With the expansion of the drainage basin development since the middle of the 20^th Century, catchment export and lake-water enrichment of nutrients have led to a continuous process of lake eutrophication in Yunnan, which has significantly impacted the structuring of biological communities and health of the lake ecosystems. On the regional scale, the time span of modern water quality monitoring records is generally short (i.e.<20 years); this has prevented a reliable evaluation of pre-disturbance level of nutrient status for the lake ecological restoration. The application of paleolimnological methods to establish diatom-nutrient quantitative model can provide important information for water quality evaluation and ecological restoration of the polluted lakes. In this study, 45 lakes in Southwest Yunnan were selected for analyzing the surface sediment diatom communities, and in combination with the water monitoring data, the key environmental gradient was then identified in driving diatom assemblages. We further identified that the lake-water TP was a significant environmental factor responsible for the spatial variations of diatom assemblages (a variance of 4.54% explained independently, P<0.01). Furthermore, the transfer function between the surface sediment diatoms and TP in 45 lakes was established through multivariate statistical analyses. Among the transfer function models established by Weighted Average Partial Least Squares regression (WA-PLS), the component 2 has the strongest prediction ability (R²_jack=0.465, RMSEP=0.396). Combining with the sediment diatom sequence of Lake Xingyun, lake-water TP values were reconstructed with this WA-PLS-2 model for the past 100 years. The null model results confirmed that lake-water TP played a significant role in driving diatom variations, and the predicted TPs show a significant correlation with the observed TPs (R²=0.79, P<0.001, n=13). The reconstruction results show that the background concentrations of TP, prior to the period of eutrophication (before 1950 AD), varied in the range of 12.8-29.3 μg/L for Lake Xingyun, which can be used as the restoration target of nutrient levels for lake restoration. This study shows that the quantitative reconstruction method of lake-water nutrient levels based on biological communities can provide reliable technical support for the construction of environmental restoration target for eutrophic lakes in Yunnan.