On the impact of shrub encroachment on microclimate conditions in the northern Chihuahuan desert



[1] Changes in vegetation cover are known for their ability to modify the surface energy balance and near-surface microclimate conditions. A major change in vegetation composition that has been occurring in many dryland regions around the world is associated with the replacement of arid grasslands by desert shrublands. The impact of shrub encroachment on regional climate conditions remains poorly investigated, and, to date, it is unclear how this shift in plant community composition may affect the microclimate. Here we used concurrent meteorological observations at two adjacent sites dominated by Larrea tridentata shrubs and native grass species, respectively, in the northern Chihuahuan desert to investigate differences in nighttime air temperatures between the shrubland and grassland vegetation covers. The nighttime air temperature was found to be substantially higher (>2°C) in the shrubland than in the grassland, especially during calm winter nights. These differences in surface air temperature were accompanied by differences in longwave radiation and sensible and ground heat fluxes. We developed a one-dimensional model to show how longwave radiation emitted by the ground at night can explain the higher nighttime air temperature over the shrubland. Because of the larger fraction of bare soil typically existing in the shrub cover, the ground surface remains less insulated and more energy flows into the ground at the shrubland site than in the grassland during daytime. This energy is then released at night mainly as longwave radiation, which causes the differences in the nighttime air temperatures between the two land covers.

1. Introduction

[2] Vegetation cover is known for its ability to influence the exchange of energy and water vapor between the land surface and the atmosphere [e.g., Fraedrich et al., 1999; Bonan, 2002]. Vegetation affects the amount of solar irradiance that is reflected by the earth surface, the partitioning of net radiation into sensible and latent heat fluxes, the rate of precipitation recycling, the entrainment of dust and other aerosols into the atmosphere, and the partitioning of precipitation into soil moisture, evapotranspiration, and runoff [Charney, 1975; Eltahir and Bras, 1996; Foley et al., 2005; Rosenfeld et al., 2001; Villegas et al., 2010]. Many studies have investigated and quantified the effects of land cover change on climate. Charney [1975] and Charney et al. [1977] first discovered the link between changes in albedo and aridification in West Africa. Replacement of seasonal forests and grasslands with desert conditions has been associated with the reduction of evapotranspiration and precipitation in the Sahel region of Africa [Xue and Shukla, 1993] and in Mongolia [Xue, 1996]. Other major land cover changes have been investigated for their potential impact on the regional climate [e.g., Bonan, 2002; Rotenberg and Yakir, 2010]. For example, in the United States the replacement of temperate forests with croplands has been shown to modify climatic conditions through its impact on albedo, stomatal conductance, surface roughness, and root depth [Bonan, 1997, 1999]. Similarly, tropical deforestation has been associated with major changes in regional temperature and rainfall regimes [Shukla et al., 1990], while the encroachment of boreal forests into regions historically dominated by tundra vegetation has been shown to lead to warming through important vegetation-albedo feedbacks [Foley et al., 1994].

[3] Woody plant encroachment into grassland ecosystems has been happening in many regions around the world over the past several decades [Van Auken, 2000; Knapp et al., 2008]. Within the southwestern United States, this dramatic shift in plant community composition has been particularly well documented for the Sonoran and Chihuahuan deserts [Buffington and Herbel, 1965; Archer et al., 1988; Archer, 1989; Van Auken, 2000, 2009]. Despite its global significance and its recognized impact on ecosystem function and services, this change in plant community composition has seldom been investigated with respect to its effect on near-surface temperature. Bhark and Small [2003] investigated the effect of vegetation and shrub cover on soil moisture conditions; Kurc and Small [2004, 2007] studied soil moisture and evapotranspiration in adjacent grassland and shrubland, while Dugas et al. [1996] examined the energy balance components over different grass and shrub species in the Chihuahuan desert. Scott et al. [2006] investigated the impact of woody plant encroachment on energy, carbon and water vapor fluxes and found that, while woody plants can take up more carbon during the dry season because of the better use of groundwater, this effect may be offset by a larger soil respiration rate during the wet season. Beltran-Przekurat et al. [2008] used numerical simulations to show that changes in land cover from grass-dominated to shrub-dominated arid landscapes in the Chihuahuan desert cause an overall decrease in sensible heat fluxes and a substantial increase in latent heat fluxes, thereby leading to a cooler and moister near-surface atmosphere during daytime. However, it is still unclear whether shrub encroachment has any impacts on nocturnal climatic conditions, and whether these effects have a feedback on vegetation dynamics. The two major shrub species in the Chihuahuan desert (i.e., Larrea tridentate and Prosopis glandulosa) are known for being sensitive to low nocturnal temperatures [Felker et al., 1982; Pockman and Sperry, 1997]. Therefore, by changing the nocturnal temperature conditions, these shrub species may affect their own survival in this area.

[4] The encroachment of mesquite (Prosopis glandulosa) and creosotebush (Larrea tridentata) in North American deserts, has been explained as an effect of large-scale drivers such as “climate warming” [e.g., Pockman and Sperry, 1997] or increased atmospheric carbon dioxide (CO2) concentrations [e.g., Polley et al., 1992], and of local controls (e.g., grazing and fire management) [Archer et al., 1995], which operate through important feedbacks between biotic and abiotic processes [e.g., Buffington and Herbel, 1965; Archer, 1990, 1994]. We argue that in the case of Larrea tridentata one of the factors contributing to this shift in vegetation composition is the change in near-surface microclimate conditions caused by the replacement of grass cover with woody vegetation. This dependence on microclimate conditions would be consistent with the observation that Larrea encroachment is sensitive to extreme negative temperatures [Pockman and Sperry, 1997]. Thus, shrub establishment in relatively warm years could modify the surface energy balance and provide warmer nocturnal microclimate conditions thereby favoring the survival of Larrea during cold winters.

[5] Hayden [1998] and Carre [2005] compared temperature measurements taken on two land covers in the northern Chihuahuan desert. They found that in the shrubland, the nighttime minimum temperatures were higher (by 4°C–6°C) than in the adjacent grasslands. This warming effect is likely to enhance the chances of establishment and survival of these cold-sensitive shrub species. These results, which are in agreement with long-term temperature records from nearby areas in the Southwestern United States [Balling, 1988; Bryant et al., 1990; Balling et al., 1998; Small and Kurc, 2003], need to be validated by field observations capable of explaining the underlying processes. Therefore, the objective of this paper is to explain the difference of nighttime temperatures over grass-dominated and shrub-dominated areas and identify the salient processes causing the climatic differences observed between these two land covers. Understanding these processes will be necessary, for example, if we want to parameterize the effects of shrub encroachment on the near-surface atmosphere in climate models.

2. Methods

2.1. Study Site

[6] To investigate the effect of shrub encroachment on near-surface microclimatic conditions, we compared surface temperatures and energy fluxes at grassland and adjacent shrubland sites. To this end, we used data from the Sevilleta National Wildlife Refuge, located in the northern Chihuahuan desert of the Rio Grande Valley, approximately 80 km south of Albuquerque, New Mexico. Because the Sevilleta National Wildlife Refuge shows a dramatic encroachment front of Larrea tridentata (creosotebush) shrubs into native desert grassland, it represents an ideal location to investigate differences in surface energy flows associated with the two different land covers, both of which exist under the same regional climate conditions.

[7] Concurrent measurements were made on two identical flux towers currently deployed over a Larrea-dominated shrubland and a Bouteloua-dominated grassland as part of the Long-Term Ecological Research (LTER) program. The grassland (34.3402°N, 106.6854°W) and shrubland (34.3338°N, 106.7340°W) tower sites are in the McKenzie Flats area of the Sevilleta National Wildlife Refuge. The distance between the towers is about 5 km, and the elevation difference is less than 10 m. The Sevilleta Refuge contains extensive semi-arid grassland dominated by C4 perennial grasses (Bouteloua gracilis, B. eriopoda, Sporobolis spp., Hilaria jamesii, Muhlenbergia spp.) located on relatively level topography along the western edge of the Los Piños Mountains. In the grassland site the total vegetation cover (live plus litter) averages 60% with 40% bare soil, while in the shrubland site the average vegetation cover is 30% with 70% of bare soil [Kurc and Small, 2004].

[8] Temperature, humidity (with HMP45C Vaisala temperature/RH probe, 0.2°C–0.35°C accuracy), and wind speed were measured (with CSAT3 sonic anemometer) at a height of 3 m above ground level since January 2007. The data we used here cover the period from July 2007 to June 2008. Turbulent energy fluxes were derived using the eddy covariance technique at 3 m height above ground (CSAT3 sonic anemometer and LI-7500 open-path IRGA). The flux data represent 30 min averages. In addition, measurements of radiation components (CNR1 four-way Kipp&Zonen net radiometer) and pressure (CS105 Vaisala PTB101B barometer) were taken at 3 m height above ground, while soil heat fluxes were measured in the ground at 5 cm depth (HFT3 REBS soil heat flux plates). Days were considered to be “clear” when the ratio of daily incoming shortwave radiation to theoretically determined extraterrestrial solar radiation was greater than 64% [Whiteman et al., 1999]. Nighttime was defined as those hours with zero incoming shortwave radiation. Wintertime was defined as the season from 1 November 2007 to 29 February 2008, while summertime included the months of May, June, July and August.

2.2. Surface Aerodynamic Characteristics

[9] To explain the differences in surface energy flows between the two vegetation types, we determined the effect of landscape conditions on surface aerodynamic characteristics such as displacement height and roughness length for momentum sink. Using the logarithmic wind speed profile and the friction velocity (estimated from the momentum flux), the roughness length for momentum sink, z0, and the displacement height, d, were estimated following the procedure described by Martano [2000]. As expected, both the roughness length and the displacement height were larger in the shrubland than in the grassland (Table 1).

Table 1. Information on Vegetation Cover, Aerodynamic Parameters, Albedo, and Emissivity for the Shrubland and Grassland Sites
Surface CharacteristicsShrublandGrassland
Vegetation cover (%)3060
Displacement height, d (m)0.30
Roughness length, z0 (m)0.04–0.060.03
Winter albedo, α0.2040.209
Summer albedo, α0.2010.184
Winter emissivity, ɛsfc  
   All day0.9630.964
Summer emissivity, ɛsfc  
   All day0.9690.965

2.3. Surface Emissivity and Albedo

[10] To investigate the processes governing the transfer of energy in the grassland or the shrubland site, we calculated surface emissivity and albedo. For the emissivity (ɛsfc) we used the Stefan-Boltzmann law, expressing the longwave radiation (Lwsfc) emitted by the ground surface as

equation image

where σ is the Stefan-Boltzmann constant and Tsfc is the ground surface temperature. Since Tsfc was not measured in this experiment, the aerodynamic method was employed to calculate Tsfc from measurements of sensible heat flux density (H). Using the Ohm's law analogy, over horizontal homogenous vegetation canopy, H can be expressed as a function of Tsfc and Tair. Solving for Tsfc, we obtain

equation image

where ρ is the density of air, cp is the specific heat of air at constant pressure, and RH is the resistance to heat transfer and consists of two components,

equation image

where RbH is the bulk boundary layer resistance and RAero is the aerodynamic resistance. RbH is expressed in terms of atmospheric turbulence levels and the intrinsic characteristics of heat transfer [Wesely and Hicks, 1977]:

equation image

where κ = 0.4 is the von-Karman constant, u* is the friction velocity, Pr is the Prandtl number (≈0.72) for air, and Sc is the Schmidt number. The Schmidt number is calculated as the ratio of kinematic viscosity to mass diffusivity with temperature dependence [Campbell and Norman, 1998]. The aerodynamic resistance RAero is a function of momentum transfer expressed as

equation image

where z is the height above the ground surface and Ψm((zd)/L) is the diabatic function [Businger et al., 1971; Dyer, 1974] expressed in terms of the Monin-Obukhov length (L). Therefore, with equations (1)(5), using the observed upwelling longwave radiation as Lwsfc, the emissivity is expressed as a function of air temperature and turbulence measurements as

equation image

where u is the wind speed measured at the same height as Tair, u*, and H.

[11] The surface albedo (α) was estimated to quantify the amount of solar energy absorbed by the surface in the presence of grass and shrub cover. The albedo was calculated by

equation image

where Kup and Kdn are the upwelling and downwelling shortwave radiation, respectively. These radiation components were measured by pyranometers at 3 m height. Only data from 0900 to 1600 Local Standard Time (LST) were used in the calculation of α to avoid the large uncertainties associated with low solar elevation angles. Values of emissivity and albedo are reported in Table 1.

2.4. Atmospheric Stability

[12] Different rates of surface cooling can occur owing to differences in atmospheric stability. In this study the gradient Richardson number (Ri) was used to assess the impact of stability on rates of surface cooling. The gradient Richardson number represents the relative contribution of buoyancy and wind shear to the production/destruction of turbulence [e.g., Stull, 1988],

equation image

where g is the gravitational acceleration and θ represents the potential temperature. The gradient Richardson number is related to the Monin-Obukhov length (L) through the Businger-Dyer formulas [Businger et al., 1971; Dyer, 1974]; large positive values of Ri correspond to stable conditions with weak turbulent mixing.

3. Results

3.1. Observations

[13] We analyzed near-surface air temperatures to assess differences in minimum winter temperatures between the two land covers. Nighttime temperatures in the shrubland were higher than those in the grassland at most times (Figure 1). These differences were particularly strong in the winter months, when the mean nighttime air temperatures in the shrubland were on average about 2°C higher than those in the grassland (Figure 2).

Figure 1.

Nighttime temperature comparison between shrubland and grassland, for the whole year and in the winter months only (inset). Analysis is based on July 2007 to June 2008 half hourly points (November 2007 to February 2008 data in the inset). The nighttime air temperature in the shrubland is significantly higher than in the grassland (p < 0.0001 both for the t test of the whole year and of the winter months). For 72% of the time (or 79% of the time if only winter months are considered) the shrubland has warmer near-surface conditions than the grassland does.

Figure 2.

Monthly mean and standard deviation of temperature differences between the two vegetation covers in the winter months (November 2007 to February 2008).

[14] Temperature differences were established between 1700 and 1900 LST and were maintained throughout the night (Figure 2). Diurnal patterns of temperature differences between the shrubland and the grassland were similar throughout the year. Maximum nighttime differences between near-surface air temperatures over the two land covers (Figure 3) ranged between 3°C and 7°C. The temperature differences between the shrubland and the grassland sites depended on near-surface stability (i.e., the Richardson number, Ri) and wind speed (Figure 4). Temperature differences increased when the surface layer became more stable (i.e., for large values of Ri; Figure 4a) or when the wind speed decreased (Figure 4b). Under these conditions, mixing was relatively weak so that the microclimate in the shrubland and the grassland was not strongly affected by air advected from the surrounding areas, but remained controlled mainly by local conditions [Geiger, 1965]. Therefore, the shrubland and the grassland can create and maintain their own microclimate, particularly during calm nights. The overall implication of this analysis is that shrub encroachment leads to warmer near-ground nighttime conditions, especially in the winter season, when shrubs are able to maintain milder microclimate conditions. Warmer winter night conditions, in turn, may favor the survival of shrub species owing to their lack of tolerance to freezing temperatures [Pockman and Sperry, 1997].

Figure 3.

Monthly mean and standard deviation of maximum daily nighttime temperature differences between the two vegetation covers.

Figure 4.

(a) Relationship between nighttime temperature differences and Richardson number under stable conditions and (b) relationship between nighttime temperature differences and wind speed for the whole year (July 2007 to June 2008) and winter months (November 2007 to February 2008).

[15] The differences in near-ground temperature between the two land covers can be explained by differences in energy fluxes and surface energy balance. The average values of each measured energy component, including sensible heat fluxes (H), ground heat fluxes (G), upwelling and downwelling longwave radiation (Lwup and Lwdn, respectively), and the net longwave radiation (Lwnet), are reported in Table 2 for the case of nocturnal wintertime conditions. To better relate surface energy fluxes to the differences in nocturnal temperature between the two land covers, we also constrained the analysis focusing in particular on (1) those nights in which shrubland temperatures exceeded by at least 2°C those in the grassland (“Constraint 1”) and (2) the time period in which the nocturnal air temperature differences were established (i.e., between sunset and 2000 LST; “Constraint 2”). These results are discussed in section 4.

Table 2. Nighttime Energy Components for Shrubland, Grassland, and the Difference Between Shrubland and Grassland, Reported as the Average Over Winter Months of November 2007 to February 2008a
Energy ComponentConstraintShrublandGrasslandDifference
  • a

    Constraint 1 conditions are defined as when temperature differences over the two land covers are equal to or larger than 2°C. Constraint 2 only concerns the measurements from sunset to 2000 LST.

H (W m−2)1−9.2−12.12.9
G (W m−2)1−17.3−18.71.5
Lwup (W m−2)1316.7307.59.2
Lwdn (W m−2)1247.8246.91.0
Lwnet (W m−2)1−68.9−60.7−7.7

3.2. Energy Budget Model

[16] To understand why nighttime temperature in the shrubland was higher than in the grassland and how differences in energy fluxes between the grassland and shrubland sites contributed to such temperature differences, we developed a one-dimensional energy budget model. The temporal variability of air temperature near the ground is driven by the surface energy budget. Assuming a negligible effect of horizontal advection during calm nights, the vertical energy transfer can be related to the temporal variability of the average air temperature within a near-surface air layer of thickness za through the energy balance equation:

equation image

which involves all the energy exchanges occurring at the upper and lower boundaries of the layer of air (Figure 5).

Figure 5.

Schematic representation of the one-dimensional energy budget model. The parameters are explained in text.

[17] The term Lwatm is the downward longwave radiation emitted by the atmosphere. This radiation is partly absorbed by the ground (ɛsfcLwatm), and partly reflected to the atmosphere ((1 − ɛsfc)Lwatm). Lwatm was estimated by the Stefan-Boltzmann law as a function of the air temperature and the emissivity of atmosphere (ɛatm). The other longwave radiation components in equation (9) include the longwave radiation emitted by the ground (Lwsfc), and the downwelling (Lwdn) and upwelling (Lwup) longwave radiation measured at 3 m height (Figure 5).

[18] The numerical integration of the energy balance equation (9) using energy flux measurements from the two sites allows us to calculate the average air temperature values within the near-surface air layer. Six clear winter nights with calm wind and pronounced temperature differences between grassland and shrubland sites were selected to test the model. To this end, we used values of ground surface emissivity calculated on the basis of winter daytime measurements so that surface emissivity can be considered an independent quantity. The performance of this model is affected by errors associated with measurements and parameter estimation. Moreover we made the approximation of using the near-ground air temperature to estimate the atmospheric longwave radiation Lwatm and we tested the model comparing the average temperature of a 3 m thick air layer with values of air temperature measured at 3 m height. The emissivity of the atmosphere (ɛatm) was therefore the only parameter that can be varied in our model. This parameter was estimated so that air temperatures calculated with this energy budget model best fit those measured during the six selected clear-sky, calm winter nights. As a result, we found that ɛatm = 0.88.

[19] Despite its approximated nature, this 1-D energy budget model reproduced the general pattern of temperature differences between shrubland and grassland observed in the experiment (Figure 6). Therefore, the energy budget model provides a process-based framework to explain the air temperature differences between the two land covers as an integrated effect of differences in all the energy fluxes.

Figure 6.

Comparison of the observed temperature differences between shrubland and grassland (solid line connecting solid circles) with the results of a one-dimensional energy budget model (solid line) in a selected sample case (the clear-sky calm winter night between 16 and 17 December 2007). Sensitivity test with respect to longwave radiation emitted by the ground surface is shown with the dashed line. (The same ground longwave radiation is used for the two vegetation covers. See text.) Each dotted line shows the result of a sensitivity test with respect to the other components of the energy balance, including (1) sensible heat fluxes, (2) ground heat fluxes, (3) measured upwelling, and (4) downwelling longwave radiation.

[20] In addition, this one-dimensional energy budget model allows us to examine the sensitivity of these temperature differences to different terms of the energy balance. To test this sensitivity, we replaced each energy flux component with the mean of values measured over the two land covers while keeping the other terms of the energy balance unchanged. As shown in Figure 6, differences in nocturnal air temperature are not very sensitive to changes in sensible heat fluxes, ground heat fluxes, measured upwelling or downwelling longwave radiation. In fact, changes in these energy components do not result in significant changes in the patterns of the calculated air temperature difference between the two land covers. Conversely, changes in ground surface temperature (hence in Lwsfc) lead to a major change in the patterns of nighttime air temperatures over the two land covers. Therefore, differences in ground longwave radiation between grassland and shrubland are the major contributor to the higher air temperatures observed at night in the shrubland with respect to the grassland. This result is consistent with the mechanistic explanation of differences in nocturnal air temperature presented in the discussion.

4. Discussion

[21] Diurnal surface temperature variation is controlled by factors involving the energy balance at the surface. The energy balance involves the net radiation, the ground heat flux, and the turbulent fluxes of latent and sensible heat. A detailed investigation of these factors is presented to explain the main processes underlying the nocturnal temperature difference between the shrubland and grassland. The net radiation is contributed by net longwave and shortwave radiation, which are affected by the albedo and emissivity of the ground surface. Differences in albedo between the shrubland and grassland were minor (Table 1) and did not cause relevant differences in the net radiation balance during daytime. At night, the shrubland lost more net longwave radiation than the grassland. The upwelling longwave radiation at night was larger in the shrubland than in the grassland in all the three cases reported in Table 2. Combining these results with the smaller average nighttime emissivity in the shrubland than in the grassland (Table 1), we can conclude that, compared to grassland, the shrubland not only had a higher nighttime air temperature but also a higher nighttime ground surface temperature.

[22] The ground heat flux at each location was calculated as a weighted average of the values measured under the bare soil and vegetated microsites, using values of bare soil fractions typical of the two land covers (Table 1, shrubland: 70% bare soil, grassland: 40% bare soil [Kurc and Small, 2004]). Cumulative ground heat fluxes (i.e., the time integral of the ground heat fluxes throughout the day or night) were larger in absolute value in the shrubland than in the grassland. These differences became more pronounced in clear-sky conditions during the day (Figure 7), presumably because vegetation was less effective in causing the differences of received energy between shaded and nonshaded ground during cloudy days. To explain the differences in ground heat fluxes between the two land covers, we looked at the ground heat fluxes from bare soil and vegetated microsites at the grassland and shrubland sites. We found that the diurnal ground heat fluxes in bare soil microsites were larger than those in microsites covered by either shrub or grass vegetation (Figure 8). Moreover, daytime ground heat fluxes were greater in the shrubland than in the grassland both in the bare soil (with 19.2 W m−2 average difference for all clear winter days) and in the vegetated (with 9.2 W m−2 average difference for all clear winter days) microsites. The greater ground heat fluxes observed at both sites in the bare soil microsites with respect to the adjacent vegetated soil plots (average difference of 24.4 W m−2 for the shrubland and 14.4 W m−2 for the grassland for all clear winter days) provide an explanation for the overall greater ground heat fluxes at the shrubland site (Figure 7), where a greater bare soil fraction existed. The different bare soil fractions existing in the two sites resulted in relatively large differences in ground heat fluxes between the two land covers corresponding with the observation and explanation by Kurc and Small [2004]. At night, the differences in cumulative ground heat flux between shrubland and grassland were smaller than during daytime (on average −0.1 MJm−2 at night and 0.8 MJm−2 during day for all clear winter days, Figure 7), consistently with the smaller difference in nocturnal ground heat fluxes observed between bare soil and vegetation microsites at both locations (Figure 8).

Figure 7.

Comparison between cumulated ground heat fluxes measured between 1 November and 31 December 2007 at the shrubland and grassland sites. Inset shows mean and standard deviation of ground heat fluxes over shrubland and grassland in clear days in the same period. At each site, ground heat flux values were calculated as weighted averages of the values measured at vegetated and unvegetated microsites using the fractional land cover reported for each site.

Figure 8.

Mean and standard deviation of ground heat fluxes under bare soil and vegetation cover in (a) the shrubland and (b) the grassland, in clear winter days (November 2007 to February 2008).

[23] Diurnal patterns of ground heat fluxes over the shrubland and the grassland showed a lag of about 2 hours between the daytime peaks in ground heat flux for the two land covers. In fact, in the shrubland the maximum ground heat flux occurred at 1300 LST, 2 hours earlier than in the grassland (Figure 7, inset). Thus, shrubland soils responded more quickly to daytime warming than grassland soils, owing to the larger bare soil fraction in the shrubland and the lack of insulation of the ground surface by grass biomass. In fact, at the patch scale we found that a lag of about 2 hours exists between the peaks of ground heat fluxes measured in vegetated and bare soil microsites (Figure 8). The downward sensible heat fluxes were greater in absolute value (>1.9 W m−2 on average) for the grassland than the shrubland (Table 2) consistently with the more rapid cooling of the air above the grassland.

[24] In summary, on the basis of this analysis of radiation and energy fluxes, we can explain the emergence of nighttime temperature differences between the two land covers as follows. In the shrubland, where a larger fraction of bare soil typically exists, the soil surface was poorly insulated by vegetation. Thus, more energy was received by the underlying soil column during the day in the form of ground heat fluxes. Moreover, owing to the limited insulation of the soil surface, soil heating occurred more rapidly in the shrubland than in the grassland. This energy was then released at night in the form of longwave radiation. Because differences in ground surface emissivity between grassland and shrubland were negligible, the higher nocturnal longwave radiation emitted by the shrubland was due to differences in soil surface temperatures. Similar findings were reported for daytime conditions at the same shrubland and grassland research sites [Small and Kurc, 2003; Kurc and Small, 2004]. Therefore, differences in bare soil fraction caused a differential diurnal heating of the soil in the two land covers. Thus, at night more energy was released from the ground in the shrubland than in the grassland, contributing (as indicated by simulations with the energy balance model) to the differential heating of the near-surface air, thereby causing the observed differences in nocturnal air temperatures between shrubland and grassland. These differences were particularly strong during calm nights, and their persistence throughout the night was favored by the relatively stable boundary layer conditions.

[25] The explanation of air temperature differences based on the different bare soil fractions existing on the two land cover is consistent with other observations from the northern Chihuahuan desert [Carre, 2005] and with similar findings from the Sonoran desert, where differences in air temperature of about 4°C were detected across the Mexico–United States border, and explained as a result of the higher bare soil fractions due to the heavy overgrazing on the Mexico side [Balling, 1988; Bryant et al., 1990; Balling et al., 1998]. However, to our knowledge this effect of warming induced by an increase in bare soil had never been explained before in the context of its relation with shrub encroachment. Although warming may also have a positive effect on grass growth, in the case of Larrea-encroached landscapes the increase in minimum temperature associated with grassland-to-shrubland conversion is likely to favor only the establishment of these freeze-sensitive shrub species, because grasses remain dormant during wintertime.


[26] This research was supported by NSF-DEB 0743678 and the NSF-DEB-0620482 to the University of New Mexico for Long-Term Ecological Research. We would like to thank the anonymous reviewers for their constructive comments, which improved the manuscript.