Automated kharif rice mapping using SAR data and machine learning techniques in GEE platform

被引：0

作者：

Vyas, Saurabh P. ^{[1
]}

Kumar, Mukesh ^{[2
]}

Kathiria, Dhaval ^{[1
]}

Jani, Mandakini ^{[1
]}

Pandya, Mehul R. ^{[2
]}

Bhattacharya, Bimal K. ^{[2
]}

机构：

[1] Anand Agr Univ, Coll Agr Informat Technol, Anand 388110, India

[2] Indian Space Res Org, Space Applicat Ctr, Ahmadabad 380058, India

来源：

CURRENT SCIENCE | 2024年 / 126卷 / 10期

关键词：

Google earth engine; large-scale rice mapping; machine learning; multi-temporal; SAR; LAND-COVER; CLASSIFICATION; PADDY; EXTRACTION; CROPS;

D O I：

10.18520/cs/v126/i10/1265-1272

中图分类号：

O [数理科学和化学]; P [天文学、地球科学]; Q [生物科学]; N [自然科学总论];

学科分类号：

07 ; 0710 ; 09 ;

摘要：

The present study employs temporal C -band Sentinel -1 synthetic aperture radar (SAR) data within the Google Earth Engine (GEE) platform to evaluate discriminability and estimate acreage of kharif rice across major Indian states. Utilizing multi -temporal Sentinel -1 Cband SAR data, including time -series cross -polarization vertical-horizontal channels, the research spanned states such as Punjab, Haryana, Uttar Pradesh, Madhya Pradesh, Bihar, Jharkhand, Chhattisgarh, Telangana, Andhra Pradesh, West Bengal, Odisha and Assam. Employing five machine learning algorithms on GEE, with random forest demonstrating high performance, achieved 98.59% accuracy and 0.92 kappa coefficient ( kappa ) in Odisha. Subsequently, the RF algorithm was applied for kharif rice acreage estimation, yielding overall accuracies from 88.48% to 97.28% and kappa between 0.87 and 0.96 with deviations from reported acreage ranging from 0.95% to 12% across diverse states. The study underscores the efficacy of SAR data and machine learning within GEE for precise large-scale automated mapping of kharif rice.

引用

页码：1265 / 1272

页数：8