Land cover, considered as an essential component of climate systems, plays a critical role in land-atmosphere interactions. Its impact on the energy budget, water cycle, and biochemical composition in a given area is reflected in surface albedo, roughness, and soil water and thermal characteristics. Because many parameters in land surface models are defined on the basis of land cover types, an accurate land cover description is fundamental for land surface simulation ( Sellers et al., 1997).

Many studies investigating the importance of realistically delineating land surface information in land surface and climate simulations (e.g., Tian et al., 2004a, b; Lawrence and Chase, 2007; Gao et al., 2008; Gao and Jia, 2013) have demonstrated the complex relationships between land cover, water, and energy balance, and overlying atmospheric processes. However, common land cover maps based on satellite-derived remote sensing data are inconsistent, especially in Asia, and do not provide accurate land cover information for land surface and climate modeling for China ( McCallum et al., 2006; Chen and Xie, 2009; Ran et al., 2010; Gao and Jia, 2012). It follows that it is necessary to collect high-quality land cover data to reduce these uncertainties in land surface simulation for China.

In this study, we employed the high-accuracy multisource integrated Chinese land cover dataset (MICLCover) developed by Ran et al. (2012), which combines multisource data based on the Dempster-Shafer evidence theory, in the Community Land Model version 4 (CLM4) to assess how these land cover parameters affect land surface simulation over China. In CLM4, plant functional types are derived from MODIS (MODerate resolution Imaging Spectroradiometer) satellite data, while glaciers, lakes, wetlands, and urban areas are obtained from various other sources. Hereafter, the default land cover data in CLM4 is simply denoted as MODIS. Two 33-yr simulations (1972-2004) using MICL data and MODIS land cover data were conducted for China.

Based on the Dempster-Shafer theory, Ran et al. (2012) developed an MICLCover map using the International Geosphere-Biosphere Program (IGBP) classification system by combining the Vegetation Atlas of China (1:1000000) ( Hou et al., 2001), a land use map of China (1:100000) for the year 2000 ( Liu et al., 2002), a glacier distribution map of China (1:100000) ( Wu and Li, 2004), a swamp-wetland map of China (1:1000000) ( Zhang, 2002), and the MODIS product for 2001 (MODIS2001) ( Friedl et al., 2002). The map was further developed to show 1-km resolution glacier, lake, wetland, and urban land use, and 16 plant functional types consistent with the classifications in the Community Land Model (downloadable from the Cold and Arid Regions Science Data Center:http://westdc.westgis.ac.cn/data/). We scaled the MICLCover map up from 1 km to 0.5° by means of area-weighted averages, and then extracted the ratios of the areas of glacier, lake, wetland, and urban land use and the 16 plant functional types from the MICLCover map. In CLM4, the default plant functional types were based on MODIS satellite data ( Lawrence and Chase, 2007) from November 2000 to October 2001 (the MODIS2001 product, mentioned above); glaciers were obtained from the IGBP Data and Information System Global 1-km Land Cover Data Set (IGBP DISCover) ( Loveland et al., 2000); lake and wetland percentages were derived from 1°×1° data for perennial freshwater lakes and swamps/marshes ( Cogley, 1991); and urban areas were derived from a population density dataset, LandScan 2004 ( Dobson et al., 2000).

Comparisons of area percentages of plant functional types between MICL data and MODIS data for six broad categories of plant functional types are shown in Fig. 1. According to the mean annual precipitation ( P_mean), we divided China into four climate zones: arid region ( P_mean≤ 200 mm); semiarid region (200 mm < P_mean≤ 400 mm); semihumid region (400 mm < P_mean ≤ 800 mm); and humid region ( P_mean > 800 mm) . Precipitation values from Qian et al. (2006) were used here, and also used for atmospheric forcing in this study, to enable comparable simulations and analyses to be performed. The detailed differences for each climate zone using the two datasets are listed in Table 1. Overall, the areas of bare soil and trees was reduced by 14.6% and 5.5% in the MICL data compared to the MODIS data over China as a whole, mainly because of increases in grassland and crop areas by 9.9% and 5.0% (Table 1). The greatest change in vegetation cover types was the large decrease in bare soil corresponding to the large increase in grass cover, both in arid regions (-18.6% corresponding to 17.0%) and in semiarid regions (-35.6% corresponding to 36.6%), specifically in areas of Inner Mongolia, the Tibetan Plateau, and mountainous areas of northwestern China (Figs. 1a and 1e). The MICL data specified smaller areas of tree and grass cover and greater crop cover over eastern China (Figs. 1b, 1c, and 1e), and greater shrub cover in southern China (Fig. 1d).

Figure 1

	Figure Option ViewDownloadNew Window
	Figure 1 Percentage differences in area of plant functional types between MICL (Multisource Integrated Chinese Land cover) data and MODIS (MODerate resolution Imaging Spectroradiometer) data in China at 0.5° grid resolution. Isolines refer to the mean annual precipitation (mm yr-1).

Table 1

Table 1

Table 1 Differences of area percentage of land cover types between MICL data and MODIS data in China overall, and in arid, semiarid, semihumid, and humid climate sub-regions in China. China Arid Semi-arid Semi-humid Humid Bare soil -14.6 -18.6 -35.6 -8.8 -4.2 Needleleaf tree -3.6 0.0 -0.5 -5.8 -6.3 Broadleaf tree -1.9 0.1 0.4 -1.0 -5.9 Shrub 1.8 -1.2 -4.6 0.6 9.0 Grass 9.9 17.0 36.6 1.3 -2.8 Crop 5.0 0.4 -0.4 9.0 8.1 Glacier 0.5 1.1 0.7 0.1 0.4 Lake 1.6 0.8 1.7 1.5 2.3 Wetland 1.1 0.5 1.4 2.4 0.0 Urban 0.0 0.1 0.3 0.7 -0.9

Table 1 Differences of area percentage of land cover types between MICL data and MODIS data in China overall, and in arid, semiarid, semihumid, and humid climate sub-regions in China.

The Community Land Model version 4.0 ( Lawrence et al., 2011) (CLM4SP, where SP = satellite phenology) was employed in this study. CLM4 is an extensive update of the previous CLM version 3.5, incorporating changes to the model structure and parameterizations, including soil hydrological and thermal processes, a revised snow model, an added urban canyon model, a carbon-nitrogen cycle modeling capability, and transient land cover changes, among other modifications.

Two integration experiments were conducted at 0.5° resolution to explore the use of MICL data on a land surface simulation over China: (1) a control run (MODIS) with default land cover data; and (2) a comparative run (MICL) with MICL data. Besides the differences in land cover data for the two experiments, other land surface information and model settings were exactly the same. Atmospheric forcing data from Qian et al. (2006) for 1972-2004 circularly forced CLM4 three times. Altogether, the experiments covered a 99-yr period, of which the first 66-yr simulation period was regarded as spin-up. The results for the final 33 years (1972-2004) were used for the following analysis.

A land cover dataset for China (MICLCover) developed by Ran et al. (2012) combining multiple information sources was used in CLM4 in this study to assess the impacts of the new land cover information on land surface simulation over China. Two 33-yr simulations (1972- 2004) were conducted, one using the MICL dataset as the input for the CLM4, and the other using the default MODIS dataset.

Compared to the MODIS data, simulations using the MICL data showed decreases in bare soil cover (reduced by 14.6%), needleleaf tree cover (3.6%), and broadleaf tree cover (1.9%), resulting from increased areas of shrubland (increased by 1.8%), grassland (9.9%), crops (5.0%), glaciers (0.5%), lakes (1.6%), wetlands (1.1%), and unchanged urban areas. These differences led to a reduction of 0.3% in the simulated mean annual surface albedo over China as a whole, with the largest change occurring across semiarid regions (-2.1%). Simulated surface albedo using MICL data was more consistent with observed albedo than previously, for arid and semiarid regions. The land surface temperature increased over Inner Mongolia and the western Tibetan Plateau using MICL data mainly due to the reduced surface albedo. A reduced number of trees over the eastern humid region resulted in an increased land surface temperature.

Sensible heat flux was shown to increase greatly in areas where the surface albedo decreased by a large amount, resulting in higher levels of incident shortwave radiation being absorbed by those surfaces. However, no significant change (0.1 W m^-2) was found in the mean annual sensible heat flux when averaged over China, due to variations in arid and semiarid regions being almost cancelled out by opposite variations in humid and semihumid regions. The latent heat flux increased by 1 W m^-2, with almost identical changes being simulated across the whole of China.

In addition, both sets of simulations captured the same spatial distribution of mean runoff as obtained from the Global Runoff Data Centre (GRDC) data. The mean simulated annual runoff over all of China decreased by 6.8 mm yr^-1, consisting of decreased surface runoff of 2.6 mm yr^-1 and decreased subsurface runoff of 4.2 mm yr^-1, with the greatest contribution to this change being from humid regions.

Accurate descriptions of land cover have the potential to improve land surface modeling and even climate modeling; therefore, more simulations and analyses are required to further investigate the detailed mechanism of climate effects.

Acknowledgements. This study was supported by the National Basic Research Program of China (Grants Nos. 2010CB951101 and 2010CB428403) and the National Natural Science Foundation of China (Grant No. 91125016).