1. Introduction
During the last 3 decades, the role of land surface processes in controlling exchanges of energy, mass, and momentum between the land surface and atmosphere has been investigated by numerous studies through field measurements as well as numerical experiments (e.g., Sellers et al. 1988; Dickinson and Henderson-Sellers 1988; Xue and Shukla 1993; Betts et al. 1996; Xue et al. 2004, 2006; among many others). These studies generally show that interactions between the atmosphere and terrestrial ecosystem play a significant role in climate variations that is as important as atmospheric dynamics and composition, ocean circulation, and solar orbit perturbations (Pielke et al. 1998). In most biophysical land surface models, the leaf area index (LAI), defined as the one-sided green leaf area per unit ground area, is an important indicator of vegetation state because it affects the radiative transfer process within the canopy and evapotranspiration from the surface and consequently modulates near-surface climate and atmospheric circulations.
In this context, more than 2 decades of the Advanced Very High Resolution Radiometer (AVHRR) measurements have provided valuable information on vegetation properties such as Normalized Difference of Vegetation Index (NDVI), LAI derived from NDVI, and their long-term and interannual variations in association with climate variability (e.g., Myneni et al. 1998; Los et al. 2000). In spite of several factors that may cause uncertainties in the retrieved vegetation properties (e.g., intersensor discontinuity, orbital drift, volcanic aerosols, clouds, and water vapor, etc.), it has been proposed that effects of climate change have been apparent in vegetation’s phenology since the early 1980s, especially in mid- and high latitudes of the Northern Hemisphere. For example, Slayback et al. (2003) have found that significant long-term trends in NDVI can be identified among several datasets that differ in their approaches to correct for sensor calibration, orbital characteristics, and aerosols. Meanwhile, they also suggested that independent validation by comparison with other measurements is needed to help in identifying uncertainties in datasets.
Despite some qualitative agreement in terms of long-term trend and interannual variation of vegetation activity, there exist significant discrepancies in remote sensing–derived LAI (RSLAI) among datasets because of their different retrieval algorithms and atmospheric corrections even though measurements are from the same source. There have been a number of analytical studies to estimate uncertainties in the RSLAI and/or NDVI fields by conducting extensive error analysis (e.g., Los et al. 2000; Buermann et al. 2002) and by evaluating consistency between RSLAI and near-surface climatic variables such as sea surface temperature (SST), precipitation, and surface air temperature (e.g., Myneni et al. 1996; Los et al. 2000; Buermann et al. 2002; Lotsch et al. 2003). In addition, efforts have been made to evaluate the quality of RSLAI values by direct comparison with field measurements (e.g., Los et al. 2000; Buermann et al. 2002). This type of comparison, however, has usually been compromised to some degree because of the difference in spatial measurement scales.
To investigate the quality of global RSLAI datasets and the implications of their use in climate models, a number of numerical experiments have been conducted by using RSLAI estimates as a surface boundary condition of GCMs (e.g., Bounoua et al. 2000; Oleson and Bonan 2000; Buermann et al. 2001; Guillevic et al. 2002; van den Hurk et al. 2003; Tian et al. 2004a, b; Lawrence and Slingo 2004a, b). All of these studies stressed the importance of the accurate designation of land surface parameters in a climate model for improvements of simulations. Up until now, these studies have focused on impacts of RSLAI on the simulated climate in terms of the overall difference pattern at global scale and/or some areas associated with the same vegetation types, as well as the annual cycle of continental and/or global averages (Buermann et al. 2001; Lawrence and Slingo 2004a; Tian et al. 2004b). Another issue in RSLAI application is to estimate the impacts of uncertainties in RSLAI data in climate simulations. There have been several sensitivity experiments regarding this issue that investigate impacts of the potential uncertainties in the LAI dataset and its seasonality (Chase et al. 1996; van den Hurk et al. 2003), the surface vegetation fraction (Oleson and Bonan 2000; Bounoua et al. 2000), and the interannual variability of vegetation (Buermann et al. 2001; Guillevic et al. 2002). These studies used a single RSLAI dataset and assumed somewhat extreme potential changes in surface vegetation to design their experiments. However, the response of the models to uncertainties associated with the currently existing RSLAI has not been addressed.
The purposes of this study are to investigate 1) the impact of two widely used RSLAI datasets retrieved from the same source (i.e., AVHRR measurements) on a GCM’s seasonal climate simulation and 2) the mechanisms that lead to improvements in regional simulations from using RSLAI fields instead of prescribed values derived from ground surveys, which are generally extrapolated spatially according to vegetation-cover type. We believe these issues, which have not been addressed in previous studies, should be among the major issues in the forthcoming research on RSLAI data applications. The selected regions for regional analysis are East Asian, North American, and West African summer monsoon areas as well as boreal forests in North America, which have been identified as the regions sensitive to vegetation biophysical processes (VBPs; Xue et al. 2004).
In this paper, section 2 describes the two RSLAI datasets used in this study. General patterns of seasonal mean RSLAI values for boreal spring and summer are also briefly explained in this section. GCM description and experimental design are presented in section 3. Section 4 discusses the results of GCM simulations, highlighting the impacts of uncertainties in RSLAI datasets and the mechanisms leading to improvements in the seasonal climate simulation over several regions by using RSLAI. Finally, concluding remarks are given in section 5.
2. RSLAI
a. Data sources
As a part of the Global Energy and Water Cycle Experiment (GEWEX), the International Satellite Land Surface Climatology Project Initiative II (ISLSCP II) has recently released a global data collection in order to support investigations of the global carbon, water, and energy cycles (see online at http://islscp2.sesda.com). Among the vegetation variables of ISLSCP II datasets, two sources of monthly NDVI datasets retrieved from the AVHRR measurements are available: one is the Fourier Adjusted, Sensor and Solar Zenith Angle Corrected, Interpolated, Reconstructed NDVI (FASIR-NDVI) for 17 yr from 1982 to 1998 (Los et al. 2000) and the other is the Global Inventory Monitoring and Modeling Studies dataset (GIMMS-NDVI) for 22 yr from 1981 to 2002 (Pinzon et al. 2005). Based on each NDVI dataset, the LAI values were derived via two different algorithms by Los et al. (2000) and Myneni et al. (1997), which produced FASIR-LAI and GIMMS/Boston University LAI (GIMMSBU-LAI), respectively. The former, based on the work of Sellers et al. (1996), used empirical relationships for deriving the Fraction of Photosynthesis Active Radiation (FPAR) from NDVI and LAI from FPAR, and the latter used a three-dimensional radiative transfer model with a vegetation map (Myneni et al. 1997). The GIMMSBU dataset does not explicitly account for atmospheric constituents and cloud frequency while the FASIR dataset does (Los et al. 2000). Therefore, despite the same source measurement from AVHRR, the final products of LAI estimates had substantial differences, which were mainly caused by different retrieval algorithms and atmospheric correction methods. Both RSLAI datasets used in this study were obtained at monthly temporal resolution and 1° × 1° spatial resolution.
In addition, in order to preliminarily compare RSLAI values with field measurements, historical estimates of LAI from field sites archived at the Oak Ridge National Laboratory (ORNL), which is also in the ISLSCP II holding, are used. This is the only currently available globally ground-measured LAI dataset, which includes approximately 1000 published estimates of LAI from nearly 400 unique field sites, covering the period 1932–2000 (Scurlock et al. 2001). In the ORNL’s LAI compilation, the most commonly measured biome/land cover types are needleleaf forests in the Northern Hemisphere. Measurements for a few managed forest (plantation) and grass sites are also included. Despite the difficulty in direct comparison because of the scale dependency, this LAI dataset provides a rough insight into the consistency between RSLAI values and field measurements.
b. General features of RSLAI fields and their uncertainty
The seasonal mean of FASIR- and GIMMSBU-LAI datasets for boreal spring [March–May (MAM)] and boreal summer [June–August (JJA)] seasons from 1982 to 1998 are shown in Figs. 1 and 2, respectively. In MAM, spatial patterns and their magnitudes of the two RSLAI fields are generally similar to each other (Figs. 1a,b,d). In both datasets, the densest vegetation with LAI greater than 5 LAI unit (m2 m−2; hereafter we will omit LAI-unit in this paper) is found in tropical rain forests along the equator such as in the Amazon basin, central Africa, and Indonesia, and in needleleaf forest regions in the West Coast of the United States (Fig. 1a). Other dense vegetation areas are regions surrounding tropical rain forests in the South American and African continents, the southeastern United States, and southeastern Asia. In addition, there exist relatively low LAIs of about 1∼2 in boreal forests of the Northern Hemisphere. In spite of similar spatial patterns and magnitude of zonal mean distribution between FASIR- and GIMMSBU-LAI values (Fig. 1d), there are substantial differences at regional scales, especially in tropical and subtropical regions, which seem to be somewhat systematic according to vegetation type (Fig. 1b). For instance, in the tropical rain forests of the Amazon and central Africa, the FASIR-LAI values are larger than GIMMSBU-LAI by about 1∼2, with a maximum more than 2.5 near the lower Amazon River. On the other hand, for the subtropical regions with short vegetation types such as savanna and grass in the northwest of the Amazon basin, north and south of central Africa, and Southeastern Asia, the FASIR-LAI values are smaller than GIMMSBU-LAI by about 1∼2. The zonal mean distribution of their LAI’s standard deviation over 17 yr also shows large differences in tropical region and some midlatitude areas in both the Southern and Northern Hemisphere. Figure 1e indicates that FASIR-LAI exhibits larger interannual variations compared to those of the GIMMSBU dataset.
During the JJA period, which is the monsoon season in the Northern Hemisphere, the growth of vegetation is profound in boreal forests between 40° and 70°N in both datasets, so that the LAI values in this region become comparable to those in the Tropics (Figs. 2a,d). The systematic discrepancies that appeared in tropical and subtropical areas between two datasets during MAM are still apparent in JJA. The most substantial disagreements between FASIR and GIMMSBU datasets in JJA are found in boreal forests of the North America and Eurasia continents, where the FASIR-LAI values are larger than those of GIMMSBU-LAI by 1∼2 (Fig. 2b). These amounts correspond to about a 25% difference in terms of zonal averages (Fig. 2d). It is worth noting that the LAI values during JJA in the tropical evergreen rain forests of central Africa have been decreased dramatically compared to spring, which seems to be one of the features in the FASIR-LAI dataset (see Figs. 1a, 2a). The differences of standard deviation between two datasets become much more substantial than in MAM, especially in the Northern Hemisphere between 30° and 70°N (Fig. 2e). These discrepancies as well as differences in long-term trends are associated with differences in retrieval algorithms as well as atmospheric corrections (e.g., Gutman 1999; Slayback et al. 2003).
Direct comparison of satellite-retrieved vegetation properties with the field-measured values is difficult because of uncertainties caused by spatial- and temporal-scale dependencies between datasets (Myneni et al. 2002; Tian et al. 2003). For example, Tian et al. (2003) have shown that LAI retrieval errors are strongly dependent on the spatial resolution due to the effect of pixel heterogeneity, which is shown to increase as the resolution of data decreases. Nevertheless, a few efforts for direct comparison have been made to provide rough insight as to the quality of the retrieved values (e.g., Los et al. 2000; Buermann et al. 2002). In this context, we also compare FASIR- and GIMMSBU-LAI values with the historical measurements archived by ORNL for 14 available sites (Table 1). Because large errors are expected for evergreen trees over snow-covered regions (Los et al. 2000; Tian et al. 2004a, b), the comparison has been made only for the warm season from June to September. The LAI values that have been prescribed previously in the lookup table of the Simplified Simple Biosphere Model Version 1 (SSiBv1; Xue et al. 1991) are also provided in Table 1 as well as Figs. 1c and 2c (TABLE-LAI) as an alternative reference showing to what extent the RSLAI differs from previous prescriptions. Overall, both the RSLAI values are closer to ORNL’s field measurements than TABLE-LAI values except for a few sites in which the vegetation types are not comparable between vegetation types used for deriving RASLAI and in the field measurements due to the spatial heterogeneity (see Table 1). For example, ORNL’s LAI value at Mount Taylor in New Mexico is 7, which is much larger than values of RSLAI (0.8 and 1.6) and TABLE-LAI (0.3) because the classification of surface biome in ORNL’s archive is temperate mixed forest, which is likely to be a small fraction of the 1° cell denoted as shrubs/bare soil in the SSiBv1 classification. In addition, LAI measurements at Alachua County in Florida are made at the managed forest in ORNL’s archive so that its LAI values show large discrepancies from the grid cell average estimates of RSLAI with needleleaf evergreen trees. Averaged over 14 sites, the TABLE-LAI value (6.1) is larger than ORNL’s archive (3.8) by about 60%, while the FASIR- and GIMMSBU-LAI values (4.4 and 3.0, respectively) are close to ORNL’s field measurements, with the relative differences being about +16% and −20%, respectively. By and large, RSLAI shows better consistency with field measurements than TABLE-LAI.
As shown in Figs. 1 and 2, Table 1 confirms that FASIR-LAI values are consistently larger than those of GIMMSBU except for the site of Alachua. This is consistent with the analyses of Buermann et al. (2002), who attributed the larger LAI values in FASIR datasets to the larger FASIR-NDVI values induced by differences in the temporal compositing method and the inappropriate application of NDVI-LAI relational parameters of Los et al. (2000), which should be dependent on spatial scales. Based on their comparison of 1° aggregated GIMMSBU- and FASIR-LAI datasets with the measurements at 4 field campaign sites, Buermann et al. (2002) stressed the superiority in the GIMMSBU-LAI dataset. However, through the comparison with ORNL archives in this study, it is hard to determine which of these two RSLAI datasets is consistently superior to the other. For example, for the boreal forest areas north of 45°N in North America, the LAI values of GIMMSBU are closer to field measurement but the reverse is true for the south of 45°N and a few European sites. Because of the scale discrepancy between field measurements and RSLAI, we should not expect these two datasets to have the exact same LAI value over the same area. Nevertheless, the ORNL archives, as the first and only available globally measured LAI dataset, provides valuable and preliminary information regarding the RSLAI quality. More comprehensive methods and field measurements are necessary to further evaluate the satellite products.
3. Model descriptions and experimental design
To assess the impacts of RSLAI on seasonal climate, the National Centers for Environmental Prediction’s (NCEP) GCM (Kalnay et al. 1990; Kanamitsu et al. 2002) coupled with the SSiBv1 (Xue et al. 1991) is used. A parameter set for each of the vegetation types is derived from a variety of sources (Dorman and Sellers 1989; Willmott and Klink 1986; Xue et al. 1996), many of which are invariant with season. Seasonally varying monthly values of some vegetation properties, such as LAI, green leaf fraction, and surface roughness length, are prescribed for most vegetation types or calculated in the model for the crop type without considering interannual variation (Xue et al. 1996).
In this study, the NCEP GCM/SSiBv1 is used with 28 vertical levels and T62 (∼2°) horizontal resolution. The simulations consist of 3 experiments using 3 different LAI datasets, and the simulation period of each experiment is 20 months starting on 1 January and ending on 31 August of the next year for 3 El Niño (1982/83, 1987/88, 1997/98) and 3 La Niña (1984/85, 1988/89, 1995/96) cases, which include many climate anomaly events in some continents. The control experiment uses the LAI values from the lookup table that has usually been prescribed according to vegetation types of the SSiBv1 (Dorman and Sellers 1989; Xue et al. 1996; hereafter referred to as the TABLE experiment). The other 2 experiments used FASIR and GIMMSBU RSLAI datasets for the 6 specific cases after interpolation from 1° resolution to model grids, which are referred to as the FASIR and GIMMS experiments, respectively. Because the FASIR dataset also provides other surface properties, such as vegetation cover fraction (VCF), green leaf fraction, surface roughness length, and zero-plane displacement height, we also use them for the GCM’s surface boundary condition in both the FASIR and GIMMS experiments in order to maintain concomitant changes in surface parameters due to the changes of RSLAI amounts. Therefore, all these surface properties are different between TABLE and FASIR as well as between the TABLE and GIMMS experiments; however, LAI is the only different variable between the FASIR and GIMMS experiments. This strategy enables us to investigate not only the impact on seasonal climate simulations by using RSLAI but also the impact of RSLAI uncertainties between FASIR and GIMMSBU datasets. This paper focuses on the ensemble average of the second summer season (hereafter JJA2) of the 6-yr cases, which is consistent with the results from the first JJA but with clearer climatic signals. Results from the second spring season (MAM2) are also shown briefly because they are generally consistent with the JJA results with less significance. Please note that the averaged LAI values for these 6 yr are almost identical compared to the 17-yr average, with minor differences in standard deviation in some latitudes (Figs. 1e, 2e).
To evaluate the simulation results, the observed monthly surface air temperature (SAT) and precipitation data are used. The SAT used in this study was aggregated from the 0.5° analyses of daily maximum and minimum temperature from the Global Telecommunicating System (GTS) archived at the Climate Prediction Center (CPC) of NCEP. The GTS-based global land temperature analyses used the same algorithm developed for precipitation interpolation as in Xie et al. (1996). The observed precipitation was obtained from the CPC Merged and Analyzed Precipitation (CMAP) with 2.5° spatial resolution (Xie and Arkin 1997).
4. Simulation results
a. GCM response to RSLAI uncertainties
A number of studies have been published that investigated the GCM response to RSLAI uncertainty by assigning a possible LAI error range in the GCM simulation (e.g., Chase et al. 1996; Buermann et al. 2001; van den Hurk et al. 2003). In this study, we investigate this issue through the comparison between the FASIR and GIMMS experiments. We present results of the near-surface climate and energy budget analysis from these two experiments in this section. Figure 3 shows the spatial patterns of latent heat flux, SAT at canopy space, and precipitation differences between the FASIR and GIMMS experiments (GIMMS minus FASIR) during JJA2. The reduction of latent heat flux in JJA2 is evident in most parts of high latitudes of the Northern Hemisphere, including the boreal forests of North America, central Europe, and northeastern Asia (Fig. 3a). In contrast, in central Africa, Southeastern Asia, and the northern edge of Russia (tundra lands), the latent heat flux is increased. These variation patterns in latent heat flux correspond well to the LAI difference shown in Fig. 2b. The changes of sensible heat fluxes exhibit similar patterns to those of latent heat fluxes except of opposite sign (not shown). The changes in the latent and sensible heat fluxes indicate the importance of VBP in partitioning surface available energy into surface heat fluxes, especially for the forests (vegetation types 2, 3, 4, and 5; see Table 2 for definition) in the mid- and high latitudes of the Northern Hemisphere (Fig. 3a). In these regions, the percent change of forest LAI difference ranges about 12%∼25% and its impact on surface climate is relatively larger than for other vegetation types and other regions (Table 2). There are several regions in which the response of the surface heat flux does not correspond directly to LAI changes such as the central United States (crops), western Europe (crops–grass), and northeast of the Caspian Sea (grass–shrubs). These features seem to be caused by complex nonlinearity in VBP with respect to LAI, especially for short vegetation types as discussed in Bounoua et al. (2000) and Buermann et al. (2001).
Except for the changes of SAT at the northern edge of Russia, the spatial patterns of differences in SAT and precipitation (Figs. 3b,c) generally correspond well to changes in LAI and surface heat fluxes. That is, the smaller LAI values in the GIMMS experiment produce slightly warmer temperature north of 30°N because of the decrease of latent heat flux, and larger LAI values produce cooler temperature in central Africa and Southeastern Asia (Fig. 3b). However, the changes in surface heat fluxes between the GIMMS and FASIR experiments are not sufficient to produce statistically significant changes in the simulated precipitation and temperature. For instance, the regions with statistically significant differences in SAT are limited only for a few spots in North America, central Africa, and Eurasia. The differences in spatial distribution of precipitation are more sporadic (Fig. 3c), with even less significant features compared to the SAT.
Table 2 summarizes the impact of the LAI difference between the FASIR and GIMMSBU datasets on the simulated surface climate variables according to SSiBv1 vegetation-cover types in the Northern Hemisphere. On average for all vegetation types, the LAI values from GIMMSBU data are smaller than FASIR by more than about 11%; however, the impacts of these differences on most surface climate variables are less than 1.5% except for the relatively large percent change in surface temperature of about 4.4% between the FASIR and GIMMS experiments, which is mainly attributed to dwarf trees with ground cover (type 10). Generally, the surface albedo is lower over the denser vegetated surface. In this study, however, changes in surface albedo are neither consistent nor significant to the changes in LAI, which is similar to the results of Buermann et al. (2001). On average, an 11% reduction of LAI in the GIMMS experiment brings a 1% decrease in surface albedo, which is mainly attributed to the short vegetation-type surface mainly in the cold region. In particular, LAI differences between the two datasets have almost no impact on the simulated net radiation because the surface albedo is hardly changed by the differences of LAI value alone in the growing season unless there is substantial change in VCF (Buermann et al. 2001).
In general, the changes of surface heat fluxes due to the LAI differences in two RSLAI datasets (reflecting LAI uncertainties) do not cause significant differences in surface temperature and precipitation for most vegetation types. The significant impacts of the LAI differences are only limited in a few regions for surface temperature and even fewer regions in Eurasia for precipitation. By and large, despite the globalwide significant differences between FASIR- and GIMMSBU-LAI values (see Fig. 2b and Table 2), simulated surface climate patterns from the two experiments generally look consistent and the difference in magnitude is still limited over only a few areas. Similar conclusions were reported by Guillevic et al. (2002) for uncertainty in LAI interannual variability. We will further discuss this issue in section 5.
b. Impacts of the RSLAI on seasonal climate
Xue et al.’s studies (2004, 2006) indicate that the VBP has the most significant impact over the monsoon regions and some of the large continental areas. In this section, we focus on the impact of using RSLAI data on the GCM-simulated climate and on the underlying mechanisms in East Asia, North America, and West Africa, where the major monsoon systems exist during the boreal summer season.
1) General statistics and significance
The most substantial differences between the RSLAI and TABLE-LAI datasets during summer appear in the East Asia Monsoon (EAM) and boreal forests of North America where the RSLAI value is significantly smaller (up to more than 4) than that of TABLE-LAI (see Fig. 2c). The difference of LAI in the midlatitudes of North America (30°∼45°N) is also significant by exhibiting positive–negative–positive patterns from the western to the eastern United States (Fig. 2c). For the West African Monsoon (WAM) area, on the other hand, the LAI difference is relatively marginal, with values around 1 (Table 3).
Global and regional averages of the JJA2 surface temperature and precipitation between observation and model experiments are listed in Table 3. For global averages, three experiments produce very similar results. The TABLE experiment exhibits slightly cold and wet biases compared to the observation, with the differences being about −0.1°C and 0.4 mm day−1, respectively. Both the FASIR and GIMMS experiments consistently reduce these biases (Table 3). The improvements in SAT are more profound than those in precipitation in global mean and three regional means (Table 3). The percent improvement (∼1%) in the simulated global mean precipitation is small compared to the 14% relative error in the TABLE experiment.
For the statistical significance test of these differences, we applied Wigley and Santer’s (1990) statistical method, which employs the pool-permutation procedure (PPP). Although this method does not alter the results from the conventional t test (i.e., providing local significance level), it accounts for the effects of multiplicity and spatial autocorrelation and provides field significance level (p value). This method does not have the a priori assumption required for the conventional t test, that is, spatial and temporal independency between sample distributions, and is useful in significance testing where the number of samples is small. The applications of this method have been well demonstrated for significance on impacts of land cover change by Narisma and Pitman (2004) and of LAI change by Buermann et al. (2002). The statistical meaning of NT5 and NT10 in Table 4 is the percent fraction of locally successful pixels from the conventional gridpoint t test for the time means between RSLAI and TABLE experiments at the 5% and 10% levels, respectively. The corresponding p value of NT5 and NT10 indicates the probability of obtaining the local t value by chance. A p value close to zero would indicate a significant result. For example, if p ≤ 0.05, the difference is statistically significant at the 95% confidence level.
Table 4 shows that the improvements in SAT shown in Table 3 are statistically significant at the 5% and 10% levels over 16%–19% and 24%–28% of all global land points, respectively. In addition, p values of these statistics for SAT give significant results at the 1%–3% level, which is a clear indication that NT5 and NT10 are more reliable than one would expect to occur by chance. Table 4 also shows that the improvements of precipitation are statistically significant at the 5% and 10% levels over 6%–8% and 12%–14% of total land grid points, respectively, and their p values are also significant within the 10% significance level.
The fraction values of NT10 for SAT and precipitation in Table 4 are comparable with the results from Buermann et al.’s (2001) sensitivity experiments for LAI changes (see their Table 5). Buermann et al. (2001) did not provide the corresponding p values. Improvements of the simulated climate in the FASIR and GIMMS experiments are more apparent at regional scales (Table 4). The fraction values are much larger than the ones for latitudinal bands shown in Buermann et al. (2001). These regional significances are meaningful except for the SAT in West Africa and precipitation in East Asia, which will be discussed in the following sections.
Because the results from GIMMS are similar to FASIR, the analyses of the model results in the following sections are focused on the comparison between the TABLE and FASIR experiments.
2) Regional precipitation
The summer precipitation of East Asia and Africa is strongly controlled by the monsoon system, which is caused by the thermal contrast between land and ocean including nonlinear interactions among land, ocean, and atmosphere. Xue et al. (2004) investigated the role of VBP in EAM and WAM regions and had proposed that the perturbations induced by VBP are important for not only the intensity and evolution of the summer monsoon system but also for their spatial precipitation distribution and atmospheric circulation at the continental scales.
The precipitation differences between the observations and the TABLE experiment, as well as between the TABLE and FASIR experiments, are shown in Fig. 4. In spite of general agreement in the spatial pattern of global precipitation between the observation and the TABLE experiment (not shown), the simulated precipitation at regional scales exhibits substantial discrepancies from observation. In the East Asian region, the main deficiencies in the TABLE experiment are strong precipitation in the East Asian continent, weak precipitation in the west part of the northern Pacific (“plum,” rainfall band of East Asia, and near the Korean Peninsula and Japan), and too strong monsoon precipitation over the South China Sea and part of the Bay of Bengal (Fig. 4a). These disagreements were also found in the simulation of the 1987 summer monsoon case using the NCEP GCM (Xue et al. 2004). The FASIR experiment brings improvements of the simulated precipitation by reducing rainfall amounts from the South China Sea to northeastern China (Fig. 4b), which corresponds well to the LAI differences between the TABLE and FASIR experiments. It also reduces the precipitation biases over parts of the west Pacific Ocean and Indian Ocean. However, the FASIR experiment produces more precipitation in the regions of northeastern India, central China (west to 110°E), and Mongolia. By and large, the simulation over East Asia is improved (Table 3). However, despite clear spatial patterns of precipitation differences consistent with the LAI differences and more than a 17% fraction of successful grids from the local t test at the 10% level, changes in the simulated precipitation in this area do not provide strong confidence in their significance (Table 4). The spatial distributions of both the precipitation difference and its significance are quite consistent with those of vertically integrated moisture flux convergence (VIMC) and vertical motion in this region (not shown). In central China, both LAI and evaporation are decreased in the FASIR experiment (note that the change in evaporation is statistically significant; Table 4, Fig. 6a); the rainfall, however, is increased. It seems the local effect is diluted via complex processes of land–atmosphere–ocean interactions in the strong monsoon system.
In the North American region, the simulated precipitation in the TABLE experiment is overestimated in the northwestern, northeastern, and southern coast regions of the continent including the adjacent Atlantic Ocean. Meanwhile, moderate underestimation of the simulated precipitation by about 1∼2 mm day−1 appears in the midlatitudes of North America, which include the North American Monsoon (NAM) area and Great Plains (Fig. 4c). Because the intensity of the NAM system is relatively weak compared to the EAM, the response of the precipitation to the local LAI changes in the FASIR experiment in the North American region is more interpretable than in the East Asian region. The smaller LAI values of the FASIR dataset in boreal forests north of 45°N lead to the reduction of overestimated precipitation by the TABLE experiment (Fig. 4d). The slightly larger LAI values in the midlatitudinal midwestern United States play a role in increasing precipitation in this area. In both areas, the FASIR experiment improves the simulated precipitation with the significance at a 10% level. There are a few regions that show the deterioration of the simulated precipitation over land such as the eastern and southeastern United States. These changes, however, are also consistent with the LAI changes; that is, the larger (smaller) LAI values of the FASIR dataset in the eastern (southeastern) U.S. region lead to an increase (decrease) in precipitation. NT5 and NT10, for the simulated precipitation in the northwestern U.S. region, show the improvement at about 4% and 9% significance levels, respectively (Table 4).
Despite the relatively small LAI differences in the West African region between 5° and 17°N, the magnitude of the response to the simulated precipitation and its significance are comparable to the other regions (Figs. 4e,f; Table 4). In the TABLE experiment, the dry bias is shown along a latitudinal band between the equator and 10°N (which is elongated from the eastern Atlantic Ocean to 25°E) and wet bias in the mountainous regions in the northeast (Fig. 4e). The increased precipitation (by about 1–2 mm day−1) in the FASIR experiment reduces dry biases in West Africa but enhances wet biases in the eastern Sahel (Fig. 4f). Furthermore, the precipitation increase in the FASIR experiment is concentrated between 11° and 17°N [hereafter referred to as the northern part of the Sahel (NSHL)], which has smaller LAI values than its southern area between 5° and 11°N [hereafter referred to as the southern part of the Sahel (SSHL)]. It should be recognized that not only LAI but also other surface properties such as VCF, green fraction, and roughness length are changed in the FASIR experiment. We will give more detailed discussions on mechanisms in section 4.
3) Regional surface air temperature
The simulations in the FASIR and GIMMS experiments reduce cold biases of the SAT for the East Asia and northwest U.S. regions and warm biases for the West African region in the TABLE experiment (Table 3). Figure 5 shows the differences of SAT between observation and the TABLE experiments (TABLE − observation) and between the FASIR and TABLE experiments (FASIR − TABLE) during JJA2. In this comparison, the difference of elevation between station measurement and model topography is corrected with the lapse rate of standard atmosphere, that is, −6.5°C km−1 (Xue et al. 1996). The TABLE experiment produces cold biases in most parts of China and the Tibetan Plateau and warm biases in the other regions (Fig. 5a). The maximum of the cold biases is greater than 4°C over the Tibetan Plateau. Because of the smaller LAI values in the FASIR experiment, the SAT simulated by the FASIR experiment (Fig. 5b) is increased more than 0.5°C compared to the TABLE experiment in most parts of East Asia, which is statistically significant at a 90% confidence level. The increase of temperature in the FASIR experiment reduces the cold biases of the TABLE experiment in the Tibetan Plateau and eastern China (east of 110°E) while it aggravates warm biases in other regions such as India and northeastern China. Over eastern China, which is the major monsoon region, the fraction values for NT5 and NT10 are greater than 56% and 63%, respectively, with the p values close to zero (Table 4) giving a robust confidence on the significance of the improvements in the SAT simulation. The patterns of temperature differences and the area with a significance level higher than 90% correspond well to the larger LAI difference areas between the FASIR and TABLE experiments. For instance, the smaller LAI values in the FASIR experiment along the coast of eastern China and the southern flank of the Tibetan Plateau produce temperature increase patterns in these regions. Also, in the areas with small changes in LAI such as western China and Mongolia, the SAT difference was small, less than 0.5°C. Despite the deterioration of SAT in the FASIR experiment over India by enhanced warm bias, it is also consistent with the smaller LAI values in FASIR dataset. It is likely that the model deficiency and/or contamination of the remote sensing data play a role in the simulation problem. The FASIR-LAI values in this region are less than 1 in both MAM and JJA (Figs. 1a, 2a).
In North America, the TABLE experiment shows significant warm biases in most U.S. regions except for the western coastal area (Fig. 5c). The center of these warm biases is located over the Great Plains with the maximum value greater than 6°C. Changes in the simulated temperature by the FASIR experiment roughly correspond to the LAI differences (Fig. 5d). The smaller LAI values in the FASIR experiment in boreal forests north of 45°N lead to the increase of surface temperature, and larger LAI values in the eastern United States result in a decrease in temperature in this region. In addition, in the monsoon region over the Midwest United States, the simulated SAT is decreased because of the increase in evapotranspiration from the surface, which also corresponds to the larger LAI values in the FASIR experiment than in TABLE. Contrary to the relatively large and significant changes in the simulated precipitation over the Sahel, differences in the simulated SAT between the FASIR and TABLE experiments are marginal and less significant (Fig. 5f; Table 4). The SAT decrease in this region is consistent with the precipitation increases.
By and large, the magnitude of SAT differences between the TABLE and RSLAI experiments has the same sign comparable to that of biases of the TABLE experiment, which means the RSLAI experiments produce the SAT closer to the observation (Table 3). A statistical significance test gives confidence in these improvements in the SAT for East Asia and the northwest United States but a certain doubt for West Africa due to the smaller differences (Table 4). The changes in precipitation, SAT, and LAI in East Asia and North America correspond well to each other. As such, the RSLAI experiment reduces wet biases in East Asia and the northwest United States and dry biases in West Africa (Fig. 4).
4) Surface energy budget
This section discusses the possible mechanisms that produce the precipitation and temperature changes. We will mainly focus on the surface energy budget, on which the surface vegetation properties have the most impact. Table 5 shows differences in near-surface climate variables between the FASIR and TABLE experiments for a few typical vegetation types in East Asia (crops), northwest America (needleleaf evergreen forest), NSHL (shrubs with bare soil), and SSHL (savanna). These differences are obtained from spatial averages for the neighboring six points of each type. For all regions in Table 5 except for the semiarid NSHL region, the changes of LAI between the FASIR and TABLE experiment bring consistent changes in surface energy budget and near surface climate. That is, the reduction of LAI leads to a decrease in surface evapotranspiration (latent heat flux) and an increase in sensible heat flux, which results in a decrease in precipitation and an increase in surface temperature. In particular, because the changes of net available radiative energy are only about 2∼6 W m−2 in East Asia and North America, improvements of the simulated near-surface climate in the FASIR experiment in these regions are mainly induced by the VBP that play an important role in partitioning surface available energy into sensible and latent heat fluxes. Figure 6 shows the spatial distribution of surface latent heat flux difference between the FASIR and TABLE experiments. The decrease in JJA2 latent heat flux over East Asia (Fig. 6a) and the boreal forests of North America (Fig. 6b) is evident, which corresponds well to the smaller LAI values in the FASIR experiments compared to the TABLE experiments (Fig. 2c). The latent heat flux over the Midwest and eastern U.S. regions is increased in the FASIR experiment because of the larger LAI values. Furthermore, these spatial patterns of LAI and latent heat fluxes are consistent with those of precipitation (Fig. 4d), which clearly indicates the role of VBP in modulating the precipitation.
For the Sahel region in Africa, it is interesting to note that significant impacts of surface vegetation on the simulated latent heat flux (Fig. 6c) and precipitation (Fig. 4f) appear in NSHL, where the LAI difference between the two experiments is small, rather than in SSHL. As presented earlier, in addition to LAI, some other vegetation properties are also changed in the FASIR experiment. The difference of VCF between the FASIR and TABLE experiments is shown in Fig. 7a. In most regions of Africa except for the tropical rain forest, desert, and east coast areas, the VCF in the FASIR experiments is larger than that of the TABLE experiment. In the NSHL region, the increased VCF (and LAI with small magnitude) in the FASIR experiment (compared to the TABLE) decreases surface albedo and consequently produces higher net radiation of about 12 W m−2, which contributes to high latent heat flux. Meanwhile, in SSHL, changes in surface albedo are very small (about 1%) because of the compensation between the reduction in LAI and increase in VCF (Fig. 7a) for the FASIR experiment, which makes small changes in surface net radiation, about 5 W m−2 (Fig. 7b; Table 5). The reduction of latent heat flux in parts of SSHL (Fig. 7c) is consistent with the lower LAI (Table 5).
The direction of changes in latent and sensible heat fluxes over both SSHL and NSHL regions is consistent with the changes in precipitation and surface temperature, that is, increased (decreased) precipitation and cooler (warmer) temperature in the NSHL (SSHL) region (Table 5) compared to TABLE. Despite the comparable magnitude of the changes in latent heat flux of about 16 W m−2 with opposite signs, the impacts on surface temperature and precipitation are more discernable in NSHL than in SSHL, which is associated with the changes in atmospheric circulation.
Clark et al. (2001) explored the regional dependence of VBP impact on climate and found that the northern Sahel (i.e., NSHL in this study) was one of two areas in northern Africa where VBP had the most significant impacts on regional precipitation. The Sahel region study of Xue (1997) indicated that less vegetation and high albedo there would induce lower net radiation and less evapotranspiration; less moisture was transferred to the atmosphere through the boundary layer. This resulted in less convection and lower atmospheric latent heating rates. The reduced total diabatic heating rate in the atmosphere was associated with relative subsidence, less moisture flux convergence, and lower rainfall. These changes further reduced the evapotranspiration, producing a positive feedback. In this case, the higher VCF in the FASIR experiment also produces a similar positive feedback with more precipitation. The net radiation is higher (Fig. 7b), and the VIMC, which is the main source for the Sahelian precipitation, in the NSHL region shows a clearly consistent pattern with the increase in precipitation (Figs. 4f and 7c) in the FASIR experiment. The magnitude of the VIMC difference is more than 1.2 mm day−1, which is comparable to changes of precipitation. The relative rising and strong westerly monsoon flow are produced in the NSHL region (around 10°N) throughout the depth of the troposphere (Fig. 8). These features are similar to the desertification experiment of Xue (1997), but with the opposite direction. Therefore, the land surface/atmosphere interactive feedback processes cause the increase of precipitation in NSHL. Meanwhile, affected by the relative rising motion over NSHL, the relative vertical motion over the SSHL region transfers from rising to subsidence (Fig. 8b). The overall VIMC has little change, and precipitation change is also small in this region (Table 5).
5) Regional SAT for the boreal spring (MAM2)
Since the differences in LAI values between FASIR and TABLE datasets are substantial in East Asia and North America (Fig. 1), we present the SAT simulation results during the second boreal spring (MAM2) in these two regions (Fig. 9). The results of the precipitation are not shown because the MAM is not the major rainy season there, so its changes in this season are limited. Overall, the smaller values in the FASIR-LAI dataset contribute to an increase in SAT both in East Asia and the boreal forests in North America, which is consistent with the results from JJA2 but with relatively weak significance due to the smaller differences compared to those of JJA2. An increase in SAT results in improvements of the simulation results for most areas in China except for southeastern China, Indochina, and west to Mongolia (Fig. 9b), as well as the western United States (Fig. 9d), while it deteriorates results for southeastern China and northeastern India. In spite of weak magnitude and its significance, the consistent response of the simulated SAT to the LAI difference appears in the eastern United States (see Figs. 1c, 9d). It is also worth noting that the increases in cold bias near Mongolia and northeastern China (Fig. 9b) are associated with higher VCF in the FASIR dataset in this region. By and large, model results for the boreal spring season are consistent with those of summer but with smaller impacts.
5. Discussion and summary
Since the launch of AVHRR on board National Oceanic and Atmospheric Administration satellite series, there have been numerous studies that retrieve NDVI and LAI from AVHRR measurements and evaluate their quality by recognizing signals and noise. This study assesses the impact of two different RSLAI datasets on a GCM’s seasonal climate simulation as well as the mechanism that leads to the improvement in simulations over several regions.
Based on the analyses of these two RSLAI datasets for 17 yr from 1982 to 1998, their spatial distribution patterns and characteristics are discussed. The FASIR-LAI values are generally larger than GIMMSBU-LAI, which is consistent with the previous results by Buermann et al. (2002). In addition, notable disagreements exist between these two datasets in terms of the magnitude of long-term trend and interannual variability. Retrieval algorithms and atmospheric correction methods have been identified as major reasons for uncertainties in the satellite-retrieved estimate (Gutman 1999). These RSLAIs are compared with the ORNL field measurements archive for a few selected sites. Although both RSLAI datasets are closer to ORNL’s field measurements than TABLE-LAI, the comparison does not allow the identification of which RSLAI dataset is superior to the other. Further investigation by comparison between satellite products and ground measurements is necessary, with careful consideration of the scale dependency issue.
We performed a 20-month-long control run with the prescribed LAI values from the SSiBv1 lookup table and two sensitivity runs with the GIMMSBU- and FASIR-LAI values for 6 ENSO years. Between sensitivity runs, other surface properties including VCF, green fraction, and roughness length are derived from FASIR datasets in order to isolate the LAI effect. Regional differences in the simulated near-surface climates between the GIMMS and FASIR experiments indicate systematic and consistent responses according to the differences in LAI for tall forest types, that is, warming and decrease of precipitation due to the smaller LAI values in the GIMMS experiment compared to the FASIR experiment. However, despite the discrepancies between FASIR- and GIMMSBU-LAI datasets, the differences generally do not cause significant changes in the simulated near-surface climate. This conclusion has important implications for the end users who are going to use RSLAI estimates as a surface boundary forcing of their climate models.
The GCM experiment using the RSLAI and other satellite-derived land surface products shows substantial improvements of the near-surface climate in the East Asia and West Africa summer monsoon areas and boreal forests of North America compared to the control experiment, in which vegetation parameter values mainly based on ground surveys are used. In this study, we focus on the mechanism of RSLAI bringing improvements in seasonal climate simulations for the regions of East Asia, North America, and West Africa. The reduced LAI signal in the East Asian summer monsoon areas and boreal forests of North America leads to increases in surface temperature and decreases in precipitation, which consistently improve the simulated climate compared to the control runs. In these regions, changes in net radiative energy at the surface in the FASIR experiment are only 2∼6 W m−2, which correspond to 1%∼3% changes compared to the TABLE experiment, while the decrease in latent heat flux due to reduced LAI values is about 10∼20 W m−2, which is more than 15%. Therefore, improvements of the simulated climate in the FASIR experiment are highly associated with the VBP that plays an important role in partitioning surface available radiation energy into sensible and latent heat fluxes. The changes of atmospheric circulation induced by surface fluxes play a major role in modulating near-surface climate in the Sahel region, especially for the semiarid NSHL, where the LAI is not different between the FASIR and TABLE experiments, but VCF change–induced lower albedo leads to high net radiation, relative rising vertical motion, and moisture flux convergence in the FASIR experiment. The spatial pattern of near-surface climate changes exhibits a consistent pattern of VIMC. In NSHL, the increase in precipitation due to the change in atmospheric circulation results in an increase of latent heat flux and produces a positive feedback.
A comprehensive statistic method (Wigley and Santer 1990) has been introduced to evaluate the model results. A significance test gives confidence on the improvements in the simulated climate for the FASIR experiment compared to the TABLE experiment, not only at the global but also at the regional scale. Weak significance for the precipitation in East Asia and SAT in West Africa is identified and may be caused by the complex land–atmosphere–ocean interaction of the strong monsoon system and changes in local circulation, respectively. It is interesting to note that although the differences between GIMMSBU- and FASIR-LAIs and between FASIR- and TABLE-LAIs both cover large spatial areas (Figs. 2b,c), the former difference does not yield a significant difference in simulated precipitation and surface air temperature, while the latter does. As emphasized earlier, in the first set of experiments (i.e., GIMMSBU-LAI and FASIR-LAI comparison), LAI is the only variable being tested. In the second set of experiments, not only LAI but also VCF and other variables are tested. Another possible cause is the magnitude of the LAI differences shown in Figs. 2b,c. In Fig. 2b, most differences are less than 2, with the average around 1 over the boreal forests (Table 2), while in Fig. 2c, the areas with significant differences in simulated surface climate have LAI differences larger than 2. Table 5 shows that the differences for East Asia and North America are around 4.
Based on results of this study, when other surface properties are the same, it may be necessary to have the LAI difference larger than around 1–2 to obtain significant climate impact in this GCM. Another study (Guillevic et al. 2002), which used the National Aeronautics and Space Administration (NASA) Seasonal-to-Interannual Prediction Project atmospheric GCM and an earlier version of ISLSCP II RSLAI from 1982 to 1990, investigated the sensitivity of the simulated climate to the interannual variability of LAI. The results showed that although variations in vegetation properties influenced transpiration and interception loss in the GCM runs, their impact on large-scale regional climate is much weaker because the impact is drowned out by atmospheric variability. After carefully checking the magnitude of LAI fluctuation at a regional scale in their runs, we found that the LAI variation range in their study is less than 1 during 9 yr from 1982 to 1990. Therefore, the LAI impact in their GCM on the climate variability is consistent with our results.
Since the launch of the Earth Observing System (EOS) satellite Terra, the use of Moderate Resolution Imaging Spectroradiometer’s (MODIS) vegetation products has been rapidly increased during the last several years. Although the MODIS dataset has been considered to show better quality than AVHRR, which was explicitly designed for land application (Tian et al. 2004b), because of the narrower spectral bands among other things, the MODIS LAI was not used in this study because of the shorter period. Meanwhile, Gallo et al. (2004) reported that the MODIS and AVHRR NDVI values from 16-day composites were quite similar to each other when sampled over similar time intervals, spatial areas, and land cover types. More simulations with different models and RSLAI estimates are necessary to further evaluate the climate sensitivity to LAI uncertainties and variability.
Acknowledgments
Funding was provided by NASA Grants NAG5-9713 and NNG04GB05G and by NSF Grants ATM-0097260 and ATM-0353606. The model runs were carried out on the NCAR supercomputers. The authors gratefully acknowledge Drs. Molly Brown and J. E. Pinzon of NASA GSFC for providing the GIMMSBU-LAI dataset and Dr. Pingping Xie of NCEP for surface temperature data. We also thank Dr. Gemma. T. Narisma of University of Wisconsin—Madison for providing the code of statistical significance test.
REFERENCES
Betts, A. K., J. H. Ball, A. C. M. Beljaars, M. J. Miller, and P. A. Viterbo, 1996: The land surface-atmosphere interaction: A review based on observational and global modeling perspectives. J. Geophys. Res., 101 , 7209–7225.
Bounoua, L., G. J. Collatz, S. O. Los, P. J. Sellers, D. A. Dazlich, C. J. Tucker, and D. A. Randall, 2000: Sensitivity of climate to changes in NDVI. J. Climate, 13 , 2277–2292.
Buermann, W., J. Dong, X. Zeng, R. B. Myneni, and R. E. Dickinson, 2001: Evaluation of the utility of satellite-based leaf area index data for climate simulation. J. Climate, 14 , 3536–3550.
Buermann, W., Y. Wang, J. Dong, L. Zhou, X. Zheng, R. E. Dickinson, C. S. Potter, and R. B. Myneni, 2002: Analysis of a multiyear global vegetation leaf area index data set. J. Geophys. Res., 107 .4646, doi:10.1029/2001JD000975.
Chase, T. N., R. A. Pielke, T. G. F. Kittel, R. Nemai, and S. W. Running, 1996: Sensitivity of a general circulation model to global changes in leaf area index. J. Geophys. Res., 101 , 7393–7408.
Clark, D. B., Y. Xue, R. J. Harding, and P. J. Valdes, 2001: Modeling the impact of land surface degradation on the climate of tropical North Africa. J. Climate, 14 , 1809–1822.
Dickinson, R. E., and A. Henderson-Sellers, 1988: Modelling tropical deforestation: A study of GCM land-surface parameterizations. Quart. J. Roy. Meteor. Soc., 114 , 439–462.
Dorman, J. L., and P. J. Sellers, 1989: A global climatology of albedo, roughness length and stomatal resistance for atmospheric general circulation model (SiB). J. Appl. Meteor., 28 , 833–855.
Gallo, K., L. Ji, B. Reed, J. Dwyer, and J. Eidenshink, 2004: Comparison of MODIS and AVHRR 16-day normalized difference vegetation index composite data. Geophys. Res. Lett., 31 .L07502, doi:10.1029/2003GL019385.
Guillevic, P., R. D. Koster, M. J. Suarez, L. Bounoua, G. J. Collatz, S. O. Los, and S. P. P. Mahanama, 2002: Influence of the interannual variability of vegetation on the surface energy balance—A global sensitivity study. J. Hydrometeor., 3 , 617–629.
Gutman, G. G., 1999: On the use of long-term global data of land reflectances and vegetation indices derived from the advanced very high resolution radiometer. J. Geophys. Res., 104 , 6241–6256.
Kalnay, E., M. Kanamitsu, and W. E. Barker, 1990: Global numerical weather prediction at the National Meteorological Center. Bull. Amer. Meteor. Soc., 71 , 1410–1428.
Kanamitsu, M., and Coauthors, 2002: NCEP dynamical seasonal forecast system 2000. Bull. Amer. Meteor. Soc., 83 , 1019–1037.
Lawrence, D. M., and J. M. Slingo, 2004a: An annual cycle of vegetation in a GCM. Part I: Implementation and impact on evaporation. Climate Dyn., 22 , 87–105.
Lawrence, D. M., and J. M. Slingo, 2004b: An annual cycle of vegetation in a GCM. Part II: Global impacts on climate and hydrology. Climate Dyn., 22 , 107–122.
Los, S. O., and Coauthors, 2000: A global 9-yr biophysical land surface dataset from NOAA AVHRR data. J. Hydrometeor., 1 , 183–199.
Lotsch, A., M. A. Friedl, B. T. Anderson, and C. J. Tucker, 2003: Coupled vegetation-precipitation variability observed from satellite and climate records. Geophys. Res. Lett., 30 .1774, doi:10.1029/2003GL017506.
Myneni, R. B., S. O. Los, and C. J. Tucker, 1996: Satellite-based identification of linked vegetation index and sea surface temperature anomaly areas from 1982–1990 for Africa, Australia and South America. Geophys. Res. Lett., 23 , 729–732.
Myneni, R. B., R. R. Nemani, and S. W. Running, 1997: Estimation of global leaf area index and absorbed PAR using radiative transfer models. IEEE Trans. Geosci. Remote Sens., 35 , 1380–1393.
Myneni, R. B., C. J. Tucker, G. Asrar, and C. D. Keeling, 1998: Interannual variations in satellite-sensed vegetation index data from 1981 to 1991. J. Geophys. Res., 103 , 6145–6160.
Myneni, R. B., and Coauthors, 2002: Global products of vegetation leaf area and fraction absorbed PAR from year one of MODIS data. Remote Sens. Environ., 83 , 214–231.
Narisma, G. T., and A. J. Pitman, 2004: The effect of including biospheric responses to CO2 on the impact of land-cover change over Australia. Earth Interactions, 8 .[Available online at http://EarthInteractions.org.].
Oleson, K. W., and G. B. Bonan, 2000: The effects of remotely sensed plant functional type and leaf area index in simulations of boreal forest surface fluxes by the NCAR land surface model. J. Hydrometeor., 1 , 431–446.
Pielke, R. A., R. Avissar, M. Raupach, A. J. Dolman, Y. Xeng, and S. Denning, 1998: Interactions between the atmosphere and terrestrial ecosystems: Influence on weather and climate. Global Change Biol., 4 , 461–475.
Pinzon, J. E., M. E. Brown, and C. J. Tucker, 2005: EMD correction of orbital drift artifacts in satellite data stream. The Hilbert-Huang Transform and Its Applications, N. Huang and S. S. P. Shen, Eds., World Scientific, 167–186.
Scurlock, J. M. O., G. P. Asner, and S. T. Gower, 2001: Worldwide historical estimates of leaf area index, 1932–2000. Oak Ridge National Laboratory Rep. ORNL/TM-2001/268, Oak Ridge, TN, 23 pp.
Sellers, P. J., F. G. Hall, G. Asrar, D. E. Strebel, and R. E. Murphy, 1988: The First ISLSCP Field Experiment (FIFE). Bull. Amer. Meteor. Soc., 69 , 22–27.
Sellers, P. J., S. O. Los, C. J. Tucker, C. O. Justice, D. A. Dazlich, G. J. Collatz, and D. A. Randall, 1996: A revised land surface parameterization (SIB2) for atmospheric GCMs. Part II: The generation of global fields of terrestrial biophysical parameters from satellite data. J. Climate, 9 , 706–737.
Slayback, D. A., J. Pinzon, S. O. Los, and C. J. Tucker, 2003: Northern hemisphere photosynthesis trends 1982–99. Global Change Biol., 9 , 1–15.
Tian, Y., Y. Wang, Y. Zhang, Y. Knyazikhin, J. Bogaert, and R. B. Myneni, 2003: Radiative transfer based scaling of LAI retrievals from reflectance data of different resolutions. Remote Sens. Environ., 84 , 143–159.
Tian, Y., R. E. Dickinson, L. Zhou, R. B. Myneni, M. Friedl, C. B. Chaaf, M. Carroll, and F. Gao, 2004a: Land boundary conditions from MODIS data and consequences for the albedo of a climate model. Geophys. Res. Lett., 31 .L05504, doi:10.1029/2003GL019104.
Tian, Y., R. E. Dickinson, L. Zhou, and M. Shaikh, 2004b: Impact of new land boundary conditions from Moderate Resolution Imaging Spectroradiometer (MODIS) data on the climatology of land surface variables. J. Geophys. Res., 109 .D20115, doi:10.1029/2003JD004499.
van den Hurk, B. J. J. M., P. Viterbo, and S. O. Los, 2003: Impact of leaf area index seasonality on the annual land surface evaporation in a general circulation model. J. Geophys. Res., 108 .4191, doi:10.1029/2002JD002846.
Wigley, T. M. L., and B. D. Santer, 1990: Statistical comparison of spatial fields in model validation, perturbation, and predictability experiments. J. Geophys. Res., 95 , 851–865.
Willmott, C. J., and K. Klink, 1986: A representation of the terrestrial biosphere for use in global climate studies. Proc. ISLSCP Conf., Rome, Italy, European Space Agency, 109–112.
Xie, P., and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 2539–2558.
Xie, P., B. Rudolf, U. Schneider, and P. A. Arkin, 1996: Gauge-based monthly analyses of global land precipitation from 1971–1994. J. Geophys. Res., 101 , 19023–19034.
Xue, Y., 1997: Biosphere feedback on regional climate in tropical north Africa. Quart. J. Roy. Meteor. Soc., 123B , 1483–1515.
Xue, Y., and J. Shukla, 1993: The influence of land surface properties on Sahel climate. Part I: Desertification. J. Climate, 6 , 2232–2245.
Xue, Y., P. J. Sellers, J. L. Kinter III, and J. Shukla, 1991: A simplified biosphere model for global climate studies. J. Climate, 4 , 345–364.
Xue, Y., M. J. Fenessy, and P. J. Sellers, 1996: Impact of vegetation properties on U.S. summer weather prediction. J. Geophys. Res., 101 , 7419–7430.
Xue, Y., H-M. H. Juang, W-P. Li, S. Prince, R. DeFries, Y. Jiao, and R. Vasic, 2004: Role of land surface processes in monsoon development: East Asia and West Africa. J. Geophys. Res., 109 .D03105, doi:10.1029/2003JD003556.
Xue, Y., F. De Sales, W. Li, C. R. Mechoso, C. Nobre, and H-M. H. Juang, 2006: Role of land surface processes in South American monsoon development. J. Climate, 19 , 741–762.
(a) Seasonal mean of FASIR-LAI for MAM and differences of LAI between (b) GIMMSBU and FASIR, and (c) FASIR and TABLE datasets. (d), (e) Zonal averages of LAI values and their std dev, respectively. Dashed lines in (d) and (e) indicate the averages for the six selected years for the GCM experiments.
Citation: Journal of Climate 20, 6; 10.1175/JCLI4054.1
Differences of (a) latent heat flux, (b) SAT, and (c) precipitation between GIMMS and FASIR experiments (GIMMS − FASIR) for JJA2. Dashed contour lines in (b) and (c) indicate the 90% confidence level from the Student’s t test.
Citation: Journal of Climate 20, 6; 10.1175/JCLI4054.1
Differences of the JJA2 precipitation between (a), (c), (e) Xie–Arkin observation and TABLE experiment and (b), (d), (f) FASIR and TABLE experiments for (a), (b) East Asia, (c), (d) North America, and (e), (f) Africa regions. Dashed lines indicate significance from the Student’s t test over the 90% confidence level.
Citation: Journal of Climate 20, 6; 10.1175/JCLI4054.1
Difference of surface latent heat flux between FASIR and TABLE experiments (FASIR−TABLE) during JJA2 for (a) East Asia, (b) North America, and (c) Africa. Dashed lines indicate significance from the Student’s t test over the 90% confidence level.
Citation: Journal of Climate 20, 6; 10.1175/JCLI4054.1
Difference of (a) vegetation cover fraction, (b) net radiation, and (c) vertically integrated moisture flux convergence between FASIR and TABLE experiments (FASIR−TABLE) during JJA2.
Citation: Journal of Climate 20, 6; 10.1175/JCLI4054.1
Differences of (a) zonal and (b) vertical winds between FASIR and TABLE experiments (FASIR − TABLE) averaged over 15°W–30°E. Negative values in (b) indicate rising motion.
Citation: Journal of Climate 20, 6; 10.1175/JCLI4054.1
Comparison of satellite LAI values with field measurement archives at ORNL.
JJA2 differences in near-surface climate variables between the GIMMS and FASIR experiments (GIMMS−FASIR) according to SSiBv1 vegetation types over the Northern Hemisphere. Values in parentheses are the number of grids used in the calculation of averages Note: SH ≡ sensible heat, and LH latent heat.
Comparison of the seasonal averages (JJA2) for the temperature and precipitation between observation and GCM experiments. Values in parentheses are % changes of LAI and precipitation.
Test statistic fraction values (%) and corresponding field significance levels (p values; in parentheses) for JJA2.
JJA2 differences of near-surface variables between FASIR and TABLE experiments (FASIR − TABLE). NSHL and SSHL indicate northern and southern parts of the Sahel region, respectively. Vegetation types 12, 4, 9, and 6 are crops, needleleaf evergreen trees, shrubs with bare soil, and savanna, respectively. Note: SWR ≡ shortwave radiation, and LWR ≡ longwave radiation.
