Study on the extraction method of Glycyrrhiza uralensis Fisch. distribution area based on Gaofen-1 remote sensing imagery: a case study of Dengkou county

<i>Glycyrrhiza uralensis</i> Fisch., a perennial medicinal plant with a robust root system, plays a significant role in mitigating land desertification when cultivated extensively. This study investigates Dengkou County, a semi-arid region, as the research area. First, the reflectance differences of feature types, and the importance of bands were evaluated by using the random forest (RF) algorithm. Second, after constructing the <i>G. uralensis</i> vegetation index (GUVI), the recognition accuracy of <i>G. uralensis</i> was compared between the RF classification model constructed based on the January-December GUVI and common vegetation indices feature set and the support vector machine (SVM) classification model constructed on the GUVI feature set. Finally, the spectral characteristics of <i>G. uralensis</i> and other feature types under the 2022 GUVI feature set were analyzed, and the historical distribution of <i>G. uralensis</i> was identified and mapped. The results demonstrated that the blue and near-infrared bands are particularly significant for distinguishing <i>G. uralensis</i>. Incorporating year-round (January-December) data significantly improved identification accuracy, achieving a producer's accuracy of 97.26%, an overall accuracy of 93.00%, a Kappa coefficient of 91.38%, and a user's accuracy of 97.32%. Spectral analysis revealed distinct differences with <i>G. uralensis</i> of different years and other feature types. From 2014 to 2022, the distribution of <i>G. uralensis</i> expanded from the northeast of Dengkou County to the central and southwestern regions, transitioning from small, scattered patches to larger, concentrated areas. This study highlights the effectiveness of GUVI and RF classification models in identifying <i>G. uralensis</i>, demonstrating superior performance compared to models using alternative feature sets or algorithms. However, the generalizability of the RF model based on the GUVI feature set may be limited due to the influence of natural and anthropogenic factors on <i>G. uralensis</i>. Therefore, regional adjustments and optimization of model parameters may be necessary. This research provides a valuable reference for employing remote sensing technology to accurately map the current and historical distribution of <i>G. uralensis</i> in regions with similar environmental conditions.