Accurate and timely mapping of agricultural products is a crucial component in management and decision-making for promoting food security and sustainable development. This study investigates the classification of different agricultural products in Mobarakeh City, utilizing a time series of Sentinel 2 images acquired between October 1, 1401 and October 15, 1402. To prepare for the classification, the vegetative calendar of different crops in the study area was gathered. The classification was based on spectral bands and indices of crop growth phenology, providing valuable insights into the growth cycles of different crops in the region. For crop classification, we used a combination of spectral bands, including red, near-infrared, and normalized difference vegetation index (NDVI) and greenness. Additionally, self-organizing maps (SOM) were used to assess and enhance the quality of satellite image time series data prior to modeling. The SOM method was applied to eva‎luate quality of the sample data and identify outliers and noisy data. After removing the outlying data points, the images were classified using random forest, gradient boosting, support vector machine, and temporal convolutional neural network methods, both before and after quality control measures were implemented. The results indicated that after quality control, the overall accuracy and Kappa coefficient of the analyses were considerably improved. Random forest and temporal convolutional neural network methods demonstrated superior prediction and modeling performance compared to gradient boosting and support vector machine methods. The overall accuracy of the four methods increased from 0.91, 0.87, 0.83, and 0.91 before quality control to 0.96, 0.94, 0.89, and 0.95 after quality control, respectively. Overall, Random forest and temporal convolutional neural network models provide the best performance among the studied models. Therefore, these two methods were used to check the cultivated area of agricultural lands. The calculated cultivated area was compared with the official agricultural statistics. Considering that the obtained statistics had a high agreement, the approach used in this research can be considered as a practical tool for future landscape planning investigations.

سعيد پورمنافي , محسن احمدي

عليرضا سفيانيان , مصطفي تركش اصفهاني