Abstract: Abstract: Reference crop evapotranspiration (ET0) forecasting is important for real time irrigation scheduling. In this paper we improved the daily average modification method (DAM) and compared 3 ET0 forecasting methods including the Hargreaves-Samani (HS) model, DAM and improved daily average modification method (iDAM) for their accuracy in Zhanghe irrigation district. The data of annual sunshine duration, annual average temperature and annual precipitation of each meteorological stations in Zhanghe irrigation district were collected. The deviation range of these 3 factors were from -2.7% to 7.7%, from -1.2% to 1.2%, from -3.7% to 23.4%. The HS model was also improved for the local use. The ET0 calculated by FAO56-Penman-Monteith (PM) model was used as the control. In the DAM model, the daily annual average value of ET0 was estimated by fitness, which could introduce fitness error in the ET0 estimation. Thus, we improved the DAM by using the real observed historical meteorological data to calculate the daily annual average value of ET0. Daily historical meteorological data of Zhongxiang and Jingzhou Station for the period from January 1, 1999 to May 24, 2014 and the public weather forecasts of 7 days ahead from May 24, 2012 to May 24, 2014 were collected. The historical data were used to calculate the value of ET0 by PM model and the ET0 calculated for the period 1999-2008 and 2002-2011 were used to calibrate HS model and get the correction factors of weather type. The weather were classified into 4 types. The 3 methods were used to forecast ET0 from May 24, 2012 to May 24, 2014. The results showed that the mean absolute error (MAE) of the HS model in the calibration period and validation period were 0.46 and 0.46 mm/d, respectively. The root mean square error (RMSE) was 0.63 and 0.64 mm/d and the correlation coefficients were 0.92 and 0.91, respectively. It indicated that the improved HS model was suitable for ET0 estimation in Zhanghe irrigation district. The correction factors of weather type in Zhongxiang station were highest in the sunny day, followed by the cloudy, overcast and rainy day. The values were higher than North China Plain. The daily annual average value of ET0 by DAM was smaller in the days of 1-150 but higher in the 250-356 days, indicating that the improvement of DAM was necessary. In the Zhongxiang station, the MAE of HS model, DAM and iDAM methods were 0.75, 0.80, 0.76 mm/d, RMSE were 1.00, 1.07, 1.05 mm/d, and correlation coefficients were 0.82, 0.80, 0.80, respectively. In Jingzhou station, the MAE of the 3 methods above were 0.72, 0.90, 0.71 mm/d, RMSE were 0.95, 1.16, 0.99 mm/d, correlation coefficients were 0.84, 0.77, 0.82. Among the 3 methods, the iDAM method had the highest accuracy for the forecast horizon of 1 day but the HS method was the best for the forecast horizon of 2-7 days. With the increase of forecast period, the MAE and RMSE increased, indicating that the forecast accuracy decreased. Overall, the 3 proposed methods were well for ET0 forecasting and the best method was the HS model. In future, we can try to forecast ET0 using the HS model for irrigation forecast in Zhanghe irrigation district.