Earth and Space Science最新文献

{"title":"The Birth of an Underground Gas Storage in a Depleted Gas Reservoir—Results From Integrated Seismic and Ground Deformation Monitoring","authors":"E. Priolo,&nbsp;I. Zinno,&nbsp;M. Guidarelli,&nbsp;M. Romanelli,&nbsp;R. Lanari,&nbsp;D. Sandron,&nbsp;M. Garbin,&nbsp;L. Peruzza,&nbsp;M. A. Romano,&nbsp;D. Zuliani,&nbsp;L. Tunini,&nbsp;A. Magrin","doi":"10.1029/2023EA003275","DOIUrl":"https://doi.org/10.1029/2023EA003275","url":null,"abstract":"<p>In the Po Valley (Italy), near Lodi, a depleted methane reservoir was recently converted into an underground gas storage (UGS) facility. We describe the new monitoring infrastructure that integrates seismic and ground deformation capabilities. We also present results obtained before and after the start of UGS operation, namely: (a) the so-called “background,” for seismicity and deformation estimated over several decades before the UGS; (b) the “seismic baseline,” assessed using the new monitoring network over nearly 2 years before the gas injection began; and (c) the seismicity and deformation measured over the first 3 years of UGS operation. In practice, we observe the phenomena associated with the “birth” of a UGS with remarkable instrumental capabilities. Following three cycles of injection/extraction, about 30 events consistent with natural, tectonically related seismicity were located within 30 km of the UGS. Moreover, the observed uplift of about 2 cm is consistent with theoretical expectations of ground deformation. Our study confirms that UGS in depleted gas reservoirs, if well managed, can result in negligible, if any, human-induced seismicity and limited ground surface deformation. It also shows the importance of measuring the undisturbed conditions prior to industrial activities over a sufficiently long period of time, to correctly interpret the phenomena observed later.</p>","PeriodicalId":54286,"journal":{"name":"Earth and Space Science","volume":"11 11","pages":""},"PeriodicalIF":2.9,"publicationDate":"2024-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1029/2023EA003275","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142737373","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep Autoencoders for Unsupervised Anomaly Detection in Wildfire Prediction","authors":"İrem Üstek,&nbsp;Miguel Arana-Catania,&nbsp;Alexander Farr,&nbsp;Ivan Petrunin","doi":"10.1029/2024EA003997","DOIUrl":"https://doi.org/10.1029/2024EA003997","url":null,"abstract":"<p>Wildfires pose a significantly increasing hazard to global ecosystems due to the climate crisis. Due to its complex nature, there is an urgent need for innovative approaches to wildfire prediction, such as machine learning. This research took a unique approach, differentiating from classical supervised learning, and addressed the gap in unsupervised wildfire prediction using autoencoders and clustering techniques for anomaly detection. Historical weather and normalized difference vegetation index data sets of Australia for 2005–2021 were utilized. Two main unsupervised approaches were analyzed. The first used a deep autoencoder to obtain latent features, which were then fed into clustering models, isolation forest, local outlier factor and one-class support vector machines for anomaly detection. The second approach used a deep autoencoder to reconstruct the input data and use reconstruction errors to identify anomalies. Long Short-Term Memory autoencoders and fully connected (FC) autoencoders were employed in this part, both in an unsupervised way learning only from nominal data. The FC autoencoder outperformed its counterparts, achieving an accuracy of 0.71, an F1-score of 0.74, and an MCC of 0.42. These findings highlight the practicality of this method, as it effectively predicts wildfires in the absence of ground truth, utilizing an unsupervised learning technique.</p>","PeriodicalId":54286,"journal":{"name":"Earth and Space Science","volume":"11 11","pages":""},"PeriodicalIF":2.9,"publicationDate":"2024-11-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1029/2024EA003997","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142708100","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"High-Resolution Morphology of Lunar Lava Tube Pits Using Photogrammetric Modeling of Multiple Stereo Images","authors":"Miyu Zhou,&nbsp;Zhen Ye,&nbsp;Rong Huang,&nbsp;Changyu Zhou,&nbsp;Chen Chen,&nbsp;Hao Chen,&nbsp;Yusheng Xu,&nbsp;Xiaohua Tong","doi":"10.1029/2024EA003532","DOIUrl":"https://doi.org/10.1029/2024EA003532","url":null,"abstract":"<p>Underground lava tubes are promising candidates for the construction of lunar bases because they are believed to offer good protection against radiation and harsh thermal environments on the lunar surface. Although the extent and structures of the underground lava tubes are uncertain, previously identified “skylights”, believed to represent places where lava tube roofs have collapsed, may provide insights into tube structures. Unfortunately, owing to the steep slopes, considerable depth, and associated difficult illumination conditions, the availability of detailed morphologic models of these pits is limited. In this study, we reconstruct the topography of lava tube pits using a refined photogrammetric approach. We use improved census cost for disparity search, refined point cloud coarse-to-fine registration, and weighted fusion of point clouds from several matched stereo image pairs to obtain a high-resolution topographic model. Experiments are conducted for two prominent skylights, that is, the Mare Tranquillitatis Pit and the Marius Hills Hole, to demonstrate the effectiveness of the proposed approach. Several LRO NAC image pairs are selected from images covering these two areas, and fused topographic models with a spatial resolution of 2 <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mi>m</mi>\u0000 </mrow>\u0000 <annotation> $mathrm{m}$</annotation>\u0000 </semantics></math> are generated. The quality of our generated topographic models is validated against terrain products provided by the LROC team. Compared to these previous models, our model is generated based on a much denser point cloud and provides better coverage and more details. Benefiting from the detailed three-dimensional models, morphological analysis is carried out to investigate the geometric dimensions (e.g., depth, diameters, slopes) of the lava tube pits.</p>","PeriodicalId":54286,"journal":{"name":"Earth and Space Science","volume":"11 11","pages":""},"PeriodicalIF":2.9,"publicationDate":"2024-11-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1029/2024EA003532","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142708080","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Statistical Methods for Interpreting Spatial and Temporal Heterogeneity of Martian Tropical Water Ice Informed by Properties of Crater Ejecta Types","authors":"Jamie D. Riggs,&nbsp;Michelle R. Kirchoff","doi":"10.1029/2024EA003796","DOIUrl":"https://doi.org/10.1029/2024EA003796","url":null,"abstract":"<p>The martian tropical water ice spatial and temporal distribution was characterized using impact crater ejecta type, location, size, and age in one of two epochs, <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mo>≤</mo>\u0000 <mn>3.4</mn>\u0000 </mrow>\u0000 <annotation> ${le} 3.4$</annotation>\u0000 </semantics></math> Ga and <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mo>&gt;</mo>\u0000 <mn>3.4</mn>\u0000 </mrow>\u0000 <annotation> ${ &gt;} 3.4$</annotation>\u0000 </semantics></math> Ga, using statistical models designed for spatial and temporal correlation structures. The indicator thought to identify the presence of ice is craters with layered ejecta, while the indicator thought to identify no ice is craters with radial ejecta. These indicators imply the location (longitude and latitude) and, potentially, depth (crater diameter as a proxy) of ice, and when the ice was present. The spatial and temporal distribution of layered ejecta versus radial ejecta may inform on the geography and evolution of ice. A statistical spatial point analysis was conducted on a 54-sample data set (craters with diameters 2.77–10.00 km) for an equatorial region (<span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mn>0</mn>\u0000 <mo>°</mo>\u0000 </mrow>\u0000 <annotation> $0{}^{circ}$</annotation>\u0000 </semantics></math> to <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mo>−</mo>\u0000 <mn>30</mn>\u0000 <mo>°</mo>\u0000 </mrow>\u0000 <annotation> ${-}30{}^{circ}$</annotation>\u0000 </semantics></math> S, and <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mn>10</mn>\u0000 <mo>°</mo>\u0000 </mrow>\u0000 <annotation> $10{}^{circ}$</annotation>\u0000 </semantics></math> E to <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mn>340</mn>\u0000 <mo>°</mo>\u0000 </mrow>\u0000 <annotation> $340{}^{circ}$</annotation>\u0000 </semantics></math> W. The analysis shows the spatial and temporal distribution of tropical ice in the study region is most likely random.</p>","PeriodicalId":54286,"journal":{"name":"Earth and Space Science","volume":"11 11","pages":""},"PeriodicalIF":2.9,"publicationDate":"2024-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1029/2024EA003796","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142707717","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Influence of Parameterization Changes on Arctic Low Cloud Properties and Cloud Radiative Effects in Two Versions of the HadGEM3 Atmospheric Model: GA7.1 and GA6","authors":"Patrick C. Taylor,&nbsp;Robyn C. Boeke,&nbsp;Alejandro Bodas-Salcedo","doi":"10.1029/2024EA003579","DOIUrl":"https://doi.org/10.1029/2024EA003579","url":null,"abstract":"<p>Arctic clouds play a key role in Arctic climate variability and change; however, contemporary climate models struggle to simulate cloud properties accurately. Model-simulated cloud properties are determined by the physical parameterizations and their interactions within the model configuration. Quantifying effects of individual parameterization changes on model-simulated clouds informs efforts to improve models and provides insights on climate system behavior. This study quantities the influence of parameterization scheme changes on Arctic low cloud properties within the Hadley Center Global Environmental Model 3 atmospheric model using a suite of experiments where individual parameterization packages are changed between the two configurations. The results indicate, surprisingly, that single parameterization changes explain most of the cloud property changes, whereas multiple parameterizations, including non-cloud schemes, contribute to cloud radiative effect differences. Non-cloud parameterizations are those not used to compute time step cloud properties. We employ a three-term decomposition to quantify contributions from (a) regime independent, (b) regime dependent, and (c) the regime frequency of occurrence changes. Decomposition results indicate that cloud property changes vary by meteorological regime, each term contributes differently to each cloud property change, and non-cloud parameterization changes make substantial contributions to the LW and SW cloud radiative effects by affecting clear-sky fluxes. The analysis provides insights on the role of non-cloud parameterizations for setting cloud radiative effects, a pathway for cloud-atmosphere circulation interactions, raises questions on the importance of infrequently occurring regimes to climate simulations, and on useful observational approaches for improving models.</p>","PeriodicalId":54286,"journal":{"name":"Earth and Space Science","volume":"11 11","pages":""},"PeriodicalIF":2.9,"publicationDate":"2024-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1029/2024EA003579","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142707662","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Can Large Strains Be Accommodated by Small Faults: “Brittle Flow of Rocks” Revised","authors":"Xiaoyu Zou,&nbsp;Yuri Fialko","doi":"10.1029/2024EA003824","DOIUrl":"https://doi.org/10.1029/2024EA003824","url":null,"abstract":"<p>Brittle deformation in the upper crust is thought to occur primarily via faulting. The fault length-frequency distribution determines how much deformation is accommodated by numerous small faults versus a few large ones. To evaluate the amount of deformation due to small faults, we analyze the fault length distribution using high-quality fault maps spanning a wide range of spatial scales from a laboratory sample to an outcrop to a tectonic domain. We find that the cumulative fault length distribution is well approximated by a power law with a negative exponent close to 2. This is in agreement with the earthquake magnitude-frequency distribution (the Gutenberg-Richter law with b-value of 1), at least for faults smaller than the thickness of the seismogenic zone. It follows that faulting is a self-similar process, and a substantial fraction of tectonic strain can be accommodated by faults that don't cut through the entire seismogenic zone, consistent with inferences of “hidden strain” from natural and laboratory observations. A continued accumulation of tectonic strain may eventually result in a transition from distributed fault networks to localized mature faults.</p>","PeriodicalId":54286,"journal":{"name":"Earth and Space Science","volume":"11 11","pages":""},"PeriodicalIF":2.9,"publicationDate":"2024-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1029/2024EA003824","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142666057","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"3-D Subsurface Geophysical Modeling of the Charity Shoal Structure: A Probable Late Proterozoic-Early Paleozoic Simple Impact Structure in Eastern Lake Ontario","authors":"Mary H. Armour,&nbsp;Joseph I. Boyce,&nbsp;Phillip Suttak,&nbsp;Doug Hrvoic","doi":"10.1029/2024EA003605","DOIUrl":"https://doi.org/10.1029/2024EA003605","url":null,"abstract":"<p>The Charity Shoal structure is a circular, ∼1.2-km-diameter, bedrock-rimmed shoal in eastern Lake Ontario with a ∼20-m-deep central basin. The structure has been proposed as a possible Middle Ordovician impact crater or volcanic intrusion. We conducted marine seismic and magnetic surveys (9-km²) and 3-D geophysical modeling to better resolve the Charity Shoal subsurface geology and possible origins. Three models were evaluated: (a) a buried (&gt;450 m) impact structure in Mesoproterozoic basement, (b) a maar-diatreme, (c) a cylindrical, zoned volcanic plug. Seismic profiles and multi-beam bathymetry revealed &gt;30 m of Quaternary sediments overlying Middle Ordovician (Trenton Group) carbonate bedrock and complex, 3-dimensional folding and faulting of the structure rim. Magnetic surveys recorded an annular magnetic high (&gt;600 nT) over the structure rim and a central magnetic low (∼500–600 nT) coincident with a ∼−1.7 mGal Bouguer gravity anomaly. The continuity of Trenton Group strata in seismic profiles rules out a previously proposed Mesozoic maar-diatreme intruded into Paleozoic strata. The zoned volcanic plug model reproduced the annular magnetic anomaly but was incompatible with Bouguer gravity profiles. The magnetic anomaly was best reproduced by a simple impact structure seated in Mesoproterozoic basement at 450–500 m depth with a rim-to-rim diameter of ∼1.2 km and rim height of ∼10–20 m. A 100-m wide and 50-m-deep channel in the Mesoproterozoic basement may record fluvial dissection of the southwestern rim. A buried (&gt;450 m), simple impact crater is most compatible with all available geophysical data at Charity Shoal.</p>","PeriodicalId":54286,"journal":{"name":"Earth and Space Science","volume":"11 11","pages":""},"PeriodicalIF":2.9,"publicationDate":"2024-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1029/2024EA003605","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142666062","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Study on Acoustic Variability Affected by Upper Ocean Dynamics in South Eastern Arabian Sea","authors":"Kotta Srinivasu,&nbsp;M. C. Sanjana,&nbsp;G. Latha,&nbsp;T. V. S. Udaya Bhaskar,&nbsp;Hasibur Rahaman,&nbsp;A. Thirunavukkarasu,&nbsp;R Venkatesan","doi":"10.1029/2023EA003497","DOIUrl":"https://doi.org/10.1029/2023EA003497","url":null,"abstract":"<p>The influence of upper ocean dynamics on the acoustic field in the South Eastern Arabian Sea (SEAS) is studied using in situ oceanographic/acoustic measurements from a moored buoy, along with satellite-derived and climatological data sets. Upper-ocean variability at the site is quantified using Mixed Layer Depth (MLD), Isothermal Layer Depth (ILD), Barrier Layer Thickness (BLT), Maximum Spice Depth (MSD), and Sonic Layer Depth (SLD), along with surface variability factors such as Sea Surface Temperature, Sea Surface Salinity, Spice, and Sea Level Anomaly. The mixed layer acoustic duct (MLAD) varies from 2 to 100 m, with BLT varying from 5 to 99 m, and a mean SLD of 43 m. A thick transition layer connects the mixed layer with the thermocline during winter. The observations reveal that maximum SLD, MSD, and BLT occurred during January–March. Unlike other seasons when SLD follows MLD, winter SLD is influenced by BLT, suggesting strong salinity stratification due to low-salinity water intrusion from the Bay of Bengal by East India Coastal Current. During these months, the SLD varies from 80 to 100 m, with the corresponding minimum cut-off frequency varying from 300 to 200 Hz. Results are correlated with estimated Sound Pressure Level (SPL) from Ambient Noise Measurements during November 2018 to November 2019. SPL variation follows SLD for low and mid-frequencies, with the highest SPL noted during January-February. Acoustic propagation simulations at 250 and 1,000 Hz revealed features like acoustic duct leakage and channeling, indicating energy transfers between the surface acoustic duct and deeper layers.</p>","PeriodicalId":54286,"journal":{"name":"Earth and Space Science","volume":"11 11","pages":""},"PeriodicalIF":2.9,"publicationDate":"2024-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1029/2023EA003497","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142665092","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A New Generation of Hydrological Condition Simulator Employing Physical Models and Satellite-Based Meteorological Data","authors":"Wenchao Ma,&nbsp;Kenshi Hibino,&nbsp;Kosuke Yamamoto,&nbsp;Misako Kachi,&nbsp;Riko Oki,&nbsp;Haruya Yoshikawa,&nbsp;Kei Yoshimura","doi":"10.1029/2023EA003228","DOIUrl":"https://doi.org/10.1029/2023EA003228","url":null,"abstract":"&lt;p&gt;Determining the distribution and dynamics of water on land at any given moment poses a significant challenge due to the constraints of observation. Consequently, as advancements in land surface models (LSMs) have been made, numerical simulation has emerged as an increasingly accurate and effective method for hydrological research. Nonetheless, systems that represent multiple land surface parameters in a near-real-time manner are scarce. In this study, we present an innovative land surface and river simulation system, termed Today's Earth (TE), which generates state and flux values for the near-surface environment with multiple outputs in near-real-time. There are currently three versions of TE, distinguished by the forcing data utilized: JRA-55 version, employing the Japanese 55-year Reanalysis (JRA-55, from 1958 to the present); GSMaP version, utilizing, the Global Satellite Mapping of Precipitation (GSMaP, from 2001 to the present), and MODIS version, utilizing the Moderate Resolution Imaging Spectroradiometer (MODIS, from 2003 to the present). These long-term forcing data set allow for outputs of the JRA-55 version from 1958, the GSMaP version from 2001, and the MODIS version from 2003. Aiming to provide water and energy values on a global scale in real-time, the TE system utilizes the LSM Minimal Advanced Treatments of Surface Interaction and Runoff (MATSIRO) (Takata et al., 2003, https://doi.org/10.1016/s0921-8181(03)00030-4; Yamazaki et al., 2011, https://doi.org/10.1029/2010wr009726) at a horizontal resolution of 0.5°, along with the river routing model CaMa-Flood (Yamazaki et al., 2011, https://doi.org/10.1029/2010wr009726) at a horizontal resolution of 0.25°. Both land surface and river products are available in 3-hourly, daily, and monthly intervals across all three versions. A notable feature of TE is its ability to release both state and flux parameters in near-real-time, offering convenience for various aspects of hydrological research. In addition to presenting the general features of TE-Global, this study examines the performance of snow depth, soil moisture, and river discharge data in daily intervals from 2003 to 2021, with validation spanning 2003 to 2016. When comparing snow depth results, the correlation coefficient ranged between 0.644 and 0.658, while for soil moisture it ranged between 0.471 and 0.494. These findings suggest that the LSM yields comparable results when utilizing JRA-55, MODIS, or GSMaP. Interestingly, river output from the three products exhibited distinct characteristics varying from GSMaP to JRA-55 and MODIS. For river discharge, the correlation coefficient ranged from 0.494 to 0.519, the root mean square error ranged from 3,730 m&lt;sup&gt;3&lt;/sup&gt;/s to 6,330 m&lt;sup&gt;3&lt;/sup&gt;/s, and the mean absolute error ranged from 3,000 m&lt;sup&gt;3&lt;/sup&gt;/s to 5,160 m&lt;sup&gt;3&lt;/sup&gt;/s among the different forcing versions. The overall bias in river discharge from GSMaP was 1,570 m&lt;sup&gt;3&lt;/sup&gt;/s, in contrast to −589 m&lt;sup&gt;3&lt;/sup&gt;/s fo","PeriodicalId":54286,"journal":{"name":"Earth and Space Science","volume":"11 11","pages":""},"PeriodicalIF":2.9,"publicationDate":"2024-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1029/2023EA003228","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142665094","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Monthly Prediction on Summer Extreme Precipitation With a Deep Learning Approach: Experiments Over the Mid-To-Lower Reaches of the Yangtze River","authors":"Yi Fan,&nbsp;Yang Lyu,&nbsp;Shoupeng Zhu,&nbsp;Zhicong Yin,&nbsp;Mingkeng Duan,&nbsp;Xiefei Zhi,&nbsp;Botao Zhou","doi":"10.1029/2024EA003926","DOIUrl":"https://doi.org/10.1029/2024EA003926","url":null,"abstract":"<p>Accurate predictions of monthly extremes assume paramount importance in enabling proactive decision-making, which however are lacked in skills even for state-of-the-art dynamical models. Taking the extreme precipitation prediction over the mid-to-lower reaches of the Yangtze River, China, as an instance, a multi-predictor U-Net deep learning approach is designed to enhance the prediction over the European Center for Medium-Range Weather Forecasts (ECMWF) model, with the single-predictor U-Net parallelly examined as the benchmark. Focusing on the precipitation extremes, an extreme associated component is incorporated into the model loss function for optimization. Besides, predictions composed by daily outputs with multiple lead times are imported as a comprehensive set in the training phase to augment the deep learning sample size and to emphasize enhancements in predictions at the monthly timescale as a whole. Results indicate that the multi-predictor U-Net effectively improves predictions of extreme summer precipitation frequency, showing distinct superiority to the raw ECMWF and the single-predictor U-Net. Multiple evaluation metrics indicate that the model shows a significant positive improvement ratio ranging from 65.1% to 80.0% across all grids compared to the raw ECMWF prediction, which has also been validated through applications in the two extreme summer precipitation cases in 2016 and 2020. Besides, a ranking analysis of feature importance reveals that factors such as humidity and temperature play even more crucial roles than precipitation itself in the multi-predictor extreme precipitation prediction model at the monthly timescale. That is, in such a deep learning approach, the monthly prediction on extreme precipitation benefits significantly from the inclusion of multiple associated predictors.</p>","PeriodicalId":54286,"journal":{"name":"Earth and Space Science","volume":"11 11","pages":""},"PeriodicalIF":2.9,"publicationDate":"2024-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1029/2024EA003926","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142665093","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}