Irrigated Crop Types Mapping in Tashkent Province of Uzbekistan with Remote Sensing-Based Classification Methods

Area of Science:

• * Earth Observation and Geosciences
• * Agricultural Remote Sensing
• * Geospatial Analysis

Background:

• * Accurate crop type mapping is crucial for land management and agricultural monitoring, particularly in developing nations lacking digital cadastral systems.
• * Remote Sensing (RS) data offers a viable alternative for monitoring and inventory where traditional methods are insufficient.
• * Optical RS data, using medium and high spatial resolution imagery, presents an opportunity for detailed crop classification.

Purpose of the Study:

• * To compare and assess the importance of optical RS data for crop type classification.
• * To evaluate the effectiveness of Landsat 8 (L8) and Sentinel-2 (S2) satellite imagery for this purpose.
• * To analyze the performance of different vegetation indices and machine learning algorithms in crop classification.

Main Methods:

• * Acquisition of L8 and S2 satellite data over Tashkent Province during the May-October crop growth period.
• * Calculation of Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and two forms of Normalized Difference Water Index (NDWI1, NDWI2).
• * Application of Support-Vector-Machine (SVM) and Random Forest (RF) classification algorithms for crop type mapping.

Main Results:

• * All tested vegetation indices achieved an Overall Accuracy (OA) above 84%.
• * The highest OA of 92% was obtained using EVI-NDVI with the RF method on L8 data.
• * Classification results showed minimal differences when compared with Official State Statistics (OSS), with differences as low as 0.1 thousand ha for S2 data using EVI with RF.

Conclusions:

• * The study demonstrates the high potential of using combined L8 and S2 data for accurate crop type mapping.
• * Joint utilization of these satellite datasets enhances classification accuracy and enables better crop separation.
• * Future research should explore the integration of high temporal resolution data from combined sensors for improved vegetation monitoring.