View in EDS HTML Full Text PDF Full Text

RaDiT: A Differential Transformer-Based Hybrid Deep Learning Model for Radar Echo Extrapolation.

Saved in:
Bibliographic Details
Title: RaDiT: A Differential Transformer-Based Hybrid Deep Learning Model for Radar Echo Extrapolation.
Authors: Zhu, Wenda1,2 (AUTHOR), Lu, Zhenyu2,3 (AUTHOR) zhenyulu@nuist.edu.cn, Zhang, Yuan3,4 (AUTHOR), Zhao, Ziqi1,4 (AUTHOR), Lu, Bingjian3,5 (AUTHOR), Li, Ruiyi1,5 (AUTHOR)
Source: Remote Sensing. Jun2025, Vol. 17 Issue 12, p1976. 30p.
Subjects: Transformer models, Differential transformers, Deep learning, Radiant intensity, Extrapolation
Abstract: Radar echo extrapolation, a critical spatiotemporal sequence forecasting task, requires precise modeling of motion trajectories and intensity evolution from sequential radar reflectivity inputs. Contemporary deep learning implementations face two operational limitations: progressive attenuation of predicted echo intensities during autoregressive inference and spectral leakage-induced diffusion at high-intensity echo boundaries. This study presents RaDiT, a hybrid architecture combining differential transformer with adversarial training for radar echo extrapolation. The framework employs a U-Net backbone augmented with vision transformer blocks, utilizing differential attention mechanisms to govern spatiotemporal interactions. Our differential attention mechanism enhances noise suppression under high-threshold conditions, effectively minimizing spurious feature generation while improving metric reliability. A conditional GAN discriminator is integrated to maintain microphysical consistency in generated sequences, simultaneously addressing spectral blurring and intensity dissipation. Comprehensive evaluations demonstrate RaDiT's superior performance in preserving spatiotemporal coherence and intensity across 0–90 min forecasting horizons. The proposed architecture achieves CSI improvements of 10.23% and 2.88% at 4 × 4 and 16 × 16 spatial pooling scales, respectively, for ≥30 dBZ thresholds on the CMARC dataset compared to PreDiff. To our knowledge, this represents the first successful implementation of differential transformers for radar echo extrapolation. [ABSTRACT FROM AUTHOR]
Copyright of Remote Sensing is the property of MDPI and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
Text:
  Availability: 1
Header DbId: egs
DbLabel: Engineering Source
An: 186260780
AccessLevel: 6
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: RaDiT: A Differential Transformer-Based Hybrid Deep Learning Model for Radar Echo Extrapolation.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Zhu%2C+Wenda%22">Zhu, Wenda</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Lu%2C+Zhenyu%22">Lu, Zhenyu</searchLink><relatesTo>2,3</relatesTo> (AUTHOR)<i> zhenyulu@nuist.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Zhang%2C+Yuan%22">Zhang, Yuan</searchLink><relatesTo>3,4</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Zhao%2C+Ziqi%22">Zhao, Ziqi</searchLink><relatesTo>1,4</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Lu%2C+Bingjian%22">Lu, Bingjian</searchLink><relatesTo>3,5</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Li%2C+Ruiyi%22">Li, Ruiyi</searchLink><relatesTo>1,5</relatesTo> (AUTHOR)
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22Remote+Sensing%22">Remote Sensing</searchLink>. Jun2025, Vol. 17 Issue 12, p1976. 30p.
– Name: Subject
  Label: Subjects
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Transformer+models%22">Transformer models</searchLink><br /><searchLink fieldCode="DE" term="%22Differential+transformers%22">Differential transformers</searchLink><br /><searchLink fieldCode="DE" term="%22Deep+learning%22">Deep learning</searchLink><br /><searchLink fieldCode="DE" term="%22Radiant+intensity%22">Radiant intensity</searchLink><br /><searchLink fieldCode="DE" term="%22Extrapolation%22">Extrapolation</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Radar echo extrapolation, a critical spatiotemporal sequence forecasting task, requires precise modeling of motion trajectories and intensity evolution from sequential radar reflectivity inputs. Contemporary deep learning implementations face two operational limitations: progressive attenuation of predicted echo intensities during autoregressive inference and spectral leakage-induced diffusion at high-intensity echo boundaries. This study presents RaDiT, a hybrid architecture combining differential transformer with adversarial training for radar echo extrapolation. The framework employs a U-Net backbone augmented with vision transformer blocks, utilizing differential attention mechanisms to govern spatiotemporal interactions. Our differential attention mechanism enhances noise suppression under high-threshold conditions, effectively minimizing spurious feature generation while improving metric reliability. A conditional GAN discriminator is integrated to maintain microphysical consistency in generated sequences, simultaneously addressing spectral blurring and intensity dissipation. Comprehensive evaluations demonstrate RaDiT's superior performance in preserving spatiotemporal coherence and intensity across 0–90 min forecasting horizons. The proposed architecture achieves CSI improvements of 10.23% and 2.88% at 4 × 4 and 16 × 16 spatial pooling scales, respectively, for ≥30 dBZ thresholds on the CMARC dataset compared to PreDiff. To our knowledge, this represents the first successful implementation of differential transformers for radar echo extrapolation. [ABSTRACT FROM AUTHOR]
– Name: AbstractSuppliedCopyright
  Label:
  Group: Ab
  Data: <i>Copyright of Remote Sensing is the property of MDPI and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.)
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=186260780
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.3390/rs17121976
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 30
        StartPage: 1976
    Subjects:
      – SubjectFull: Transformer models
        Type: general
      – SubjectFull: Differential transformers
        Type: general
      – SubjectFull: Deep learning
        Type: general
      – SubjectFull: Radiant intensity
        Type: general
      – SubjectFull: Extrapolation
        Type: general
    Titles:
      – TitleFull: RaDiT: A Differential Transformer-Based Hybrid Deep Learning Model for Radar Echo Extrapolation.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Zhu, Wenda
      – PersonEntity:
          Name:
            NameFull: Lu, Zhenyu
      – PersonEntity:
          Name:
            NameFull: Zhang, Yuan
      – PersonEntity:
          Name:
            NameFull: Zhao, Ziqi
      – PersonEntity:
          Name:
            NameFull: Lu, Bingjian
      – PersonEntity:
          Name:
            NameFull: Li, Ruiyi
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 15
              M: 06
              Text: Jun2025
              Type: published
              Y: 2025
          Identifiers:
            – Type: issn-print
              Value: 20724292
          Numbering:
            – Type: volume
              Value: 17
            – Type: issue
              Value: 12
          Titles:
            – TitleFull: Remote Sensing
              Type: main
ResultId 1