Assessing future rainfall uncertainties of climate change in Taiwan with a bootstrapped neural network‐based downscaling model

Li, Chi‐Yu; Lin, Shy‐Der; Chuang, C. K.; Hu, Yen‐Li

doi:10.1111/wej.12443

Cited by 6 publications

(3 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Statistical downscaling methods (SDMs) are commonly used for this purpose. Generally, SDMs use ground rainfall observations over a given region to calibrate the co-located GCM output such that the calibrated model can be used to predict future rainfall with the impact of climate change [7][8][9].…”

Section: Introductionmentioning

confidence: 99%

“…With advances in machine learning, recent developments in statistical downscaling methods have shifted from using traditional statistical methods to machine learning (or deep learning)-based methods in the past two decades [7][8][9][10]. In addition, in recent years, with improved hardware computing speed, machine learning has come to be widely applied in various fields [11,12].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Deep Learning-Based Downscaling Method Considering the Impact on Typhoons to Future Precipitation in Taiwan

Lin,

Zhu,

Wang

2024

Atmosphere

Self Cite

View full text Add to dashboard Cite

This study proposes a deep neural network (DNN)-based downscaling model incorporating kernel principal component analysis (KPCA) to investigate the precipitation uncertainty influenced by typhoons in Taiwan, which has a complex island topography. The best tracking data of tropical cyclones from the Joint Typhoon Warning Center (JTWC) are utilized to calculate typhoon and non-typhoon precipitation. KPCA is applied to extract nonlinear features of the BCC-CSM1-1 (Beijing Climate Center Climate System Model version 1.1) and CanESM2 (second-generation Canadian Earth System Model) GCM models. The length of the data used in the two GCM models span from January 1950 to December 2005 (historical data) and from January 2006 to December 2099 (scenario out data). The rainfall data are collected from the weather stations in Taichung and Hualien (cities of Taiwan) operated by the Central Weather Administration (CWA), Taiwan. The period of rainfall data in Taichung and in Hualien spans from January 1950 to December 2005. The proposed model is constructed with features extracted from the GCMs and historical monthly precipitation from Taichung and Hualien. The model we have built is used to estimate monthly precipitation and uncertainty in both Taichung and Hualien for future scenarios (rcp 4.5 and 8.5) of the GCMs. The results suggest that, in Taichung and Hualien, the summer precipitation is mostly within the normal range. The rainfall in the long term (January 2071 to December 2080) for both Taichung and Hualien typically fall between 100 mm and 200 mm. In the long term, the dry season (January to April, November, and December) precipitation for Taichung and that in the wet season (May to October) for Hualien are less and more affected by typhoons, respectively. The dry season precipitation is more affected by typhoons in Taichung than Hualien. In both Taichung and Hualien, the long-term probability of rainfall exceeding the historical average in the dry season is higher than that in the wet season.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Deep Learning-Based Downscaling Method Considering the Impact on Typhoons to Future Precipitation in Taiwan

Lin,

Zhu,

Wang

2024

Atmosphere

Self Cite

View full text Add to dashboard Cite

show abstract

“…Focusing on Taiwan, evidence of altering rainfall regime induced by climate change was indicated by several recent studies [32][33][34][35][36][37][38]. However, little research has examined the changes in streamflow regime in Taiwan, with the exception of Yeh et al [39] who investigated the long-term streamflow trend in northern Taiwan using the Mann-Kendall test.…”

Section: Introductionmentioning

confidence: 99%

Nonstationary Analyses of the Maximum and Minimum Streamflow in Tamsui River Basin, Taiwan

Shiau

Liu

2021

Water

View full text Add to dashboard Cite

This study aims to detect non-stationarity of the maximum and minimum streamflow regime in Tamsui River basin, northern Taiwan. Seven streamflow gauge stations, with at least 27-year daily records, are used to characterize annual maximum 1- and 2-day flows and annual minimum 1-, 7-, and 30-day flows. The generalized additive models for location, scale, and shape (GAMLSS) are used to dynamically detect evolution of probability distributions of the maximum and minimum flow indices with time. Results of time-covariate models indicate that stationarity is only noted in the 4 maximum flow indices out of 35 indices. This phenomenon indicates that the minimum flow indices are vulnerable to changing environments. A 16-category distributional-change scheme is employed to classify distributional changes of flow indices. A probabilistic distribution with complex variations of mean and variance is prevalent in the Tamsui River basin since approximate one third of flow indices (34.3%) belong to this category. To evaluate impacts of dams on streamflow regime, a dimensionless index called the reservoir index (RI) serves as an alternative covariate to model nonstationary probability distribution. Results of RI-covariate models indicate that 7 out of 15 flow indices are independent of RI and 80% of the best-fitted RI-covariate models are generally worse than the time-covariate models. This fact reveals that the dam is not the only factor in altering the streamflow regime in the Tamsui River, which is a significant alteration, especially the minimum flow indices. The obtained distributional changes of flow indices clearly indicate changes in probability distributions with time. Non-stationarity in the Tamsui River is induced by climate change and complex anthropogenic interferences.

show abstract

Applying the Deep Neural Network to Estimate Future Trend and Uncertainty of Rainfall under Climate Change

Lin

Yeh

Zhu

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

This study proposes a deep neural network (DNN) as a downscaling framework to compare original variables and nonlinear data features extracted by kernel principal component analysis (KPCA). It uses them as learning data for DNN downscaling models to assess future regional rainfall trends and uncertainties in islands with complex terrain. This study takes Taichung and Hualien in Taiwan as examples. It collects data in different emission scenarios (RCP 4.5, RCP 8.5) simulated by two Global Climate Models: ACCESS and CSMK3, in the Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC), and monthly rainfall data of case regions from January 1950 to December 2005 in the Central Weather Bureau in Taiwan. DNN model parameters are optimized based on historical scenarios to estimate the trends and uncertainties of future monthly rainfall in the case regions. A multivariate linear regression is used as a baseline model to compare their effectiveness. The simulated results show that by both ACCESS and CSMK3, the dimensionless root mean squared error (RMSE) of KPCA was higher than that of the original variables in Taichung and Hualien. According to the analysis of three-class classification (according to the arrangement in descending power of historical rainfall, the predicted rainfall is divided into three ranges, low, normal, and high, marked by 30% and 70% of monthly rainfall), the wet season rainfall at the two stations is concentrated in the normal range. The probability of rainfall increase will improve in the dry season and will reduce in the wet season in the mid-term to long-term. The future wet season rainfall in Hualien has the highest variability. It ranges from 201 mm to 300 mm, with representative concentration pathways RCP 4.5 much higher than RCP 8.5. The median percentage increase and decrease in RCP 8.5 are higher than in RCP 4.5. This indicates that RCP 8.5 has a greater impact on future monthly rainfall.

show abstract

Assessing future rainfall uncertainties of climate change in Taiwan with a bootstrapped neural network‐based downscaling model

Cited by 6 publications

References 61 publications

A Deep Learning-Based Downscaling Method Considering the Impact on Typhoons to Future Precipitation in Taiwan

A Deep Learning-Based Downscaling Method Considering the Impact on Typhoons to Future Precipitation in Taiwan

Nonstationary Analyses of the Maximum and Minimum Streamflow in Tamsui River Basin, Taiwan

Applying the Deep Neural Network to Estimate Future Trend and Uncertainty of Rainfall under Climate Change

Contact Info

Product

Resources

About