Study of The Relationship Between El Niño–Southern Oscillation (ENSO) And Extreme Heat in Taiwan
(2022) In Master's Theses in Mathematical Sciences MASM02 20221Mathematical Statistics
- Abstract
- El Niño-Southern Oscillation (ENSO) drives global weather and can be forecasted with a 6-12 months lead time. It’s also widely believed that there’s a relationship between phases of ENSO and extreme weather events such as extreme heat or precipitation.
This thesis focuses on the relationship between Oceanic Nino Index (ONI) in Nino region 3.4 and monthly temperature data in Taiwan, from 1980 to 2019. The temperature data is obtained from the NOAA Global Historical Climatology Network.
First an EM-like algorithm is used to complete the weather data matrix with missing values. Then an SVD decomposition separates the data matrix into temporal patterns and spatial patterns. The temporal patterns are modelled by a long short-term memory... (More) - El Niño-Southern Oscillation (ENSO) drives global weather and can be forecasted with a 6-12 months lead time. It’s also widely believed that there’s a relationship between phases of ENSO and extreme weather events such as extreme heat or precipitation.
This thesis focuses on the relationship between Oceanic Nino Index (ONI) in Nino region 3.4 and monthly temperature data in Taiwan, from 1980 to 2019. The temperature data is obtained from the NOAA Global Historical Climatology Network.
First an EM-like algorithm is used to complete the weather data matrix with missing values. Then an SVD decomposition separates the data matrix into temporal patterns and spatial patterns. The temporal patterns are modelled by a long short-term memory (LSTM) neural network with ONI as an external input signal. The spatial patterns are modelled separately by Universal Kriging with selected covariates. The space-time residuals are obtained and modelled by a Gaussian random field with semi-parametric non-stationary spatial covariance structures. Markov Chain Monte Carlo method is used together with Bayesian estimation to compute parameters in the non-stationary spatial covariance structures. Finally, extreme temperatures can be predicted by combining spatial-temporal patterns and prediction of the time series. Predictions with and without ONI were evaluated showing improved performance when using future ONI values. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9091875
- author
- Tang, Kean LU
- supervisor
- organization
- course
- MASM02 20221
- year
- 2022
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- ENSO, SVD, Gaussian random field, LSTM, Universal Kriging, Markov Chain Monte Carlo
- publication/series
- Master's Theses in Mathematical Sciences
- report number
- LUNFMS-3111-2022
- ISSN
- 1404-6342
- other publication id
- 2022:E52
- language
- English
- id
- 9091875
- date added to LUP
- 2022-08-15 17:57:59
- date last changed
- 2022-08-15 18:31:11
@misc{9091875, abstract = {{El Niño-Southern Oscillation (ENSO) drives global weather and can be forecasted with a 6-12 months lead time. It’s also widely believed that there’s a relationship between phases of ENSO and extreme weather events such as extreme heat or precipitation. This thesis focuses on the relationship between Oceanic Nino Index (ONI) in Nino region 3.4 and monthly temperature data in Taiwan, from 1980 to 2019. The temperature data is obtained from the NOAA Global Historical Climatology Network. First an EM-like algorithm is used to complete the weather data matrix with missing values. Then an SVD decomposition separates the data matrix into temporal patterns and spatial patterns. The temporal patterns are modelled by a long short-term memory (LSTM) neural network with ONI as an external input signal. The spatial patterns are modelled separately by Universal Kriging with selected covariates. The space-time residuals are obtained and modelled by a Gaussian random field with semi-parametric non-stationary spatial covariance structures. Markov Chain Monte Carlo method is used together with Bayesian estimation to compute parameters in the non-stationary spatial covariance structures. Finally, extreme temperatures can be predicted by combining spatial-temporal patterns and prediction of the time series. Predictions with and without ONI were evaluated showing improved performance when using future ONI values.}}, author = {{Tang, Kean}}, issn = {{1404-6342}}, language = {{eng}}, note = {{Student Paper}}, series = {{Master's Theses in Mathematical Sciences}}, title = {{Study of The Relationship Between El Niño–Southern Oscillation (ENSO) And Extreme Heat in Taiwan}}, year = {{2022}}, }