土地多功能利用是提高土地利用效率、缓解人地矛盾、促进区域可持续发展的重要途径。研究构建了“生产-生活-生态”功能的土地多功能性评价指标体系,基于3 km×3 km格网,融合了遥感、统计、POI等多源地理数据,利用投影寻踪模型对黑河中游土地多功能性进行评价,通过变量相关分析和双变量局部空间自相关分析,揭示了土地多功能权衡与协同关系,并采用RGB三通道合成与二阶聚类分析相结合进行土地功能分区。结果表明：（1） 黑河中游在不同利用方式下呈现不同功能,在走廊平原区土地功能以生产功能为主导,生活功能和生态功能为辅。生产功能在走廊绿洲农业区优势明显,生活功能高值区集中于城镇等基础设施条件较好的地区,祁连山、龙首山发挥着生态屏障作用。（2） 研究时段内,土地多功能性稳中增强。土地多功能性与一级功能的空间格局保持稳定,土地功能间协调性普遍增强,不同功能在空间上日益重叠。（3） 按照主导利用方向,土地可划分为以农业生产与城镇空间为主体的重点开发区、优化开发区和适度开发区,以及以生态空间为主的生态屏障区、生态缓冲区和生态修复区。（4） 土地多功能利用仍有提升空间,尤其要加强对非主导功能的关注,协调功能间的固有矛盾;统筹推进生态修复,关注自然资本增值,从构建区域绿水青山的格局挖掘土地价值新的增长点。

Multifunctional land use is an important way to improve land-use efficiency, alleviate the contradiction between people and land, and promote regional sustainable development. The Heihe River’s middle reaches in the middle of the Hexi Corridor, Gansu Province, China is the most densely populated and economically developed area, characterized by the most intense conflict between humans and land in the Heihe River Basin. However, the lack of knowledge about multifunctional land-use conditions hinders sustainable land-use planning and management in this area. In this study, we developed a land multi-function evaluation index system based on the production-life-ecology framework. We then applied a projection pursuit model based on a genetic algorithm to make a grid-based land multifunctional assessment at a 3 km×3 km resolution by assimilating geographical datasets from multiple sources, including remote sensing, socio-economic statistics, and point-of-interest (POI) data. To reveal the trade-offs and synergies of land multi-functions,we performed Spearman’s correlation and bivariate local spatial autocorrelation analysis. We also conducted a functional land zoning using RGB composite and clustering analysis. We found the following results: (1) The land functions in the study area varied with land-use patterns. The primary land function was for production, and the secondary land functions were for living and ecological functions. The land functions also showed spatial variability, characterized by the dominant production functions in the oasis agricultural lands, the useful living functions in the densely populated areas such as cities and towns with better facilities, and the most important ecological functions in Qilian Mountain and Longshou Mountain. (2) Land multi-functionality was enhanced during the study period. The spatial pattern of land multi-functionality and the dominant functions were stable, and the coordination among multiple functions was universally enhanced, indicating that various functions were more overlapped spatially. (3) On the basis of the principle of dominant use direction, the land in the Heihe River’s middle reaches can be divided into six zones: key development zones, optimizing development zones, moderate development zones, ecological barrier zones, ecological buffer zones, and ecological restoration zones. The former three zones were mainly used for agricultural production and urban space functions, whereas the latter three zones acted as ecological spaces. (4) There was still some room for improvement in land multi-functionality. Attention should be paid to non-dominant functions and the coordination of the inherent contradictions among functions. Efforts should be made to collaboratively promote ecological restoration and appreciation of natural capitalsand discover new growth points by constructing the regional pattern of lucid waters and lush mountains. This study highlights the POI data acquired from the Internet map open platforms. Compared with the conventional indicators, POI data can be updated more readily and can reflect the locations and types of facilities more accurately, capturing the levels of each land function more effectively. When making functional land zoning, this study also emphasizes using the RGB composite method to obtain dominant land functional patterns while simultaneously considering the cold and hot spots of land function and the trade-offs and synergies among functions. This approach reflected the current spatial pattern of land function clustering and the internal relationships among functions simultaneously, improving the classified zones’ accuracy and operationality.