Impact of vegetation properties on U.S. summer weather prediction

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 101, NO. D3, PAGES 7419-7430, MARCH 20, 1996 Impact of vegetation properties on U.S. summer weather prediction Yongkang Xue and Michael J. Fennessy Center for Ocean-Land-Atmosphere Studies, Calverton, Maryland Piers J. Sellers NASA Goddard Space Flight Center, Greenbelt, Maryland Abstract. Systematic biases in U.S. summer integrations with the Center for Ocean- Land-Atmosphere Studies (COLA) atmospheric general circulation model (GCM) have been identified and analyzed. Positive surface air temperature biases of 2°—4°K occurred over the central United States. The temperature biases were coincident with the agricultural region of the central United States, where negative precipitation biases also occurred. The biases developed in‘June and became very significant during July and August. The impact of the crop area vegetation and soil properties on the biases was investigated in a series of numerical experiments. The biases were largely caused by the erroneous prescription of crop vegetation phenology in the surface model of the GCM. The prescribed crop soil properties also contributed to the biases. On the basis of these results the crop model has been improved and the systematic errors in the U.S. summer simulations have been reduced. The numerical experiments also revealed that land surface effects on the atmospheric variables at and near the surface during the North American summer are very pronounced and persistent but are largely limited to the area of the anomalous land surface forcing. In this regard, the midlatitude land surface effects described here are similar to those previously found for tropical regions. 1. Introduction Land surface processes have been shown to have substantial effects on short-term weather predictions and long-term cli- mate projections. Changes in land surface conditions influence the atmospheric circulation by modifying the surface energy balance and hydrological cycle. For example, Rowell and Blondin [1990] showed that the 5-day weather forecast for West Africa from the European Center for Medium Range Weather Forecasts (ECMWF) operational forecasting model was sensitive to the surface moisture distribution. Xue and Shukla [1993] used the Center for Ocean-Land-Atmosphere (COLA) Studies general circulation model (GCM) to simulate one of the observed African drought anomaly patterns in re- sponse to changes in the land surface characteristics. This GCM includes a simplified version of Sellers et al.’s [1986] simple biosphere model (SSiB) [Xue et al., 1991]. The simple biosphere model (SiB) [Sellers et al., 1986] was designed to simulate the interactions between the Earth’s land surface and the atmosphere by treating the vegetation explic- itly and realistically. A comparison between simulations with the COLA GCM coupled to SiB and the same GCM coupled with a conventional hydrological model shows that the coupled biosphere-atmosphere model produces a more realistic parti- tioning of energy at the land surface [Sato et al., 1989]. Both SiB and SSiB have been validated by using observational data from many field experiments, including the Amazon tropical rainforest experiment [Sellers er al., 1989; Xue et al., 1991], the First ISLSCP International Satellite Land-Surface Climatology Copyright 1996 by the American Geophysical Union. Paper number 9SJD02169. 0l48-0227/96/95JD-02169$05.00 Project Field Experiment (FIFE) [Sellers et al., 1992; Chen et al., this issue], the Anglo—Brazilian Amazonian Climate Obser- vation Study [Xue et al., 1995a], the HAPEX-Mobilhy experi- ment on a crop—grassland site in France [Shao and Henderson- Sellers, 1995; Xue et al., 1995b]. These calibrations have provided vegetation and soil property information for some vegetation types, and have led to improvements in the surface biosphere model, resulting in more realistic simulations. How- ever, these field measurements were made at a few sites for only about one third of the SiB vegetation types. Because of differences in spatial scales, the application of data from these site studies to GCM simulations needs further investigation. For vegetation types with little or no observational data, indi- rect information on vegetation and soil properties from scien- tific literature has to be used. We will address this problem in more detail in section 2. Although land surface modeling can enhance our ability to understand land surface-atmosphere interactions, poor or inadequate representation of surface pro- cesses or land surface conditions may have a negative impact on weather prediction and climate studies. In this study, systematic errors in U.S. summer seasonal predictions with the COLA GCM have been identified. The simulated June, July, and August (JJA) mean surface temper- ature in the central U.S. was 2°—4°K higher than observations, and the JJA mean precipitation was about 1 mm d‘1 (30%) lower than observations in the same region. Similar systematic errors have been noted in other GCMs and weather forecast models. In a 10-year integration using the National Center for Atmospheric Research (NCAR) Community Climate Model (CCM2), Bonan [1994] found large systematic warm temper- ature biases (10°—15°K) in central North America, which A. Hahmann and R. E. Dickinson (personal communication, 7419