Direct and indirect effects of atmospheric conditions and soil moisture on surface energy partitioning revealed by a prolonged drought at a temperate forest site

Authors


Abstract

[1] The purpose of this paper is to examine the mechanism that controls the variation of surface energy partitioning between latent and sensible heat fluxes at a temperate deciduous forest site in central Missouri, USA. Taking advantage of multiple micrometeorological and ecophysiological measurements and a prolonged drought in the middle of the 2005 growing season at this site, we studied how soil moisture, atmospheric vapor pressure deficit (VPD), and net radiation affected surface energy partitioning. We stratified these factors to minimize potential confounding effects of correlation among them. We found that all three factors had direct effects on surface energy partitioning, but more important, all three factors also had crucial indirect effects. The direct effect of soil moisture was characterized by a rapid decrease in Bowen ratio with increasing soil moisture when the soil was dry and by insensitivity of Bowen ratio to variations in soil moisture when the soil was wet. However, the rate of decrease in Bowen ratio when the soil was dry and the level of soil moisture above which Bowen ratio became insensitive to changes in soil moisture depended on atmospheric conditions. The direct effect of increased net radiation was to increase Bowen ratio. The direct effect of VPD was very nonlinear: Increased VPD decreased Bowen ratio at low VPD but increased Bowen ratio at high VPD. The indirect effects were much more complicated. Reduced soil moisture weakened the influence of VPD but enhanced the influence of net radiation on surface energy partitioning. Soil moisture also controlled how net radiation influenced the relationship between surface energy partitioning and VPD and how VPD affected the relationship between surface energy partitioning and net radiation. Furthermore, both increased VPD and increased net radiation enhanced the sensitivity of Bowen ratio to changes in soil moisture and the effect of drought on surface energy partitioning. The direct and indirect effects of atmospheric conditions and soil moisture on surface energy partitioning identified in this paper provide a target for testing atmospheric general circulation models in their representation of land-atmosphere coupling.

1. Introduction

[2] The exchange of sensible and latent heat fluxes at the surface is one of the most important aspects of the land-atmosphere coupling. These energy fluxes, which draw from the partitioning of net radiation absorbed by the surface, are the direct fuel for atmospheric dynamics [Garstang and Fitzjarrald, 1999]. For a given amount of available energy at the surface, change in the relative proportions of sensible heat and latent heat fluxes critically influences boundary layer structures, cloud development [Rabin et al., 1990; Segal et al., 1995; Ray et al., 2003], rainfall [Betts and Ball, 1998; Pielke, 2001], and ultimately contributes to how strong the land and atmosphere are coupled together [Koster et al., 2004]. Therefore it is crucial to understand the variability of land surface energy partitioning and the mechanism behind that variability.

[3] Previous studies have examined the variation of land surface energy partitioning. Data from Betts and Ball [1995] indicate that under mostly cloud-free conditions, there is a clear diurnal cycle in Bowen ratio (sensible heat flux divided by latent heat flux) with the maximum occurring in the midday. The same data also indicate that Bowen ratio is higher over dry soil than over wet soil. Fitzjarrald et al. [2001] found that Bowen ratio also has seasonal patterns that correspond with vegetation phenology; this correspondence is usually characterized by a decrease in Bowen ratio around the time of leaf emergence and an increase following leaf senescence. Wilson et al. [2002] examined how Bowen ratio varied across sites with different vegetation structures and climatic conditions. A conclusion that can be drawn from these and other previous studies is that land surface energy partitioning can be affected by a suite of atmospheric, soil, and ecophysiological factors, including available energy, atmospheric humidity, temperature, soil moisture, vegetation structure, and stomatal behavior [Baldocchi, 1997].

[4] However, it is often difficult to discern how these abiotic and biotic factors affect land surface energy partitioning. This difficulty arises not only because the variations of these factors are highly correlated under natural conditions [Gu et al., 1999] but also because the result of energy partitioning feeds back to their variations [Entekhabi et al., 1996; Eltahir, 1998]. The case of soil moisture demonstrates this difficulty. Conceptually, the effect of soil moisture on land surface energy partitioning is clear: Increased soil moisture increases the amount of water available for transpiration and physical evaporation and therefore should decrease Bowen ratio. However, soil moisture is related to a myriad of physical and biological processes that control surface energy balance. As discussed in detail by Eltahir [1998], a wetter soil could increase net radiation due to decreases in surface albedo and Bowen ratio. Also, a decrease in Bowen ratio could lead to accumulation of water vapor in the boundary layer, which could decrease the evaporative power of the surface air and influence the stomatal dynamics. Additionally, changes in soil moisture could alter the convective processes and affect the development of clouds and precipitation [Rabin et al., 1990; Findell and Eltahir, 2003; Ray et al., 2003]. All these factors complicate the relationship between soil moisture and surface energy partitioning.

[5] Thus the mechanism controlling surface energy partitioning has to be examined in a network of interacting physical and biological processes and feedback loops. An opportunity to gain some insight into this network arose in the mid growing season of 2005 when a prolonged drought occurred at the Missouri Ozark AmeriFlux site where we have been conducting continuous measurements of fluxes of radiative energy, sensible heat, latent heat, and carbon dioxide as well as complementary vegetation and soil water status. As the drought developed, soil and predawn leaf water potentials declined; however, wide ranges of atmospheric humidity and radiation variations were observed during the drought period. With careful stratification, we were able to analyze how soil moisture, atmospheric humidity, and net radiation interactively influence surface energy partitioning. We observed that the effects of one factor on surface energy partitioning depend on the other factors, reflecting dynamic controls of atmospheric, soil and ecophysiological processes on land-atmosphere energy exchanges. Specifically, with this study, we answered the following two groups of questions:

[6] The first group consists of direct-effect questions:

[7] 1. How does soil moisture affect surface energy partitioning for a given set of values of atmospheric vapor pressure deficit (VPD) and net radiation?

[8] 2. How does VPD affect surface energy partitioning for a given soil moisture condition and a given value of net radiation?

[9] 3. How does net radiation affect surface energy partitioning for a given soil moisture condition and a given value of VPD?

[10] The second group consists of indirect-effect questions:

[11] 4. How does soil moisture affect the relationship between VPD and surface energy partitioning and that between net radiation and surface energy partitioning?

[12] 5. How does VPD affect the relationship between soil moisture and surface energy partitioning and that between net radiation and surface energy partitioning?

[13] 6. How does net radiation affect the relationship between soil moisture and surface energy partitioning and that between VPD and surface energy partitioning?

2. Study Site, Measurements, and Method of Analysis

2.1. Study Site

[14] We carried out the study at the Missouri Ozark AmeriFlux site. This site, which has been operating since spring of 2004 with a suite of meteorological and ecological instrumentation, is located in the University of Missouri's Baskett Wildlife Research and Education Area (BREA, Lat. 38°40′N, Long. 92°12′W). BREA is within the Ozark border region of central Missouri. Second-growth upland oak-hickory forests constitute the major vegetation type at the BREA [Rochow, 1972; Pallardy et al., 1988]. Major tree species include white oak (Quercus alba L.), black oak (Q. velutina Lam.), shagbark hickory (Carya ovata (Mill.) K.Koch), sugar maple (Acer saccharum Marsh.), and eastern redcedar (Juniperus virginiana L.). The climate of the area is warm, humid, and continental [Critchfield, 1966], with mean January and July temperatures of −2.2°C and 25.2°C, respectively. Annual precipitation averages 940 mm, and moderate to severe droughts commonly occur between July and September [e.g., Hinckley et al., 1979; Bahari et al.,1985]. Dominant soils at BREA are Weller silt loam and a broad type classified as “Steep Stony Land” which includes all rocky slopes with a thin soil covering [Garrett and Cox, 1973]. The comparatively thin soils of these oak forests often exacerbate plant water stress when droughts occur.

2.2. Measurements

[15] Measurements used in this study included fluxes of sensible and latent heat, common meteorological variables, soil water potential, and predawn leaf water potential. These data were from the summer of 2005.

2.2.1. Measurements of Flux and Meteorological Variables

[16] We obtained the flux and meteorological measurements from a 32-m walkup scaffold tower. We installed the instruments at the top of the tower, about 15 m above the top of the canopy (the canopy height is about 17m). We measured the sensible and latent heat fluxes with the eddy covariance technique [Baldocchi, 2003]. The eddy covariance system consisted of a three-dimensional ultrasonic anemometer (Model 81000, RM Young) and a fast responding, open-path infrared gas analyzer (LI7500, Li-Cor). Outputs from the ultrasonic anemometer and the gas analyzer were sampled at 10 Hz using a computer-controlled system. We used Reynolds averaging over half-hour periods to compute scalar fluctuations and flux covariances and applied the density corrections of Webb et al. [1980]. In addition, we estimated the canopy air space energy storages with an eight-level temperature/humidity profile system. We also made parallel observations of routine meteorological variables. We measured rainfall with tipping bucket rain gauges. The shortwave and longwave radiation balance was monitored with a 4-way net radiometer (CNR 1, Kipp and Zonen) installed at the top of the tower. This net radiometer includes two pyranometers and two pyrgeometers (upward and downward looking) and outputs incoming and reflected solar radiation and incoming and outgoing longwave radiation simultaneously.

2.2.2. Measurements of Soil Water Potential

[17] Measurements of soil water potential were obtained from heat-dissipation-based water potential sensors (CS229; Campbell Scientific, Inc., Logan, Utah). We installed the CS229 sensors at approximately 0.05, 0.15, 0.3 and 0.7m in two replicate soil pedons to yield information on the availability of water within the zone of maximum root density. The probes were calibrated in situ by comparison to soil water content data obtained from time domain reflectometry buried probes (20 cm long three pronged probes; Soil Moisture Equipment Corp., Galeta, California) installed side by side with the heat dissipation sensors. All soil water content measurements were converted to soil water potentials using moisture retention curves generated for the A and B horizons as described by Hanson et al. [2003], but using a cooled mirror dewpoint-sensor-based method (Aqualab Series 3, Decagon, Pullman, Washington). To represent the overall soil water status, the eight samples of soil water potential (four samples in each of the two pedons) were averaged and the mean soil water potential was used in this study.

2.2.3. Measurements of Predawn Leaf Water Potential

[18] We measured predawn leaf water potential of lower crown leaves. Leaf samples for measurement of water potential were collected before dawn periodically through the growing season for canopy and sapling individuals of important tree species at the site. A total of 20–21 samples were obtained each day with 6–7 taken from Quercus alba, and the rest distributed among Q. velutina, Acer saccharum, Carya ovata, Fraxinus americana, and Juniperus virginiana. Leaf water potential was estimated with a pressure chamber [Turner, 1981]. Predawn estimates of leaf water potential quantify the maximum rehydration state of plants overnight after the diurnal minimum experienced the previous day. Unless soil is very dry or anomalous nocturnal transpiration is substantial, predawn estimates leaf water potential of the lower crown also provide a good proxy for soil water potential in the rooting zone [Kozlowski and Pallardy, 1997].

2.3. Method of Analysis

[19] We used a combinatorial stratification method to analyzing the direct and indirect effects of multiple biotic and abiotic factors on surface energy partitioning. In this method, the relationship between one factor and surface energy partitioning is examined successively at different, narrow intervals of other factors. The surface energy partitioning was characterized with the Bowen ratio as a function of (1) soil and plant water status, (2) atmospheric water vapor pressure deficit (VPD), and (3) net radiation. A combinatorial stratification method was needed because we wanted to avoid potential complication which may arise from correlation among VPD, net radiation, and soil moisture [Gu et al., 1999] in explaining the effects of these factors on the Bowen ratio. Initially, we included wind speed in the analysis but did not find a significant dependence of the Bowen ratio on wind speed; therefore we dropped it from further analysis. Temperature can also affect the Bowen ratio but it would be extremely difficult to separate its effect from that of VPD. Furthermore, stomata, which strongly regulate latent heat flux, respond more directly to atmospheric humidity than to temperature [Ball et al., 1987]. Therefore we did not include temperature in this analysis.

[20] To stratify, we divided soil and plant water status, VPD, and net radiation into two, four, and four levels, respectively. Soil and plant water status was categorized into drought and nondrought conditions based on observations of soil water potential and predawn leaf water potential. We divided the middle of the growing season (days 130–270) of 2005 into drought and nondrought periods. We used the middle growing season to avoid potentially confounding effects of leaf expansion early and leaf senescence late in the growing season on sensible heat and latent heat exchanges. There is no commonly adopted criterion for separating drought and nondrought conditions. In a widely cited article, Hsiao [1973] considered plants under mild stress when plant water potential is several tenths of a MPa “below corresponding values in well watered plants under mild evaporative demand,” under moderate stress when water potential declines below this range but remains above −1.5 MPa, and under severe water stress as water potential declines below −1.5 MPa. As shown later, our study site experienced mild, moderate and severe drought stress from late June to early August, according to the guideline suggested by Hsiao [1973]. Therefore we could have considered different stages in the drought development of 2005. However, in so doing, we found that there was no sufficient number of data in the moderate and severe drought stages for us to make robust statistical analysis of variability in energy partitioning. We thus decided to consider just two conditions for soil and plant water status: drought versus nondrought with the cutoff soil water potential set at −0.7MPa. This point of separation was consistent with that observed at the onset of drought-related stomatal closure observed in previous studies of canopy trees of dominant species at the BREA [Loewenstein and Pallardy, 1998]. Using measurements of predawn leaf water potential as a basis for separating drought and nondrought conditions did not lead to much difference because soil water potential and predawn leaf water potential tracked each other well.

[21] The four levels of VPD were <1000, 1000–2000, 2000–3000, and >3000 Pa while the four levels of net radiation were <250, 250–500, 500–700, and >700 Wm−2. Similar to the separation of soil and plant water status into drought and nondrought conditions, these levels of VPD and net radiation represented a tradeoff between the desire to have a fine-grained stratification and the requirement of adequate number of data for statistical analysis.

[22] With this stratification scheme, we proceeded with the analysis of the direct and indirect effects of soil moisture, VPD, and net radiation on surface energy partitioning. The effects of soil moisture were examined at 4 × 4 combinations of VPD and net radiation levels, that of VPD at 2 × 4 combinations of water status and net radiation levels, and that of net radiation at 2 × 4 combinations of water status and VPD levels.

3. Results

3.1. Assessment of the Drought

[23] The study site received a mere 19mm of precipitation in July of 2005. This amount was only one fifth of the climatic average (the 1971–2000 mean of the July precipitation was 95 mm, according to the long-term data from the nearby weather station at Columbia, Missouri, archived in the National Climate Data Center). The July mean temperature at the top of the tower was also higher than the climatic average (26°C versus 25.2°C). Lack of significant precipitation events from roughly mid-June to mid-August (Figure 1a) caused steady decline in soil water potential and predawn leaf water potential (Figure 1b). This declining trend was only briefly interrupted by a light shower in mid-July; it resumed subsequently until a major precipitation event occurred in mid-August. The predawn leaf water potential declined from above −0.5 MPa before the drought to near −2.5 MPa at the peak of the drought while soil water potential declined from about −0.5 MPa to −2.0 MPa, indicating drought progression from mild, moderate to severe conditions [Hsiao, 1973]. A second drought also developed in mid-September, but it lasted only briefly. Using soil water potential of −0.7 MPa as a separation criterion, we grouped the days from day 130 to 270 into two categories, the drought periods (day 177 to 225 and 251 to 257) and the nondrought periods (the rest), which are indicated in Figure 1b.

Figure 1.

Measurements of (a) precipitation and (b) predawn leaf water potential and soil water potential during the growing season of 2005.

[24] Although the overall patterns of soil water potential and predawn leaf water potential tracked each other well, some systematic differences between them existed (Figure 1b). During the early part of the major drought period, the predawn leaf water potential was less negative than the soil water potential; however, during the later part of the same period, the predawn leaf water potential was more negative. Predawn leaf water potential is a good integrative indicator of the overall soil water status if roots keep close contact with the soil. As soil dries, soil water content increases with the depth in the soil. Since the soil water potential was measured near the soil surface while plant roots can access water in deeper soil, predawn leaf water potential was higher than the measured soil water potential. As the soil becomes very dry, however, hydraulic conductivity can decline dramatically and soil-root contact may be lost [Eagleson, 1978]. These factors can prevent overnight equilibration of the plant with the soil and account for the pattern of lower predawn leaf than soil water potential under severe drought observed in the present study.

3.2. Effects of Soil Moisture on Surface Energy Partitioning Under Different Levels of VPD and Net Radiation

[25] Figure 2 shows the relationship between soil moisture and Bowen ratio at two levels of VPD (1000–2000 Pa and 2000–3000 Pa) and all four levels of net radiation. We did not show the data for the other two levels of VPD (<1000 Pa and >3000 Pa) because the relationship revealed by Figure 2 can be broadly extended to these two levels of VPD.

Figure 2.

Effects of soil moisture on Bowen ratio at different levels of VPD ((a–d) 1000–2000 Pa and (e–h) 2000–3000 Pa) and net radiation (<250 Wm−2 (Figures 2a and 2e), 250–500 Wm−2 (Figures 2b and 2f), 500–700 Wm−2 (Figures 2c and 2g), and >700 Wm−2 (Figures 2d and 2h)).

[26] Overall, Bowen ratio decreased with increasing soil moisture; however, this overall pattern was greatly affected by the levels of soil moisture, VPD, and net radiation (Figure 2). When the soil was dry, soil moisture exerted strong controls on Bowen ratio (if VPD and net radiation were not too low); however, as the soil water availability improved, Bowen ratio eventually became insensitive to soil moisture. VPD and net radiation affected the rate of decrease of Bowen ratio with increasing soil moisture when the soil was dry and the transitional level of soil moisture that separated sensitive and insensitive soil moisture conditions for Bowen ratio. Increased VPD and net radiation tended to enhance the sensitivity of Bowen ratio to changes in soil moisture and cause the transitional soil moisture to be higher (for example, compare Figure 2a with Figure 2h); at low levels of VPD and net radiation (VPD < 1000Pa and net radiation < 250 Wm−2; data not shown), soil moisture had a very weak influence on Bowen ratio even when the soil was dry.

3.3. Effects of VPD on Surface Energy Partitioning Under Different Levels of Soil Moisture and Net Radiation

[27] The values of VPD observed under nondrought conditions were mostly less than 3000 Pa while those under drought conditions were up to 5000 Pa (Figure 3). However, over the range of 0–3000 Pa, there was substantial overlapping in VPD between drought and nondrought conditions at all levels of net radiation. Similarly, there was substantial overlapping in VPD among different levels of net radiation under both drought and nondrought conditions. We emphasize these common ranges in VPD because it is only within these common ranges that we could validly examine how soil water status and net radiation affected the Bowen ratio–VPD relationship.

Figure 3.

Effects of atmospheric water vapor pressure deficit (VPD) on Bowen ratio at different levels of net radiation ((a and b) <250 Wm−2, (c and d) 250–500 Wm−2, (e and f) 500–700 Wm−2, and (g and h) >700 Wm−2) under nondrought (Figures 3a, 3c, 3e, and 3g) and drought (Figures 3b, 3d, 3f, and 3h) conditions.

[28] Under nondrought conditions, the Bowen ratio decreased with increased VPD at all levels of net radiation (Figures 3a, 3c, 3e, and 3g). Furthermore, the decrease was sharper at lower values of VPD. Under drought conditions, the Bowen ratio also appeared to decrease with VPD over the range of VPD shared with nondrought conditions (<3000 Pa) and at low levels of net radiation (<500 Wm−2) but the relationship was much less clear (Figures 3b and 3d) while at high levels of net radiation (>500 Wm−2, Figures 3f and 3h), the Bowen ratio appeared to increase with VPD. Note that at high levels of net radiation (Figures 3e, 3f, 3g, and 3h), we cannot confidently attribute the apparent reversal of the Bowen ratio–VPD relationship from a decreasing trend to an increasing trend to changes in soil water status because few points had VPD values larger than 3000 Pa under nondrought conditions (Figures 3e and 3g). What we can draw from Figure 3 regarding the role of soil moisture in the Bowen ratio–VPD relationship, given the much increased scatter in the plots under drought conditions, is that drought weakened the dependence of the Bowen ratio on VPD, at least for VPD less than 3000Pa. For net radiation and under drought conditions, we can attribute, with relative confidence, the increasingly recognizable pattern of increasing Bowen ratio with VPD at high values of VPD to the increase in net radiation since there was substantial overlapping in the ranges of VPD at all levels of net radiation.

[29] To better understand the Bowen ratio–VPD relationship observed above, we examined how latent and sensible heat fluxes changed with VPD under different levels of soil moisture and net radiation (Figure 4). Under nondrought conditions, latent heat flux initially increased with VPD but tended to level off or even decrease as VPD continued to increase; the leveling off occurred at higher values of VPD or was absent as net radiation increased (Figure 4a). Compared with the latent heat flux–VPD relationship under nondrought conditions, latent heat flux under drought conditions showed a much subdued initial increase with VPD; moreover, at nearly all levels of net radiation, it reached a peak at some intermediate value of VPD and decreased with VPD as VPD continued to increase (Figure 4b). Albeit these important differences between drought and nondrought conditions, one might reasonably assume, given the overall patterns of the latent heat flux–VPD relationship, that under nondrought conditions, a prominent decrease of latent heat flux with VPD could eventually occur had values of VPD gone much higher (that is, larger than 3000 Pa), particularly at lower levels of net radiation.

Figure 4.

Changes of (a and b) latent and (c and d) sensible heat fluxes with VPD under drought and nondrought conditions at different levels of net radiation. Error bars indicate 95% confidence intervals.

[30] The sensible heat flux–VPD relationship (Figures 4c and 4d) was complementary to the latent heat flux–VPD relationship under both drought and nondrought conditions and at all levels of net radiation, for the reason of energy conservation.

3.4. Effects of Net Radiation on Surface Energy Partitioning Under Different Levels of VPD and Soil Moisture

[31] The Bowen ratio generally increased with net radiation, but drought and VPD modulated this general relationship (Figure 5). Under drought conditions, which increased the Bowen ratio for a given level of net radiation, the Bowen ratio tended to increase faster with net radiation than under nondrought conditions, particularly at high levels of VPD. Consequently, the difference in the Bowen ratio between drought and nondrought conditions increased with net radiation and VPD as can be seen from Figures 5c and 5d.

Figure 5.

Effects of net radiation on Bowen ratio under drought and nondrought conditions at different levels of VPD ((a) <1000 Pa, (b) 1000–2000 Pa, (c) 2000–3000 Pa, and (d) >3000 Pa). Error bars indicate 95% confidence intervals. The 95% confidence interval for Bowen ratio at VPD > 3000 Pa under the nondrought condition (Figure 5d) was not provided because an insufficient number of data were available.

[32] How VPD affected the Bowen ratio–net radiation relationship can be better seen from Figure 6, which is a rearrangement of Figure 5. Under nondrought conditions, the increase in the Bowen ratio with net radiation was strongest at low VPD and high net radiation. In contrast, under drought it was strongest at high VPD. In addition, under nondrought conditions, the Bowen ratio–net radiation relationship could be distinguished at all levels of VPD (Figure 6a) while under drought conditions, the relationship was sensitive to change in VPD only at high levels of VPD (Figure 6b).

Figure 6.

Data of Figure 5 replotted to show the manner in which the increase in Bowen ratio with net radiation is modulated by VPD under (a) nondrought and (b) drought conditions.

[33] Figure 7 provides a basis for understanding Figure 6. Although latent heat flux and sensible heat flux all tended to increase with net radiation under both drought and nondrought conditions and at all levels of VPD, the trend of increase differed between the drought and nondrought conditions and varied with VPD. The increase in latent heat flux with net radiation was greater under nondrought conditions (Figure 7a) than under drought conditions (Figure 7b), while the opposite was true for sensible heat flux (Figures 7c and 7d). This pattern was responsible for the greater increase in the Bowen ratio with net radiation under drought (Figure 6b) than under nondrought conditions (Figure 6a). Further, under nondrought conditions, the effect of net radiation on latent heat and sensible heat fluxes depended sensitively on VPD (Figures 7a and 7c), but under drought the relationship between net radiation and latent heat or sensible heat flux became sensitive to VPD only at high values of VPD (Figures 7b and 7d).

Figure 7.

Changes of (a and b) latent and (c and d) sensible heat fluxes with net radiation under drought and nondrought conditions and at different levels of VPD.

4. Discussion

[34] Our results showed that soil moisture, net radiation, and VPD have both direct and indirect effects on surface energy partitioning. These results allow us to depict a picture of land surface energy partitioning from ecophysiological and micrometeorological perspectives. To do so, we now discuss underlying processes that may be responsible for these direct and indirect effects. The convoluted nature of these effects makes the discussion difficult. We will first discuss the direct effects of the three factors individually. When one factor is discussed, the other factors are assumed to be constant unless otherwise stated. Afterward, we will address the indirect effects jointly and all factors can then change simultaneously.

4.1. Direct Effects of Soil Moisture

[35] Soil water plays a central role in the exchange of water to the atmosphere by vegetation. Through its influence on the production of abscisic acid by plants, soil water status affects stomatal conductance directly [Liu et al., 2005]. An ample soil water supply allows stomata to respond freely to variations in biophysical factors such as radiation and VPD. Under this situation, latent heat flux dominates over sensible heat flux (compare Figures 4a and 4c) and vegetation has a strong influence on the boundary layer condition. Under drought stress, however, reduced soil moisture brings about an increase in the production of abscisic acid, which stimulates the closure of stomata. Consequently, the capacity of stomata to respond to variations in atmospheric conditions becomes limited; this limitation weakens the dominance of latent heat flux and eventually leads to the dominance of sensible heat flux under drought conditions.

[36] More insights are gained by examining the transitional effects of soil moisture on surface energy partitioning (Figure 2). As discussed above, when there is plenty of moisture in the soil, stoma can respond freely to changes in atmospheric conditions. Under this situation, the latent heat flux is affected more by the atmospheric evaporative power and net available energy, subject to the constraints of the water transport capacity of the root and xylem system and the maximal stomatal openness, than by the available soil water. Therefore with a moist soil, surface energy partitioning is primarily controlled by atmospheric conditions and is insensitive to perturbation in soil moisture. As the soil dries, however, the influence of soil moisture on stomatal conductance increases, which leads to stronger control by soil moisture on surface energy partitioning. Therefore the overall relationship between Bowen ratio and soil water potential is characterized by a rapid decrease in Bowen ratio with increasing soil water potential when soil water potential is low and by nearly constant Bowen ratio with varying soil water potential when soil water potential is high. However, the range of soil moisture within which Bowen ratio is insensitive to soil moisture and the slope of the Bowen ratio–soil moisture relationship when soil moisture does affect Bowen ratio depend on atmospheric conditions (Figure 2).

4.2. Direct Effects of Net Radiation

[37] Net radiation determines the total amount of heat that can be transported from surface to the atmosphere (the residual is ground heat flux and vegetation heat storages). Our results indicate that net radiation also affects how the total heat transport is partitioned between latent heat flux and sensible heat flux. Both sensible and latent heat fluxes increase with net radiation (Figure 7); however, their increases are not in proportion. In theory, sensible heat flux could be as high as the surface available energy, while latent heat flux is ultimately limited by the amount of water available in the soil, the water transport capacity of the root and xylem system, and the stomatal openness. Therefore the pace of increase in latent heat flux with net radiation must slow down as net radiation increases, causing the curve of latent heat flux against net radiation to be convex and that of sensible heat flux against net radiation to be concave (Figure 7). Consequently, the Bowen ratio tends to increase with net radiation (Figures 5 and 6).

4.3. Direct Effects of VPD

[38] The influence of atmospheric VPD on surface energy partitioning is very nonlinear (Figures 3 and 4). This nonlinearity stems from the two simultaneous but opposing effects that atmospheric VPD exerts on latent heat flux. On the one hand, atmospheric VPD is a measure of atmospheric evaporative power; it essentially amounts to the driving force for water vapor diffusion across the stomatal pore. Therefore the higher it is, the larger the latent heat flux is. This is the positive effect of increased VPD on latent heat flux. On the other hand, stomatal conductance is influenced by VPD such that higher VPD leads to reduced stomatal conductance and therefore smaller latent heat flux [Ball et al., 1987; Leuning, 1995]. This is the negative effect of increased VPD on latent heat flux. How latent heat flux and therefore the Bowen ratio change with VPD depends on the balance between these two opposing effects, other conditions being equal. At low levels of VPD, the positive effect of increased VPD outweighs its negative effect and therefore latent heat flux increases and the Bowen ratio decreases with VPD. However, at high levels of VPD, the constraint on stomatal conductance becomes stronger and the growth in the negative effect eventually outpaces the growth in the positive effect on latent heat flux as VPD continues to increase. Consequently, at some transitional value of VPD, latent heat flux reaches a peak and starts to decrease with VPD while the Bowen ratio reaches a minimum and starts to increase with VPD. However, this transitional value of VPD is not a constant; it depends on soil moisture and other atmospheric conditions. Also note that soil moisture and VPD tend to be correlated such that very high (low) VPD may not occur when soil is wet (dry) [Gu et al., 1999]. As a result, a given data set may only show a partial relationship between the Bowen ratio and VPD (for example, Figures 3a, 3c, 3e, and 3g).

4.4. Indirect Effects of Soil Moisture, Net Radiation, and VPD

[39] The relationship between any of the three factors (soil moisture, net radiation, and VPD) and surface energy partitioning is not static; instead, the effect of one factor depends on variations in other factors. Soil moisture interferes with both the positive and negative effects of increased VPD on latent heat flux. VPD as a measure of the atmospheric evaporative power functions only if there is ample soil moisture. With a dry soil, the positive effect of increased VPD diminishes because of the reduced water supply in the soil. Furthermore, reduced soil moisture exerts a strong constraint on stomatal conductance and consequently, the responsiveness of stomatal conductance to changes in VPD may be limited. Therefore drought tends to stop the initial increase in latent heat flux with VPD at a smaller value of VPD as can be inferred by comparing Figure 4a with Figure 4b; it also weakens the overall dependence of surface energy partitioning on VPD, which is indicated both by the increased scatter in the Bowen ratio–VPD relationship under drought (compare Figures 3a, 3c, 3e, and 3g with Figures 3b, 3d, 3f, and 3h) and by the reduced sensitivity of latent heat and sensible heat fluxes to changes in VPD at low levels of VPD under drought (compare Figures 4a and 4c with Figures 4b and 4d).

[40] Soil moisture also interferes with how net radiation affects the relationship between surface energy partitioning and VPD. With ample soil water, increased net radiation enhances the role of VPD as a measure of atmospheric evaporative power, causing more rapid initial increase of latent heat flux with VPD at low levels of VPD and the transitional VPD at which latent heat flux peaks to move higher (Figure 4a). With a dry soil, the effect of net radiation on the relationship between surface energy partitioning and VPD is mostly seen at high levels of VPD because at low levels of VPD, energy partitioning is not sensitive to changes in VPD. Under this situation, increased net radiation tends to amplify the limiting effect of increased VPD on stomatal conductance (the negative effect of increased VPD) and enhance the tendency for the Bowen ratio to increase with VPD at high levels of VPD (compare Figures 3b and 3d with Figures 3f and 3h).

[41] While soil moisture interferes with how VPD affects surface energy partitioning, VPD also interferes with how soil moisture affects surface energy partitioning. Increased VPD tends to enhance the sensitivity of Bowen ratio to changes in soil moisture and narrow the range of soil moisture within which Bowen ratio is not sensitive to it (Figure 2). Perhaps not surprising, increased VPD causes the Bowen ratio to go even higher under drought conditions, which is indicated by the increasing difference between the drought and nondrought curves from Figures 5a–5d. Furthermore, the relationship between soil moisture and surface energy partitioning also depends on the level of net radiation. For the transitional effects of soil moisture (Figure 2), increased net radiation affects the relationship between Bowen ratio and soil moisture in a way qualitatively similar to that of increased VPD, that is, increased net radiation tends to enhance the sensitivity of Bowen ratio to changes in soil moisture and narrow the range of soil moisture within which Bowen ratio is invariant. As net radiation increases, the effect of drought on the Bowen ratio increases, particularly at high levels of VPD. This effect of net radiation is indicated by the increasing difference between the drought and nondrought curves with net radiation (for example, examine Figure 5c).

[42] Likewise, the relationship between net radiation and land surface energy partitioning also depends on soil moisture and VPD. Drought suppresses the increase in latent heat flux (compare Figures 7a and 7b) and causes the Bowen ratio to rise faster with net radiation (compare Figure 6a with Figure 6b). At the meantime, under nondrought conditions, the increase in VPD tends to dampen the influence of net radiation on the Bowen ratio (Figure 6a), possibly through its positive effect on latent heat flux (that is, VPD as a measure of atmospheric evaporative power; compare different curves in Figure 7a) whereas under drought conditions, the increase in VPD enhances the dependence of Bowen ratio on net radiation (Figure 6b) through its negative effect on latent heat flux (that is, the increased pressure on stomatal closure; Figure 7b).

5. Conclusion

[43] To conclude, we answer the questions listed in the beginning of the paper.

[44] The answers to the direct-effect questions are as follows:

[45] 1. How does soil moisture affect surface energy partitioning for a given set of values of VPD and net radiation?

[46] The answer is that typically, Bowen ratio decreases rapidly with increasing soil moisture when the soil is dry but becomes insensitive to variations in soil moisture when the soil is wet. However, the rate of decrease and the level of soil moisture above which Bowen ratio is invariant depend on atmospheric conditions.

[47] 2. How does VPD affect surface energy partitioning for a given soil moisture condition and a given value of net radiation?

[48] The answer is that the effect of VPD on surface energy partitioning is strongly nonlinear because VPD can both positively and negatively affect latent heat flux. At low VPD, an increase in VPD enhances latent heat flux and decreases Bowen ratio; at high VPD, an increase in VPD decreases latent heat flux and increases Bowen ratio. However, this overall relationship is sensitively affected by net radiation and soil moisture conditions. Furthermore, one may only see part of this relationship if the observed range of VPD is limited.

[49] 3. How does net radiation affect surface energy partitioning for a given soil moisture condition and a given value of VPD?

[50] The answer is that Bowen ratio increases with net radiation.

[51] The answers to the indirect-effect questions are as follows:

[52] 4. How does soil moisture affect the relationship between VPD and surface energy partitioning and that between net radiation and surface energy partitioning?

[53] The answer is that reduced soil moisture weakens the influence of VPD but enhances the influence of net radiation on surface energy partitioning. Soil moisture also controls how net radiation influences the relationship between surface energy partitioning and VPD. When the soil is wet, increased net radiation enhances the positive effect of VPD on latent heat flux; however, when the soil is dry, increased net radiation enhances the negative effect of VPD on latent heat flux. Furthermore, how VPD affects the relationship between surface energy partitioning and net radiation depends on soil moisture. When the soil is wet, the increase in VPD dampens the increase of Bowen ratio with net radiation; when the soil is dry, the increase in VPD enhances the increase of Bowen ratio with net radiation at high levels of VPD.

[54] 5. How does VPD affect the relationship between soil moisture and surface energy partitioning and that between net radiation and surface energy partitioning?

[55] The answer is that increased VPD enhances the sensitivity of Bowen ratio to changes in soil moisture. Low VPD coupled with low net radiation may cause Bowen ratio to change very little with soil moisture even when the soil is dry. As VPD increases, however, Bowen ratio decreases faster with increasing soil moisture and the level of soil moisture above which Bowen ratio is insensitive to soil moisture becomes higher. Consequently, at high VPD, the effect of drought on the surface energy partitioning is much stronger. The answer to the question how VPD affects the relationship between net radiation and surface energy partitioning is included in the answer to the Question 4.

[56] 6. How does net radiation affect the relationship between soil moisture and surface energy partitioning and that between VPD and surface energy partitioning?

[57] The answer is that increased net radiation enhances the sensitivity of Bowen ratio to changes in soil moisture. Low net radiation coupled with low VPD may cause Bowen ratio to change very little with soil moisture even when the soil is dry. As net radiation increases, however, Bowen ratio decreases faster with increasing soil moisture and the level of soil moisture above which Bowen ratio is insensitive to soil moisture becomes higher. Furthermore, increased net radiation enhances the effect of drought on the surface energy partitioning. The answer to the question how net radiation affects the relationship between VPD and surface energy partitioning is included in the answer to the Question 4.

[58] Our answers to the six direct and indirect-effect questions have been developed on the basis of data from a single vegetation type (a deciduous forest with a closed canopy). Therefore it is legitimate to ask how general these answers are with respect to different vegetation types and canopy structures. In developing the answers to these questions, we did not have to invoke species-specific properties; general ecophysiological and micrometeorological principles appear to be sufficient for interpreting the observed patterns (section 4). Thus we expect that our answers will be at least qualitatively applicable to other vegetation types. However, species have diverse behaviors of stomatal dynamics and different rooting depths and levels of drought tolerance; plant canopies may have different leaf area indices and architectures. These idiosyncrasies may well mean that the answers we developed here will have to be revised quantitatively for different vegetation types and land surface characteristics. For example, as vegetation structure varies from dense to sparse, one can expect VPD plays increasingly more a role of a measure of atmospheric evaporative power than that of a factor regulating stomatal conductance, thus changing the direct and indirect effects of VPD on surface energy partitioning. Only through analyses of actual data from different sites can we know definitively how general our answers to the six questions are.

[59] Our finding that multiple soil and atmospheric factors have both direct and indirect effects on land surface energy partitioning has important implication for understanding many boundary layer phenomena. For example, Rabin et al. [1990] observed that when the atmosphere is relatively dry, clouds develop first over adjacent drier areas, but when the atmosphere is relatively moist, clouds form earlier over more moist areas. In contrast, Ray et al. [2003] showed that clouds occur more frequently over more moist areas regardless of atmospheric humidity. The apparent discrepancies between these observations may result from the complicated direct and indirect effects of soil and atmospheric conditions on land surface energy partitioning.

[60] Our results also provide an opportunity for the atmospheric modeling community to improve the representation of the coupling between land surface and the atmosphere in current atmospheric general circulation models (AGCMs). Current AGCMs show a wide range of land-atmosphere coupling strengths [Koster et al., 2004] and the causes for this large variability are not clear [Lawrence and Slingo, 2005]. The existence of this variability and our inability to account for it indicate that our current understanding of processes controlling land-atmosphere interactions is still poor. To advance, AGCM modelers could examine their simulation results to see whether their AGCMs are capable of reproducing the direct and indirect effects of soil moisture and atmospheric conditions on land surface energy partitioning that we reported in this paper. This exercise could yield important information on not only the causes for the large variability in the modeled land-atmosphere coupling strengths but also how the representations of land-atmosphere interactions and convective processes can be improved in current AGCMs [Arakawa, 2004]. Ideally all the direct and indirect effects identified here should be reproduced by AGCMs. At a minimum, AGCMs should strive to match the observed effects of soil moisture on surface energy partitioning. Soil moisture is a variable that contains memories of the past and changes systematically. Thus errors caused by inaccurate representation of effects of soil moisture may accumulate overtime and lead to eventual breakdown of AGCM predictions.

Acknowledgments

[61] We thank Roger A. Pielke, Dennis D. Baldocchi, and Qing Liu for commenting on an earlier draft of this paper. We also acknowledge contributions from two anonymous reviewers who challenged us to think deeper and make maximal use of our data. In particular, Figure 2 resulted directly from suggestions made by one reviewer whose prediction was almost in perfect agreement with our data. Because of their specific suggestions, we are able to provide more insightful answers to our questions. This paper is a contribution to the Missouri Ozark AmeriFlux Project (MOFLUX, 3ERKP483), a joint effort between ORNL, University of Missouri, and NOAA/ATDD. MOFLUX is supported by the U.S. Department of Energy, Office of Science, Biological and Environmental Research Program, Environmental Science Division. ORNL is managed by UT-Battelle, LLC, for the U.S. Department of Energy under the contract DE-AC05-00OR22725. U.S. Department of Energy support for the University of Missouri (grant DE-FG02-03ER63683) is gratefully acknowledged.

Ancillary