Gao et al. (2018) Irrigation Mapping Using Sentinel-1 Time Series at Field Scale

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2018
• Authors: Qi Gao, Mehrez Zribi, Maria‐José Escorihuela, Nicolas Baghdadi, Pere Quintana Seguí
• DOI: 10.3390/rs10091495

Research Groups

Not explicitly listed in the text.

Short Summary

This study proposes and validates a methodology using Sentinel-1 SAR time series metrics (VV and VH polarization) combined with machine learning (SVM, RF) to accurately map irrigated and non-irrigated agricultural fields, achieving an overall accuracy exceeding 81%.

Objective

• To assess the potential of various statistical and geometrical metrics derived from Sentinel-1 Synthetic Aperture Radar (SAR) time series for accurately mapping irrigated fields.
• To develop a robust irrigation mapping methodology based solely on SAR data, ensuring applicability regardless of cloud cover conditions.

Study Configuration

• Spatial Scale: Field level (agricultural parcels). Study site: Urgell, Catalunya, Spain.
• Temporal Scale: Time series analysis of Sentinel-1 data (specific duration not specified, but covers an agricultural season).

Methodology and Data

• Models used: Support Vector Machine (SVM), Random Forest (RF) (classification algorithms).
• Data sources: Sentinel-1 SAR satellite data (VV and VH polarizations), SIGPAC (Geographic Information System for Agricultural Parcels) used for field segmentation and ground truthing.
• Metrics derived: Mean backscatter value, variance of the signal, correlation length, and fractal dimension of the Sentinel-1 time series.

Main Results

• The proposed methodology successfully classified irrigated crops, irrigated trees, and non-irrigated fields using a vector of SAR-derived metrics.
• The Support Vector Machine (SVM) classification achieved a good overall accuracy of 81.08% when validated against SIGPAC ground truthing.
• The Random Forest (RF) classification performed slightly better, achieving an overall accuracy of approximately 82.2% (when the tree depth was set at three).
• The methodology's robustness may decrease in areas where irrigation practices do not strongly influence soil moisture changes.

Contributions

• Development of a novel irrigation mapping methodology relying exclusively on Sentinel-1 SAR data, which ensures operational capability under all weather conditions (e.g., frequent cloud cover), overcoming a major limitation of optical remote sensing methods.
• Quantification and validation of the utility of various SAR time series metrics (including correlation length and fractal dimension) for distinguishing irrigation practices at the field scale.

Funding

Not listed in the text.

Citation

@article{Gao2018Irrigation,
  author = {Gao, Qi and Zribi, Mehrez and Escorihuela, Maria‐José and Baghdadi, Nicolas and Quintana‐Seguí, Pere},
  title = {Irrigation Mapping Using Sentinel-1 Time Series at Field Scale},
  journal = {Remote Sensing},
  year = {2018},
  doi = {10.3390/rs10091495},
  url = {https://doi.org/10.3390/rs10091495}
}

Generated by BiblioAssistant using gemini-flash-latest (Google API)