On the coupling strength between the land surface and the atmosphere: From viewpoint of surface exchange coefficients

Authors


Abstract

[1] This study addresses the land-atmospheric coupling strength by using long-term AmeriFlux data from a wide range of land covers and climate regimes to reconstitute the surface exchange coefficient, Ch, which governs the total surface heat fluxes. For spring and summer, results show stronger coupling for tall canopy with Ch values ten times larger than for shorter vegetation. Observed Ch are then compared to values from the Noah land model. Results indicate that Noah underestimated (overestimated) Ch for forest (grass and crops), implying an insufficient (too efficient) coupling for tall canopy (short canopy). This discrepancy is attributed to the treatment of the roughness length for heat. With modest adjustments, the Noah model can reproduce the observed Ch. This study highlights the crucial role of treating the surface exchange processes in coupled land/weather/climate models and the need to use long-term flux data for different vegetation types and climate regimes to assess and mitigate their deficiencies.

1. Introduction

[2] Land-atmospheric interactions (e.g., feedback between soil moisture and precipitation) may hold the key for improving the predictability of weather and climate [e.g., Betts et al., 1996; Pielke et al., 1999; Chen et al., 2001; Trier et al., 2004; Los et al., 2006]. In particular, analysis of simulations using coupled land-surface/climate models by Koster et al. [2004] revealed several “hot spots” in terms of strong coupling between soil moisture and summer rainfall. Such studies, however, depend on the reliability of land surface models (LSMs) in predicting the strength of surface-atmosphere coupling, as expressed by the surface exchange coefficients. For example, Ruiz-Barradas and Nigam [2005] argued that excessively land-atmospheric coupling in models (manifested in too large latent heat flux) might lead to an incorrect relationship between soil moisture and precipitation, and the results of Zhang et al. [2008] did not support the hot spot hypothesis of Koster et al. [2004] for the central Great Plains. These results underline the critical importance of representing land-atmospheric interactions in atmospheric models and naturally raise a question: what is the right coupling between the land surface and the atmosphere?

[3] Although the coupling issue was investigated in previous studies [Ruiz-Barradas and Nigam, 2005; Dirmeyer et al., 2006; Zhang et al., 2008], only the relationship between soil moisture, evaporation, and precipitation was assessed. Different ways of characterizing the land-atmospheric feedback were proposed and included, for instance, the calculation of feedback (recycling) numbers based on atmospheric budget analysis [Trenberth, 1999] and the diagnosis of a coupling coefficient from ensemble model experiments [Koster et al., 2004]. However. these methods, in general, heavily relied on model or reanalysis results. The more fundamental coupling issue regarding the efficiencies of exchanging energy and water vapor between the land surface and the atmosphere still remains poorly understood. Therefore, the main objective of this study is to develop a framework based on analysis of surface exchange coefficients to address this issue by employing long-term observations at diverse locations. As a first step, this paper investigates two specific questions: 1) what is the coupling strength between land surface and the atmosphere over different land-cover types? and 2) how well is this coupling represented by the Noah land surface model (LSM), which is widely used for examining land-atmospheric interactions in regional and global models?

2. Methods and Observational Evidences

[4] The AmeriFlux network, established in 1996, provides continuous observations of surface fluxes of water vapor and energy and currently comprises measurement sites from North America, Central America, and South America. Our goal is to explore the land-atmospheric coupling in spring and summer during vegetation growing season, when the land surface plays a more prominent role in transporting heat and water vapor to the atmosphere due to higher incoming solar radiation and photosynthetically active vegetation. After a careful inspection of data quality and length (at least two years of data) for variables required (surface energy budgets and near-surface weather variables) in our study, we selected 12 AmeriFlux sites spanning different land-cover types (snow, cropland, grassland, shrub, forest) and climate regimes (wet, semi arid and arid regions). Figure 1 shows the geographical locations of these sites and the general information about these sites is given in Table 1. More information about the AmeriFlux network can be found at http://public.ornl.gov/ameriflux/.

Figure 1.

Locations of 12 AmeriFlux sites (in dark circles) selected for this study. Also shown is the distribution of vegetation based on the IGBP/MODIS land cover classification.

Table 1. General Information About the 12 AmeriFlux Sites Used in This Study
Site LocationLatitude, LongitudeElevation (m)Land-Cover TypeCanopy Height (m)Years of Data Used
Ivotuk (AK)68.49, −155.75568open shrub0.12004, 2005, 2006
Brookings (SD)44.35, −96.83510temperate grass0.2 to 0.42005, 2006, 2007
Audubon Research Ranch (AZ)31.59, −110.511469desert grassland0.1 to 0.22004, 2005, 2006, 2007
Fort Peck (MT)48.31, −105.10634grass0.2 to 0.42004, 2005, 2006, 2007
Kendal Grassland (AZ)31.74, −109.941531warm C4 grass0.52005, 2006, 2007
Vaira Ranch (CA)38.41, −120.95129grazed C3 grass0.55 ± 0.122004, 2005, 2006, 2007
ARM SGP Main (OK)36.61, −97.49311winter wheat0 to 0.52003, 2004, 2005, 2006
Mead (NE)41.16, −96.47362maize-soybean rotation2.92002, 2003, 2004, 2005
Bondville (IL)40.01, −88.29219maize-soybean rotation3.0 (maize), 0.9 (soybean)2003, 2004, 2005, 2006
Flagstaff (AZ)35.09, −111.762215ponderosa pine forest182006, 2007
Niwot Ridge (CO)40.03, −105.553050subalpine coniferous forest11.52002, 2003, 2006, 2007
Ozark (MO)38.74, −92.20219.4oak hickory forest24.22005, 2006, 2007

[5] One primary function of LSMs is to provide sensible heat flux (SH) and latent heat flux (or surface evaporation, LE) as lower boundary layer conditions to the coupled atmospheric models. These fluxes are responsible for driving the diurnal evolution of the boundary layer, modifying its stability, and subsequently affecting the formation of clouds and precipitation. Surface heat fluxes are, as a common practice in LSMs, related to mean properties of the flow through the use of bulk transfer relations [Garratt, 1992] such as

equation image
equation image

where ρ is the air density and Cp the air heat capacity. The atmospheric quantities ∣Ua∣ (wind speed), θa (air potential temperature), and qa (air specific humidity) are evaluated at the lowest model level or at a specific measurement height above the ground, while θs and qs are surface values. Ch and Ce are the surface exchange coefficient for heat and moisture, respectively, which are directly related to the coupling strength. LeMone et al. [2008] pointed out that modifying these surface exchange coefficients allows modeled SH and LE to match observations, while changing soil moisture and vegetation phenology input only has minor impact on LSM performance. Therefore, our approach is to explore the land-atmospheric coupling strength through examining the variations of Ch and Ce for different land-cover types and climate regimes because they control the total amount of energy going back to the atmosphere. In this investigation, we make the common assumption that CeCh and hereafter focus on Ch.

[6] Instruments at AmeriFlux sites directly provide SH and ∣Ua∣; θa and θs are calculated from observed air temperature and outgoing longwave radiation flux. Note that some AmeriFlux sites provide SH at several levels above the ground, but we elect to use the data measured at/above the canopy top, because they are more representative of the energy transported to the atmosphere. Ch can be reconstituted by using equation (1) and AmeriFlux 30-minute data, and then averaged from 1000 to 1500 local time to obtain midday values, similar to the analysis of Trier et al. [2004].

[7] Figure 2 shows the reconstituted midday values of Ch for the 12 AmeriFlux sites through spring (March–April–May) and summer (June–July–August) for the years documented in Table 1. For the Ivotuk, Alaska, site the land cover changes from the predominant snow in spring to low shrubs in summer, leading to a large increase in summertime Ch due to the rougher surface. That aside, the values of summer Ch are comparable to or slightly larger than that for spring, and spring Ch varies more than summertime values. More importantly, the variability of Ch across land cover types becomes immediately clear and can be roughly divided into three categories in order of increasing Ch: very smooth surface (snow), short vegetation (grass, crop, shrub), and tall vegetation (forests). Tall vegetation has rougher surfaces, larger Ch, and hence stronger coupling. For instance, Ch for forests can be 10 times larger than that for short vegetation (crops, grassland, and shrubs).

Figure 2.

Ch (plotted at log10 scale) derived from AmeriFlux observations for different land-cover types. These are midday (1000–1500 LST) values and averaged for spring (March–April–May) and summer (June–July–August). The median values of spring (summer) average Ch are represented by triangles (stars). The blue (cyan) bars comprise 75% of all midday values Ch for spring (summer) for each site.

3. Evaluation and Discussion of the Noah Model Results

[8] Because the Noah LSM [Chen and Dudhia, 2001; Ek et al., 2003] has been widely used in mesoscale and global models for investigating the feedback between soil moisture and precipitation [Chen et al., 2001; Koster et al., 2004; Trier et al., 2004, 2008; Dirmeyer et al., 2006; Zhang et al., 2008], we will next evaluate its Ch calculation. The Noah model uses an extension of the similarity-theory-based stability functions of Paulson [1970] to calculate Ch [Chen et al., 1997], which uses different roughness lengths for momentum (zom) and for heat (zot). It is well documented that zot is different from zom because heat and momentum transfer are determined by different resistances and mechanisms in the roughness layer [e.g., Brutsaert, 1982; Sun and Mahrt, 1995]. In Noah, zot is related to zom as a function of atmospheric flow proposed by Zilitinkevich [1995]:

equation image

where k = 0.4 is the von Kármán constant, and Re is the roughness Reynolds number. Czil is an unknown empirical coefficient and currently specified as 0.1 in Noah based on calibration with field data measured over grassland [Chen et al., 1997]. For a given AmeriFlux site, we assuming zom is 7% of the canopy height [Molder and Lindroth, 1999], and 30-minute fluxes, air temperature, humidity, pressure, and wind speed measured at the site are used to obtain Ch.

[9] Figure 3 shows midday Ch calculated by Noah and averaged for spring and summer. Comparing to observation-derived values in Figure 2, the modeled Ch has much smaller variability across land-cover types. It illustrates two deficiencies in Noah: overestimating Ch for short vegetation and substantially underestimating it for tall vegetation. This finding seems to agree with Ruiz-Barradas and Nigam [2005] in that LSMs may have an overly strong coupling and hence provide too much water vapor for the U.S. Great Plains, where the short vegetation (grass and crops) is predominant. On the other hand, land surface models may significantly underestimate the coupling for forested regions.

Figure 3.

Same as in Figure 2 but for Ch calculated by the Noah LSM.

[10] There is a rich literature investigating the complex relationship of zot/zom, also known as parameter kB−1 = ln(zom/zot) in the agricultural and boundary layer community [e.g., Duynkerke, 1992; Stewart et al., 1994; Troufleau et al., 1997; Verhoef et al., 1997]. Recent numerical simulations demonstrated the important role of zot in land surface modeling, boundary layer development, and summer convective initiation [LeMone et al., 2008; Trier et al., 2004]. Hence, we also tested the Brutsaert [1982] approach to calculating zot a) for smooth surfaces (e.g., snow, ice): zot = 0.395 ν/u*; b) for bluff-rough surfaces and short vegetation: zot = 7.4 zom exp (−2.46 Re1/4); and c) for tall trees: zot = βzom (1/7 < β < 1/3). The other approach tested here is to still employ equation (3) but calibrate the constant Czil for each given site following the method of Chen et al. [1997]. For most sites, the calibrated summer Czil values are close to the spring values. Also note that Czil is close to zero for the Ozark forest site with the tallest canopy among the 12 sites, and it is argued that zot can be larger than zom (thus a zero or negative Czil) for tall forest [Molder and Lindroth, 1999]. Using the least-squares regression method, these calibrated Czil can be related to the canopy height h (in meters) as:

equation image

[11] Ch, calculated with the above methods, is shown in Figures 4a and 4b (only the median values are plotted for clarity). When compared to the results using the default Czil = 0.1 in Zilitinkevich's formulation, using Brutsaert's different zot formulations for different canopy types significantly improves Ch calculations for short and tall vegetation, only it underestimates Ch for crops. Using the Czil-h relationship in equation (4) produces results similar to Brutsaert's scheme. This exercise demonstrates that a modest adjustment in constant Czil values in equation (3) or in the treatment of zot can substantially alter or improve the land-atmospheric coupling strength for different land cover types. However, systematic research is still needed to understand the underlying physics and to improve the representation of Ch for different land-cover types and climate regimes.

Figure 4.

The median values of Ch (a) for spring season and (b) for summer season. Observations are represented by circles; for Ch calculated by Noah, using the default Czil = 0.1 are represented by stars; using the Czil-h relationship of equation (4) are represented by triangles; x symbols represent using Brutsaert scheme.

4. Conclusions

[12] This study has sought to develop a framework, using multiple-year observed surface flux and weather variables and the Noah LSM as an example, to assess the land-atmospheric coupling strength for different land-cover types and climate regimes. Multiple-year AmeriFlux data are used to reconstitute the surface exchange coefficients Ch for spring and summer seasons. Ch is a critical parameter controlling the total energy transported from the land surface to the atmosphere and directly reflects the land-atmospheric coupling strength. Observations show higher Ch and stronger coupling for tall vegetation than that for short vegetation, but the Noah model tends to overestimate Ch for short vegetation such as grass, shrubs, and crops. This seems to confirm the finding of Ruiz-Barradas and Nigam [2005] in that LSMs may be too efficiently coupled to the atmospheric models and lead to an overly strong soil moisture-precipitation feedback for the U.S. Great Plains, where the short vegetation (grass and crops) is predominant. Equally important and less known is that the model may substantially underestimate the coupling strength for forested regions. Therefore, it highlights the importance of correctly determining the coupling strength, which is in turn related to defining the roughness length for heat/moisture zot in LSMs and demonstrates that assigning different Czil values for different land-cover types in Zilitinkevich's formulation will allow Noah to reasonably reproduce the observed Ch. Note that this study was conducted with an offline model and the issue of defining Czil constant may be specific to Noah, so these results should not be pushed very far. Nevertheless, it highly complements previous model-based investigations and provides a potentially valuable framework for analyzing, through evaluating modeled Ch against long-term observations, the correctness of representing land-atmospheric coupling strength in other LSMs used for weather and climate models.

Acknowledgments

[13] This work benefited from inspirational discussions with Margaret LeMone, Kenneth Mitchell, and Sergej Zilitinkevich. We thank the principal investigators of the AmeriFlux observation sites for their permission to use their data, Mukul Tewari for plotting Figure 1, and Copy Phillips for internal review. This research was supported by the NCAR Water System Program, the NASA Cooperative Agreement Notice (CAN) Applied Science Program (grant NNA06CN03A), and the NASA–Terrestrial Hydrology Program (grant NNG04GI84G). The National Center for Atmospheric Research is sponsored by the National Science Foundation.

Ancillary