Precipitation Projection with Statistical Downscaling along the Heihe River Basin for the 21st Century

This paper focuses on the precipitation projection of the Heihe River basin with downscaling for 2011–2100 using Coupled Model Intercomparison Project Phase 5 multimodel ensemble, combined with European Center for Medium-Range Weather Forecasts reanalysis data and meteorological stations observation in the Heihe River basin. Grid precipitation projection is mapped onto observatory sites through three downscaling methods for bias corrections, which include the model drift removal (MDR), multivariate linear regression (MLR), and Bayesian model average (BMA). Results show that an overestimate of the 15-model ensemble precipitation in the Heihe River basin has not yet been totally removed after MDR is removed, owing to the presence of a nonstationary bias. However, it works well on bias correction if MLR and BMA are used in downscaling with the factors of v-wind, specific humidity, and geopotential height on 700 hPa. The test demonstrates that BMA has a good estimate on averaged precipitation, but it gives a low variance and correlation coefficient with the meteorological stations observation. Conversely, MLR can produce a good variance in precipitation and a high correlation coefficient, but a negative precipitation estimate often appears in the lower reaches of the river, especially in cold and dry seasons. These problems have been overcome to a great extent as soon as the model precipitation is introduced into the downscaling models. Moreover, the test also shows that BMA is in favor of the bias correction in the upper reaches of the river, whereas MLR is good at the site-precipitation estimate in the middle and lower reaches or the whole river basin. The precipitation projection with downscaling shows that the averaged precipitation at 14 sites of the basin would decrease in comparison with that of the 1971–2000 observation, with the rates of −9.7%, −12.5, and −12.1% for 2011–2040, 2041–2070, and 2071–2100, respectively. The rates of projected precipitations are 1.4%, 1.6%, and 2.3% in the upper reaches; −16.3%, −21.4%, and −22.6% in the middle reaches; and 13.0%, 4.2%, and 21.4% in the lower reaches for the three periods of projection, respectively. The projection shows that the precipitation would be increasing slowly in the upper reaches of the Heihe River basin, decreasing significantly in the middle reaches and for 2011–2040, and then increasing remarkably for 2041–2100 in the lower reaches. This implies that the water shortage problem would be intensified in the middle reaches, a farmland area with climate warming under RCP4.5 scenarios. A strategic adjustment is recommended to the structure of the agriculture and economics around the middle reaches for adapting to future climate change along the Heihe River basin.