Abstract: 【Objective】 To construct a biomass estimation model for the key growth stages of winter wheat and analyze the transferability of the estimation model under different water treatments and in different years scenarios, which could provide technical support for the rapid estimation of winter wheat above-ground biomass, phenotypic research, and crop water and fertilizer decision-making. 【Method】 In this study, by setting different water and nitrogen treatments, the DJI M600 Pro unmanned aerial vehicle（UAV） equipped with the Anzhou Technology K6 multispectral imager was used to acquire images of winter wheat during the key growth stages. The digital surface model（DSM） of the images was extracted, and the plant height was extracted based on the UAV images. The winter wheat above-ground biomass estimation model was constructed and improved through the BP neural network method. 【Result】 Under the natural condition of water-nitrogen coupling, the change in the measured plant height of winter wheat was relatively small, but irrigation under nitrogen-sufficient conditions could increase the measured plant height of winter wheat. The linear determination coefficient（R²） between the plant height extracted by the UAV and the measured plant height was 0.81, indicating that the plant height extracted by the UAV could explain 81% of the plant height variation. For the winter wheat above-ground biomass estimation model constructed based on the plant height extracted from UAV remote sensing images, R², rootmean-square error（RMSE） and relative performance deviation（RPD） were 0.58, 4528.23 kg/ha and 1.25 respectively. This showed that the model could rapidly estimate the winter wheat above-ground biomass, but the model had poor robustness（RPD＜1.4）. The estimated value（16198.27 kg/ha） was smaller than the measured value（16960.23 kg/ha）, and the estimated values were relatively scattered. Through data transformation, for the winter wheat biomass estimation model constructed based on the ratio( above-graund biomass/plant height extracted by UVA ration, R², RMSE and RPD were 0.88, 2291.90 kg/ha and 2.75 respectively. The improved model had strong robustness（RPD＞2.0）. The estimated value（17478.21 kg/ha） was close to the measured value（17222.59 kg/ha）, and the model estimation accuracy has increased by 51.72%. It was verified that the improved winter wheat above-ground biomass estimation model had strong transferability under different water treatments and in different years. R² of the transfer estimation model was above 0.85, achieving accurate and rapid estimation of winter wheat above-ground biomass.【Conclusion】 Extracting plant height information using UAV images and improving the winter wheat above-ground biomass estimation model through data transformation can effectively improve the estimation accuracy of wheat biomass estimation model. The improved winter wheat above-ground biomass estimation model shows strong transfer ability under different water treatments and in different years scenarios. However, there are differences in its transferability under different nitrogen-level scenarios. Therefore, before applying the model for transfer estimation, histogram feature analysis should be carried out on the datasets of different scenarios, and various influencing factors should be comprehensively considered to enhance the generalization ability and robustness of the model.