A distributed water-heat coupled model for mountainous watershed of an inland river basin in Northwest China (III) using the outputs from Mesoscale model version 5

Abstract

The distributed water-heat coupled (DWHC) model is calibrated, with the help of the Mesoscale model version 5 (MM5), by calculating the daily precipitation, the daily average air temperature at the 2.0 m heights and the daily potential evaporation in Heihe mountainous watershed area and its vicinity (96.786°∼102.284°E, 37.328°∼40.601°N, 17 × 10⁴ km²), from February 11 to June 30, 2003. The MM5 model periodically ran every 10 days in 3 km × 3 km grid resolution with an integral time step of 3 s. In the MM5 model, many scheme or options are consulted or adopted, such as the Grell scheme cumulus parameterization method, the Dudhia option, the cloud-radiation scheme, MRF PBL option and the modified Oregon State University Land-surface model (OSULSM). According to the projection transform methods, the MM5 outputs are interpolated to the 1 km × 1 km grid in Alberts projection by using triangle-based cubic interpolation (Cubic) and nearest neighbor interpolation (Nearest) methods, with which the DWHC model shares the same method. The result shows that, when the Nearest method is used, the Nash-Sutcliffe equation value of the daily average runoff is 0.79, the balance error is −0.79% and the goodness of fit R ² value is 0.81. Meanwhile, when the Cubic method is used, the Nash-Sutcliffe equation value, the balance error and the R ² value are 0.79, −0.65% and 0.80, respectively. Though the runoff simulation result is not favorable, it is still better than that using measured data at the meteorological and hydrological stations; the latter has a Nash-Sutcliffe equation value of 0.61. The MM5-DWHC model results also show that runoff mainly occurs on land surfaces and from shallow soil layers. According to model calibration results, certain outputs of MM5 are singular to some extent and the DWHC model is very sensitive to the initial values.