Construction of regional rural landscape character identification system: A case of a riverine area along the middle reaches of the Yangtze River, China

• The traditional LC factor was defined as the component factor. • This study constructed a new LC composition 3C structure. • Categorical regression was used to screen for landscape characterization factors. • This study improves the K-prototype clustering by writing a weighted and elbow code. Landscape character assessment (LCA) can outline the distinctions between landscapes and aid in identifying and preserving a sense of place, with the core result being landscape character (LC) mapping. Despite LCA being popular, challenges, such as a single LC factor structure and its inherent and subjective selection, remain. LC identification methods are constantly being updated; however, limitations by sample size, data type, and lack of object-specificity remain to be addressed. In this study, the traditional LC factor was defined as the part of the factor component based on which the configuration and coordinate factors were introduced to construct the 3C structure of the LC composition. Considering the area along the middle reaches of the Yangtze River, the bottom-up induction method was used to screen the five composition factors and their weight with a greater influence on the rural landscape through CATREG, and obtained 329 types of landscape description units and 382,279 characterized patches through fuzzy superposition. K-prototype clustering with improved weighting and elbow method coding was introduced to determine the optimal K-value for the 106 LC types. Boundary identification and unique segmentation using eCognition yielded 456 LC areas. Using a top-down deductive approach combined with an overall landscape view, the three types and seven LC zones were obtained using large-scale geohydrological name types and categorized by the LC area as a boundary. Three aspects of the LCA identification stage were improved: composition of LC factors, assignment of LC factor weights, and improvement of identification techniques and methods to construct a more comprehensive, accurate, and geographically specific LC identification system.