CMADS Verification
Evaluation and Hydrological Simulation of CMADS and CFSR Reanalysis Datasets in the Qinghai-Tibet Pl
Next
Verification

  Multisource reanalysis datasets provide an effective way to help us understand hydrological processes in inland alpine regions with sparsely distributed weather stations. The accuracy and quality of two widely used datasets, the China Meteorological Assimilation Driving Datasets to force the SWAT model (CMADS), and the Climate Forecast System Reanalysis (CFSR) in the Qinghai-Tibet Plateau (TP), were evaluated in this paper. The accuracy of daily precipitation, max/min temperature, relative humidity and wind speed from CMADS and CFSR are firstly evaluated by comparing them with results obtained from 131 meteorological stations in the TP. Statistical results show that most elements of CMADS are superior to those of CFSR. The average correlation coefficient (R) between the maximum temperature and the minimum temperature of CMADS and CFSR ranged from 0.93 to 0.97. The root mean square error (RMSE) for CMADS and CFSR ranged from 3.16 to 3.18 °C, and ranged from 5.19 °C to 8.14 °C respectively. The average R of precipitation, relative humidity, and wind speed for CMADS are 0.46; 0.88 and 0.64 respectively, while they are 0.43, 0.52, and 0.37 for CFSR. Gridded observation data is obtained using the professional interpolation software, ANUSPLIN. Meteorological elements from three gridded data have a similar overall distribution but have a different partial distribution. The Soil and Water Assessment Tool (SWAT) is used to simulate hydrological processes in the Yellow River Source Basin of the TP. The Nash Sutcliffe coefficients (NSE) of CMADS+SWAT in calibration and validation period are 0.78 and 0.68 for the monthly scale respectively, which are better than those of CFSR+SWAT and OBS+SWAT in the Yellow River Source Basin. The relationship between snowmelt and other variables is measured by GeoDetector. Air temperature, soil moisture, and soil temperature at 1.038 m has a greater influence on snowmelt than others (Fig.1).

Fig.1 The Locations of TP and the Digital Elevation Model of Yellow River Source Basin

    In the monthly scale, NSE for CMADS+SWAT and CFSR+SWAT range from 0.42 to 0.68, which are improvements compared to OBS+SWAT. OBS+SWAT underestimate runoff: the NSE is −0.8 and −0.72 in calibration and validation over the period of a monthly scale. CFSR+SWAT overestimate runoff in most years and seriously overestimated the summer runoff in 2009, 2011 and 2013. NSE of CFSR+SWAT is 0.59 and 0.42 in calibration and validation over the period of a monthly scale. CMADS+SWAT do well in forcing the hydrological model and the simulation results are better, NSE of CMADS+SWAT is 0.68 to 0.58 in the monthly scale (Fig.2).

water-10-00513-g006.png

Fig.2 Simulation result of OBS+SWAT, CMADS+SWAT and CFSR+SWAT in Yellow River Source Basin in monthly scale (2009–2013)


For detail, see ref: Liu, J.; Shanguan, D.; Liu, S.; Ding, Y. Evaluation and Hydrological Simulation of CMADS and CFSR Reanalysis Datasets in the Qinghai-Tibet Plateau. Water 2018, 10, 513.   http://www.mdpi.com/2073-4441/10/4/513.