Objective  The aims were to study the scale effects on soil fertility assessment in tobacco fields in Yunnan-Guizhou Plateau, so as to provide scientific basis for precise fertilization and different soil management decisions.

  Method  Soil samples of plough layers (0 - 20 cm) of 100, 180 and 220 typical tobacco fields were collected respectively in Zhongshan, Shuicheng and Panzhou regions of Liupanshui city, Guizhou Province in Yunan-Guizhou Plateau. Nine fertility indicators including pH, organic matter (OM), alkali-hydrolyzed nitrogen (AN), available phosphorus (AP), rapid available potassium (AK), water-soluble chlorine (Cl⁻), exchangeable magnesium (Mg^{2 +} ), available boron (B) and available zinc (Zn) were determined. Based on the scale of the whole city (L scale, all 500 samples) and the scale of the 3 tobacco-growing area (S scale, samples of each region), the weights of fertility indicators were determined by principal component analysis (PCA), soil integrated fertility index (IFI) was calculated by weighted sum method, and the spatial distribution of IFI was formed by the interpolation method of simple kriging.

  Result  ① Soil fertility indicators were affected by the region, environmental variables (altitude, annual mean temperature and precipitation), soil type and parent material. There were significant differences between the weights of soil fertility indicators calculated at different scales. The relative differences in weights of fertility indicators were increased with the increase of the number of tobacco fields. ② At L and S scales, the mean IFI of the three regions were all higher than 0.5. IFI of L scale in Zhongshan and Panzhou were significantly higher than that of S scale (P < 0.05), but no significant difference was found in IFI at L and S scales in Shuicheng. Below the higher level of IFI (< 0.6), the number of tobacco fields at L scale was less than that at S scale. ③ At L and S scales, the spatial distribution of IFI was different in different regions. Compared with L scale, for Zhongshan, the high IFI region (0.8 -1) in the northeast was decreased while the low IFI region (0 - 0.4) in the northwest and southeast was increased at S scale. For Shuicheng and Panzhou, the spatial distribution of IFI was similar at L and S scales, but the pattern of IFI was more fragmented at S scale in Shuicheng, and the area of the median region (0.4 - 0.6) at S scale in Panzhou. ④ The number or density of samples affected the difference of map-spot numbers at L and S scales. The more samples and the more suitable L scale were for evaluation, which could better reflect the spatial differentiation of IFI and were beneficial to guide the precise fertilization and soil improvement.

  Conclusion  Soil fertility evaluation in Yunnan-Guizhou Plateau has obvious scale effect, and the spatial scale should be considered in soil fertility evaluation so as to reveal the spatial differentiation of soil fertility. It could serve for precision fertilization and soil improvement.