-02 Last updated: 2016-01-13Bibliographically approved
6. A biologically based machine learning approach to tropical cyclone intensity forecasting
Open this publication in new window or tab >>A biologically based machine learning approach to tropical cyclone intensity forecasting
(English)Manuscript (preprint) (Other academic)
Abstract [en]

A biologically based ANN using four hierarchical levels, is trained and tested using temporal sequences of 2D inputs to forecast Tropical Cyclone (TC) intensity12, and 24 hours ahead in the Atlantic basin. We use five parallel input layers to feed infrared, ocean heat content, sea-level pressure, wind direction and wind speed images into the network. Forecasts are produced in the Saffir-Simpson hurricane intensity scale and are compared to the observed wind speeds in the TC best track data on two separate test datasets for validation. Forecasting accuracy is more than 95% for the test dataset containing temporal continuations of the TC lifecycle time-step images that are excluded from training, whereas, forecasting accuracy is between 30% and 55%, when images of a novel TC are used for testing. This result reveals that biologically inspired ANNs have a potential to be further developed into an effective TC intensity forecasting technique.

Keywords
Biologically based artificial neural networks; bi-directionally connected networks; Markov chain; temporal sequence learning; simple recurrent networks
National Category
Other Computer and Information Science
Identifiers
urn:nbn:se:liu:diva-123196 (URN)
Available from: 2015-12-06 Created: 2015-12-06 Last updated: 2018-01-10Bibliographically approved

Open Access in DiVA

fulltext(7028 kB)1969 downloads
File information
File name FULLTEXT02.pdfFile size 7028 kBChecksum SHA-512
b2ca80a8b341e7a7aa3fbae68ddde8bf6cea724f0e438549f31ac26f4362308de684f2a23a3526474ad8eeedce5ba53366b1c833af70dd706bdfed19eb23fed2
Type fulltextMimetype application/pdf
omslag(2786 kB)59 downloads
File information
File name COVER01.pdfFile size 2786 kBChecksum SHA-512
5580cc85b30c22164cd8443b1265c3ecdcba0d6b1bceb6d4f9724a82b80710b5d8e05787bea9eaa3f8568c5a2b2591bb0a9bb7db400541812c3f60b915909847
Type coverMimetype application/pdf
Order online >>

Other links

Publisher's full text

Authority records

Roy, Chandan

Search in DiVA

By author/editor
Roy, Chandan
By organisation
Human-Centered systemsFaculty of Science & Engineering
Other Computer and Information Science

Search outside of DiVA

GoogleGoogle Scholar
Total: 1970 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
Total: 3556 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • oxford
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf