Abstract: The preservation rate of farmland shelterbelts reflects the integrity of the shelterbelt structure, determining the protection effectiveness of shelterbelts. Although traditional field measured data of shelterbelt preservation rate are accurate, this method could not meet the needs for rapid and large-scale monitoring. It is of great importance to obtain the information of shelterbelt preservation rate by remote sensing, which could be significantly helpful for the structural and functional research and precise management of farmland shelterbelts. In this study, using the shelterbelt vector line and vegetation fraction data extracted from Jilin-1 high-resolution multispectral remote sensing images, a set of ordered points were extracted along the belt orientation and belt width directions by considering each shelterbelt as a two-dimensional matrix. The shelterbelt boundary was recognized by analyzing the waveform characteristics, and the shelterbelt pixels were determined by the threshold value. Finally, the preservation rate of farmland shelterbelts was identified from the remote sensing image, which were further verified by the reference data (sample number was 160). There were three main findings. (1) The remote sensing identification results and the reference data was significantly correlated (R²=0.936), and the average absolute error was 5.4%, indicating a high overall accuracy. (2) The identification results in different belt directions were highly correlated with the reference data. The correlation in north-south direction was relatively low, which may be affected by belt shadows. (3) The absolute error of identification results increased first and then decreased with the increases of preservation rate. The absolute error was the smallest in high preservation rate (>80%) and low preservation rate (<20%), and the largest in the middle state, especially in the interval of 50%-60%, because these shelterbelts were more fragmented, with small and numerous gaps. In summary, the preservation rate of farmland shelterbelt can be accurately identified based on high-resolution multispectral satellite images. This method promotes the ability of remote sensing monitoring for shelterbelt structure, which could provide technical support for accurate monitoring the structure and function of farmland shelterbelts.

Key words: farmland shelterbelt, belt structure, remote sensing, agroforestry system