考古遗址预测模型（SPM）在考古学研究和文化资源管理中有着广泛的应用，但传统的构建方法过度依赖数字高程模型（DEM），对多光谱遥感数据的应用较少。本项研究以陇东黄土高原地区的区域考古调查为例，利用第三次全国文物普查数据、商业DEM数据和Landsat 8 OLI多光谱遥感数据，构建3种考古遗址预测模型（DEM模型、多光谱遥感模型和混合模型）。研究表明，黄土高原地区先秦遗址分布对坡向、纵向曲率、地形开阔度等地貌因素具有明显的相关性，而引入主成因素分析法（PCA）提取的多光谱因子不仅可以有效提高单纯基于DEM派生数据构建的遗址预测模型的精度和预测效率，而且还能够很好地改善遗址预测模型的空间结构，增强遗址预测模型的应用效果。基于上述混合模型的分析，陇东黄土高原地区先秦时期的土地资源开发，从仰韶文化时期的不足10%，迅速提升至龙山文化时期的43.1%，先秦时期整体达到了57.3%的规模，土地资源的加速开发奠定了该地区早期文明发展的基础。

This paper aims to explore a new method on the traditional archaeological Site Predictive Model (SPM) by integrating Multi-Spectral Remote Sensing (MSRS) resources. The application on the case study from the Loess Plateau of east Gansu Province shows great prospect of the combination of SPM and MSRS on the regional archaeological studies and cultural resource managements.The SPM is a useful statistic model in archaeological research and cultural heritage management practices. Previous archaeological applications of the SPMs relied heavily on the Digital Elevation Model (DEM) while overlooked the MSRS. This research attempts to integrate the MSRS, mainly the Landsat 8 OLI data, into the SPMs to increase the model’s predicting efficiency and to acquire high-quality predictive models, which can be used to evaluate the ancient land use degree in statistics.This research compares three SPMs that were based on the DEM-derived terrain factors, the factors from MSRS, and the mixed factors. The terrain factors include the following. First, the landform indices, such as slope, aspect, curvature, texture, and convexity. Second, the hydrological indices, such as channel network base level, valley depth, and vertical distance to rivers. Third, the topographic position indices, such as relative slope, topographic openness, and wind exposition. The MSRS factors are derived by the reduction of the PCA method on Landsat 8 OLI’s band 1—7. The mixed factors are the combination of terrain and MSRS factors totally. Archaeological data dated to the pre-Qin period from the systematic archaeological survey in east Gansu Province are incorporated into the three SPMs by using the Logistic Regression provided by Kvamme (1983) and AIC gradient optimization. Three assessment indices, including Goodness-of-Fit R², AIC and, G value are calculated to test the three models on their predictive effects. Two spatial analysis methods, including cross raster statistics and cross PCF on sites vs. models are incorporated to evaluate the models on their spatial structures.The comparison of the three models indicates that the MSRS can significantly improve the validity of the modular prediction with the greater R² (model1=0.265, model2=0.312, model3=0.414), G value (model1=0.51, model2=0.61, model3=0.64) and smaller AIC (model1=444.6, model2=420.0, model3=379.1). Meanwhile, models that combined MSRS data can achieve a better spatial structure than the DEM. Further analysis based on the mixed SPM (model3) indicates that the total land-use areas in the eastern part of the Loess Plateau increased from less than 10% during the Yangshao period to up to 43.1% during the Longshan period. The latter laid a solid foundation for the development of social complexity in the Pre-Qin period. The analysis on the MSRS-based model also reveals that the combination of bands 5-2-1 in OLI can provide more information on ancient sites.The land-use of Loess Plateau in east Gansu Province has demonstrated strong continuity since the prehistoric period. Thus, the modern multi-spectral remote sensing data can be well utilized into traditional archaeological site predictive model not least to improve its spatial structure and raise the model’s efficiency. Landsat 8 OLI can provide effective multi-spectral data for site identification and land-use classification. The band 5 of OLI potentially contains information on archaeological sites, which must be studied further.