|
Classification of agricultural lands in Khuzestan province based on the type of cropping system using remote sensing data and random forest model
|
Esmaiel Abdalli   , Armin Moghimi * , Mohammad javad Valadan zoej |
|
|
|
Abstract: (7 Views) |
| Accurate and timely identification of cropping systems in agricultural lands is one of the fundamental challenges in modern agricultural research. This study aims to separate the areas under wheat and barley cultivation in a part of Khuzestan province in the period 1401-1402 into three groups including irrigated, dryland and others (other agricultural uses, barren land and urban areas), and presents an innovative framework based on multitemporal remote sensing data and machine learning algorithms. In this study, Sentinel-2, Sentinel-1, Landsat-8 and SRTM satellite data have been used in an integrated manner to analyze different dimensions of land cover from a structural, temporal and spatial perspective. Also, a set of specific phenological indices have been designed based on temporal and spectral differences between crop classes. These indices have performed well alone; Thus, they recorded an overall accuracy of 84.31% and a kappa coefficient of 0.75, which provided more accurate results than other simpler combinations. However, combining these phenological indices with other temporal, spectral, radar, thermal, and topographic features led to a significant improvement in model performance, achieving an overall accuracy of 91.57% and a kappa coefficient of 0.87. In this regard, the Random Forest algorithm was used as the main classification method. The results show that phenological indices, especially in sensitive bands such as RED and NIR, play a key role in increasing classification accuracy, and their integration with other remote sensing data can improve the separation of similar covers. This study emphasizes the importance of data-driven design of temporal indices. |
Article number: 7 |
| Keywords: Remote sensing, Phenological Indices, Cropping systems, Classification, Machine Learning |
|
|
Full-Text [PDF 1170 kb]
(5 Downloads)
|
Type of Study: Research |
Subject:
Photo&RS
Received: 2025/07/10
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
