Dynamic Global Vegetation Models (DGVMs) desc-ribe the structure, distribution, and succession of vegetation by applying basic ecological and physiological principles ( Prentice et al., 2007). DGVMs have successfully reproduced the global distribution of vegetation and simu-lated changes of the carbon, water, and nutrient cycles. They have also been employed to study many theoretical and applied ecological problems, e.g., to reveal the impact of fire disturbance on forest coverage ( Bond et al., 2005), and to reproduce the past and predict the future distribution of vegetation ( Cox et al., 2000; Friedlingstein et al., 2006; Prentice et al., 2011; Scheiter et al., 2012). However, there are fewer DGVMs than Earth System Models, and several Earth System Models apply the same DGVM. Consequently, many model parameters have to be reoptimized, so maintaining consistency among different component models requires care during coupling. This paper presents a coupling of Dynamic Global Vegetation Model of the Institute of Atmospheric Physics Chinese Academy of Sciences (IAP-DGVM) to the Common Land Model (CoLM), and presents a preliminary assessment of the coupled model in reproducing the global distribution of ecosystems.

IAP-DGVM ( Zeng et al., 2014) is a Dynamic Global Vegetation Model developed by the Institute of Atmospheric Physics, Chinese Academy of Sciences. It incorporates several recently developed parameterization sche-mes, including the shrub sub-model ( Zeng et al., 2008; Zeng, 2010), a process-based fire parameterization ( Li et al., 2012), and establishment and competition para-meteri-zation schemes ( Song and Zeng, 2014). IAP-DGVM applies the concepts of plant functional types (PFTs) and divides global vegetation into 12 PFTs according to their physiological characteristics. The performance of IAP- DGVM has previously been evaluated by offline simulations, by coupling to the Community Land Model (CLM3.0) as a test-bed. The capability of IAP-DGVM has been demonstrated in reproducing the global distribution of trees, shrubs, grasses, and bare soil under observed atmospheric conditions ( Zeng et al., 2014), as well as the dependence of vegetation distribution on climate conditions ( Zeng, 2010).

CoLM is a land surface model developed by Beijing Normal University. The initial version of CoLM ( Dai et al., 2003) combined the best features of the Land Surface Model of the National Center for Atmospheric Research (NCAR LSM) ( Bonan, 1996), the IAP land model (IAP94) ( Dai and Zeng, 1997), and the Biosphere-Atmosphere Transfer Scheme (BATS) ( Dickinson et al., 1993), and was eventually adopted as the Community Land Model ( Oleson et al., 2004). CoLM continues to undergo further development: improvements have included calculations of leaf temperature, photosynthesis, stomatal density, thermal, and hydrological processes in the soil, and energy and water balance ( Ji and Dai, 2010). The most recent version of CoLM includes a DGVM, originating from Lund-Potsdam-Jena (LPJ)-DGVM ( Sitch et al., 2003), and incorporates an early version of shrubs ( Zeng et al., 2008).

Figure 1

	Figure Option ViewDownloadNew Window
	Figure 1 The geographical distribution of the 10 regions.

Both IAP-DGVM and CoLM have become component models of the Chinese Academy of Sciences’ Earth System Model (CAS-ESM). First, IAP-DGVM was directly coupled to CoLM by replacing the original DGVM, while preserving both models‘ original parameters. This work investigates the performance of the coupled model, by finding inconsistencies in the coupling, as well further optimizing the model parameters.

Two global offline simulations were conducted: one using the coupled model, CoLM-IAP, and the other using CoLM with its own DGVM, CoLM-LPJ. Both simulations were circularly forced using the atmospheric data of Qian et al. (2006), and run for longer than a 1000 year period with a T42 resolution (128 × 64 grid cells) until a state of equilibrium was reached. The last 50 years of the simulation were analyzed. Observed global vegetation distributions came from the CLM4 surface vegetation dataset, developed by Lawrence and Chase (2007) and Ramankutty et al. (2008). The results were compared with our previous simulations of IAP-DGVM coupled with CLM3 (denoted as CLM-IAP) ( Zeng et al., 2014), in which the parameters of IAP-DGVM were optimized.

This paper studies the models’ capabilities in simulating the fractional coverage (FC) of different vegetation types. Besides the comparison of global distribution, 10 regions (Fig. 1) were selected, based on the work of Xue et al. (2010), to focus on different vegetation types, e.g.,evergreen trees in the Amazon and Central Africa, deciduous trees in eastern North America and Europe, shrubs in Siberia, and grasses in the South American savanna and West Africa.

Figure 2 shows the differences in FC of evergreen trees over the 10 regions for all of the models we compared. Except for the Amazon and Central Africa, CoLM-LPJ overestimated tropical evergreen tree coverage over most low latitude regions (Fig. 2a), with average errors up to 47.9% and 36.3% over South American and West African savanna respectively (Table 1). It underestimated needleleaf evergreen trees over mid-high latitudes. Globally, CoLM-IAP simulated higher coverage of evergreen trees (Fig. 2b). The simulation over the Amazonian tropical forest was improved, but the biases over South American savanna, the southern edge of the Sahara desert and also South Asia were slightly enlarged. However, CoLM-IAP did not show the improvement in boreal mid-high latitudes in contrast with CoLM-LPJ. CLM-IAP had the fewest simulation biases for evergreen tree coverage, reflected in the fact that it found higher FC over the Amazon and boreal regions, and lower FC over the South American savanna, West Africa, and South Asia (Table 1).

Table 1

Table 1

Table 1 The averaged fractional coverage percentage of evergreen trees, deciduous trees, shrubs, grasses, and bare soil over the 10 study regions as simulated using the Common Land Model coupled with its own dynamic global vegetation model (Co), Common Land Model coupled with dynamic global vegetation model of Institute of Atmospheric Physics (I), Community Land Model coupled with dynamic global vegetation model of Institute of Atmospheric Physics (C), and observation (O). 1 2 3 4 5 6 7 8 9 10 Evergreen tree Co 73.8 57.1 32.7 36.9 22.7 21.3 64.9 24.1 30.2 4.3 I 82.8 74.7 41.2 54.8 32.3 32.1 72.1 35.1 33.5 4.2 C 86.7 59.4 27.7 15.6 27.8 28.4 53.7 22.5 31.4 21.3 O 82.4 35.7 19.1 0.6 30.1 35.2 17.0 2.8 23.1 16.7 Deciduous tree Co 18.2 17.2 26.5 32.3 46.3 37.0 14.8 12.5 33.8 29.3 I 9.2 12.4 12.6 23.8 30.9 23.1 14.9 16.8 27.4 21.7 C 7.2 13.0 20.0 11.0 30.1 29.7 19.1 14.5 28.1 12.6 O 11.0 17.7 13.8 9.7 15.7 28.2 27.3 18.0 14.8 15.0 Shrub Co 0.0 0.0 5.5 0.0 0.0 0.0 0.0 3.7 2.7 0.7 I 0.0 0.0 6.9 0.0 0.2 0.0 1.5 3.7 3.4 1.0 C 0.1 4.3 11.7 0.0 6.1 0.2 1.8 8.8 7.7 31.3 O 0.0 0.3 2.4 0.0 9.3 4.5 0.8 6.2 5.3 37.6 Grass Co 4.8 4.2 10.0 3.0 19.4 38.6 5.4 5.8 24.8 53.6 I 4.8 4.9 11.8 2.5 22.9 41.6 6.4 11.6 26.2 60.8 C 5.5 11.3 13.3 25.5 14.5 26.4 18.8 35.3 21.6 18.1 O 5.2 33.3 15.8 33.6 15.5 17.5 41.9 48.4 23.9 11.7 Bare soil Co 0.0 18.2 23.9 27.2 1.0 0.5 14.0 53.3 4.6 4.5 I 0.0 4.6 26.0 18.5 3.2 0.5 4.1 32.0 5.7 4.6 C 0.0 7.4 23.6 46.9 7.5 4.1 1.2 17.3 9.1 8.3 O 0.3 4.7 24.8 42.1 4.6 2.0 2.1 18.2 7.9 12.1

Table 1 The averaged fractional coverage percentage of evergreen trees, deciduous trees, shrubs, grasses, and bare soil over the 10 study regions as simulated using the Common Land Model coupled with its own dynamic global vegetation model (Co), Common Land Model coupled with dynamic global vegetation model of Institute of Atmospheric Physics (I), Community Land Model coupled with dynamic global vegetation model of Institute of Atmospheric Physics (C), and observation (O).

The difference in FC of deciduous trees showed a high spatial heterogeneity (Fig. 3). The FC of deciduous trees in CoLM-LPJ was overestimated over the Amazon, West Africa, and South Asia, but underestimated over South Ame-rican savanna and the south of Africa. Over mid-high latitudes, the FC of deciduous trees predicted by CoLM- LPJ was larger, particularly over eastern North America and Europe and had errors greater than 30%. In CoLM- IAP, the errors in FC of deciduous trees were reduced over most of the globe. For example, the average errors over the Amazon, West Africa, and South Asia were reduced by 5.4%, 8.5%, and 11.5%, respectively, while they were reduced by 15.4% and 6.4% over eastern North America and Europe. CLM-IAP improved the simulated deciduous tree FC over South American savanna and West Africa, with average errors reduced by 4.2% and 12.8%, whereas over eastern North America and Europe the biases were a little enlarged (Table 1).

Figure 4 shows the distributions of shrubs and grasses. In CoLM-LPJ and CoLM-IAP, the FC of shrubs was significantly underestimated, particularly for boreal shrubs. For example, over Siberia boreal shrub FC was predicted to be less than 1% in both CoLM-LPJ and CoLM-IAP, whereas CLM-IAP predicted an FC of shrubs exceeding 37% over Siberia, which is in closer agreement with observations. Correspondingly, CoLM-LPJ and CoLM-IAP overestimated the FC of grasses over mid-high latitudes. However, over low latitudes it underestimated because of the larger simulated FC of trees. Over South American savanna, and the west and south of Africa, the average FC of grasses was predicted to be less than 10%, while observations show that it can exceed 40%.

Figure 2

	Figure Option ViewDownloadNew Window
	Figure 2 The differences in percentage of fractional coverage of deciduous trees between (a) CoLM-LPJ and OBS, (b) CoLM-IAP and CoLM-LPJ, (c) CoLM-IAP and OBS, and (d) CLM-IAP and CoLM-IAP.

Figure 3

	Figure Option ViewDownloadNew Window
	Figure 3 The differences in percentage of fractional coverage of deciduous trees between (a) CoLM-LPJ and OBS, (b) CoLM-IAP and CoLM-LPJ, (c) CoLM-IAP and OBS, and (d) CLM-IAP and CoLM-IAP.

Figure 4

	Figure Option ViewDownloadNew Window
	Figure 4 The fractional coverage of (a) shrubs and (b) grasses in CoLM-LPJ. Panels (c-h) show the fractional coverage percentage of shrubs and grasses in CoLM-IAP, CLM-IAP, and OBS.

IAP-DGVM was coupled with CoLM, and its capability in simulating the FC of different plant functional types was compared with the default CoLM, the coupling of IAP-DGVM with CLM3, as well as observations.

The coupled model significantly improved on the ability of CoLM to simulate the global distribution of deciduous trees. The average errors over eastern North America and South Asia were reduced by 15.4% and 11.5% respectively. In addition, the increase of evergreen tree coverage over the Amazon and Central Africa also reduced the bias of overestimating deciduous trees in these regions by CoLM.

Some significant biases still remain in CoLM-IAP. For example, evergreen trees extended to the tropical savanna, and boreal evergreen trees and boreal shrubs were replaced by C3 arctic grass. However, such biases were not presented in our previous studies coupling IAP-DGVM with CLM3 ( Zeng et al., 2014). This is because CoLM and CLM employ different treatments of photosynthesis, respiration, and other related biophysical processes, resulting in differences in simulated gross primary production (GPP) and net primary production (NPP) between CoLM-IAP and CLM-IAP. Both CoLM-LPJ and CoLM- IAP had a higher NPP to GPP ratio over subtropical reg-ions, and a higher GPP but lower NPP to GPP ratio over boreal regions, compared to CLM-IAP. Although such differences may be not significant for the simulation of the global carbon cycle, they can be large enough to drive an ecosystem towards different equilibrium states. We continue to investigate the sensitivities of IAP-DGVM to the differences between CoLM and CLM, and hope to be able to reoptimize the model parameters in IAP-DGVM, and CoLM if necessary.

Acknowledgments. This work was supported by Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDA05110103) and the National Basic Research Program of China (Grant No. 2010CB951801).