Identification of cropland in Tibetan Plateau based on time series remote sensing features

Abstract

Cropland is crucial for regional food security, especially in vulnerable areas like the Tibetan Plateau. Accurate monitoring was hindered of cropland distribution due to complex topography and diverse crop phenology, making it challenging to assess its agricultural sustainability. To address this, this study aimed to develop a cropland identification approach based on an optimal identification feature knowledge graph (OIFKG) derived from time series remote sensing data. Cropland OIFKG (C_OIFKG) enhanced cropland identification accuracy by 96.6%, with producer’s accuracy and user’s accuracy of 98.1% and 89.9% respectively for cropland. The total cropland area in the Tibetan Plateau for 2022 was estimated at 1,800,160 hectares, representing about 1% of the total land area, with a significant concentration in the northeastern Qinghai province and the Yarlung Zangbo River Valley of Tibet Autonomous Region. The total cropland area estimated in this study for the Tibetan Plateau lied within the range provided by two published land cover datasets, being 3.56% lower than one dataset and 16.4% higher than the other. The cropland identification approach proposed by this study reduced reliance on known samples, improving spatiotemporal generalization capability. In the Tibetan Plateau, where cropland distribution was exceedingly rare, the method still achieved promising performance in cropland identification, demonstrating its effectiveness on the assessment of agriculture sustainability in high-altitude regions with intricate landscapes. Moreover, further assessment of C_OIFKG's applicability in different regions and compatibility with multi-source remote sensing data is needed.

Keywords

• Automated cropland mapping
• optimal identification feature
• knowledge graph
• remote sensing
• Tibetan Plateau