A review of predictive uncertainty estimation with machine learning

Artificial Intelligence Review, Journal Year: 2024, Volume and Issue: 57(4)

Published: March 18, 2024

Language: Английский

---

A New Benchmark on Machine Learning Methodologies for Hydrological Processes Modelling: A Comprehensive Review for Limitations and Future Research Directions

Knowledge-Based Engineering and Sciences, Journal Year: 2023, Volume and Issue: 4(3), P. 65 - 103

Published: Dec. 31, 2023

The best practice of watershed management is through the understanding hydrological processes. As a matter fact, processes are highly associated with stochastic, non-linear, and non-stationary phenomena. Hydrological simulation modeling challenging issues in domains hydrology, climate environment. Hence, development machine learning (ML) models for solving those complex problems took essential place over past couple decades. It can be observed, data availability has increased remarkably, thus computational resources led to resurgence ML models’ development. been witnessed huge efforts on using facility several review researches have conducted. Literature studies approved capacity field hydrology classical “traditional models” based their forecastability, flexibility, precision, generalization, execution convergence speed. However, although potential merits were observed model’s development, limitations allied such as interpretability black-box models, practicality management, difficulty explain physical In this survey, an exhibition all published articles recognize research gaps direction. ultimate aim current survey establish new milestone interested environment researchers applications models.

Language: Английский

---

Quantitative improvement of streamflow forecasting accuracy in the Atlantic zones of Canada based on hydro-meteorological signals: A multi-level advanced intelligent expert framework

Ecological Informatics, Journal Year: 2024, Volume and Issue: 80, P. 102455 - 102455

Published: Jan. 4, 2024

Developing reliable streamflow forecasting models is critical for hydrological tasks such as improving water resource management, analyzing river patterns, and flood forecasting. In this research, the first time, an emerging multi-level TOPSIS (technique order preference by similarity to ideal solution) based hybridization comprised of Boruta classification regression tree (Boruta-CART) feature selection, multivariate variational mode decomposition (MVMD), a hybrid Convolutional Neural Network (CNN) Bidirectional Gated Recurrent Unit (CNN-BiGRU) deep learning was adopted multi-temporal (one three days ahead) forecast daily in Rivers Prince Edward Island, Canada. For aim, step, Boruta-CART selection technique determines most effective lagged components among all antecedent two-day information (i.e., t-1 t-2) hydro-meteorological features (from 2015 2020), including level, mean air temperature, heat degree days, total precipitation, dew point relative humidity Bear Winter Afterwards, (MVMD) decomposes input time series decrease complexity non-linearity non-stationary ones before feeding (DL) models. Here, CNN-GRU employed primary DL model, along with kernel extreme machine method (KELM), random function link (RVFL), CNN bidirectional recurrent neural network (CNN-BiRNN) comparative A scheme applying several performance measures like correlation coefficient (R), root square error (RMSE), reliability designed robustness assessment (MVM-CNN-BiGRU, MVM-CNN-BiRNN, MVM-RVFL, MVM-KELM) standalone The computational outcomes revealed that River, MVM-CNN-BiGRU, owing its best day ahead: score 1, R = 0.960, RMSE 0.098, 65.082; 0.999, 0.924, 0.33) outperformed other models, followed MVM-KELM, respectively. Moreover, MVM-CNN-BiGRU terms (one-day 0.890, 0.955, 0.274, 34.004; three-days 0.686, 0.330) superior provided expert system could be vital local decision-making process, absence modeling, during seasons reduce damage residential areas.

Language: Английский

---

Enhancing Streamflow Prediction Physically Consistently Using Process-Based Modeling and Domain Knowledge: A Review

Sustainability, Journal Year: 2024, Volume and Issue: 16(4), P. 1376 - 1376

Published: Feb. 6, 2024

Streamflow prediction (SFP) constitutes a fundamental basis for reliable drought and flood forecasting, optimal reservoir management, equitable water allocation. Despite significant advancements in the field, accurately predicting extreme events continues to be persistent challenge due complex surface subsurface watershed processes. Therefore, addition framework, numerous techniques have been used enhance accuracy physical consistency. This work provides well-organized review of more than two decades efforts SFP physically consistent way using process modeling flow domain knowledge. covers hydrograph analysis, baseflow separation, process-based (PBM) approaches. paper an in-depth analysis each technique discussion their applications. Additionally, existing are categorized, revealing research gaps promising avenues future research. Overall, this offers valuable insights into current state enhanced within consistent, knowledge-informed data-driven framework.

Language: Английский

---

Comparison of Machine Learning Algorithms for Merging Gridded Satellite and Earth-Observed Precipitation Data

Water, Journal Year: 2023, Volume and Issue: 15(4), P. 634 - 634

Published: Feb. 6, 2023

Gridded satellite precipitation datasets are useful in hydrological applications as they cover large regions with high density. However, not accurate the sense that do agree ground-based measurements. An established means for improving their accuracy is to correct them by adopting machine learning algorithms. This correction takes form of a regression problem, which measurements have role dependent variable and data predictor variables, together topography factors (e.g., elevation). Most studies this kind involve limited number algorithms conducted small region time period. Thus, results obtained through local importance provide more general guidance best practices. To generalizable contribute delivery practices, we here compare eight state-of-the-art correcting entire contiguous United States 15-year We use monthly from PERSIANN (Precipitation Estimation Remotely Sensed Information using Artificial Neural Networks) gridded dataset, earth-observed Global Historical Climatology Network database, version 2 (GHCNm). The suggest extreme gradient boosting (XGBoost) random forests most terms squared error scoring function. remaining can be ordered follows, worst: Bayesian regularized feed-forward neural networks, multivariate adaptive polynomial splines (poly-MARS), machines (gbm), (MARS), networks linear regression.

Language: Английский

---

Multi-Step Ahead Probabilistic Forecasting of Daily Streamflow Using Bayesian Deep Learning: A Multiple Case Study

Water, Journal Year: 2022, Volume and Issue: 14(22), P. 3672 - 3672

Published: Nov. 14, 2022

In recent decades, natural calamities such as drought and flood have caused widespread economic social damage. Climate change rapid urbanization contribute to the occurrence of disasters. addition, their destructive impact has been altered, posing significant challenges efficiency, equity, sustainability water resources allocation management. Uncertainty estimation in hydrology is essential for By quantifying associated uncertainty reliable hydrological forecasting, an efficient management plan obtained. Moreover, forecasting provides future information assist risk assessment. Currently, majority forecasts utilize deterministic approaches. Nevertheless, models cannot account intrinsic forecasted values. Using Bayesian deep learning approach, this study developed a probabilistic model that covers pertinent subproblem univariate time series multi-step ahead daily streamflow quantify epistemic aleatory uncertainty. The new implements sampling Long short-term memory (LSTM) neural network by using variational inference approximate posterior distribution. proposed method verified with three case studies USA horizons. LSTM point models, LSTM-BNN, BNN, Monte Carlo (MC) dropout (LSTM-MC), were applied comparison model. results show long (BLSTM) outperforms other terms reliability, sharpness, overall performance. reveal all outperformed lower RMSE value. Furthermore, BLSTM can handle data higher variation peak, particularly long-term compared models.

Language: Английский

---

Comparison of Tree-Based Ensemble Algorithms for Merging Satellite and Earth-Observed Precipitation Data at the Daily Time Scale

Hydrology, Journal Year: 2023, Volume and Issue: 10(2), P. 50 - 50

Published: Feb. 12, 2023

Merging satellite products and ground-based measurements is often required for obtaining precipitation datasets that simultaneously cover large regions with high density are more accurate than pure products. Machine statistical learning regression algorithms regularly utilized in this endeavor. At the same time, tree-based ensemble adopted various fields solving problems accuracy low computational costs. Still, information on which algorithm to select correcting contiguous United States (US) at daily time scale missing from literature. In study, we worked towards filling methodological gap by conducting an extensive comparison between three of category interest, specifically random forests, gradient boosting machines (gbm) extreme (XGBoost). We used data PERSIANN (Precipitation Estimation Remotely Sensed Information using Artificial Neural Networks) IMERG (Integrated Multi-satellitE Retrievals GPM) gridded datasets. also earth-observed Global Historical Climatology Network (GHCNd) database. The experiments referred entire US additionally included application linear benchmarking purposes. results suggest XGBoost best-performing among those compared. Indeed, mean relative improvements it provided respect (for case latter was run predictors as XGBoost) equal 52.66%, 56.26% 64.55% different predictor sets), while respective values 37.57%, 53.99% 54.39% 34.72%, 47.99% 62.61% gbm. Lastly, useful context investigated.

Language: Английский

---

Deep Huber quantile regression networks

Neural Networks, Journal Year: 2025, Volume and Issue: 187, P. 107364 - 107364

Published: March 5, 2025

Language: Английский

---

A Geographical Appraisal of Hydrological Drought—A Case Study

Springer geography, Journal Year: 2025, Volume and Issue: unknown, P. 29 - 50

Published: Jan. 1, 2025

Language: Английский

---

Hydrological post-processing for predicting extreme quantiles

Journal of Hydrology, Journal Year: 2023, Volume and Issue: 617, P. 129082 - 129082

Published: Jan. 6, 2023

Language: Английский

---