東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Modeling Groundwater Recharge and St...

Achieng, Kevin Otieno.

FindBook

Google Book

Amazon

博客來

Modeling Groundwater Recharge and Streamflow and Potential Changes in the Future Climate: Bayesian Averaging and Machine Learning Paradigm.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Modeling Groundwater Recharge and Streamflow and Potential Changes in the Future Climate: Bayesian Averaging and Machine Learning Paradigm./
作者:	Achieng, Kevin Otieno.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2019,
面頁冊數:	275 p.
附註:	Source: Dissertations Abstracts International, Volume: 81-05, Section: B.
Contained By:	Dissertations Abstracts International81-05B.
標題:	Civil engineering. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=22616953
ISBN:	9781088394946

Modeling Groundwater Recharge and Streamflow and Potential Changes in the Future Climate: Bayesian Averaging and Machine Learning Paradigm.
Achieng, Kevin Otieno.

Modeling Groundwater Recharge and Streamflow and Potential Changes in the Future Climate: Bayesian Averaging and Machine Learning Paradigm. - Ann Arbor : ProQuest Dissertations & Theses, 2019 - 275 p.

Source: Dissertations Abstracts International, Volume: 81-05, Section: B.

Thesis (Ph.D.)--University of Wyoming, 2019.

This item must not be sold to any third party vendors.

Groundwater sustains nearly one-third of the global population, whereas the surface water (especially the streamflow) supplies the rest of water demand. About 90% of the United States' rural population depends on groundwater. About 70% of global irrigation water demand comes from groundwater. Climate models have been used to predict streamflow and groundwater recharge for the current and future periods. However, due to parametric and structural differences among the climate models, they often don't produce uncertainty free hydrological variables. In the face of climate change and the uncertainties associated with the climate models, these hydrological variables from the climate models are increasingly being averaged in Bayesian frameworks. However, these Bayesian-based hydrological studies often use an assumed single Bayesian prior which may lead to ill-poised posterior and predictions. Besides the Bayesian frameworks, machine learning is also increasingly becoming popular in water resource based studies. However, most machine learning based studies have focused on surface water processes with limited studies on groundwater recharge and its associated changes in future climate. Therefore, this dissertation consists of two main scientific themes. The first one is to investigate and recommend appropriate Bayesian prior for averaging climate models based streamflow and groundwater recharge predictions. The second one is to investigate the plausibility of using machine learning framework to average groundwater recharge from multiple climate models, and predict future changes in groundwater recharge. The groundwater recharge is averaged in twelve Bayesian frameworks and thirteen machine learning frameworks from ten regional climate models. The changes in groundwater recharge in future climate (2038-2069) with respect to the current period (1968-1998) are also modeled using both the Bayesian framework and machine learning approach based on ten regional climate models. The developed modeling approaches are illustrated and tested in two case study river basins of relatively small and large size. The results suggest that the non-Empirical Bayes g-Local (non-EBL)-based Bayesian priors are more suitable for averaging streamflow and groundwater recharge within the Bayesian framework. The predicted future groundwater recharge may increase or decreases depending on the size and location of basins. The results further suggest that the deep neural networks perform better in combining recharge from multiple climate models than the rest of the machine learning models. It is demonstrated that the choice of the prior probability affects posterior inclusion probability values of the individual climate models that are averaged within the Bayesian framework, and increasing the complexity of the neural networks by increasing the number of hidden layers does not significantly increase their performance in groundwater recharge simulation.

ISBN: 9781088394946Subjects--Topical Terms:

860360
Civil engineering.
Subjects--Index Terms:

Artificial/Deep Neural Networks (A/DNNs)

Modeling Groundwater Recharge and Streamflow and Potential Changes in the Future Climate: Bayesian Averaging and Machine Learning Paradigm.
LDR:04356nmm a2200409 4500 001 2265897
005 20200529130312.5
008 220629s2019 ||||||||||||||||| ||eng d
020 $a 9781088394946
035 $a (MiAaPQ)AAI22616953
035 $a AAI22616953
040 $a MiAaPQ $c MiAaPQ
100 1 $a Achieng, Kevin Otieno. $3 3543078
245 1 0 $a Modeling Groundwater Recharge and Streamflow and Potential Changes in the Future Climate: Bayesian Averaging and Machine Learning Paradigm.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2019
300 $a 275 p.
500 $a Source: Dissertations Abstracts International, Volume: 81-05, Section: B.
500 $a Advisor: Jianting, Zhu.
502 $a Thesis (Ph.D.)--University of Wyoming, 2019.
506 $a This item must not be sold to any third party vendors.
506 $a This item must not be added to any third party search indexes.
520 $a Groundwater sustains nearly one-third of the global population, whereas the surface water (especially the streamflow) supplies the rest of water demand. About 90% of the United States' rural population depends on groundwater. About 70% of global irrigation water demand comes from groundwater. Climate models have been used to predict streamflow and groundwater recharge for the current and future periods. However, due to parametric and structural differences among the climate models, they often don't produce uncertainty free hydrological variables. In the face of climate change and the uncertainties associated with the climate models, these hydrological variables from the climate models are increasingly being averaged in Bayesian frameworks. However, these Bayesian-based hydrological studies often use an assumed single Bayesian prior which may lead to ill-poised posterior and predictions. Besides the Bayesian frameworks, machine learning is also increasingly becoming popular in water resource based studies. However, most machine learning based studies have focused on surface water processes with limited studies on groundwater recharge and its associated changes in future climate. Therefore, this dissertation consists of two main scientific themes. The first one is to investigate and recommend appropriate Bayesian prior for averaging climate models based streamflow and groundwater recharge predictions. The second one is to investigate the plausibility of using machine learning framework to average groundwater recharge from multiple climate models, and predict future changes in groundwater recharge. The groundwater recharge is averaged in twelve Bayesian frameworks and thirteen machine learning frameworks from ten regional climate models. The changes in groundwater recharge in future climate (2038-2069) with respect to the current period (1968-1998) are also modeled using both the Bayesian framework and machine learning approach based on ten regional climate models. The developed modeling approaches are illustrated and tested in two case study river basins of relatively small and large size. The results suggest that the non-Empirical Bayes g-Local (non-EBL)-based Bayesian priors are more suitable for averaging streamflow and groundwater recharge within the Bayesian framework. The predicted future groundwater recharge may increase or decreases depending on the size and location of basins. The results further suggest that the deep neural networks perform better in combining recharge from multiple climate models than the rest of the machine learning models. It is demonstrated that the choice of the prior probability affects posterior inclusion probability values of the individual climate models that are averaged within the Bayesian framework, and increasing the complexity of the neural networks by increasing the number of hidden layers does not significantly increase their performance in groundwater recharge simulation.
590 $a School code: 0264.
650 4 $a Civil engineering. $3 860360
650 4 $a Hydrologic sciences. $3 3168407
650 4 $a Natural resource management. $3 589570
650 4 $a Water resources management. $3 794747
653 $a Artificial/Deep Neural Networks (A/DNNs)
653 $a Bayesian Formulations
653 $a Climate Models
653 $a Groundwater
653 $a Machine Learning
653 $a Support Vector Regressions
690 $a 0543
690 $a 0388
690 $a 0528
690 $a 0595
710 2 $a University of Wyoming. $b Civil & Architectural Engineering. $3 1023756
773 0 $t Dissertations Abstracts International $g 81-05B.
790 $a 0264
791 $a Ph.D.
792 $a 2019
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=22616953