1. Introduction
The cryosphere plays an important role in the earth’s climate system. In particular, snow cover, with its high reflectivity, high emissivity, and low thermal conductivity, has a significant impact on the surface radiation budget, turbulent energy fluxes, and local hydrological fluxes to the atmosphere and can further influence climate both locally and downstream (Cohen and Rind 1991; Mote 2008). In a proof-of-concept model experiment, Xu and Dirmeyer (2011) first identified the “cold spots” of snow–atmosphere coupling where the atmosphere is most sensitive to snow variability.
There are two main mechanisms that contribute to this snow–atmosphere coupling: the instantaneous snow albedo effect (Dickinson 1983; Hall et al. 2008) and the delayed hydrologic effect (Cohen and Rind 1991). Both effects influence the atmosphere through the energy and water balances. However, how these two mechanisms contribute to the snow–atmosphere coupling has not been determined previously. In particular, the delayed hydrological effect is difficult to study from observations. However, this hydrological effect triggers a soil moisture, evapotranspiration, and precipitation feedback that can play a critical role for weather and climate after the snow has melted. The hydrological effect acts as a bridge linking the cold-season processes (including snow, ice, and frozen ground) to summer land–climate interactions. Zhao (1999) hypothesized that the snow hydrological effect over the Tibetan Plateau plays a major role in affecting the interannual variability of the East Asian monsoon.
In Xu and Dirmeyer (2013, hereafter Part I), we improved upon the experiment of Xu and Dirmeyer (2011) to quantify snow–atmosphere coupling strength (i.e., the degree to which the atmosphere responds in a consistent manner to anomalies in snowpack) with prescribed realistic snow variability. In this paper, we extend the analysis to investigate the separate roles of snow albedo effect and snow hydrological effect in contributing to the overall coupling strength.
We have pioneered the utilization of observed snow cover fraction (SCF) information from Moderate Resolution Imaging Spectroradiometer (MODIS) observations to replace the model’s SCF parameterization (see Part I for a description of the model) and to study the separate impacts of prescribed snow water equivalent (SWE) and SCF anomalies on the atmosphere, based on satellite observations and reanalysis data. Under the assumptions that the snow hydrological effect is initiated by snowmelt in spring and the snow albedo effect is a linear function of SCF, the snow albedo effect and hydrological effect have been separated in the experiment design. We describe these two components of snow–atmosphere feedback in detail in section 2. Section 3 describes the model experiments used to isolate these components. Results are presented in sections 4 and 5. Section 6 provides a discussion.
2. Snow albedo effect and snow hydrological effect
a. Background
Snow surfaces have the highest albedos in nature. The shortwave albedo ranges from 0.60 in wet and melting snow to greater than 0.85 for fresh snow. The presence of snow cover sharply increases the net surface albedo by 30%–60%. The high albedo of snow cover dramatically reduces absorbed and net surface shortwave radiation. This direct impact is known as the snow albedo effect (SAE). It dramatically affects the land surface energy budget and influences air temperature, density, and surface pressure. Walsh et al. (1982) demonstrated that the presence of snow cover is associated with near-surface cooling of 5–10 K in the lower troposphere.
The indirect impact, known as the snow hydrological effect (SHE), is a result of soil moisture anomalies from snowmelt that have a delayed impact on the atmosphere through land–atmosphere interactions (soil moisture, evapotranspiration, and precipitation feedbacks). During spring, the snowmelt infiltrates the soil triggering initial anomalies in soil moisture, depending on anomalies in the snowpack and the original soil moisture state. Runoff from rapid snowmelt in spring is hydrologically a two-step process. The excess will first lead to Hortonian runoff due to snowmelt excess over infiltration. Once the soil becomes saturated, Dunne runoff due to saturation excess occurs and more water is discharged into river networks. These processes are highly nonlinear. The snowmelt and maximum soil infiltration capacity are critical parameters for representing this process. For the hydrological modeling of snowmelt, the maximum of SWE prior to snowmelt is the key variable to quantify how much water is available to affect soil moisture. On the other hand, the maximum soil infiltration capacity is determined from soil texture and preexisting soil moisture (Entekhabi and Eagleson 1989), which are computed based on the effective soil porosity and saturated fraction.
Very few studies have investigated these issues, likely because of the complicated snowmelt and runoff processes and the general unavailability of accurate snow water content datasets. In one study, Quiring and Kluver (2009) found the maximum snow depth during the previous winter shows a good correlation with soil moisture anomalies in the following summer, based on snowfall observations. Positive snowfall anomalies are associated with wetter than normal soil conditions during the summer (June–August) in the northern Great Plains and negative anomalies with drier soils. However, the strength of the relationship between winter/spring snowfall and summer soil moisture varies significantly over space and time, which limits its utility for seasonal forecasting.
Besides the direct and indirect snow effects, positive and negative snow–atmosphere feedbacks further amplify or ameliorate anomalies. The most important positive feedback is the snow albedo feedback. With warmer temperatures, the area of snow cover decreases and land surfaces absorb an increasing fraction of solar radiation as net albedo drops. This increase of total absorbed solar radiation contributes to continued and accelerated melting and warming. On the other hand, a colder climate will keep more snow cover in place and sustain lower air temperatures (Wiscombe and Warren 1980).
Another important but less known negative (self-regulating) feedback is the snowfall-stability feedback, first suggested by Walland and Simmonds (1996). With the sudden increase in snow cover after a snowstorm, the air temperature in the lower troposphere decreases and static stability of the atmosphere increases; this reduces the probability of subsequent snowfall. This short-term reduction in snowfall allows decreasing snow cover by snow sublimation and redistribution by wind. The reduced snow cover increases the sensible heat fluxes to the atmosphere and decreases static stability, creating an environment again favorable for snowstorms. This negative feedback can keep the snow cover relatively stable over high latitudes in the winter.
b. Separation based on snowmelt
To separate the two mechanisms of snow–atmosphere coupling, Xu and Dirmeyer (2011) proposed a simple classification based on the local timing of snowmelt. According to the snow ablation curve, it was found that the snow cover is relatively stable, with low variability during the weeks before the spring snowmelt. With increasing solar radiation and air temperature, a threshold is reached, after which the snow begins to melt rapidly. The rate of melt becomes greatest between times when SWE drops from roughly 80% to 20% of its peak winter value (Xu 2011). Thus, we can define the snowmelt begin date as the date when SWE falls to 80% of its peak value and the snowmelt end date as the date when it reduces to 20% of peak.
On the basis of this definition, the Northern Hemisphere snows begin to melt at the southern fringes of the maximum winter snow extent at the end of February, progressing to higher latitudes and elevations over the next few months (Fig. 1). The Tibetan Plateau shows a relatively late snowmelt relative to other terrain at the same latitude because of its high elevation. The snowmelt period, the interval between snowmelt begin and end dates, is short in middle latitudes and increases toward the pole. If the snowmelt period is less than 10 days in a grid box, we extend the interval both before and after to at least 10 days for stability of the calculations. Globally, almost all seasonal snow disappears by the end of May, except for a few regions at very high altitudes or along the Arctic coast.
The climatological (top) snowmelt begin date (80% of peak SWE) and (bottom) end date (20% of peak SWE) based on the SWE ablation curve obtained from GLDAS. The color scale indicates the Julian day. Note that the snowmelt end date over much of Greenland is after Julian day 200 (no color).
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
The climatological (top) snowmelt begin date (80% of peak SWE) and (bottom) end date (20% of peak SWE) based on the SWE ablation curve obtained from GLDAS. The color scale indicates the Julian day. Note that the snowmelt end date over much of Greenland is after Julian day 200 (no color).
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
The climatological (top) snowmelt begin date (80% of peak SWE) and (bottom) end date (20% of peak SWE) based on the SWE ablation curve obtained from GLDAS. The color scale indicates the Julian day. Note that the snowmelt end date over much of Greenland is after Julian day 200 (no color).
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
On the basis of these two thresholds, we can define three distinct stages in the simulations: stable snow before snowmelt (from 1 March to begin date), the snowmelt period (between snowmelt begin date and end date), and after snowmelt (from snowmelt end date through 31 May). Before the snowmelt stage, the albedo effect is the main mechanism for snow–atmosphere coupling. After snowmelt, the snow hydrological effect is dominant. Both effects are active during the snowmelt stage.
c. Sensitivity to model parameterization
In Part I, we demonstrated areas of strong snow–atmosphere coupling within the National Center for Atmospheric Research (NCAR) Community Climate System Model (CCSM). However, we also found that the coupling strength in this model is sensitive to the model SCF parameterization. Figure 2 demonstrates the sensitivity of the albedo effect to the snow depth derived from SWE. Figures 2a and 2b show the mean and standard deviation of snow depth within the 10 ensemble members for March in the Control experiment (described in Part I). The snow depths increase northward from the middle latitudes, except over high terrain such as the Tibetan Plateau and Rocky Mountains. However, the standard deviations are relatively large over middle latitudes, indicating that there is considerable variability over these regions. Figure 2c shows the change of snow cover fraction with snow depth (df/dh, the slope of the parameterization curve; for full equation, see Part I) at the value of the climatological mean snow depth of March. With relatively small values of snow depth in the middle latitudes, the snow cover fraction here experiences large sensitivity to snow depth because the slope of the curve df/dh is steep. Multiplied by the mean incident shortwave radiation (Fig. 2d), the sensitivities of snow albedo effect to the change in snow depth (the potential variability of total shortwave energy input to the ground) are demonstrated in Fig. 2e. The highly sensitive regions are concentrated in the middle latitudes, and those sensitivities gradual weaken toward the pole and tropics. This distribution pattern is consistent with the snow–atmosphere coupling pattern quantified by the numerical experiments shown in Part I.
The sensitivity of the SAE to the snow depth in the model: (a) the monthly mean snow depth in March from model, (b) interannual variability of monthly mean snow depth represented by standard deviation, (c) the slope of SCF parameterization curve, (d) incident solar radiation, and (e) SASI.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
The sensitivity of the SAE to the snow depth in the model: (a) the monthly mean snow depth in March from model, (b) interannual variability of monthly mean snow depth represented by standard deviation, (c) the slope of SCF parameterization curve, (d) incident solar radiation, and (e) SASI.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
The sensitivity of the SAE to the snow depth in the model: (a) the monthly mean snow depth in March from model, (b) interannual variability of monthly mean snow depth represented by standard deviation, (c) the slope of SCF parameterization curve, (d) incident solar radiation, and (e) SASI.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
The effective slope of the SCF parameterization curve df/dh is a key factor controlling the variability of the snow albedo effect in any snow scheme. Meanwhile, the coupling strength computed in ideal snow simulations is strongly impacted by this parameterization. Importing realistic SCF observed by satellite removes this model-dependent sensitivity, as was shown in Part I. Here we will apply this constraint when examining snow albedo sensitivity.
d. Snow albedo sensitivity
The second term, 
Figure 3 shows the zonal mean (line) and interannual variability (shading, shown as a range) of the first term (term 1, red) and the second term (term 2, blue) of the snow albedo effect [W m−2 (°)−1] based on the data from March to June. The original data for computation are obtained from MODIS and the Control simulations. The magnitude of term 1 in the mean is generally larger than term 2 for all months; however, the interannual variability (shading) of term 1 is far less than for term 2. During March, term 1 is generally strengthening with latitude and reaches a maximum at roughly 70°N. From April to June, this term decreases quickly in magnitude as the peak moves northward toward the pole. After May, there is almost no impact by the first term south of 60°N. However, the second term is generally small in the mean but shows relatively larger range (variability). The maximum interannual variability moves from middle latitudes northward to high latitudes from March to June. The Tibetan Plateau near 35°N still shows large interannual variability even after May. The variability in the meridional SCF gradient 
Zonal mean (lines) and interannual variability (shading, shown as the range) of snow albedo forcing [W m−2 (°)−1] in term 1 and term 2 from March to June. The data for the computation are obtained from MODIS and the Control simulation.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
Zonal mean (lines) and interannual variability (shading, shown as the range) of snow albedo forcing [W m−2 (°)−1] in term 1 and term 2 from March to June. The data for the computation are obtained from MODIS and the Control simulation.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
Zonal mean (lines) and interannual variability (shading, shown as the range) of snow albedo forcing [W m−2 (°)−1] in term 1 and term 2 from March to June. The data for the computation are obtained from MODIS and the Control simulation.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
e. Snow hydrological sensitivity
Similarly, we can define a snow hydrological sensitivity index (SHSI) to quantify the climate forcing due to the hydrological effect of snowmelt. Snowmelt infiltration is a key process that strongly depends on the surface infiltration–runoff parameterization. The central concept underlying this parameterization is that of fractional saturated or impermeable area, which is determined by topographic characteristics and soil moisture, generating overland flow due to both saturation excess (Dunne runoff) and infiltration excess (Hortonian runoff).
3. Model experiments
On the basis of the previous snow–atmosphere coupling experiments, the analyses are extended to separate the contributions to snow–atmosphere coupling strength from snow albedo and hydrological effects. The details of the experiment design have been described in section 3 of Part I. Here we give a brief summary. A portion of CCSM is selected to concentrate on land–climate interactions. To focus on the land–atmosphere signal, the ocean and sea ice components are prescribed to the observed annual cycle. A recent version of the Community Atmospheric Model (CAM, version 3.6.48) coupled with Community Land Model (CLM, version 3.5) comprise the main components in this study. Ten-member ensembles of 6-month simulations are generated during the boreal snow depletion phase, beginning on 1 March.
The first ensemble experiment, called Control, is a typical climate simulation in which the 10 different atmosphere and land initializations are obtained from model restart files at 1 March from 2000 to 2009 from a long Atmospheric Model Intercomparison Project (Hurrell et al. 2008) style simulation. In the second ensemble experiment, named ModBoth (prescribed model snow states, both snow variables), all ensemble members are constrained to maintain precisely the same model time series of the key snow states (SWE and SCF). To achieve this, the snow states in each simulation are read from the previously recorded file from one arbitrary Control ensemble member. The atmosphere initial states and sea surface temperature boundary conditions are the same as for the ensemble members of the Control experiment. The effect and, thus, the land–atmosphere coupling strength due only to the snow states can be investigated. The realistic snow cover fraction (RealSCF) and realistic snow water equivalent (RealSWE) experiments provide an excellent means to investigate the SAE and SHE separately.
a. RealSCF
The RealSCF experiment is like the experiment ModBoth, except that the SCF in the 10 ensemble members are read from gridded observed MODIS SCF from 2000 to 2009. Each ensemble member takes specified realistic surface SCF that is prescribed by the remote sensing retrieval from MODIS, scaled to account for model bias. In the meantime, the SWE at each time step is still read from the output of one of member of the Control ensemble. The main purpose of RealSCF is to investigate how the simulation is improved and land–atmosphere coupling is changed by using “realistic” SFC without the errors of the model parameterization. Since ground albedo is a linear function of SCF and all ensemble members are forced by the same SWE, the RealSCF experiment precisely quantifies the atmospheric variability contributed only by snow albedo effect. Any predictability enhancement of ensemble RealSCF compared with ensemble ModBoth is due to the contribution of the more accurate representation of the snow albedo effect.
b. RealSWE
As opposed to RealSCF, all members in experiment RealSWE are prescribed to have SWE obtained from the Global Land Data Assimilation System (GLDAS), but with the same SCF from the one member of the Control ensemble. The only difference between ensemble RealSWE and ensemble ModBoth is the SWE that is obtained from offline GLDAS analysis, which eliminates the possible biases in the SWE simulation. As the most important factor to determine the potential available water for melt, infiltration, and runoff, a more accurate representation of the SWE depletion curve should result in more precise simulation of snow hydrological processes. Since the initial soil moisture of each ensemble member is set to the climatological mean (for details, see Part I) and SCF is forced to the same model output state, the snowmelt contribution to soil moisture anomalies can emerge and the atmospheric variability from the SHE can be separately investigated.
If we define the SHE to include all composited climate effects due to snowmelt in the soil and subsequent soil moisture, evapotranspiration, and precipitation feedbacks, the RealSWE experiment can be used to isolate the climate impact due to snowmelt by excluding the climate effect associated with snow albedo. Since the atmosphere initial and boundary forcings, except for SWE, are the same in ensembles RealSWE and ModBoth, the coupling strength and predictability difference between them is largely contributed by the hydrological effect.
4. Snow–atmosphere coupling strength
a. Coupling strength due to snow albedo effects
As mentioned previously, RealSCF differs from ModBoth in that the specified SCF is from observational sources. Figure 4 shows the relative coupling strength (the difference of intraensemble similarity within ModBoth and RealSCF experiments) for air temperature (2-m height) from March to June. During March and April, the impact of realistic SCF is relatively weak and concentrated in the middle latitudes. Only a few sizeable patches of higher values are located in eastern Europe, Tibet, and the Great Lakes. Beginning in May, the stronger couplings move to high latitudes and altitudes, including Siberia and the Far East. One notable feature is the Tibetan Plateau, which displays enhanced coupling during May with realistic SCF. There is stronger coupling over the Middle East in May and June that cannot be explained by a local snow albedo effect. Detailed investigation indicates those false signals may be triggered by commission errors in MODIS retrievals that mistake some clouds as snow.
Coupling strength (the difference of intraensemble similarity between ModBoth and RealSCF experiments; dimensionless) for 2-m air temperature from (top to bottom) March through June.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
Coupling strength (the difference of intraensemble similarity between ModBoth and RealSCF experiments; dimensionless) for 2-m air temperature from (top to bottom) March through June.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
Coupling strength (the difference of intraensemble similarity between ModBoth and RealSCF experiments; dimensionless) for 2-m air temperature from (top to bottom) March through June.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
Figure 5 shows the SASI [from Eq. (2)] based on the MODIS SCF retrievals from March to June. The map displays a similar progression to the SCF–atmosphere coupling strength obtained from numerical experiments (Fig. 4). The regions of strong coupling strength show reasonable migration from middle latitudes to the polar region during the snow ablation period from March to June. The SASI obtained from MODIS SCF retrievals agrees much more with the snow–atmosphere coupling strength obtained from RealSCF than the model-simulated pattern (Part I).
(top to bottom) SASI from March through June based on the MODIS retrievals.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
(top to bottom) SASI from March through June based on the MODIS retrievals.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
(top to bottom) SASI from March through June based on the MODIS retrievals.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
This simple snow albedo sensitivity index captures major patterns of the effects of snow albedo simulated by the RealSCF experiment. However, a shortcoming is evident in that the northward migration of coupling strength (Fig. 4) is damped over the boreal forests. Our simple assumption of a 0.4 difference in albedo between snow-covered and snow-free land does not hold well under forest canopies (cf. Barlange et al. 2005). Nevertheless, this index, based on either model or observed data, can be used to gain insight into strength of the albedo effect. The coupling strength of precipitation is relatively weak and the pattern is scattered (not shown), being far less robust than the air temperature response.
b. Coupling strength due to hydrological effects
Like the previous experiment, the only difference between RealSWE and ModBoth is the SWE that is specified from GLDAS from 2000 to 2009, which will trigger different soil moisture anomalies after the snow melts. The coupling strength and predictability differences between ModBoth and RealSWE are contributed only by the snow hydrological effect.
Figure 6 shows the SWE-driven coupling strength (the change of intraensemble similarity from ModBoth to RealSWE experiments) for air temperature (2-m height) from March to June. During March, the coupling strengths are much weaker than for the albedo effect (Fig. 4) and are mainly concentrated over ephemeral snow regions where the snow begins to melt in March. The snow hydrological effect is strongest in May, especially near the Tibetan Plateau. There are large areas in both Eurasia and North America in June that show strong snow–atmosphere coupling due to the lagged hydrological effect.
As in Fig. 4, but for ModBoth vs RealSWE experiments.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
As in Fig. 4, but for ModBoth vs RealSWE experiments.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
As in Fig. 4, but for ModBoth vs RealSWE experiments.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
Comparing the SHSI of Eq. (9) (Fig. 7) to the coupling strength (Fig. 6), there are similarities in pattern, but some large differences in magnitude. The hydrological forcing due to snowmelt gradually decreases with snow depletion, but the coupling strengths are mostly increasing as the atmospheric response strengthens.
(top to bottom) SHSI obtained from GLDAS from March through June.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
(top to bottom) SHSI obtained from GLDAS from March through June.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
(top to bottom) SHSI obtained from GLDAS from March through June.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
As with the ideal snow evolution Control experiment, the coupling strength in precipitation due to SHE is very scattered and patchy (not shown). However, the hemisphere average coupling strength increases gradually with time from March to June. This feature agrees with the general hypothesis that the hydrological effect is delayed and gradually increases after snowmelt.
The snow hydrological effect is essential to the soil moisture, evapotranspiration, and precipitation interaction and feedback after snow has melted. Classical land–climate interaction studies provide a useful conceptual framework to explain the evapotranspiration regimes as a function of soil moisture, based on the original idea of Budyko (1961) and recently improved by Koster et al. (2002). Three main evapotranspiration regimes, characterized by the evaporative fraction and soil moisture, can be defined: two soil moisture–limited regimes (dry and transitional regimes) and an energy-limited evapotranspiration regime (wet regime).
In the energy-limited evapotranspiration regime, the evaporative fraction is independent of soil moisture content above a given critical soil moisture value. Below this critical value, soil moisture content provides a first-order constraint on evapotranspiration (soil moisture–limited evapotranspiration regime).
On the basis of the wilting point, another important threshold for whether or not evapotranspiration takes place separates the soil moisture–limited regimes into dry and transitional regimes. In the wet (above a critical high soil moisture value) and dry (below the wilting point) climate regimes, soil moisture anomalies do not greatly impact evapotranspiration variability. In the transitional regime (between wilting point and the critical high value), however, soil moisture can strongly determine evapotranspiration variability and, thus, the resulting feedbacks to the atmosphere when shortwave radiation is plentiful. Much of the Northern Hemisphere transitions from energy-limited to moisture-limited conditions during spring (Dirmeyer et al. 2009).
We can compare the land–atmosphere coupling strength estimated from CAM coupled to CLM using the same method as in the Global Land Atmosphere Coupling Experiment from Guo et al. (2006), although in that experiment land–atmosphere coupling strength is estimated during summer. The “hot spots” of land–atmosphere interaction (strong coupling regions of soil moisture, evapotranspiration, and precipitation feedback) generally agree with the regions with strong snow hydrological effects, as shown in Fig. 6. SHE is a land–climate interaction driven by available snowmelt infiltration in spring, manifested as soil moisture anomalies.
5. Comparison of albedo effect and hydrological effect
The snow albedo effect directly changes the energy available at the land surface. The anomaly in albedo is a strong and immediate forcing to weather and climate. It can have no effect when and where snow is both climatologically and actually absent. On the other hand, the hydrological effect is a slow and persistent forcing to the atmosphere. Even after snowmelt, the hydrological effect continues to affect the atmosphere through soil moisture, evapotranspiration, and precipitation feedback. Table 1 summarizes the differences between the snow albedo and hydrological effects.
Characteristics of SAE and SHE.
To compare the contribution of the albedo and hydrological effects to snow–atmosphere coupling strength, the relative ratios of temperature coupling strength contributed by each, [ΩT(ModBoth) − ΩT(ModSCF)]/[ ΩT(ModBoth) − ΩT(ModSWE)], are shown in Fig. 8. When SCF < 0.01, the ratio is not plotted, as the SAE can lead to meaningless nonzero ratios. During March, almost all strongly coupled regions are due to the albedo effect (ratio > 1) except for a few regions along the southern edge of the snow cover, where snow is beginning to melt. By April, the albedo effect is significantly reduced, with rapid snowmelt toward the north, and the hydrological effect increases accordingly. By May, most of the apparent snow–atmosphere coupling strength is contributed by the hydrological effect. The Tibetan Plateau retains some strong albedo effect because of persistent snow anomalies at high elevations. Similar evolution is found in the precipitation coupling strength, but with smaller-scale local variations (not shown). This ratio cannot show the continuing hydrological effect in places where snowmelt is complete.
The ratio of temperature coupling strength contributed by SAE to SHE from (top to bottom) March–June.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
The ratio of temperature coupling strength contributed by SAE to SHE from (top to bottom) March–June.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
The ratio of temperature coupling strength contributed by SAE to SHE from (top to bottom) March–June.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
Figure 9 shows the temperature coupling strength due to the albedo effect, defined by ΩT(ModBoth) − ΩT(RealSCF), averaged over the stages “before snowmelt” and “during snowmelt,” classified locally by the snow depletion curve described in section 2b. The snow and atmosphere show weak coupling before snowmelt, but quite strong coupling during snowmelt. Before the snowmelt, there are only a few regions, mainly located in Tibet, Alaska, and central and eastern Europe, which show even weak coupling with the atmosphere. Most stable snow regions, such as Siberia and Canada, do not show substantial snow–atmosphere coupling. After the snowmelt begins, the coupling strengths are significantly increased. The coupling strengths generally increase with latitude as shortwave radiation increases above critical thresholds. This implies that the large variability in SCF during snowmelt will produce substantial albedo variations that exert a strong forcing to the atmosphere above.
The temperature coupling strength (dimensionless) due to SAE for (top) before snowmelt and (bottom) during snowmelt stages.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
The temperature coupling strength (dimensionless) due to SAE for (top) before snowmelt and (bottom) during snowmelt stages.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
The temperature coupling strength (dimensionless) due to SAE for (top) before snowmelt and (bottom) during snowmelt stages.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
On the other hand, the infiltrated water during snowmelt also produces a strong land–atmosphere coupling response, as shown in Fig. 10. The snow hydrological effects are stronger during snowmelt than after snowmelt. The snowmelt process requires a large amount of latent heat that maintains temperatures at the freezing point. Europe also shows strong coupling strength due to the hydrological effect. The strong coupling in Siberia for the albedo effect is significantly weaker than that due to the hydrological effect.
As in Fig. 9, but for the SHE during and after snowmelt.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
As in Fig. 9, but for the SHE during and after snowmelt.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
As in Fig. 9, but for the SHE during and after snowmelt.
Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-11-0103.1
6. Discussion
Because of nonlinear interactions between snow and other processes at the land surface, it is difficult to study the albedo changes and hydrological effects due to snow variations and their impact on the energy and water balance at the lower boundary of atmosphere. In the real world, the albedo effect and hydrological effect can be difficult to distinguish. By the design of experiments RealSCF and RealSWE and their comparison to simulations described in Part I, we have separated these two effects and compared their individual contribution to the snow–atmosphere coupling strength in a global climate model.
Surface albedo is very sensitive to changes in snow cover fraction. Any deficiencies in the snow simulation or snowpack parameterization introduce large biases in surface albedo, affecting air temperature and other atmosphere states. For example, a rapid reduction in SCF (such as premature snowmelt) will lead to a subsequently lower albedo at the surface and quickly warm near-surface air temperatures, affecting the vertical stability and the general circulation of the atmosphere. Prescribing realistic SCF observed from MODIS in RealSCF eliminates biases in the snow albedo effect caused by imperfect snow modeling and SCF parameterization. Forcing all ensemble members to the same SWE in RealSCF also provides an approach to isolate the snow hydrological effect. The RealSCF experiment provides a reasonable estimation of snow–atmosphere coupling strength and potential predictability due solely to the snow albedo effect.
Although MODIS provides the most accurate SCF monitoring currently available, it is still far from perfect. Errors in SCF across the Northern Hemisphere in the absence of cloud are roughly 8% (Hall et al. 2002) and are highest in the middle-latitude forest regions. The commission of errors due to misidentification of clouds as snow adds some artificial noise to the RealSCF simulation. On the other hand, omission errors, where the sensor misses shallow and ephemeral snow, weakens the simulated snow effects compared to reality. In addition, the 8-day composited MODIS snow maps used in this study lack snow variability on shorter time scales, acting as a temporal filter to modulate the snow–atmosphere coupling.
Besides the SCF, snow albedo itself is also an important regulator of the albedo effect. Although snow albedo at small spatial scales has been well-studied, grid-scale spectrally averaged snow albedo is much more complicated. It has been pointed out (Marshall and Warren 1986; Marshall 1989) that the spectrally averaged snow albedo varies with snow grain size, snow cover depth, underlying background albedo (for thin snow), solar zenith angle, and impurities absorbed in the snow (such as black carbon). Among these factors, grain size is the most critical variable controlling snow albedo. However, the snow grain size is so difficult to predict or simulate that it can be only roughly parameterized by snow age and temperature history (Anderson 1976; Koh and Jordan 1995; Lehning et al. 2002). On the basis of a comparison of CLM with four other weather and climate models (Wang and Zeng 2010), CLM version 3.5 has substantially improved the snow albedo simulation by use of a two-stream scheme for radiative transfer, separating the calculation between direct beam and diffuse radiation. The improved snow albedo estimation should improve the simulation of the snow albedo effect over earlier versions of the model. Furthermore, the vegetation fraction covered by intercepted snow has also been reduced, addressing a problem of albedo overestimation over grass.
In the climate model, soil hydrologic processes are strongly affected by the heterogeneity of soil hydraulic conditions. At the top soil layer, infiltration, exfiltration, and runoff depend on the nature of soil properties and states, including soil composition, texture, permeability, slope, elevation, and water content. The more accurate the representation of heterogeneous soil properties in the model, the more precise the model will be in representing grid-scale SHE. However, most land or hydrologic modelers have focused efforts upon improvement of the runoff scheme, with very little attention on snowmelt infiltration (Zhao and Gray 1999).
Snowmelt infiltration is the key process to determine SHE, which is also strongly and nonlinearly dependent on the soil water content. The Hortonian component of surface runoff is the residual of snowmelt intensity in excess of soil infiltration and is highly sensitive to soil moisture conditions. This infiltration will further determine whether the topsoil becomes saturated, thus producing the Dunne component of surface runoff. In this RealSWE experiment, we compare SHE by unifying the model initial soil moisture to the climatological mean. This setup effectively removes the possible impact of the soil moisture anomalies, but it also removes the possible variability in infiltration due to different soil moisture conditions. How variations of soil moisture conditions impact snowmelt infiltration, such as by initializing the soil with offline analyses (Dirmeyer et al. 2006), needs to be investigated in future studies.
Finally, frozen ground or permafrost is another key factor for SHE. The presence of ice in the soil dramatically alters soil hydrological properties. If the topsoil is frozen, the surface soil is almost impermeable. The frozen ground prevents infiltration of snowmelt, resulting in early and above-normal spring runoff. The CLM 3.5 frozen soil scheme has been greatly improved recently and now has a representation of fractional permeable area (Niu and Yang 2006). Compared with the previous frozen soil scheme where ice content is solely determined by available energy, this parameterization of hydraulic properties for frozen soil has shown superior results. Because of the complicated nonlinear interaction between atmosphere, snowmelt, and underlying frozen soil, there are more uncertainties in the SHE at high latitudes. The RealSWE experiment is still far from perfect as a means to investigate SHE. Experiments with different frozen ground and soil moisture conditions need to be implemented to further quantify these uncertainties.
The three-stage snowmelt classification applied here provides a simple system to study the two coupling mechanisms. Because of variations in latitude and elevation, different grid points show large variability in snow duration and melt timing. Different lengths in each stage make it somewhat problematic to compare across regions. Furthermore, during the snowmelt stage, both coupling mechanisms can be active at once. In a model experiment such as this, we can distinguish that the strength of each mechanism inasmuch as they operate separately, is a somewhat linear fashion. Nonlinearities may cause interplays between the albedo and hydrologic effects that we ignore here.
In summary, the previous snow–atmosphere coupling experiment described in Part I was extended to investigate the separate impacts on the atmosphere of the radiatively driven snow albedo effect and the snow hydrological effect that operates through soil moisture–surface flux feedbacks. From the simple theoretical analyses, the albedo effect is governed by the snow cover fraction, while the hydrological effect is controlled by anomalies in snow water equivalent. To isolate the snow albedo effect, realistic SCF from satellite estimates was prescribed and compared with model-generated values in the modeling experiment. Similarly, imparting realistic SWE from the Global Land Data Assimilation System in the model allowed for estimation of the snow hydrological effect. On the basis of a separation by snow ablation stages (before, during, and after snowmelt), the snow albedo effect was found to be active before and during the snowmelt period. The regions of strong albedo-driven coupling move northward during spring, with the retreating edge of the snowpack in the Northern Hemisphere. The snow hydrological effect appears first during snowmelt and can persist for months, well after snowmelt has concluded.
Acknowledgments
We are grateful to NCAR for providing the open source model and computer facilities to complete these simulations. This work is part of the lead author’s Ph.D. dissertation published by George Mason University. The lead author would like to acknowledge Dr. J. Shukla for his guidance and support during his Ph.D. study, as well as Dr. Bohua Huang and Dr. John Qu, who served as dissertation committee members. This study was supported by the George Mason presidential scholarship and NASA Earth System Science fellowship.
REFERENCES
Anderson, E. A., 1976: A point energy and mass balance model of a snow cover. NOAA Tech. Rep. NWS 19, 150 pp.
Barlange, M., Zeng X. , Wei H. , and Mitchell K. E. , 2005: A global 0.05° maximum albedo dataset of snow-covered land based on MODIS observations. Geophys. Res. Lett., 32, L17405, doi:10.1029/2005GL022881.
Budyko, M. I., 1961: The determination of evaporation from the land surface. Izv. Akad. Nauk SSSR, Ser. Geogr., 6, 3–17.
Cohen, J., and Rind D. , 1991: The effect of snow cover on the climate. J. Climate, 4, 689–706.
Dickinson, R. E., 1983: Land surface processes and climate-surface albedos and energy balance. Advances in Geophysics, Vol. 25, Academic Press, 305–353.
Dirmeyer, P. A., Gao X. , Zhao M. , Guo Z. , Oki T. , and Hanasaki N. , 2006: GSWP-2: Multimodel analysis and implications for our perception of the land surface. Bull. Amer. Meteor. Soc., 87, 1381–1397.
Dirmeyer, P. A., Schlosser C. A. , and Brubaker K. L. , 2009: Precipitation, recycling and land memory: An integrated analysis. J. Hydrometeor., 10, 278–288.
Entekhabi, D., and Eagleson P. S. , 1989: Land surface hydrology parameterization for atmospheric general circulation models including subgrid scale spatial variability. J. Climate, 2, 816–831.
Guo, Z., and Coauthors, 2006: GLACE: The Global Land–Atmosphere Coupling Experiment. Part II: Analysis. J. Hydrometeor., 7, 611–625.
Hall, A., Qu X. , and Neelin J. D. , 2008: Improving predictions of summer climate change in the United States. Geophys. Res. Lett., 35, L01702, doi:10.1029/2007GL032012.
Hall, D. K., Kelly R. E. J. , Riggs G. A. , Chang A. T. C. , and Foster J. L. , 2002: Assessment of the relative accuracy of hemispheric-scale snow-cover maps. Ann. Glaciol., 34, 24–30.
Hurrell, J. W., Hack J. J. , Shea D. , Caron J. M. , and Rosinski J. , 2008: A new sea surface temperature and sea ice boundary dataset for the Community Atmosphere Model. J. Climate, 21, 5145–5153.
Koh, G., and Jordan R. , 1995: Subsurface melting in a seasonal snow cover. J. Glaciol., 41, 474–482.
Koster, R. D., Dirmeyer P. A. , Hahmann A. N. , Ijpelaar R. , Tyahla L. , Cox P. , and Suarez M. J. , 2002: Comparing the degree of land–atmosphere interaction in four atmospheric general circulation models. J. Hydrometeor., 3, 363–375.
Lehning, M., Bartelt P. , Brown B. , Fierz C. , and Satyawali P. , 2002: A physical SNOWPACK model for the Swiss avalanche warning. Part II: Snow microstructure. Cold Reg. Sci. Technol., 35, 147–167.
Marshall, S., 1989: A physical parameterization of snow albedo for use in climate models. Rep. NCAR/CT-123, 160 pp.
Marshall, S., and Warren S. G. , 1986: Parameterization of snow albedo for climate models. Snow Watch ‘85, Rep. GD-18, World Data Center for Glaciology, Boulder, CO, 215–223. [Available online at http://www.atmos.washington.edu/~sgw/PAPERS/1986_Marshall.pdf.]
Mote, T. L., 2008: On the role of snow cover in depressing air temperature. J. Appl. Meteor. Climatol., 47, 2008–2022.
Niu, G. Y., and Yang Z. L. , 2006: Effects of frozen soil on snowmelt runoff and soil water storage at a continental scale. J. Hydrometeor., 7, 937–952.
Quiring, S. M., and Kluver D. B. , 2009: Relationship between winter/spring snowfall and summer precipitation in the northern Great Plains of North America. J. Hydrometeor., 10, 1203–1217.
Walland, D. J., and Simmonds I. , 1996: Modelled atmospheric response to changes in Northern Hemisphere snow cover. Climate Dyn., 13, 25–34.
Walsh, J. E., Tucek D. R. , and Peterson M. R. , 1982: Seasonal snow cover and short-term climatic fluctuations over the United States. Mon. Wea. Rev., 110, 1474–1485.
Wang, Z., and Zeng X. , 2010: Evaluation of snow albedo in land models for weather and climate studies. J. Appl. Meteor. Climatol., 49, 363–380.
Wiscombe, W. J., and Warren S. G. , 1980: A model for the spectral albedo of snow. I: Pure snow. J. Atmos. Sci., 37, 2712–2733.
Xu, L., 2011: Snow cover as a source of climate predictability: Mechanisms of snow-atmosphere coupling. Ph.D. dissertation, George Mason University, 241 pp. [Available online at http://digilib.gmu.edu:8080/xmlui/handle/1920/6261.]
Xu, L., and Dirmeyer P. A. , 2011: Snow–atmosphere coupling strength in a global atmospheric model. Geophys. Res. Lett., 38, L13401, doi:10.1029/2011GL048049.
Xu, L., and Dirmeyer P. A. , 2013: Snow–atmosphere coupling strength. Part I: Effect of model biases. J. Hydrometeor., 14, 389–403.
Zhao, L. T., and Gray D. M. , 1999: Estimating snowmelt infiltration into frozen soils. Hydrol. Processes, 13, 1827–1842.
Zhao, Z., 1999: The Summer Rainfall in China and Its Environment Field. China Meteorological Press, 204 pp.
