Deep learning super-resolution for temperature data downscaling: a comprehensive study using residual networks DOI Creative Commons
Sumant Jha, Vivek Gupta, Priyank J. Sharma

et al.

Frontiers in Climate, Journal Year: 2025, Volume and Issue: 7

Published: May 13, 2025

Extreme weather events such as heatwaves, cyclones, floods, wildfires, and droughts are becoming more frequent due to climate change. Climate change causes shifts in biodiversity impacts agriculture, forest ecosystems, water resources at a regional scale. However, study those the scale, spatial resolution provided by general circulation models (GCMs) reanalysis products is inadequate. This evaluates advanced deep learning for downscaling European Center Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) 2-m temperature data factor of 10 (i.e., ranging approximately from 250 25 km resolution) region spanning 50° 100° E 0° N. We concentrate on gradually improving with help residual networks. compare baseline Super-Resolution Convolutional Neural Network (SRCNN) model two models: Very Deep (VDSR) Enhanced (EDSR) assess impact networks architectural improvements. The results indicate that VDSR EDSR significantly outperform SRCNN. Specifically, increases Peak Signal-to-Noise Ratio (PSNR) 4.27 dB 5.23 dB. These also enhance Structural Similarity Index Measure (SSIM) 0.1263 0.1163, respectively, indicating better image quality. Furthermore, improvements 3°C error threshold observed, showing 2.10 2.16%, respectively. An explainable artificial intelligence (AI) technique called saliency map analysis insights into performance. Complex terrain areas, Himalayas Tibetan Plateau, benefit most these advancements. findings suggest employing networks, EDSR, accuracy over approach holds promise future applications other atmospheric variables.

Language: Английский

Co-benefits of carbon neutrality in enhancing and stabilizing solar and wind energy DOI Open Access
Yadong Lei, Zhili Wang,

Deying Wang

et al.

Nature Climate Change, Journal Year: 2023, Volume and Issue: 13(7), P. 693 - 700

Published: June 5, 2023

Language: Английский

Citations

75
Machine Learning Methods in Weather and Climate Applications: A Survey DOI Creative Commons

Liuyi Chen,

Bocheng Han, Xuesong Wang

et al.

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(21), P. 12019 - 12019

Published: Nov. 3, 2023

With the rapid development of artificial intelligence, machine learning is gradually becoming popular for predictions in all walks life. In meteorology, it competing with traditional climate dominated by physical models. This survey aims to consolidate current understanding Machine Learning (ML) applications weather and prediction—a field growing importance across multiple sectors, including agriculture disaster management. Building upon an exhaustive review more than 20 methods highlighted existing literature, this pinpointed eight techniques that show particular promise improving accuracy both short-term medium-to-long-term forecasts. According survey, while ML demonstrates significant capabilities prediction, its application forecasting remains limited, constrained factors such as intricate variables data limitations. Current literature tends focus narrowly on either or forecasting, often neglecting relationship between two, well general neglect modeling structure recent advances. By providing integrated analysis models spanning different time scales, bridge these gaps, thereby serving a meaningful guide future interdisciplinary research rapidly evolving field.

Language: Английский

Citations

49
Customized deep learning for precipitation bias correction and downscaling DOI Creative Commons
Fang Wang, Di Tian, Mark Carroll

et al.

Geoscientific model development, Journal Year: 2023, Volume and Issue: 16(2), P. 535 - 556

Published: Jan. 25, 2023

Abstract. Systematic biases and coarse resolutions are major limitations of current precipitation datasets. Many deep learning (DL)-based studies have been conducted for bias correction downscaling. However, it is still challenging the approaches to handle complex features hourly precipitation, resulting in incapability reproducing small-scale features, such as extreme events. This study developed a customized DL model by incorporating loss functions, multitask physically relevant covariates correct downscale data. We designed six scenarios systematically evaluate added values weighted learning, atmospheric compared regular statistical approaches. The models were trained tested using Modern-era Retrospective Analysis Research Applications version 2 (MERRA2) reanalysis Stage IV radar observations over northern coastal region Gulf Mexico on an time scale. found that all with functions performed notably better than other conventional quantile mapping-based approach at hourly, daily, monthly scales well extremes. Multitask showed improved performance capturing fine events accounting highly aggregated scales, while improvement not large from functions. show can datasets provide estimates spatial temporal where methods experience challenges.

Language: Английский

Citations

26
Improving CMIP6 Atmospheric River Precipitation Estimation by Cycle‐Consistent Generative Adversarial Networks DOI
Yuan Tian, Yang Zhao, Jianping Li

et al.

Journal of Geophysical Research Atmospheres, Journal Year: 2024, Volume and Issue: 129(14)

Published: July 12, 2024

Abstract Given the important role of Atmospheric River precipitation (ARP) in global hydrological cycle, accurate representation ARP is significant. However, general circulation models (GCMs) demonstrate bias simulating ARP. The target this study to quantify performance intensity/frequency for CMIP6 simulations, and further improve estimation using Cycle‐Consistent Generative Adversarial Networks (CycleGAN) with highlighting more features under warming background. findings are as follows: (a) although reserved‐optimal overall reproduces observation, it still underestimated at stronger river (AR) scales, particularly AR highly active mid‐latitude regions. (b) CycleGAN‐based correction approach markedly diminishes simulations within most scales among both four Moreover, regions significant improvement, which mainly due reduction strongest scale. (c) Relative reference period (1986–2005), scale increase notably 3°C level, an average value 373.3% intensity 415.9% frequency key before correction, 451.9% 492.5% after correction. results illustrate that CycleGAN can effectively GCMs, early warning implies future strong extreme should potentially surpass current expected.

Language: Английский

Citations

16
Recently emerging trends in big data analytic methods for modeling and combating climate change effects DOI Creative Commons
Anayo Chukwu Ikegwu, Henry Friday Nweke, Emmanuel O.C. Mkpojiogu

et al.

Energy Informatics, Journal Year: 2024, Volume and Issue: 7(1)

Published: Feb. 7, 2024

Abstract Big climate change data have become a pressing issue that organizations face with methods to analyze generated from various types. Moreover, storage, processing, and analysis of activities are becoming very massive, challenging for the current algorithms handle. Therefore, big analytics designed significantly large amounts required enhance seasonal monitoring understand ascertain health risks change. In addition, would improve allocation, utilisation natural resources. This paper provides an extensive discussion analytic investigates how sustainability issues can be analyzed through these approaches. We further present methods, strengths, weaknesses, essence analyzing using methods. The common datasets, implementation frameworks modeling, future research directions were also presented clarity compelling challenges. method is well-timed solve inherent easy realization sustainable development goals.

Language: Английский

Citations

10
Storm surge modeling in the AI era: Using LSTM-based machine learning for enhancing forecasting accuracy DOI
Stefanos Giaremis, Noujoud Nader, Clint Dawson

et al.

Coastal Engineering, Journal Year: 2024, Volume and Issue: 191, P. 104532 - 104532

Published: April 20, 2024

Language: Английский

Citations

9
Impacts of Climate Change on Extreme Climate Indices in Türkiye Driven by High-Resolution Downscaled CMIP6 Climate Models DOI Open Access
Berkin Gümüş, Sertaç Oruç, İsmail Yücel

et al.

Sustainability, Journal Year: 2023, Volume and Issue: 15(9), P. 7202 - 7202

Published: April 26, 2023

In this study, the latest release of all available Coupled Model Intercomparison Project Phase 6 (CMIP6) climate models with two future scenarios Shared Socio-Economic Pathways, SSP2-4.5 and SSP5-8.5, over period 2015–2100 are utilized in diagnosing extremes Türkiye. Coarse-resolution were downscaled to a 0.1° × (~9 km) spatial resolution using European Centre for Medium-Range Weather Forecasts Reanalysis 5-Land (ERA5-Land) dataset based on three types quantile mapping: mapping, detrended delta mapping. The temporal variations 12 extreme precipitation indices (EPIs) temperature (ETIs) from 2015 2100 consistently suggest drier conditions, addition more frequent severe warming Türkiye, under scenarios. SSP5-8.5 scenario indicates water stress than scenario; total decreases up 20% Aegean Mediterranean regions Precipitation indicate decrease frequency heavy rains but an increase very also increasing amount rain days. Temperature such as coldest, warmest, mean daily maximum expected across indicating conditions by 7.5 °C end century. Additionally, coldest maximums exhibit higher variability change subregions Aegean, Southeastern Anatolia, Marmara, Türkiye while showed greater sensitivity Black Sea, Central Eastern Anatolia regions.

Language: Английский

Citations

21
Spatiotemporal changes in future precipitation of Afghanistan for shared socioeconomic pathways DOI Creative Commons

Sayed Tamim Rahimi,

Ziauddin Safari, Shamsuddin Shahid

et al.

Heliyon, Journal Year: 2024, Volume and Issue: 10(7), P. e28433 - e28433

Published: March 21, 2024

Global warming induces spatially heterogeneous changes in precipitation patterns, highlighting the need to assess these at regional scales. This assessment is particularly critical for Afghanistan, where agriculture serves as primary livelihood population. New global climate model (GCM) simulations have recently been released established shared socioeconomic pathways (SSPs). requires evaluating projected under new scenarios and subsequent policy updates. research employed six GCMs from CMIP6 project spatial temporal across Afghanistan all SSPs, including SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5. The were bias-corrected using Precipitation Climatological Center's (GPCC) monthly gridded data with a 1.0° resolution. Subsequently, change factor was calculated both near future (2020-2059) distant (2060-2099). projections' multi-model ensemble (MME) revealed increased most of SSPs higher emissions scenarios. showed substantial increase summer around 50%, SSP1-1.9 southwestern region, while decline over 50% northwestern region until 2100. annual northwest up 15% SSP1-2.6. SSP2-4.5 20% certain eastern regions far future. Furthermore, rise approximately SSP3-7.0 expected central western However, it crucial note that exhibit considerable uncertainty among different GCMs.

Language: Английский

Citations

7
Evaluation of bias correction techniques for generating high-resolution daily temperature projections from CMIP6 models DOI
Xinyi Li, Zhong Li

Climate Dynamics, Journal Year: 2023, Volume and Issue: 61(7-8), P. 3893 - 3910

Published: April 13, 2023

Language: Английский

Citations

15
Challenges of typical inter-basin water transfer projects in China: Anticipated impacts of climate change on streamflow and hydrological drought under CMIP6 DOI
Lianzhou Wu, Xiaoling Su, Te Zhang

et al.

Journal of Hydrology, Journal Year: 2023, Volume and Issue: 627, P. 130437 - 130437

Published: Nov. 6, 2023

Language: Английский

Citations

15