Tropical cyclones are among the most devastating natural disasters for human beings and the natural and manmade assets near to Atlantic basin. Estimating the current and future intensity of these powerful storms is crucial to protect life and property. Many methods and models exist for predicting the evolution of Atlantic basin cyclones, including numerical weather prediction models that simulate the dynamics of the atmosphere which require accurate measurements of the current state of the atmosphere (NHC, 2019). Often these models fail to capture dangerous aspects of storm evolution, such as rapid intensification (RI), in which a storm undergoes a steep increase in intensity over a short time. To improve prediction of these events, scientists have turned to statistical models to predict current and future intensity using readily collected satellite image data (Pradhan, in generalizing to unseen data, a result we confirm in this study. Therefore, building models for the estimating the current and future intensity of hurricanes is valuable and challenging.
In this study we focus on to estimating cyclone intensity using Geostationary Operational Environmental Satellite images. These images represent five spectral bands covering the visible and infrared spectrum. We have built and compared various types of deep neural models, including convolutional networks based on long short term memory models and convolutional regression models that have been trained to predict the intensity, as measured by maximum sustained wind speed.
By simply splitting the data into randomized training and testing sets, we are able to achieve high accuracy in predicting current storm intensity—similar to results reported by other deep-learning models (Pradhan,2018). However, we find that this result may be due primarily to having high degrees of similarity between the images in the training and testing set. To alleviate this issue, and provide a better estimate of the generalizability of our model to new storms, we instead consider a leave-one-storm-out training-testing split. This better simulates the expected performance of our models in practice, being deployed on a future storm. The 80/20 split models and leave at one storm models are dropped since similar image features over time provide the lower accuracy as well as lower loss.
We begin with the simplest model based only on a single image and attempt to predict the current intensity. After that, made a large tensor with combining all image arrays and made another model without considering about channel features. Moreover, another model have created using convolutional regression. Comparing all accuracies, low value models have identified since single image could not figure out the features of the cyclone categories. Therefore, move on to the capture the dynamic features of data using 2018). However, even the current-intensity prediction models have shown limited success sequence of images. Finnaly, preformed a long short term memory model to estimate the intensity using dynamic behaviors of the images.
By computing the mean absolute error between the storm wind speed and the model prediction, the long short term memory model experienced the lowest error values among the regression models. The long short term memory model resulted in mean absolute error values of 9.6478 knots, 11.6828 knots, 9.9167 knots and 8.2025 knots for the last four channels, which are 50%, 60%, 10% and 35% lower than the regression model.
These results show the importance of considering meaningful training testing splits to appropriately assess model performance as well as the importance of including dynamic behavior of the storm when creating intensity prediction models based on satellite images. They also lay the groundwork for developing models of more difficult problems such as predicting short-term evolution of storm intensity and rapid intensification events.
|Commitee:||Long, Hongwei, Hahn, William|
|School:||Florida Atlantic University|
|School Location:||United States -- Florida|
|Source:||MAI 82/6(E), Masters Abstracts International|
|Subjects:||Applied Mathematics, Statistics, Biostatistics|
|Keywords:||Tropical cyclones, Natural disasters, Werather prediction|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be