Abstract

Because of a lack of observations, historical simulations of land surface conditions using land surface models are needed for studying variability and changes in the continental water cycle and for providing initial conditions for seasonal climate predictions. Atmospheric forcing datasets are als... Show moreBecause of a lack of observations, historical simulations of land surface conditions using land surface models are needed for studying variability and changes in the continental water cycle and for providing initial conditions for seasonal climate predictions. Atmospheric forcing datasets are also needed for land surface model development. The quality of atmospheric forcing data greatly affects the ability of land surface models to realistically simulate land surface conditions. Here a carefully constructed global forcing dataset for 1948-2004 with 3-hourly and T62 (~1.875°) resolution is described, and historical simulations using the latest version of the Community Land Model version 3.0 (CLM3) are evaluated using available observations of streamflow, continental freshwater discharge, surface runoff, and soil moisture. The forcing dataset was derived by combining observation-based analyses of monthly precipitation and surface air temperature with intramonthly variations from the National Centers for Environmental Prediction--National Center for Atmospheric Research (NCEP-NCAR) reanalysis, which is shown to have spurious trends and biases in surface temperature and precipitation. Surface downward solar radiation from the reanalysis was first adjusted for variations and trends using monthly station records of cloud cover anomaly and then for mean biases using satellite observations during recent decades. Surface specific humidity from the reanalysis was adjusted using the adjusted surface air temperature and reanalysis relative humidity. Surface wind speed and air pressure were interpolated directly from the 6-hourly reanalysis data. Sensitivity experiments show that the precipitation adjustment (to the reanalysis data) leads to the largest improvement, while the temperature and radiation adjustments have only small effects. When forced by this dataset, the CLM3 reproduces many aspects of the long-term mean, annual cycle, interannual and decadal variations, and trends of streamflow for many large rivers (e.g., the Orinoco, Changjiang, Mississippi, etc.), although substantial biases exist. The simulated long-term-mean freshwater discharge into the global and individual oceans is comparable to 921 river-based observational estimates. Observed soil moisture variations over Illinois and parts of Eurasia are generally simulated well, with the dominant influence coming from precipitation. The results suggest that the CLM3 simulations are useful for climate change analysis. It is also shown that unrealistically low intensity and high frequency of precipitation, as in most model-simulated precipitation or observed time-averaged fields, result in too much evaporation and too little runoff, which leads to lower than observed river flows. This problem can be reduced by adjusting the precipitation rates using observed-precipitation frequency maps. Show less