Article
Environmental Sciences
Quanzhou Li, Yun Pan, Chong Zhang, Huili Gong
Summary: This study quantifies the uncertainties of groundwater storage (GWS) estimates in mainland China using GRACE satellites. It utilizes multiple data sources and applies the Bayesian model averaging approach to derive optimal estimates of GWS changes. The results show that the annual GWS trend in mainland China is -1.93 mm/yr with an uncertainty of 4.50 mm/yr, highlighting the importance of considering multi-source uncertainties when using GRACE data.
Article
Environmental Sciences
Maryam Gharekhani, Ata Allah Nadiri, Rahman Khatibi, Sina Sadeghfam, Asghar Asghari Moghaddam
Summary: Bayesian Model Averaging (BMA) was used in this study to assess groundwater vulnerability in a study area related to Lake Urmia. The results showed higher uncertainties in areas with higher vulnerability.
JOURNAL OF ENVIRONMENTAL MANAGEMENT
(2022)
Article
Engineering, Civil
Jina Yin, Frank T-C Tsai, Shih-Chieh Kao
Summary: Alluvial aquifers are complex in nature and developing a reliable groundwater model for them is challenging. This study presents a Bayesian multi-model uncertainty quantification framework to account for model parameter uncertainty in alluvial groundwater modeling, improving our understanding of groundwater dynamics and prediction reliability. The methodology was applied to the agriculturally intensive Mississippi River alluvial aquifer in Northeast Louisiana, demonstrating the importance of explicitly quantifying model uncertainty in improving groundwater level predictions.
JOURNAL OF HYDROLOGY
(2021)
Article
Environmental Sciences
Ting Zhou, Xiaohu Wen, Qi Feng, Haijiao Yu, Haiyang Xi
Summary: In this study, the ability of various data sources and machine learning models in predicting groundwater levels was investigated. The results showed that the applied data inputs and the ensemble BMA model significantly improved the predictive accuracy of the single models.
Article
Environmental Sciences
Jina Yin, Josue Medellin-Azuara, Alvar Escriva-Bou, Zhu Liu
Summary: This study introduces a novel machine learning-based groundwater ensemble modeling framework combined with Bayesian model averaging to predict groundwater storage change in agricultural regions with improved reliability.
SCIENCE OF THE TOTAL ENVIRONMENT
(2021)
Article
Environmental Sciences
Pedram Darbandsari, Paulin Coulibaly
Summary: This study evaluates the impact of different hydrologic models on the performance of the hydrologic uncertainty processor (HUP) and proposes a multimodel Bayesian postprocessor (HUP-BMA). Results demonstrate the superiority of HUP-BMA in quantifying hydrologic uncertainty and forecasting compared to traditional HUP and BMA methods.
WATER RESOURCES RESEARCH
(2021)
Article
Engineering, Civil
Mahrokh Moknatian, Rajith Mukundan
Summary: This study conducted uncertainty analysis on SWAT-HS model using multiple objective functions and found that using multiple objective functions can better capture prediction uncertainty. The Bayesian Model Averaging method was applied to quantify overall prediction uncertainty, and it was observed that uncertainty intervals estimated using multiple objective functions were wider than those using a single objective function. The study demonstrates that using multiple objective functions is an effective option in streamflow modeling and uncertainty analysis.
JOURNAL OF HYDROLOGY
(2023)
Article
Meteorology & Atmospheric Sciences
Titas Ganguly, Dhyan S. Arya
Summary: This study proposes a comprehensive framework for ranking and generating ensemble data for GCMs and validates it using precipitation and temperature data from India. The results show that Bayesian framework-based rankings outperform other methods and the orthonormal distribution-based Bayesian ranking performs well in precipitation. The weighted ensemble has closer proximity to the observed data distribution compared to the traditional mathematical average. The assessment of projected extremes shows varying levels of confidence in the attribution of precipitation extremes to anthropogenic causes under different climate scenarios.
INTERNATIONAL JOURNAL OF CLIMATOLOGY
(2023)
Article
Environmental Sciences
Gang Li, Zhangjun Liu, Jingwen Zhang, Huiming Han, Zhangkang Shu
Summary: This study used three deep learning models to predict water levels at five stations of Poyang Lake. The results were then synthesized using Bayesian model averaging to improve prediction accuracy. The uncertainty of the models was also analyzed. The proposed framework achieved satisfactory accuracy in water level prediction and can be easily applied to other hydrological variables.
SCIENCE OF THE TOTAL ENVIRONMENT
(2024)
Article
Environmental Sciences
Jose-Luis Molina, Jose-Luis Garcia-Arostegui
Summary: This research aims to analyze and model the relationship between binomial rainfall and groundwater levels. It uses Bayesian Causal Reasoning (BCR) based on Bayesian Theorem to capture the inherent causality in the data. The methodology includes classic regression analysis and Bayesian Causal Modelling Translation (BCMT) with iterative steps. This innovative methodology has been successfully applied to aquifer management in the Campo de Cartagena groundwater body in Spain.
SCIENCE OF THE TOTAL ENVIRONMENT
(2023)
Article
Agronomy
Ommolbanin Bazrafshan, Mohammad Ehteram, Zahra Gerkaninezhad Moshizi, Sajad Jamshidi
Summary: Accurate crop yield prediction is of great importance for ensuring food security. This study adopts a Multilayer Perceptron (MLP) model with optimization algorithms to predict wheat yield and demonstrates the effectiveness of ensemble modeling in improving prediction accuracy.
AGRICULTURAL WATER MANAGEMENT
(2022)
Article
Environmental Sciences
Kaj-Ivar van der Wijst, Francesco Bosello, Shouro Dasgupta, Laurent Drouet, Johannes Emmerling, Andries Hof, Marian Leimbach, Ramiro Parrado, Franziska Piontek, Gabriele Standardi, Detlef van Vuuren
Summary: The cost-benefit analysis of climate change heavily relies on the choice of damage function, which is difficult to obtain credible information for. Comparison of multiple models reveals that the optimum temperature for climate change may be lower than previously estimated. Economic analyses of global climate change have been criticized for their inadequate representation of the damages caused. The study developed and applied aggregate damage functions in three different Integrated Assessment Models (IAMs) with varying levels of complexity, including a wide but incomplete range of climate change impacts. The results show that global damages, based on medium estimates for damage functions, are projected to be approximately 10% to 12% of the GDP by 2100 under a baseline scenario with 3 degrees C temperature increase, and about 2% under a well-below 2 degrees C scenario. These damages exceed previous estimates from benefit-cost studies and indicate that optimal temperatures might be below 2 degrees C when considering damages and discount rates. Furthermore, the benefit-cost ratios range from 1.5 to 3.9, even without factoring in unaccounted damages such as biodiversity losses, health impacts, and tipping points.
NATURE CLIMATE CHANGE
(2023)
Article
Computer Science, Information Systems
Yong Yu, Xiaosheng Si, Changhua Hu, Jianfei Zheng, Jianxun Zhang
Summary: The study utilizes the Bayesian-updated ECM algorithm and modified Bayesian-model-averaging method to effectively address the uncertainties of model parameters and the degradation model in online RUL estimation. Simulation studies demonstrate that the proposed fusion algorithm significantly improves the prediction of gyroscope RUL.
SCIENCE CHINA-INFORMATION SCIENCES
(2021)
Article
Engineering, Environmental
Mi Tian, Hao Fan, Zimin Xiong, Lihua Li
Summary: Accurate and reliable predictions of debris-flow volume are crucial for assessing potential hazards and risks. This paper proposes a data-driven ensemble model that combines deterministic machine learning methods and Bayesian model averaging to probabilistically forecast debris-flow volume. The feasibility of the approach is demonstrated using rainfall-induced debris flows in Taiwan, and the performance of individual models and the ensemble model is evaluated.
BULLETIN OF ENGINEERING GEOLOGY AND THE ENVIRONMENT
(2023)
Article
Engineering, Civil
Sharvil Alex Faroz, Siddhartha Ghosh
Summary: This paper introduces a method to estimate the corrosion rate of reinforced concrete structures through instrument calibration using probabilistic measurement error models within a Bayesian framework and hyper-robust calibration approach. The proposed approach is demonstrated for a linear polarisation resistance instrument and found to be suitable for the study case, as well as general enough to be applied to other NDT instruments.