Ursu et al. (2025) Hail Damage Detection: Integrating Sentinel-2 Images with Weather Radar Hail Kinetic Energy

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-11-27
• Authors: Adrian Ursu, Istrate Vasilică, Vasile Jitariu, István Lázár
• DOI: 10.3390/rs17233850

Research Groups

• Faculty of Geography and Geology, Alexandru Ioan Cuza University, Iasi, Romania
• SC Legendary Team Security SRL, Focșani, Romania

Short Summary

This study integrates weather radar-derived Hail Kinetic Energy (HKE) with Sentinel-2 Normalized Difference Vegetation Index (NDVI) differencing to accurately assess and map short-term vegetation damage from hailstorms in northeastern Romania. The research demonstrates a strong spatial correspondence between high HKE values and significant NDVI reductions, highlighting that damage detection is most effective shortly after the event and varies by land use type.

Objective

• To assess and map short-term vegetation damage caused by hailstorms by integrating weather radar parameters (specifically Hail Kinetic Energy, HKE) and high-resolution Sentinel-2 multispectral imagery.
• To analyze the temporal evolution of vegetation response to hail damage and evaluate the differential sensitivity of various land use classes.

Study Configuration

• Spatial Scale: Regional (northeastern Romania), focusing on two case study areas: Rădăuți (Suceava county) and Dolhasca (Suceava county). Analysis performed at parcel-level with 10 meter resolution for satellite data. Radar data resolution is typically 0.5 km near the radar to several kilometers at longer ranges, with HKE isolines around 500 m.
• Temporal Scale: Two specific hail events: 17 July 2016 (Rădăuți) and 30 July 2020 (Dolhasca). Sentinel-2 imagery acquired pre-event and at multiple post-event intervals (e.g., 5, 8, and 15 days for Dolhasca; 6 days for Rădăuți). Radar data provided near real-time monitoring during storm evolution.

Methodology and Data

• Models used:
  • ASU-MRL software (2022 version) for processing weather radar data and computing Hail Kinetic Energy (HKE).
  • Non-parametric Friedman tests and Wilcoxon signed-rank tests for statistical analysis of temporal differences in ΔNDVI across land use categories.
• Data sources:
  • Weather Radar Data: S-band MRL-5 radars (Soroca and Cornești, Republic of Moldova) providing parameters like maximum reflectivity (Zmax), height of maximum reflectivity (H_Zmax), height of 45 dBZ echo above melting level (dH45), Vertically Integrated Liquid (VIL), 35 dBZ storm echo top (H35), and Hail Kinetic Energy (HKE).
  • Satellite Imagery: Sentinel-2 Level-1C multispectral imagery (Copernicus Data Space Browser) for Normalized Difference Vegetation Index (NDVI) computation using red (B4) and near-infrared (B8) bands.
  • Reanalysis Data: ERA5 (Copernicus Climate Data Store) for synoptic drivers and instability parameters (e.g., CAPE, CIN, bulk shear).
  • Ground Truth/Reports: European Severe Weather Database (ESWD) hail reports and local observer photographs (from "Meteo Nord-Est" Facebook group).
  • Land Use Classification: Derived from Sentinel-2 imagery, categorizing areas into arable land, complex agriculture, pastures, orchards, and forests.

Main Results

• A strong spatial correspondence was observed between radar-derived Hail Kinetic Energy (HKE) cores (threshold > 300 J/m²) and significant reductions in Sentinel-2 Normalized Difference Vegetation Index (ΔNDVI).
• For the Rădăuți hailstorm (17 July 2016), ΔNDVI thresholds identified between 2236 hectares (ΔNDVI > 0.20) and 5856 hectares (ΔNDVI > 0.10) of affected vegetation within the HKE > 300 J/m² zone, based on an image 6 days post-event.
• For the Dolhasca hailstorm (30 July 2020), the estimated affected area (ΔNDVI > 0.10) was 2581 hectares at 5 days post-event, decreasing to 1891 hectares at 8 days, and further to 1511 hectares at 15 days. For ΔNDVI > 0.20, the affected area decreased from 895 hectares at 5 days to 560 hectares at 15 days. This temporal decline indicates both vegetation recovery and diminishing sensitivity of the ΔNDVI signal over time.
• Analysis by land use class revealed differential sensitivity: arable fields and complex agriculture were most sensitive to hail impacts, followed by orchards and pastures. Forests exhibited the weakest but still measurable and persistent declines in NDVI.
• Radar parameters (e.g., Zmax > 60 dBZ, VIL > 30-40 kg/m², H35 > 8-9 km, dH45 > 4 km) consistently indicated severe hail potential during the events.

Contributions

• Demonstrates a robust and effective methodology for hail damage assessment by integrating radar-derived Hail Kinetic Energy (HKE) fields with high-resolution Sentinel-2 NDVI differencing.
• Establishes radar-based HKE thresholds (specifically > 300 J/m²) as a reliable and rapid temporal indicator for delineating potentially affected areas, thereby optimizing the focus for subsequent optical satellite analysis.
• Quantifies the critical role of timing in post-storm satellite acquisitions, showing that hail-induced vegetation stress is most detectable in the immediate days (within one week) following the event, with the spectral signal attenuating over time due to recovery or stabilization.
• Provides insights into the differential sensitivity and resilience of various land use categories (arable land, complex agriculture, orchards, pastures, forests) to hail damage, enhancing the understanding of ecosystem response.
• Offers a scalable framework with significant implications for operational crop monitoring, disaster response, and agricultural insurance assessment in hail-prone regions under changing climate conditions.

Funding

• Department of Geography, Faculty of Geography and Geology, ‘Alexandru Ioan Cuza’ University, of Iasi, Romania.
• POSCCEO 2.2.1, SMIS-CSNR 13984-901, No 257/28.09.2010 Project, CERNESIM.
• Grant of the “Alexandru Ioan Cuza” University of Iași, within the Research Grants program, Grant UAIC, code [GI-UAIC-2022-05].

Citation

@article{Ursu2025Hail,
  author = {Ursu, Adrian and Vasilică, Istrate and Jitariu, Vasile and Lázár, István},
  title = {Hail Damage Detection: Integrating Sentinel-2 Images with Weather Radar Hail Kinetic Energy},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17233850},
  url = {https://doi.org/10.3390/rs17233850}
}