HTML
--> --> -->The impact of SM on ET is considered to be the first segment of the connection from land surface states to the atmosphere (Dirmeyer, 2011). First of all, SM is an important source of the atmospheric water. Through ET, a major component of the continental water cycle, SM returns approximately 60% of the total precipitation that falls on land back to the atmosphere (Oki and Kanae, 2006). Secondly, SM also plays an important role in the land surface energy balance, since ET consumes more than half of the total solar energy absorbed by the land surface (Trenberth et al., 2009). Thus, SM can evidently influence the thermal condition of the land surface through ET, consequently affecting the atmosphere (Zhang and Zuo, 2011; Gao et al., 2014). So, the relationship between SM and ET is a key process in the land-atmosphere interface and is also the focus of our study.
In addition, the key regions or "hot spots" of land-atmosphere coupling are unevenly distributed. By analyzing the mean states of multi-model simulations in summer (Koster et al., 2004), those "hot spots" mainly appear in transitional climates between dry and wet zones, such as the central Great Plains of North America, the Sahel, and India. For each of the individual models, the land-atmosphere coupling intensity and spatial pattern are distinctly different, indicating that there are also numerous uncertainties in the SM feedback. More studies have further confirmed that the SM effects on climate change are mainly detected in the dry-wet transition zones (Wei et al., 2008; Dirmeyer, 2011; Hua et al., 2013). The reason is that SM and ET are not simply linearly related. For example, SM does not constrain ET if soil water content fully meets the land surface evaporation and demand of plants (i.e., an "extreme wet condition"). On the other hand, if the soil is too dry (i.e., an "extreme dry condition"), then ET stops completely (Seneviratne et al., 2010; Bellucci et al., 2015). SM does not control ET under either extreme wet or dry conditions. For this reason, most of the inferred effects of SM on climate systems (strong land-atmosphere coupling) are induced by its constraints on ET under the transitional conditions between dry and wet zones.
Compared to SM-precipitation coupling, SM-temperature coupling is more distinct (Zhang et al., 2011; Zhang and Zuo, 2011; Wu and Zhang, 2013). For instance, lower SM usually results in less ET and weaker evaporative cooling, which leads to higher ST (Mei and Wang, 2012; Zittis et al., 2013). Meanwhile, another interpretation of this negative correlation exists: high ST acts as a primary forcing factor, thereby increasing ET and drying the soil (Nicholls, 2004; Cai and Cowan, 2008). This hypothesis suggests that high (low) ST causes dry (wet) soil. In summary, SM-ST relationships can be used to classify the conditions for ET processes as water-limited (SM constrains ET) or energy-limited (ET is controlled by atmospheric backgrounds, and no land-atmosphere coupling exists under such a condition) (Koster et al., 2009; Seneviratne et al., 2010; Yin et al., 2014a), which is also closely related to dry-wet climate conditions.
Previous studies have emphasized the spatial pattern of land-atmosphere coupling and identified "hot spots" with strong SM-climate feedback. However, less attention has been paid to seasonal changes of land-atmosphere coupling (Wei and Dirmeyer, 2012; Wei et al., 2012), especially in China, where complex spatial and seasonal climate variations exist. In this study, we attempt to identify hot spots of land-atmosphere coupling and explore the seasonal changes of such coupling over China.
The remaining parts of this work are organized as follows: Section 2 introduces the datasets and methods; section 3 presents the relationships among SM, ST and ET over the study area; sections 4 and 5 discuss and summarize the findings, respectively.
2.1. Study area
China has a vast territory and is a typical monsoon region, exhibiting varying climatic conditions (Guo et al., 2003; He et al., 2007). Figure 1 shows eastern China (east of 95°E) and its major geographical regions. In general, the elevation decreases from west to east, and a typical East Asian monsoon climate system prevails across this area (Jiang et al., 2015). Additionally, previous studies (e.g., Ge and Feng, 2009; Hu et al., 2015) have pointed out that more than 96% of China's population lives in this area. Therefore, the local land-atmosphere coupling influences human life and needs to be explored extensively.
Figure1. Topography (units: m) and major geographical regions in eastern China. The geographical regions are based on Zhao (1995). The elevation data are from the Global 30 Arc-Second Elevation (GTOPO30_10min) dataset: https://lta.cr.usgs.gov/GTOPO30.2
2.2. Reanalysis dataset
Because of a lack of observations, studies related to SM on larger spatial and longer time scales are still very limited. However, the recent development and availability of various satellite-derived and reanalysis SM datasets make it possible to understand relevant scientific questions more comprehensively. The uncertainties in satellite-derived SM data and the weaknesses of current quality controls (Yin et al., 2014b) are likely to introduce some uncertainties into our results. According to Zuo and Zhang (2007, 2009), the spring SM of the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA-40 dataset (1948-2002) can reproduce the temporal and spatial features of observed SM in eastern China reasonably well. Moreover, (Zhang et al., 2008c) compared the ERA product with other multi-source datasets (e.g., observations, NCEP/NCAR, GSWP2 and CLM) and found that ERA-40 can better capture the interannual and interdecadal variabilities of observed SM.Recently, the ECMWF released a new product: ERA-Interim (Balsamo et al., 2015). In this dataset, the Tiled ECMWF Scheme used for the surface exchanges within the Land Surface Model (van den Hurk et al., 2000) is the same as that employed by ERA-40. The assimilation method, however, was upgraded from the 3D variational method of ERA-40 to a 4D variational approach in ERA-Interim (Dee et al., 2011). To further reduce the biases between the ERA-simulated variables and observations, the observed relative humidity and temperature are continuously used to adjust the SM (Douville et al., 2000; Mahfouf et al., 2000). By comparing several sets of reanalysis data (e.g., ERA-Interim, MERRA, JRA, CFSR, and NCEP), (Liu et al., 2014) noted that ERA-Interim performs best in reproducing the spatial and temporal features of the SM over eastern China. This dataset also captures the major characteristics of precipitation and evaporation, which are the two main factors affecting SM. Overall, ERA-Interim is preferable to other datasets in representing the spatial and temporal characteristics of the actual SM. Thus, this SM dataset is selected for this study, which is available at http://www.ecmwf.int/en/ research/climate-reanalysis/era-interim/land. The soil consists of four layers of thickness, i.e., 7, 21, 72 and 189 cm from top to bottom, with a horizontal resolution of 1°× 1°. Considering the greater accuracy of the near-surface ERA SM and its stronger interaction with ET (Zhang et al., 2008c; Zuo and Zhang, 2009), we select the first layer of ERA SM for further analyses. Besides, the surface latent heat flux and top layer soil temperature are chosen to represent ET and ST in our study. The study period is from March 1979 to February 2014.
2
2.3. Model simulation
To enhance confidence in our results, a group of numerical experiments simulated by version 4.0 of NCAR's Community Land Model (CLM4.0, http://www.cesm.ucar.edu/models/cesm1.1/clm/) are used. This model is a well-developed land surface model and has been widely used for studies related to land surface processes (Chen et al., 2010; Xiong et al., 2011; Zhu et al., 2013). Recent studies (e.g., Li et al., 2011; Lai et al., 2014) have stated that CLM4.0 is capable of capturing the spatiotemporal characteristics of SM from in situ observations over China. The land cover type, soil color, texture, sand ratio and other information come from the land characteristic parameter data of the model, and the soil column is vertically represented by 15 layers (0.007, 0.028, 0.062, 0.119, 0.212, 0.366, 0.620, 1.038, 1.728, 2.865, 4.739, 7.830, 12.925, 21.327, and 35.178 m). The global land surface conditions (e.g., SM, ST, ET and heat fluxes) in the last five decades are produced by offline simulations, which are based on the global 1°× 1° and 3-h atmospheric forcing dataset including near surface meteorological elements (i.e., 2-m temperature, wind speed, specific humidity, precipitation, surface pressure, downward longwave and shortwave radiation) from 1948 to 2010 developed at Princeton University (Sheffield et al., 2006). The forcing dataset is available at http://hydrology.princeton.edu/data.pgf.php. The third layer SM (6.2 cm——close to the selected layer of ERA SM), first layer soil temperature and ET from March 1979 to February 2010 are selected for our study.2
2.4. Methods
Our analyses are mainly based on Pearson correlation analysis. The monthly correlation coefficient between SM and ET (SM-ET relationship——hereafter referred to as R SM-ET) can be used to describe the extent to which SM affects ET (Dirmeyer et al., 2009; Dirmeyer, 2011). A positive R SM-ET suggests that ET changes are primarily controlled by SM, e.g., higher (lower) SM causes more (less) ET. This usually happens in arid regions under water-limited conditions. On the contrary, a negative R SM-ET implies that ET is mainly affected by the atmospheric environmental variables (e.g., temperature, humidity and wind speed) and controls the changes of SM. In other words, increased (decreased) ET tends to decrease (increase) SM. Such a negative R SM-ET typically occurs in humid regions under energy-limited conditions. For this reason, R SM-ET is often employed to determine the direction of land-atmosphere interactions, i.e., how the land affects the atmosphere, and vice versa. Another widely used approach to diagnose land-atmosphere coupling is the correlation coefficient between ST and ET (R ST-ET) (Zittis et al., 2013). This is also a measure of the interaction between soil and temperature through ET (Ruscica et al., 2014). R ST-ET is generally positive and negative under energy-limited and water-limited conditions, respectively. For example, when R ST-ET is positive, higher (lower) temperature (rather than SM) leads to more (less) ET. In contrast, a negative R ST-ET indicates that increased (decreased) ET results in lower (higher) temperature, which is very likely induced by an SM anomaly. To sum up, when both of the positive R SM-ET and negative R ST-ET are significant, it is likely that SM regulates the ET process, thereby affecting ST. This situation suggests that there is a strong land-atmosphere coupling.In this study, all data are linearly detrended before performing statistical analyses.
3.1. Seasonal mean SM and ST
Figure 2 shows the seasonal mean states of SM in the two datasets. In general, both the ERA and CLM datasets are in agreement that the soil dries from southeast to northwest, and some differences (e.g., the magnitude, location and timing of the maximum SM) exist among the four seasons, i.e., spring (March-May, MAM), summer (June-August, JJA), autumn (September-November, SON) and winter (December-February, DJF). For ERA, the SM over North China is generally lower than 22%, with a dry center located west of Hetao (bend of the Yellow River), and partially reaches 25% during summer in the east. Northeast China maintains a relatively wet condition (over 25%) throughout the year. Meanwhile, the seasonal variability is larger in the southern part of the country. For the upstream region of the Yangtze River and central China, the SM in spring-autumn and winter are above and below 31%, respectively. The SM over East China and South China ranges from 28% and 25% during winter to 31% during summer, respectively. In particular, Southwest China is a dry center (<22%) during spring but a wet center (>31%) during summer and autumn. CLM basically reproduces the spatial patterns of seasonal SM. However, there are some systematic biases: wetter conditions in wet regions and drier conditions in dry regions, which is consistent with the findings of (Lai et al., 2014). Besides, the seasonal variability of SM for CLM is relatively low. Most obviously, there is a dry center near Hetao (<10%) and two wet centers in East China and Northeast China (>37%) that maintain their magnitudes and positions during the four seasons. Moreover, other regions demonstrate corresponding seasonal variations, but much less so than in the ERA dataset. All in all, the dry-wet patterns from the two datasets match the dry and wet climate divisions in China over recent decades, insofar as humid/semi-humid and arid/semi-arid regions are located in southern and northern parts of China, and the Northeast China is an exceptional semi-humid region (Ma and Fu, 2005; Wu et al., 2005; Zhang et al., 2016).
Figure2. Seasonal mean SM for the period 1979-2013 in the ERA (7 cm, top row) and CLM (6.2 cm, bottom row) datasets (units: volume %).
Figure3. As in Fig. 2 but for ST (units: °C).Figure 3 presents the climatological ST of each season. The spatial distributions and seasonal changes of the ST in the two datasets are in close agreement. Generally, ST decreases from south to north. It is warmer in summer (ranging from 10°C to 30°C) but colder in winter (ranging from -15°C to 20°C). Comparing Figs. 2 and 3, the ability of CLM to reproduce the ST climatology is apparently better than that of SM.
The above results show that complex spatial and seasonal changes exist in both SM and ST. In the following sections, the connections between SM and ST (through ET) are further investigated. These investigations are expected to reveal the possible pathway of land-atmosphere interactions and the "hot spots" of land-atmosphere coupling over eastern China.
2
3.2. Spatial distributions and seasonal changes of R SM-ET and R ST-ET
As shown in Fig. 4, the extent to which SM affects ET, i.e., the R SM-ET, is mainly opposite to the SM distribution in each season (Fig. 2). Generally, positive and negative R SM-ET are roughly distributed in the dry north and wet south, respectively. Significant positive values are found over North China during the four seasons for both datasets, indicating that SM evidently affects ET because of the relatively dry soil condition. For other regions, differences exist not only in different seasons, but also in the different datasets. Firstly, for ERA, Northeast China exhibits lower R SM-ET values, especially during summer and winter. In contrast, the seasonal changes in negative values over humid/semi-humid regions are much more complex. For example, the Yangtze River basin shows significant negative R SM-ET during spring and summer, whereas its downstream region (East China) shows positive values during autumn and winter. Such a reversal in the changes also appears in South and Southwest China: the R SM-ET over South China is negative in summer but positive in winter, and the R SM-ET over Southwest China changes from positive to negative between winter/spring and summer/autumn. For the CLM dataset, Northeast China shows significant negative R SM-ET throughout the year, except for the winter season. Besides, the seasonal variations of R SM-ET for humid/semi-humid regions in CLM are much weaker: East China and South China maintain negative values during autumn/winter. In particular, the significant positive values over Southwest China are only shown in spring. The negative R SM-ET indicates ET is not constrained by SM under wet soil conditions, and these differences between the two datasets may be due to the systematic biases and weaker seasonal changes in the SM simulation, as discussed in section 3.1. Overall, the positive R SM-ET, a necessary (but not sufficient) condition for land-atmosphere coupling (Dirmeyer et al., 2009), is found over North China throughout the year and over Southwest China during spring in both datasets.Figure 5 demonstrates the spatial distribution of R ST-ET, which is basically opposite to that of R SM-ET (Fig. 4). For ERA, positive values are mostly found in the humid/semi-humid regions over the south and northeast, implying that the energy supply is the primary factor causing the ET anomaly. If the ST is higher than normal, the ET increases and dries the soil. Most of this positive R ST-ET can be sustained for a whole year. On the contrary, the R ST-ET over North China exhibits relatively larger seasonality. The values are mainly negative during the four seasons, but the most significant and robustly negative ones only appear in summer. Moreover, a reverse change in R ST-ET is also shown over Southwest China: it is negative in spring and changes to positive during summer and autumn. CLM basically reproduces the R ST-ET pattern of ERA.
Figure4. Spatial distributions of correlation coefficients between SM and ET (R SM-ET) in the ERA (top row) and CLM (bottom row) datasets. The dotted grid points are significant at the 95% confidence level. All data are detrended.
Figure5. As in Fig. 4 but for the correlation coefficients between ST and ET (R ST-ET).2
3.3. Seasonal evolution of land-atmosphere coupling
To further investigate the relationships between R SM-ET/ R ST-ET and SM, we produce scatterplots of the two correlation coefficients at each grid point over eastern China for the ERA and CLM datasets (Fig. 6). To reduce the influence of cold conditions (e.g., ice, snow and frozen soil) in high latitude areas (Fig. 2), only the summer pattern is chosen for analysis. It is found that both R SM-ET and R ST-ET are linearly related to SM. Under dry soil conditions, the grid points with positive R SM-ET are primarily accompanied by negative R ST-ET, which implies that SM affects ET and ST, and a strong land-atmosphere coupling exists. With increased SM, the positive R SM-ET is reduced and even becomes negative. Meanwhile, the negative R ST-ET is turning positive. These findings indicate that the influence of SM on ET and ST is reduced, and the land-atmosphere coupling intensity is weakened, with the transition from arid to humid regions. It is worth mentioning that when SM is relatively lower, a close linear relationship exists between R SM-ET and R ST-ET, implying that the coupling changes with SM can be reflected by each of them. However, with higher SM, this relationship becomes weak, and suggests that ST may have different influences on R SM-ET and R ST-ET, which can be found in their seasonal cycles and is further discussed in the following text.
Figure6. Relationship between R SM-ET and R ST-ET in summer at each grid point over eastern China for the (a) ERA and (b) CLM dataset. The different colors denote different SM ranges (units: volume %). All data are detrended before the correlation coefficient is calculated.
Figure7. SM intra-annual variability (multi-year average of maximum SM minus minimum SM within a year) in the (a) ERA and (b) CLM dataset (units: volume %). The four selected sub-areas are shown in (a): area I (21°-28°N, 97°-104°E); area II (21°-33°N, 106°-122°E); area III (35°-50°N, 96°-121°E); and area IV (40°-53°N, 123°-135°E).
Figure8. Monthly variations of SM (blue lines; units: volume %), R SM-ET (red lines) and R ST-ET (green lines) in areas (a, b) I, (c, d) II, (e, f) III and (g, h) IV from March to February for the ERA (left) and CLM (right) datasets. The dashed lines are significant at the 95% confidence level. All data are detrended before the correlation coefficient is calculated.Additionally, SM differs not only by region but also by season (Fig. 2), as does the pathway of land-atmosphere interactions (i.e., SM affects ST and vice versa; Figs. 4 and 5). To further examine the intra-annual variability of SM, the mean difference between maximum and minimum SM within a year is shown in Fig. 7. Large variations (>4%) are demonstrated in the southern half, southeastern North China and Northeast China for the ERA dataset. This is possibly due to the influence of the transformation of the monsoon system (Jiang et al., 2015). Notably, Southwest China is a large SM-variation center. For the CLM dataset, as discussed in section 3.1, the seasonal variability is relatively lower, especially in the Yangtze River basin and Northeast China. Still, the largest intra-annual change is exhibited in Southwest China. Therefore, considering the evident seasonal variations in R SM-ET, R ST-ET and SM, we first take area I (Southwest China: 21°-28°N, 97°-104°E) as an example to explore the monthly change at the regional scale. We also select the humid/semi-humid area II (including South China, East China and Central China: 21°-33°N, 106°-122°E), arid/semi-arid area III (North China: 35°-50°N, 96°-121°E) and semi-humid area IV (Northeast China: 40°-53°N, 123°-135°E) for the comparative analyses. These four sub-areas are shown in the boxed areas in Fig. 7a, and the dry-wet definitions are based on (Ma and Fu, 2005).
Figure9. Annual cycle (top row) of ST in areas I, II, III and IV, from March to February, in the (a) ERA and (b) CLM dataset. The solid dots and short lines (bottom row) are the annual mean, maximum and minimum ST, respectively. Units: °C.
Figure10. Absolute value of positive R SM-ET multiplied by negative R ST-ET. The dotted grid points indicate that both of the two correlations are significant at the 95% confidence level. All data are detrended.In Southwest China, precipitation during the wet season (May-October) accounts for about 70% to 80% of the annual total (Zhang et al., 2014). Accordingly, as shown in Figs. 8a and b, the SM in area I rapidly increases beginning in May and remains at a high level (>30%) from June to November. During the months with higher SM, negative R SM-ET and positive R ST-ET are significant (p<0.05), indicating that energy primarily controls ET. During the dry season (November-April), the soil does not receive enough water; thus, SM decreases to slightly below 25% in March. Meanwhile, the R SM-ET and R ST-ET become positive and negative, respectively. The alternating relationship signs imply that the feedback from SM to the atmosphere is enhanced (shifts from an energy-limited condition during the wet season to a water-limited condition during the dry season), particularly in early spring (March and April), characterized by the maximum R SM-ET. Similarly, in area II, R SM-ET increases and R ST-ET decreases as the soil turns from wet to dry in the ERA dataset (Fig. 8c). For the CLM dataset, however, the weaker intra-annual SM variability in area II leads to a smaller corresponding change in R SM-ET (Fig. 8d).
In area III, the significant (p<0.05) positive R SM-ET and lower SM throughout the year indicate that this area is under a water-limited condition in all seasons for both datasets (Figs. 8e and f). However, the significant (p<0.05) negative R ST-ET (strong land-atmosphere coupling) occurs only from May to August and becomes positive from September to April. The conditions in area IV are more complicated (Figs. 8g and h), as compared with the other three areas. R SM-ET changes with the ERA SM, but R ST-ET is basically positive throughout the year. The CLM SM, by contrast, is high, and weaker variability exists during the 12 months, with the R SM-ET and R ST-ET changing differently.
To further understand the monthly changes of the correlations in areas III and IV (Fig. 8), we additionally examine the regional differences in ST, which also exhibit seasonal variations among the four sub-areas (Fig. 3). The annual ST cycles based on the ERA and CLM datasets are shown in Fig. 9. Areas I and II show low intra-annual ST variabilities, whereas areas III and IV show larger ones (bottom row of Fig. 9). In areas III and IV, the ST is close to and often below 0°C from November to March, indicating that an insufficient energy supply exists for ET. In contrast, areas I and II are relatively warm throughout the year. These findings provide a potential explanation for the abnormal positive R ST-ET observed during the cold seasons in areas III and IV (Fig. 8).
To sum up, the above results show that differences exist in the land-atmosphere coupling in both space and seasonality. The differences are reflected in the spatial patterns of R SM-ET and R ST-ET and their seasonal variations. In consideration of the strong coupling only existing under the condition with positive R SM-ET and negative R ST-ET, we additionally specify an indicator, which is expressed as the absolute value of the positive R SM-ET multiplied by the negative R ST-ET, to identify the "hot spots". The results for both the ERA and CLM datasets are shown in Fig. 10. For spring, the hot spots are in North China and Southwest China; during summer and autumn, however, these hot spots are primarily in North China. In addition, land-atmosphere coupling barely exists in winter.
Previous studies have demonstrated that abnormal SM can affect the atmospheric general circulation mainly through altering the surface energy balance and land surface thermal forcing of the atmosphere, and further lead to precipitation anomalies in China (Zuo and Zhang, 2007, 2016; Zhang and Zuo, 2011; Meng et al., 2014). (Zhang and Zuo, 2011) found that anomalous spring SM from the lower and middle reaches of the Yangtze River Valley to North China is highly correlated with the summer precipitation in China. Abnormally wet spring soil tends to increase the surface evaporation and lower the surface air temperature. As a result, the reduced temperature in late spring reduces the land-sea thermal contrast, resulting in a weakened East Asian monsoon. SM is definitely an important factor reflecting the land-atmosphere interaction, and ET is a key component of the land-atmosphere feedback cycle. Our results further map the "hot spots" of land-atmosphere coupling over eastern China. In addition, we emphasize the seasonal variabilities of these hot spots. In the humid/semi-humid regions of China, strong land-atmosphere coupling mainly happens in the relatively drier seasons, which is consistent with the conclusions of (Wei et al., 2012) and (Wei and Dirmeyer, 2012), who showed that SM-precipitation coupling is stronger during dry periods for climatologically wet regions. However, in arid/semi-arid regions of China, studies should focus on the relatively warmer seasons when significant negative SM feedbacks exist with relatively higher temperature (Zhang and Dong, 2010; Zhang et al., 2011).
Our study has certain limitations. For example, land cover exerts an important effect on the exchanges of energy, water, carbon dioxide and other greenhouse gases between the land surface and atmosphere (Yin et al., 2015); thus, land-use and land-cover change (LUCC) might affect processes between the land and atmosphere. Moreover, remarkable LUCC has occurred alongside rapid economic development over the past 30 years. Our study does not consider LUCC, which might lead to some uncertainties in the results. Although the analytical process may be subjective, it still provides a reference to further study land-atmosphere interactions. It is difficult to directly measure the strength of land-atmosphere coupling in the real world, and the specific land feedback on atmospheric circulation (e.g., monsoon intensity) still needs to be explored in depth. With an increasing number of tools available to determine how the land affects the climate, researchers will have more opportunities to improve their understanding of climate predictions.
Further analysis suggests that the coupling over North China is closely associated with temperature (partly reflecting the energy supply). In North China, the dry soil condition (water-limited) is maintained throughout the year; and summer is relatively warmer comparing to other seasons, which means a sufficient energy supply for ET, consequently leading to the strongest land-atmosphere coupling. On the other hand, the SM (water supply) plays a decisive role in the coupling over humid/semi-humid regions. In Southwest China, the seasonal climate conditions are warm due to the lower latitude, meaning that the energy supply is sufficient throughout the year. However, the precipitation varies across the four seasons. By the end of the dry season (i.e., spring), the SM is evidently lower, constraining ET.
The uncertainties of this study should be noted. We use a high-quality reanalysis dataset (ERA-Interim) and model simulation (CLM4.0) to maximize the robustness of the results, and the results from both datasets generally agree with each other, albeit with some differences observed in some areas. The methods applied in this study also have some limitations. Therefore, the findings of this study should be regarded as preliminary, and additional studies are needed. The specific feedback mechanisms from the land to the atmosphere over the aforementioned "hot spots" require more thorough investigation.
