Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
1
1
Effects of soot-induced snow albedo change on snowpack and 2
hydrological cycle in western U.S. based on WRF chemistry and 3
regional climate simulations 4
5
6
7
8
9
10
11
Yun Qian, William I. Gustafson Jr., L. Ruby Leung, Steven J. Ghan 12
13
Atmospheric Science and Global Change Division 14
Pacific Northwest National Laboratory, Richland, WA 15
16
17
18
19
20
21
22
23
(To be submitted to J. Geophysical Research, 2008) 24
Corresponding author: Yun Qian, Pacific Northwest National Laboratory, Richland, WA 25
E-mail: [email protected]
2
Abstract 1
2 Radiative forcing induced by soot on snow is a major anthropogenic forcing affecting the 3
global climate. In this study we simulated the deposition of soot aerosol on snow and the 4
resulting impact on snowpack and the hydrological cycle in the western United States. A 5
yearlong simulation was performed using the chemistry version of the Weather Research and 6
Forecasting model (WRF-Chem) to determine the soot deposition, followed by two simulations 7
using WRF in meteorology-only mode, with and without the soot-induced snow albedo 8
perturbations. 9
The chemistry simulation shows large spatial variability in soot deposition that reflects 10
the localized emissions and the influence of the complex terrain. The soot-induced snow albedo 11
perturbations increase the surface net solar radiation flux during late winter to early spring, 12
increase the surface air temperature, and reduce the snow accumulation and spring snowmelt. 13
These effects are stronger over the central Rockies and southern Alberta, where soot deposition 14
and snowpack overlap the most. The indirect forcing of soot accelerates snowmelt and alters 15
stream flows, including a trend toward earlier melt dates in the western United States. 16
The soot-induced albedo reduction initiates a positive feedback process whereby dirty 17
snow absorbs more solar radiation, heating the surface and warming the air. This warming causes 18
reduced snow depth and fraction, which further reduces the regional surface albedo for the snow 19
covered regions. Our simulations indicate that the soot-induced change of maximum snow 20
albedo contributes to 60% of the net albedo reduction over the central Rockies; snowpack 21
reduction accounts for the other 40%. 22
23
3
1
1. Introduction 2
The presence of soot particles in snow can reduce the snow albedo and affect snowmelt 3
[Warren and Wiscombe, 1980, 1985; Hansen and Nazarenko, 2004; Jacobson, 2004; Flanner et 4
al., 2007]. The 2007 IPCC report listed the radiative forcing induced by “soot on snow” as one of 5
the major anthropogenic forcings affecting climate change between 1750 and 2005 [IPCC, 2007]. 6
Therefore, quantifying and reducing the albedo errors due to this effect is a priority for 7
improving simulations of climate change and the hydrological cycle using climate models. 8
9
Soot in snow 10
Soot is produced by incomplete combustion of carbonaceous material, mainly fossil fuels 11
and biomass. Black Carbon (BC) is the main component of atmospheric soot particles, which 12
typically have a diameter of around 0.1µm. Because of their dark color, these particles absorb 13
solar radiation, convert it into internal energy, and emit thermal-infrared radiation, therefore 14
warming the air [Jacobson, 2001]. Soot particles generally are hydrophobic but can mix 15
internally or externally with other hydrophilic aerosols, such as sulfate. Soot particles are 16
removed from the atmosphere within days to weeks by rainout, washout, and dry deposition. 17
18
Optical and electron microscopes show that a typical snow crystal contains thousands of 19
particles, including soot and dust. It is hypothesized that wet deposition (via snow and rain) is the 20
primary mechanism for transferring soot from the atmosphere into the snow pack. Dry deposition 21
also can be significant, accounting for several tens of percent of the total deposition [Davidson et 22
al., 1985]. The deposition source of soot into the snow pack is countered by the sink of snowmelt. 23
4
Typical mixing ratio values for soot in snow range from 5 to more than 60 ngsoot g-1snow, as shown 1
in Table 1. 2
3
Snow Albedo 4
Surface albedo, more accurately a combination of the directional reflectance and the 5
diffuse albedo weighted by the direct and diffuse fraction of incident radiation, is an important 6
surface characteristic that determines energy transfer at the surface. Major perturbations to the 7
albedo can be caused by snow over very short time frames. Snow cover can change the albedo of 8
grassland by a factor of 3–4 [Betts and Ball, 1997; Jin et al., 2002] and forested regions by a 9
factor of 2–3 [Betts and Ball, 1997; Thomas and Rowntree, 1992]. This has been emphasized by 10
the results of many albedo sensitivity inter-comparisons [Cess et al., 1991; Randall et al., 1994]. 11
Additional studies, [e.g. Barnett et al., 1988; Yeh et al., 1983; Walland and Simmonds, 1996] 12
have shown that the long-term average of snow accumulation or melt patterns may significantly 13
alter regional climate and have a strong impact upon the general circulation. Current land surface 14
models used in climate models have a wide range of methods for parameterizing snow albedo 15
[Barry, 1996; Schlosser et al., 2000]. Different snow albedo parameterizations can produce 16
significantly different snow mass and snowmelt timing as demonstrated in both off-line studies 17
[e.g., Yang and Niu, 2000] and regional climate simulations [e.g., Lynch et al., 1998]. 18
19
Many factors affect snow albedo, with the effect of soot significantly stronger than the 20
effect of snow grain size [Conway et al., 1996]. The visible snowpack albedo can be reduced 21
from 0.95 for pure snow to 0.1 for dirty snow with 100 parts per million by weight (ppmw) of 22
soot. For optically thick snowpack, the broadband visible albedo depends most strongly on the 23
5
soot content. Snow grain size and solar zenith angle have roughly one-twentieth the influence of 1
soot content [Marshall 1994, 2003]. 2
3
Effects of soot in snow on radiative forcing 4
Hansen and Nazarenko [2004] used observations to highlight the importance of soot-5
snow forcing. They used global-scale estimates of soot concentrations within snow and ice to 6
deduce the surface albedo perturbation, which results in a positive radiative forcing (RF) of 0.15 7
W m-2 and global warming of 0.24 oC, yielding a higher forcing efficacy [Hansen et al., 2005]. 8
Jacobson [2004] developed a global model that allows the soot to enter snow via precipitation 9
and dry deposition, thereby modifying the snow albedo and emissivity. He found that the 10
simulated concentrations of soot within snow were in reasonable agreement with those from 11
observations. His modeling study showed that soot on snow and sea ice can cause a decrease in 12
the surface albedo of 0.4% globally and 1% in the northern hemisphere. Flanner et al. [2007] 13
coupled a snow, ice, and aerosol radiative model to a GCM with prognostic carbon aerosol 14
transport to estimate present-day climate forcing and response from BC in snow. They estimated 15
a global annual mean BC-snow surface RF of 0.049 to 0.054 W m-2 and changes to 2-meter air 16
temperature of 0.10 - 0.15 oC. They suggested that the anthropogenic contribution to the total 17
soot-snow forcing is at least 80%. 18
19
However, all these studies are global in scale and employ coarse spatial resolutions, 20
which cannot resolve the significant orography-related features of snowpack and aerosol 21
deposition. It is well known that snowpack has a strong regional dependence on surface elevation 22
(see Figure 9). Furthermore, these studies only estimated the soot-snow induced RF and resulting 23
6
temperature response and did not investigate the hydrological impacts, such as snowpack and 1
runoff perturbations over regions with complex topography and inhomogeneous land cover such 2
as the western United States (WUS). Understanding the dynamic physical processes controlling 3
snowpack in these regions is of utmost importance for managing water resources [Leung et al., 4
2003]. In the WUS, mountain snowmelt accounts for more than 70% of the annual stream flows 5
that support irrigation in the semi-arid Central Valley and Columbia Basin and hydropower 6
generation and navigation in the major river basins. Analysis of measurements has revealed that 7
snowpack levels have dropped considerably throughout the WUS since the 1950’s [Mote et al., 8
2005]. Further reductions in snowpack amount are predicted under plausible climate change 9
scenarios [e.g., Leung et al., 2004]. 10
11
Some studies have attributed the snow retreat and reduction in the snowfall/precipitation 12
ratio to warming trends in the western U.S. over the past 50 years [Mote et al. 2005; Feng and 13
Hu 2007]. The warming trends are consistent with global climate simulations that included 14
greenhouse gas forcings in the past [Christensen et al. 2007]. In this study a potential secondary 15
contributing cause of snowpack change is investigated - the heating of snow through decreased 16
albedo caused by the deposition of soot. This accelerates snow melting and reduces snowpack. In 17
the WUS, much of the soot available for deposition into snow comes from populated regions 18
west of the mountains. During winter, the prevailing westerly to southwesterly flows transport 19
moisture from the Pacific Ocean as well as anthropogenic aerosols from coastal metropolitan 20
areas into the higher elevations where they are deposited. Vehicular emissions also contribute 21
much of the soot production in the WUS [Cooke et al., 1999]. 22
23
7
This study consists of a series of model simulations. It begins with a yearlong simulation 1
using the chemistry version of the Weather Research and Forecasting model (WRF-Chem) to 2
simulate an annual cycle of soot aerosol deposition on snow. This is then used to estimate snow 3
albedo perturbations induced by the soot within the western United States. Then, two regional 4
climate simulations are performed using WRF with the chemistry turned off, but with or without 5
the perturbed snow albedo based on the WRF-Chem simulation. 6
7
2. WRF-Chem simulation for soot deposition 8
9
WRF-Chem 10
In order to address scientific questions related to meteorological-aerosol-radiation-cloud 11
feedbacks at the urban to regional scale, scientists at the Pacific Northwest National Laboratory 12
(PNNL) have made substantial contributions to the chemistry version [Grell et al. 2005] of the 13
Weather Research and Forecasting model [Skamarock et al. 2005] (WRF-Chem) during the past 14
three years [Fast et al. 2006; Gustafson et al. 2007]. These contributions include: the CBM-Z 15
gas-phase chemistry mechanism [Zaveri and Peters 1999]; the MOSAIC sectional aerosol 16
module [Zaveri et al. 2005a, 2005b, 2008]; the Fast-J photolysis module [Wild et al. 2000]; 17
feedbacks between aerosols, clouds, and radiation [Fast et al. 2006]; activation/resuspension and 18
wet scavenging [Easter et al. 2004; Ghan et al. 2001]; aqueous chemistry [Fahey and Pandis, 19
2001]; and extending the nesting capability of WRF to include the chemistry scalars. Using 20
CBM-Z and MOSAIC, WRF-Chem simulates 55 prognostic gas species and eight aerosol 21
species plus the aerosol number, water content, and an estimate of hypsometric water content. 22
Aerosol species are assumed to be internally mixed and are explicitly carried in two forms, one 23
representing dry/interstitial conditions and one representing particles inside cloud droplets. For 24
8
this study, WRF-Chem has been further modified to track the deposition budget of BC. Variables 1
have been added to save the deposition mass of BC by size bin for three categories: BC wet 2
deposition concurrent with rain, BC wet deposition concurrent with snow, and BC dry deposition. 3
Each of these quantities was archived hourly during the simulation. The partitioning between 4
rain and snow is determined from the surface temperature, as adopted by the Noah land surface 5
model [Chen and Dudhia 2003], which is used in the WRF simulations with or without chemistry. 6
Precipitation is assumed to be rain if the surface temperature is above freezing. Otherwise, the 7
precipitation is assumed to be snow. 8
9
For the WRF-Chem simulation presented here, the MOSAIC aerosol module was 10
configured with four sectional size bins: three size bins for particles up to 2.5 µm diameter and 11
one bin for particles with diameters between 2.5 and 10 µm. This is half the number of bins used 12
in previous WRF-Chem research applications using the MOSAIC sectional aerosol module [Fast 13
et al. 2006; Gustafson et al. 2007], but is necessary to reduce computational cost for the climate 14
study presented in this paper. Even with this compromise, the simulation cost is substantial and 15
requires limiting the run to a single year. This is the first WRF-Chem study using the fully 16
coupled aerosol-cloud parameterizations for extended, climate-length simulations. In order to get 17
proper climate statistics, another two simulations lasting for five years are done using the 18
meteorological-only settings of WRF (called WRF-RCM to denote regional climate simulations 19
with WRF alone), as described in Section 3, to generate climatologically relevant statistics for 20
the impact of the soot-induced albedo perturbation impact on climate and hydrological cycle. 21
22
9
The WRF-Chem domain extends over the western United States from the Pacific Ocean 1
to the eastern side of the Rockies (see Figure 1). Meteorological boundary conditions have been 2
provided as in the extended climate simulations described in Section 3.2. Coupling between the 3
aerosol and cloud fields is accomplished via feedbacks in the Lin et al. [1983] microphysics 4
scheme and aerosols with radiation in the Goddard Space Flight Center shortwave scheme [Fast 5
et al 2006]. Trace gas and particulate emissions were compiled for a typical summer day from 6
the National Emission Inventory 1999 (NEI99), an Environmental Protection Agency emissions 7
product, and regridded to the WRF-Chem domain as a repeating diurnal cycle (see Figure 1). 8
Emissions from forest fires have not been included since the fire locations vary substantially 9
from year-to-year. This results in an underestimate of BC emissions, but is preferable to biasing 10
the soot deposition to particular regions with a year specific fire pattern. Also, most of the fires 11
occur during summer and fall when snow has already melted and is not accumulating. 12
13
Figure 1 14
15
Soot concentration in the atmosphere 16
The continuous WRF-Chem simulation, from 1 September 2003 to 31 October 2004, 17
provides a complete annual cycle of atmospheric soot concentration and deposition after 18
neglecting the first month as spin-up. Figures 2 and 3 present a comparison of the simulated 19
near-surface concentration of BC and observations from the IMPROVE network 20
(http://vista.cira.colostate.edu/IMPROVE/). The model values are from the lowest model level. 21
The spatial plots for each season in Figure 2 show the strong BC gradients that exist in the WUS. 22
Much of this is due to high emission regions near areas of complex topography. The complex 23
10
terrain preferentially forces the pollutants into areas such as the California Central Valley where 1
they accumulate. In general, WRF-Chem reproduces the spatial variation seen in the 2
observations, but it also reveals how gaps in the IMPROVE network make a thorough 3
comparison difficult because of the gradients. For example, many of the stations in California are 4
at higher elevations in the Sierra Nevada along the edge of the strong concentration gradient. 5
Small location errors in the simulated gradient lead to large differences compared to the 6
observations. 7
8
Figure 2 9
10
The spatial plots also reveal that the DJF season is simulated less accurately than the 11
other seasons. This is presumably due to a combination of several factors, such as changing 12
meteorological conditions in winter compared to summer, and the use of summer emissions for 13
the entire year. The modeled BC concentration is lower in DJF than for the other seasons over 14
much of the domain with the exception of the Central Valley, which has higher concentrations. 15
By comparison, the IMPROVE data also has lower concentrations during DJF, but the difference 16
between the model and data is greater during this season. In general, the model overpredicts BC 17
every month, as shown in Figure 3, which shows monthly averaged BC concentrations for the 18
IMPROVE stations within the domain along with the coincident WRF-Chem grid point averages. 19
The minimum, maximum , 25th percentile, median, and 75th percentile values are all higher in the 20
model than in the observations with the exception of the 75th percentile in October and the 21
maximum in June. The high maximum value observed in June is due to a forest fire, which is not 22
included in the model. 23
11
Figure 3 1
2
Soot deposition 3
To calculate the soot content within snow, black carbon (BC) deposition is tracked in 4
WRF-Chem as the model integrates forward in time using the deposition budget variables 5
described above. The separation of wet deposition into two categories, with rain or with snow, 6
provides information for parameterizing how carbon accumulates within the snowpack, either 7
depositing on the surface, possibly below the surface, or in the snow crystal. 8
9
Figure 4 shows the BC deposited for DJF and MAM. The total deposition of BC is 10
spatially correlated with BC emissions; the largest deposition occurs around large cities, e.g. in 11
coastal Southern California, the San Francisco Bay area, and Calgary. Nevada has the least 12
deposition of the states included in the model domain; this is due to both lower emissions and 13
very little rain. In general, the dry deposition (not shown) is larger over the coastal metropolitan 14
areas, e.g. Seattle, Portland, Los Angeles, and San Francisco, as well as a few large cities in 15
southern Canada. The wet deposition (not shown) is more evenly distributed spatially than the 16
dry deposition. However, the wet deposition has a strong seasonal cycle mimicking the 17
precipitation cycle, whereas the dry deposition is more similar from month-to-month because we 18
prescribed constant emissions over time except for the diurnal cycle. Also, strong orographic 19
influences are seen in the wet deposition due to orographically forced precipitation, e.g. over the 20
coastal mountains of the Pacific Northwest and portions of the Rockies. The partitioning of the 21
precipitation between rain and snow also makes a difference, particularly when determining how 22
much of the BC remains in the snowpack. 23
12
1
Figure 4 2
3
Soot content in snow and induced albedo perturbation 4
The soot content in snow is calculated based on the simulated time series of snow depth 5
and soot deposition using a methodology derived from the algorithms of Jacobson [2004]. 6
Archived hourly output from WRF-Chem is used in an offline program to determine the soot 7
mixing ratio in snow. No attempt has been made to treat different particle diameters differently; 8
deposition for all size bins has been combined into net deposition values. 9
10
The algorithm for determining the mixing ratio of soot in snow, rBC with units of 11
ngBC gsnow−1 , is as follows. This description uses an algorithmic approach whereby the formulas 12
represent the process as when coded. So, the right side of the equations represent the variables at 13
the current state, which are then applied to the variable on the left hand side of the equation. First, 14
the mass of snow deposited during the previous period is determined. For the hourly data here, 15
this time increment equates to ∆t = 3600 s. For dry deposition, the incremental mass addition of 16
soot per unit area, ∆mBCd , is simply the sum of the dry deposition from the model output over the 17
time ∆t . The incremental mass addition for wet deposition per unit area, ∆mBCw , is dependent on 18
whether the snow depth increased or decreased during ∆t . If snow is added or stayed the same 19
during ∆t , ∆mBCw is the sum of wet deposition concurrent with snow during ∆t . If the snowpack 20
thinned, then any BC that deposited with snow during ∆t is instead added to ∆mBCd . Any BC that 21
deposited with rain is assumed to instantly wash out of the snowpack and is neglected in the 22
following formulas. 23
13
Next, the change in mixing ratio of soot in snow due to ∆mBCd versus ∆mBCw is 1
determined. The mixing ratio of soot in snow due to dry and wet deposition, rBCd and rBCw , 2
respectively, is tracked separately over time and the sum of the two mixing ratios results in rBC . 3
The value rBC represents the soot mixing ratio for the radiatively important portion of the 4
snowpack, which we refer to as the effective snow depth, ′ D , and is defined as the minimum of 5
the actual snow depth, D , or ′ D max =10-2 m following Jacobson [2004]. Also, rBC is only 6
determined for cells with D>10-3 m to maintain stability in the algorithm. So, if D>10-3 m, then 7
the increased concentration due to ∆mBCd is determined from 8
1
1−
⎟⎟⎠
⎞⎜⎜⎝
⎛′
∆+⎟⎟
⎠
⎞⎜⎜⎝
⎛′
∆+=
Dtv
DmDmrr fall
sw
BCdBCdBCd (1) 9
where msw is the snow water equivalent (kg m-2) of the existing snowpack, and the right hand set 10
of parenthesis accounts for the slow fall speed of BC through the snowpack. The fall speed, v fall , 11
is set to 10-7 m s-1, following Jacobson [2004]. 12
13
The mixing ratio increase due to ∆mBCw depends on whether snow accumulated during 14
the previous ∆t . If the change in snow depth, ∆D , is negative, then rBCw is not modified at this 15
point in the procedure since BC mass from snow is added to ∆mBCd as explained above, and the 16
concentrating effect of reduced snow mass will be treated below. For ∆D >0, the mixing ratio 17
will dilute based on the fraction of new snow within the effective snow depth. If more snow falls 18
than the effective snow depth ′ D , then rBCw is reset to 0 since the existing soot will lie below ′ D 19
and becomes radiatively unimportant, and the soot mixing ratio within the effective snow depth 20
will depend on ∆mBCw during ∆t . So, if ∆D >0 we set 21
14
rBCw = rBCw
max( ′ D − ∆D, 0)′ D
⎛ ⎝ ⎜
⎞ ⎠ ⎟ (2) 1
This is then followed by 2
rBCw = rBCw + ∆mBCw
msw ′ D
⎛
⎝ ⎜
⎞
⎠ ⎟ 1+
∆tv fall
′ D
⎛
⎝ ⎜
⎞
⎠ ⎟
−1
(3) 3
to account for the increased BC mass from ∆mBCw and the BC fall speed through snow in this 4
portion of the mixing ratio. 5
6
For the case of decreasing snow depth, the mixing ratios are handled differently. If ∆D <0 7
and the temperature is above freezing, the mixing ratios are not modified, based on the 8
assumption that the liquid snowmelt washes out any BC that was in the melted snow, resulting in 9
no change in concentration. It is assumed that the deeper snow has the same mixing ratio as the 10
snow immediately above it if ∆D > ′ D . If ∆D <0 and the temperature is below freezing, then the 11
reduction is due at least in part to sublimation, which would concentrate the BC within the 12
snowpack. The fraction of ∆D due to sublimation, α sub , was not available from our model run, 13
so we assumed that 10% of this reduction is due to sublimation based on Essery et al. [2003]. 14
15
The sublimation is handled differently based on whether or not D is greater or less than 16
′ D max . If D > ′ D max , then additional snow from below ′ D enters the effective snow depth to replace 17
the snow mass lost to sublimation at the surface. Again, it is assumed that the deeper snow has 18
the same mixing ratio as the snow immediately above it. In this case, the mixing ratios are 19
adjusted using 20
rBCd = rBCd 1− αsub∆D′ D
⎛ ⎝ ⎜
⎞ ⎠ ⎟ (4) 21
15
and 1
rBCw = rBCw 1− αsub∆D′ D
⎛ ⎝ ⎜
⎞ ⎠ ⎟ . (5) 2
However, if D ≤ ′ D max , then no new snow enters into the effective snow depth so the 3
concentration increases based on the fractional reduction of snow. This fractional reduction is 4
capped when the snow depth reduces to 2 cm to prevent overly high concentrations. In reality, 5
when the average snow depth is this thin over the area of the model grid cell, the snow coverage 6
will be patchy so the potentially very high concentrations would be unrealistic without the cap. 7
So, for 2 cm < D ≤ ′ D max 8
rBCd = rBCd ′ D
max( ′ D +α sub∆D, 2 ⋅10-2 m) (6) 9
and 10
rBCw = rBCw ′ D
max( ′ D +αsub∆D, 2 ⋅10-2 m). (7) 11
Once all the above steps are complete, the final soot mixing ratio in snow is given by 12
rBC = rBCd + rBCw . (8) 13
The resulting values for the soot mixing ratio in snow is shown in Figure 5. The values are 14
around 10-80 ng g-1 in winter over the western U.S. and southern Alberta with the maximum 15
simulated values larger than 100 ng g-1 over some small regions such as eastern Idaho and 16
Calgary. These values are inline with observations and GCM simulations as shown in Table 1. 17
18
Figure 5 19
Once rBC is known, it can be used to estimate the change in snow albedo due to BC. First 20
the DJF average of rBC is calculated for each model point. Then, a 3x3 point average is 21
16
calculated for the 9 cells surrounding each grid point. This is intended to smooth the field and 1
reduce its noisiness caused by the strong gradients in both BC and precipitation in the complex 2
topography to account for their interannual variations. Finally, for the few points that have very 3
unrealistic rBC values, the smoothed seasonal average is capped with a value of 500 ng g-1, which 4
is the upper limit of the values used to determine the albedo formula below. The result, rBC,DJF , is 5
then used to estimate the snow albedo perturbation using the formula 6
DJFBCDJFBCDJFBC rrrA ,42
,63
,9 105369.8102314.2102883.2 −−− ⋅+⋅−⋅=′ (9) 7
This formula is a 3rd order polynomial fit to the points from Jacobson’s [2004] Figure 1c at a 8
wavelength of 500 nm. The perturbation represents the difference between the snow albedo for a 9
given soot-snow mixing ratio and when no soot is present. No attempt is made to represent 10
albedo perturbations from snow aging since the Noah soil model does not track snow age. Since 11
both WRF-RCM simulations, with and without the snow albedo perturbation, neglect this 12
physical process, comparing the difference between the simulations still provides meaningful 13
results for the impact of BC. It should also be noted that only the DJF season was used to 14
determine the soot content in snow due to the time independent handling of maximum snow 15
albedo in the Noah LSM (described below). The net result, based on the BC mixing ratios 16
(Figure 5a) and Equation (9), is a 1-5% perturbation to the snow albedo over most of the 17
northwestern states during winter (Figure 5b). The maximum simulated reduction in snow albedo 18
reaches 7-10% over areas coincident with the maximum soot content in snow. 19
20
3. WRF-RCM control simulation and evaluation 21
3.1 Evaluation data 22
17
To simulate the impacts of soot-induced snow albedo change on the hydrological cycle, it 1
is important to realistically capture precipitation, snowpack, and runoff in the control simulation. 2
The following datasets are particularly useful for evaluating model simulations because of the 3
relatively high spatial resolution and in some datasets, an adjustment for orographic effects. 4
5
UW/PRISM surface air temperature and precipitation 6
The UW/PRISM dataset consists of daily maximum and minimum surface temperature 7
and precipitation for 1949–2000. It was developed by the Surface Water Modeling Group at the 8
University of Washington following the methodology outlined by Maurer et al. [2001] based on 9
daily observations made at the National Oceanic and Atmospheric Administration (NOAA) 10
Cooperative Observer (Co-op) stations, with topographic adjustment based on the PRISM 11
precipitation climatology (Daly et al.). Compared to the other datasets described below, this 1/8o 12
dataset depicts a more pronounced pattern of orographic precipitation, especially over data 13
sparse areas such as the Great Basin and the intermountain West. 14
15
SNOTEL SWE 16
Point measurements of SWE are available at snow telemetry (SNOTEL) stations in the 17
western U.S. These stations are typically located in remote mountain sites above the snow line, at 18
elevations higher than that of the Co-op sites, between 610 m – 3500 m with a mean elevation of 19
2300 m. There are more than 550 stations with at least ten years of recorded precipitation and 20
SWE between 1981-2000 [Leung and Qian, 2003]. 21
22
NOHRSC SWE 23
18
Another SWE dataset developed by the National Operational Hydrologic Remote Sensing 1
Center (NOHRSC) is also used to evaluate our simulations. NOHRSC uses a statistical 2
methodology [Hartman et al. 1995] to combine station measurements of snow water, satellite 3
estimates of the snow line, and a digital surface elevation model to produce a gridded distribution 4
of snow water at a resolution of 1.5’ for WUS. We averaged the twice-weekly NOHRSC product 5
over each month for the years 1995-2000 [Ghan et al., 2006]. 6
7
CMC SWE 8
A third SWE dataset is daily snow depth in water equivalent at 0.25° resolution over 9
North America for the period 1979-1996 [Niu and Yang, 2007]. The gridded snow depth 10
combines in situ daily observations from ~8,000 U.S. cooperative stations and Canadian climate 11
stations and first-guess fields with an optimum interpolation scheme developed by Brown et al. 12
[2003], which is employed operationally at the Canadian Meteorological Centre (CMC). 13
14
GRDC Runoff 15
The runoff dataset used in this study is the University of New Hampshire (UNH)-Global 16
Runoff Data Centre (GRDC) global runoff climatology. The UNH-GRDC dataset provides 17
monthly climatological runoff fields, which are runoff outputs from a water balance model 18
driven by observed meteorological data and then corrected with the runoff fields that are 19
disaggregated from the observed river discharges. Although a no-time-delay assumption is 20
applied when the gauge-observed discharge is distributed uniformly over a catchment, the 21
resulting runoff fields over a large river basin approximate the real runoff, especially in river 22
basins that contain a sufficiently dense network of rain gauges. The UNH-GRDC dataset 23
19
preserves the accuracy of the observed discharge measurements and maintains the spatial and 1
temporal distribution of the simulated runoff, thereby providing the ‘‘best estimate’’ of terrestrial 2
runoff over large domains [Fekete et al., 2000]. 3
4
3.2 WRF-RCM Configuration 5
The regional climate model used in this study is based on the Weather Research and 6
Forecasting (WRF) model [Skamarock et al., 2005], with modifications described by Leung et al. 7
[2005], which includes the use of a wider buffer zone for the lateral boundaries, with the 8
relaxation coefficients of the nudging boundary conditions following a linear-exponential 9
function, updating of the background surface albedo to include seasonal changes, updating of sea 10
surface temperature and sea ice from the lower boundary conditions, and estimation of cloud 11
fraction based on Xu and Randall [1996]. A series of sensitivity experiments are done to 12
determine the parameterization schemes most fit for regional climate simulation in WUS. The 13
resulting parameterization choices include the CAM3 shortwave and longwave radiation [Kiehl 14
et al., 1996], the modified Kain-Fritsch convection parameterization [Kain and Fritsch, 1993; 15
Kain 2004], the WRF Single-Moment 6-class cloud microphysics scheme [Hong et al., 2004], 16
the Yonsei University (YSU) boundary layer scheme [Hong et al. 2006], and the Noah land 17
surface model [Chen and Dudhia 2001]. 18
19
The model domain encompasses all states over the western U.S. and extends into the 20
Pacific Ocean (see Figure 1a). It is centered at 25.3oN and 128.5oW, with a horizontal grid 21
spacing of 15 km. The model includes 35 vertical levels extending up to 100 hPa. The boxes in 22
Figure 1a encompass the regions analyzed in this paper, and roughly contain the boundaries of 23
the Columbia River Basin (CRB), the Sacramento-San Joaquin (SSJ) River Basin, the Central 24
20
Rockies (CR), and the Sierra Nevada (SN) mountains. Figure 1b shows the model topography. 1
At 15-km grid spacing, we can see that the Rocky Mountain and coastal ranges such as the Sierra 2
Nevada and the Cascades Mountain are well resolved. Several sensitivity experiments were 3
performed to determine the impact of snow emissitivity (SE) on the simulation of climate over 4
the western US. The simulated snowpack and temperature were found to be sensitive to the value 5
of SE (e.g. set from 0.90 to 1.0). SE is assumed to be 0.98 in this study, since the typical values 6
of snow emissivity range from 0.96 to 1.00 [e.g. Pielke 1984]. This is larger than the default 7
value of 0.90 prescribed in WRF v2. The default SE leads to an unrealistically small snowpack 8
and warm 2-m temperature biases for large portions of our domain. Using 0.98 for SE reduced 9
the warm bias from up to 5 degC to about 1 degC for the equivalent areas. 10
11
Initial and lateral boundary conditions of the simulation were derived from the NCEP-12
DOE Global Reanalysis II [Kanamitsu et al., 2002]. Sea surface temperature (SST) and sea ice 13
from AMIP-II [Taylor et al., 2000] were used to construct the lower boundary conditions. The 14
simulation was initialized on 1 September 1993 with the lateral and lower boundary conditions 15
updated every 6 hours through 31 December 1998. 16
17
3.3 Surface albedo in Noah LSM 18
The Noah LSM, developed based on the OSU LSM described by Chen and Dudhia [2001], 19
was used in this study [Skamarock et al., 2005]. As described in the Introduction, surface albedo 20
is a critical parameter that affects the net solar radiation that drives the terrestrial surface energy 21
and water budget [Xu and Hall, 2006]. A transition from no snow to a snow covered state 22
abruptly modifies the land surface albedo. The following formula is used in the Noah LSM to 23
calculate the surface albedo, ALBEDO: 24
21
)(**))(0.1( ALBSNOALBSNCOVRSHDMINSHDFACALBALBEDO −−−+= 1
ALB refers to the background snow-free surface albedo (in fraction), which is updated by 2
temporally interpolating between the prescribed vegetation-dependent background surface 3
albedos for the summer and winter. SNOALB is the maximum albedo over deep snow. SHDFAC 4
is the fractional areal coverage of green vegetation. SHDMIN is the minimum fractional areal 5
coverage of green vegetation. SNCOVR is the fractional snow cover defined as: 6
)*(0.1 * SALPRSNOWSALP eRSNOWeSNCOVR −− −−= SNEQV<SNUP 7
where SALP is a tuning parameter; RSNOW = SNEQV/SNUP; SNEQV is the snow depth in SWE; 8
and SNUP is a threshold snow depth in SWE that implies 100 percent snow cover. Hence, 9
surface albedo (ALBEDO) is calculated based on a simple weighting of the background albedo 10
(ALB) and the maximum albedo over deep snow (SNOALB) based on the fractional snow free 11
and snow cover area within a model grid cell. 12
13
To limit the values of surface albedo when snow is present, the Noah LSM inputs the 14
measured maximum albedo over deep snow [Robinson and Kukla, 1985, referred to as RK]. 15
Figure 6a shows the spatial distribution of the maximum snow albedo (SNOALB) used in the 16
default configuration of WRF. The RK dataset has a spatial resolution of 1o by 1o and is 17
produced from 109 measured scenes from the winter of 1978–1979. In RK, scene brightness was 18
converted to surface albedo by linear interpolation between the brightest tundra (albedo of 0.8) 19
and the darkest snow-covered forest (albedo of 0.18). Barlage et al. [2005] updated the RK 20
dataset by using over four years of MODIS products at 0.05º spatial resolution to produce a new 21
high resolution maximum albedo dataset of snow-covered land. The default SNOALB shown in 22
Figure 6a is what was used for the control simulation described below. 23
22
1
Figure 6 2
3
3.4 Results of control simulation 4
Precipitation 5
Figure 7 (a & b) shows the observed and simulated seasonal mean precipitation averaged 6
between December 1993 and November 1998. During the cold season, the observed precipitation 7
at 1/8o resolution shows distinct spatial distributions that resemble the complex orography. The 8
two precipitation bands along the West Coast correspond to orographic precipitation associated 9
with the coastal range and the Cascades and the Sierra Nevada that are further inland. East of 10
these mountains, precipitation decreases sharply in the basins and the intermountain zone, before 11
it increases slightly again over the Rockies. 12
13
Figure 7 14
15
The simulated cold season precipitation clearly has a spatial pattern similar to the 16
observations. The two precipitation bands near the coast are very distinguishable, with 17
precipitation decreasing in the valleys between the two elevated regions. In addition, 18
precipitation on the east side of the Cascades and the Sierra Nevada is significantly less. 19
20
Although there is a one-to-one correspondence between most areas of maximum 21
precipitation in the observations and the control simulation, the simulated orographic 22
precipitation along the windward slopes of the Cascades and Sierra Nevada is too strong. This is 23
consistent with results based on real time weather forecasts for the Pacific Northwest [e.g., Colle 24
23
et al. 2000; Westrick and Mass 2001] and MM5 regional climate simulation [Leung and Qian 1
2003] that show an increasing amount of area averaged precipitation as spatial resolution 2
increases. 3
4
During spring, the precipitation pattern is similar but with a smaller magnitude than in 5
winter. Again, WRF overpredicts the orographic precipitation along the Cascades and Sierra 6
Nevada as well as the Rockies (not shown). Figure 8 shows the monthly time series of 7
precipitation from December 1993 to November 1998 averaged over CRB and SSJ. WRF 8
captures the seasonal cycle and inter-annual variability of precipitation very well over the two 9
basins. However, it overpredicts precipitation during winter and spring by 25-35% over the two 10
basins. This behavior is similar to the MM5 model that tends to have higher large-scale moisture 11
convergence during the cold season when driven by global reanalysis at high spatial resolution 12
[Leung et al., 2003]. 13
14
Figure 8 15
16
Temperature 17
Figure 7 (c & d) shows the seasonal mean surface air temperature averaged for December, 18
January and February (DJF) based on observation and WRF control simulation. The spatial 19
distribution of large-scale features is very consistent between the two. The mesoscale details of 20
temperature are also captured over the mountains in the WRF simulation at relatively high spatial 21
resolution. The magnitude of simulated temperature during the DJF and MAM seasons (not 22
shown) is very close to the observed, with the exception of a small cold bias over the Rockies. 23
As shown in Figure 8 the basin-averaged temperature is slightly underpredicted during winter 24
24
and spring, but the cold bias is generally less than 1oC over CRB and SSJ basin. Surface air 1
temperature is simulated very well during summer and fall. 2
3
Snow water 4
Figure 9 (a, b, c & d) shows the simulated and observed distribution of DJF mean snow 5
depth in terms of snow water equivalent. We present three observational datasets of SWE to 6
provide an estimate of uncertainty in the observed quantity. Because of the improved simulation 7
of surface temperature gained through the modified snow emissivity in the model, the phase of 8
the precipitation is more accurate than with a default WRF configuration. This results in 9
improved snow accumulation and melting processes at the surface and the SWE is generally well 10
simulated. The simulation reproduces many of the features of the observed snow distribution, 11
with heavy snow cover over mountains (e.g. Cascades and the Sierra Ranges, and Rockies) and 12
light snow in the valleys (e.g. central valley) or basins (e.g. CRB). Based on the SNOTEL station 13
data, the CMC data likely have a negative bias over both high and low elevation areas. Since the 14
NOHRSC data combined station measurements of snow water in its product, the magnitude of 15
SWE from NOHRSC is very close to the SNOTEL data where measurements are available. 16
17
Figure 9 18
19
Figure 9 (e & f) shows the simulated and observed spatial distribution of mean SWE in 20
MAM. Compared to DJF, the snow cover area has significantly decreased in spring due to 21
snowmelt at the lower elevations. However, the SWE is much higher in MAM than in DJF over 22
the mountains (e.g. the Cascades, Sierra Nevada, and Rockies). During MAM, the model 23
25
captured the maximum SWE over the Sierra Nevada range and northern Rockies but 1
underpredicted SWE over the Cascades and southern Rockies. Since the surface temperature is 2
well simulated and precipitation is overpredicted in the Cascades and southern Rockies, the 3
overpredicted liquid rain on snow may have increased the melt process and decreased the SWE 4
during spring. 5
6
Runoff 7
The spatial distribution of modeled runoff generally agrees with the GRDC runoff in 8
winter (not shown). Both the modeled and observed data show higher runoff in wet regions such 9
as the high elevations, which is at least partially related to the overestimated orographic 10
precipitation. Further inland, runoff rapidly decreases coincident with decreasing precipitation in 11
the rain shadows. The main contribution to total runoff during winter is surface runoff generated 12
by liquid rain. The discrepancy between the modeled and observed runoff may also result from 13
the inconsistency of time periods used for averaging. GRDC runoff is estimated through longer 14
past discharge records, while the modeled runoff is an average over the years 1993-1998. 15
16
During spring both precipitation and snowmelt contribute to runoff. Generally, the spatial 17
distribution of large-scale features is very consistent between the observations and simulation 18
except for an overprediction of runoff resulting from the wet bias over the Cascades and Sierra 19
Nevada ranges (see Figure 10). The larger runoff over the hilly coastal areas overlaps with the 20
maximum precipitation over the central Rockies and maximum snowpack in the Cascades/Sierra 21
Nevada ranges (see Figure 9). 22
23
26
Figure 10 1
2
4. Simulation of soot-induced snow albedo change 3
To determine the impact of soot induced changes to snow albedo, a second WRF-RCM 4
simulation was done using the same settings as the control simulation except that the snow 5
albedo was modified based on soot deposition determined from the WRF-Chem simulation. This 6
was done by modifying the map of maximum snow albedo (Figure 6) used by the Noah LSM, 7
which was described in Section 3.3. The albedo perturbation, A′ , is subtracted from SNOALB. 8
We compared the results of these two 5-year long WRF regional climate simulations with and 9
without soot-induced snow albedo perturbations, to investigate the effects of soot on the surface 10
energy and water budget in the western U.S. The following analyses focus on the changes of 11
variables such as surface solar radiation, temperature, precipitation, evaporation, soil moisture, 12
snowpack, and runoff. These variables are the most relevant to the hydrological cycle and water 13
resources in the western U.S. 14
15
Surface solar radiation 16
Figure 11a shows the change in net surface shortwave radiation flux (NSW) for March 17
between the control and soot-sensitivity simulations. The change of NSW is spatially correlated 18
with the change of surface albedo (Figure 11d). The largest changes are over the central Rockies 19
and southern Alberta, where NSW increases 4-10 W m-2 in late winter and 8-14 W m-2 in early 20
spring. As seen in Figure 12, the NSW increases in winter and spring, and the maximum change 21
occurs in March. The NSW in March increases 6.4 W m-2 over the Central Rockies (CR) and 2.1 22
W m-2 over the Sierra Nevada (SN), respectively (see Table 2). The change of NSW is caused 23
27
mainly by the perturbed surface albedo rather than changes to the downward solar radiation flux. 1
As shown in Table 2 the surface NSW increases 4.8% in March over CR and 1.4% over SN, 2
respectively. However, downward solar radiation at the surface only changes -0.5% over CR and 3
-0.2% over SN, respectively. It can be seen from Figure 12 that the albedo change is similar 4
between December and March over both river basins; however, NSW increases significantly 5
during this period and reaches a peak in March because of increased downward solar radiation 6
resulting from the increased solar height angle. 7
8
Figure 11 9
10
Surface air and skin temperature 11
Surface skin (air) temperature increases by 0.3 - 2.0oC (0.2 - 1.5oC) over the majority of 12
the snow covered areas in the western U.S. during late winter to early spring. The temporal and 13
spatial distributions of temperature changes, mainly driven by the change of surface absorbed 14
solar radiation, which is correlated with the NSW change. Similar to the NSW changes, the 15
warming period is from November to April and the maximum change occurs in March. Figure 11 16
(b&c) shows the spatial distribution of changes for surface skin and air temperature, respectively, 17
averaged for March. Regional averaged increases of surface skin (air) temperature are 0.7oC 18
(0.6oC) over CR and 0.2oC (0.1oC) over SN, respectively. The increase in surface skin 19
temperature is 20-50% higher than that of surface air temperature since the incoming energy is 20
partitioned between both sensible and latent heat, as well as melting the snow and warming the 21
sub-surface soil layers. 22
23
Figure 12 24
28
1
Snowpack 2
As a result of soot-induced reduction in NSW and surface warming, there are significant 3
reductions in snowpack between December and May. Figure 13 shows the spatial distribution of 4
SWE change from January to April between the control and soot-perturbed simulations. The 5
SWE decreases 10-50 mm over mountain areas in the western US during late winter to early 6
spring. Maximum reductions of SWE are over CR, SN, and the mountainous areas of western 7
Canada. The precipitation difference between the two simulations is small (not shown), less than 8
0.2% over the two river basins (Table 2). However, the warmer temperature as a result of soot 9
perturbations lead to more precipitation comes in the form of rain rather than snow, while total 10
precipitation amount has no change, resulting in less snow accumulation during the peak snow 11
season. Meanwhile, warmer surface temperatures speed up snowmelt during spring. Driven by 12
reduced snow accumulation in the winter and increased snowmelt in spring, SWE reduction 13
reaches a maximum in March over most of the snow-covered areas except for SN, where the 14
SWE reduction reaches a maximum in April. The mean SWE in March is reduced by 8.8 mm (-15
6.2%) over the CRB basin, 10.4 mm (-11.7%) over CR, 2.7 mm (-3.2%) over the SSJ basin, and 16
8.1 mm (-3.1%) over SN, respectively. Comparing the monthly change of SWE over the two 17
river basins (Figure 12), the SWE reduction lasts longer in SSJ than in CRB because SWE was 18
overpredicted in SSJ and the incoming solar radiation is higher at the lower latitudes, both of 19
which allow changes in NSW, ALB, T2/TS, and SWE to extend into the summer. 20
21
Figure 13 22
23
29
Runoff 1
Figure 14 shows the spatial distribution of runoff from February to May. It can be seen 2
that the runoff increases during late winter because in the soot-perturbed simulation more 3
precipitation comes in the form of rain rather than snow. Figure 14a shows that runoff increases 4
by 0.1-1.0 mm per day in February over the majority of the snow covered areas. Runoff 5
increases in February by 11.0% averaged over CR and by 1.6% averaged over SN, respectively. 6
However, runoff decreases by 0.1-0.7 mm per day in May over major snow covered areas as 7
shown in Figure 14d. Runoff decreases in May by 4.4% averaged over CR and by 1.3% averaged 8
over SN, respectively. Since snow accumulation is less during winter (see Figure 13), runoff 9
from snowmelt is reduced in spring. 10
11
Figure 14 12
13
The spatial distribution of runoff changes shows an interesting mixed pattern in March 14
and April. Driven by the higher rain to snow ratio in the soot-perturbed simulation, the runoff 15
increases in March over about half of the areas. Over the other half less snowmelt from the 16
reduced snowpack leads to runoff decreases for March (Figure 14b). The resulting net mean 17
runoff change is very small (less than 0.6%) in March averaged over the CRB or SSJ basins (see 18
Table 2). Generally, the areas with increased runoff correspond to regions with higher elevation 19
(See Figure 1b) and larger SWE (see Figure 9), whereas the areas with decreased runoff 20
correspond to regions with lower elevation and smaller SWE where snow begins to melt earlier 21
and faster. Runoff changes still show a mixed spatial pattern in April but the ratio of decreases to 22
30
increases grows to 70%; the runoff increases only over a small portion of mountain areas such as 1
SN range and eastern CR ranges. 2
3
Surface water budget 4
To illustrate changes in the seasonal cycle of water partitioning, Figure 15 shows the 5
basin mean monthly changes of precipitation, total runoff, evapotranspiration, soil moisture, and 6
snowpack accumulation for the surface water budget over CRB and SSJ. In both basins, 7
precipitation changes are small throughout the seasonal cycle. This suggests that changes in other 8
components of the water budget are mostly driven by changes in snowmelt or accumulation 9
rather than precipitation amount. As a result of warming, there are significant reductions in 10
snowpack between November and April (Figure 12), which are reflected in reduced snow 11
accumulation between November and February (peak in February) and less snow melt between 12
March and May (peak in April) in Figure 15. Note that the snow accumulation rate (SP) is 13
denoted positive and snowmelt is negative. In snow-dominated basins such as CRB, inter-annual 14
variability of runoff is mainly controlled by the precipitation amount, which is not significantly 15
changed in this study. However, on a monthly basis the runoff is affected by the changes not 16
only in precipitation but also in snow accumulation or melt. As shown in Figure 15, runoff 17
increases during January and March. This is because the higher surface temperature in the soot-18
perturbed simulation causes more precipitation to come in the form of rain rather than snow. By 19
contributing directly to runoff or by causing snowmelt, a higher percentage of rainfall versus 20
snowfall during the cold season increases runoff. As less snow accumulates during winter, runoff 21
as a result of snowmelt between April and June decreases. Soil moisture shows an abrupt change, 22
in Figure 15, from February to March. During winter, increased rainfall and runoff enhanced soil 23
31
moisture. In March and April, soil moisture accumulation is reduced because of reduced 1
snowmelt and increased evapotranspiration caused by warmer temperatures. 2
3
Figure 15 4
5
Changes in the water budget over SSJ are generally similar to changes over CRB, except 6
for a smaller magnitude and delayed timing in changes of snowmelt and runoff. Warmer 7
temperatures in the soot-perturbed simulation also cause snowpack reductions in SSJ. 8
Additionally, snowmelt decreases occur later in the year than for CRB and last a month longer. 9
As a result of snowmelt change, the runoff decreases between May and August rather than 10
between April and June in CRB basin. The change of soil moisture accumulation is small during 11
most months in SSJ. 12
13
Feedbacks between albedo, solar radiation, temperature, and snowpack 14
The soot-induced decrease of maximum snow albedo causes the land surface to absorb 15
more solar radiation, increasing the skin and air temperatures. As a result of the perturbed 16
radiation and warming, snow depth and fraction decrease during winter and spring, further 17
bringing down the albedo over areas with at least some snow cover. This is a positive feedback 18
process initialized by soot deposition on snow [see Figure 16]. 19
20
Figure 11d shows the spatial distribution of mean albedo change in March, which is the 21
month with maximum albedo change. It can be seen that the albdeo decreased 0.02-0.12 over 22
snow covered areas of the Central Rockies. Numerically in the Noah LSM, the simulated albedo 23
32
reduction in the soot-perturbation simulation is caused by two processes: 1) the reduction in 1
maximum albedo (SNOALB) due to the soot perturbations, and 2) the reduction of snowpack 2
from the warmer skin temperature. Comparing the SNOALB and mean albedo changes shown in 3
Table 2, we can find that the SNOALB decreased 5.43% but albedo decreased 9.07% over CR, 4
which indicates that only 60% of the albedo reduction is directly caused by the SNOALB change. 5
Snowpack reduction accounts for the remaining 40%. In the SN region the decrease of mean 6
albedo is only slightly larger than the decrease of SNOALB, which implies that the major 7
contributor to albdeo change over SN was SNOALB instead of the snowpack reduction. In the 8
SN region, snowpack is deep and confined to the high terrain because of the sharp topographic 9
gradients, and the model overpredicted precipitation as well as snowpack. In deep snow, the 10
change in snow fractional coverage due to warming is small. Therefore, the change in ALB is 11
dominated by the change in SNOALB, and the percentage changes of both are smaller in SN and 12
the SSJ basin because the high terrain only accounts for a small fraction of the basin area defined 13
in this study. 14
15
Figure 16 16
17
5. Conclusion and discussion 18
The 2007 IPCC report listed the radiative forcing induced by “soot on snow” as a major 19
anthropogenic forcing affecting climate change between 1750 and 2005. However, it is still 20
uncertain how the soot-induced snow albedo perturbation affects regional snowpack and the 21
hydrological cycle in mountain areas. In this paper we simulate the deposition of soot aerosol on 22
snow using a coupled regional aerosol/chemistry model, WRF-Chem, to estimate the soot-23
33
induced snow albedo perturbation. This is then used to modify the maximum snow albedo for 1
sensitivity tests using WRF as a regional climate model. Investigation of the WRF regional 2
climate simulations focuses primarily on the impact of the soot-induced snow albedo 3
perturbation on snowpack and the hydrological cycle in the western United States. 4
5
The WRF-Chem simulated large spatial variability in BC concentrations that reflect the 6
localized emissions, as well as the influence of the complex terrain and the resulting 7
meteorological processes that affect chemical reactions and transport of pollutants. The 8
simulated BC concentrations are generally higher than the observed values derived from the 9
IMPROVE networks. The climate simulations have been compared against observations for 10
surface air temperature, precipitation, runoff, and snow water. The WRF-RCM model captured 11
the seasonal cycle and inter-annual variability of precipitation very well. The spatial pattern of 12
precipitation is also reasonably simulated during the cold season except for an overestimate of 13
the orographic precipitation along the Cascades and Sierra Nevada mountains. The surface air 14
temperature is spatially consistent between the observations and simulation and the bias is less 15
than 1oC over the CRB and SSJ basins. Through modification of the snow emissivity, surface 16
temperatures were substantially improved in the model, resulting in an accurate reproduction of 17
the phase of precipitation and the snow accumulation and melt processes at the surface. This 18
results in SWE being well simulated in WRF-RCM. The simulation reproduces many of the 19
features of the observed snowpack distribution. The runoff is also evaluated against observation 20
and the spatial distribution of the modeled runoff generally agrees with that of the GRDC runoff 21
in winter and spring. 22
23
34
The WRF-RCM simulations show that soot-induced snow albedo perturbations 1
significantly change the regional climate and water budget in western US. NSW, which is very 2
well spatially correlated with the change of surface albedo, increases 2-14 W m-2 from late 3
winter to early spring over the central Rockies and southern Alberta, which drives a surface skin 4
(air) temperature increase of 0.3 - 2.0oC (0.2 - 1.5oC) over the majority of the snow covered areas 5
in the western U.S. The SWE decreases 10-50 mm over the mountains during late winter to early 6
spring, which is reflected in reduced snow accumulation in winter and less snowmelt in spring. 7
The maximum change occurs in March for albedo, NSW, temperature, and snowpack. Our 8
simulations show that the precipitation difference is negligible between the perturbed and control 9
simulations. In a warmer perturbed climate more precipitation comes in the form of rain rather 10
than snow, which results in less snow accumulation but more runoff during the snowy winter 11
period. The runoff decreased in late spring, driven by the reduced snowmelt from the reduced 12
snowpack. Therefore this indirect forcing of soot has accelerated snowmelt and altered stream 13
flows, including a trend toward earlier melt dates in the western United States. 14
15
In summary, soot-induced albedo reduction causes the snow to absorb more solar 16
radiation, thus heating the snow/land surface and warming the surface air more than if there were 17
no soot. As a result of this warming, snow depth and fraction reduce during winter and spring, 18
which further reduce the albedo for areas fully or partially covered by snow. This is a positive 19
feedback process initialized by the soot deposition on snow. Our simulations indicate that over 20
the central Rockies only 60% of the albedo reduction is directly caused by the change in 21
maximum snow albedo induced by the soot in snow. Snowpack reduction accounts for the 22
remaining 40% of the surface albedo change. 23
35
1
The same feedback process in Figure 16 can also be initialized by increases in 2
greenhouse gases, which lead to skin/air warming and the subsequent changes in the feedback 3
loop. Hence, the suite of changes described in Section 5 are very similar to what have been 4
discussed by studies that investigated changes in mountain hydrology due to greenhouse 5
warming (e.g., Leung et al. [2004]). But while greenhouse forcings are more uniformly 6
distributed spatially (though the response in snowpack is still highly regional), soot forcing on 7
snow is more regional and depends on the co-existence of snowpack and soot. Therefore, larger 8
uncertainties may be expected in estimating both the forcing and response related to soot-9
induced snow albedo effects simply because of the inherent spatial scale of processes involved. 10
Additionally, other uncertainties are introduced by the methodology used in this study, which we 11
plan to address in future studies. 12
13
(a) WRF-Chem simulation and emission data. Because of the extreme computational cost, the 14
WRF-Chem simulation is run only for one year. The interannual variability of concentration and 15
deposition of soot therefore is not accounted for in this study. We used the EPA NEI99 emission 16
dataset in the WRF-Chem simulations and no adjustments were made to bring the 1999 values 17
closer to the 2003 values and no seasonal cycle was included. Also, the NEI99 estimates do not 18
include forest fire or biogenic emissions. However, most of the fires occur during summer and 19
fall, after the snow has melted, so the impact on snowpack may be not significant. 20
21
Since WRF-Chem is a regional model, assumptions must be made regarding trace gases 22
and particulates entering the model boundaries. Fixed profiles were used as inflow boundary 23
36
conditions, so the model does not reproduce the impact of trans-Pacific transport. The net effect 1
of these assumptions is that the deposition in these simulations should be considered a minimum 2
estimate. Yet, our simulated BC concentrations are noticeably higher than the observed values. 3
This suggests a need for more analysis and development to understand and reduce model biases. 4
The WRF-Chem simulation, however, provided representative magnitude and spatial distribution 5
of BC for investigating soot affects on snow in a mountainous region. 6
7
(b) Estimation for snow albedo perturbation. In the current Noah LSM, the snow albedo is 8
calculated based on snow fraction, which is parameterized according to simulated snow depth, 9
and the maximum snow albedo (SNOALB), which is based on the measured snow albedo over 10
deep snow. The SNOALB input has included spatial variability, albeit at a coarser resolution 11
than the model resolution, but not included temporal variation, which is related to other possible 12
effects of snow properties (e.g. grain size) on snow albedo and emissivity. Flanner and Zender 13
[2006] suggested several positive feedbacks amplifying the first-order warming effect of BC in 14
snow. In this study we estimated soot induced-snow albedo reduction based on very limited 15
measurements and an empirical relationship from Jacobson [2004]. We plan to incorporate an 16
interactive aerosol-snow surface radiative model [e.g. Flanner and Zender, 2006] in the land 17
surface model in WRF. Sensitivity experiments using the present model framework (e.g. by 18
halving or doubling the current snow albedo perturbation) could also provide a range for the 19
uncertainties of soot induced snow albedo effects. 20
21
(c) WRF-RCM simulations. The control and sensitivity WRF-RCM simulations are only run 22
for five years due to limited resources. It would be better to extend the simulations to better 23
37
represent the interannual variations of mountain hydroclimate and the impacts of soot. More 1
importantly, although the spatial pattern and seasonal cycle of precipitation are simulated 2
reasonably well by WRF-RCM during the cold season, an overestimation of the orographic 3
precipitation is also apparent along the Cascades and Sierra Nevada as well as CRB and SSJ 4
basins. These errors result in biases in simulating snowpack and runoff. The overpredictions of 5
both precipitation (hence snowpack) and BC concentrations have likely biased our results in the 6
same direction toward larger impacts of soot-induced snow effects. Lastly, WRF-Chem and 7
WRF-RCM are run separately (offline) so interactions between the concentration and deposition 8
of soot and climate that would modify the soot-induced snow albedo perturbation are not 9
addressed. 10
11
12
13
Acknowledgments. This research is supported by a Pacific Northwest National Laboratory 14
(PNNL) Laboratory Directed Research and Development (LDRD) project. PNNL is operated for 15
the U.S. DOE by Battelle Memorial Institute under Contract DE-AC06-76RLO1830. This 16
research used resources of the National Center for Computational Sciences at Oak Ridge 17
National Laboratory, which is supported by the Office of Science of the U.S. Department of 18
Energy under Contract No. DE-AC05-00OR22725. 19
20
38
1
2
References 3
Barlage, M., X. Zeng, H. Wei, and K. E. Mitchell (2005), A global 0.05 maximum albedo dataset of 4 snow-covered land based on MODIS observations, Geophys. Res. Lett., 32, L17405, 5 doi:10.1029/2005GL022881. 6
Barnett, T.P., L. Dumenil, U. Schlese, and E. Roeckner (1988), The effect of Eurasian snow cover on 7 global climate, Science, 239, 504-507. 8
Barry, R.G. (1996), The parameterization of surface albedo for sea ice and its snow cover, Progress in 9 Physical Geography, 20, 63-79. 10
Betts, A. K., J. H. Ball (1997), Albedo over the boreal forest, J. Geophys. Res.,102 (D24): 28901-28909. 11 Brown, R. D., B. Brasnett, and D. Robinson (2003), Gridded North American monthly snow depth and 12
snow water equivalent for GCM evaluation, Atmosphere-Ocean 41(1), 1-14. 13 Cess, R.D., G.L. Potter, et al. (1991), Intercomparison and interpretation of snow-climate feedback 14
processes in seventeen atmospheric general circulation models, Science, 253, 888-892. 15 Chen, F, J. Dudhia (2001), Coupling an advanced land surface-hydrology model with the Penn State-16
NCAR MM5 modeling system. Part I: Model implementation and sensitivity, Mon. Wea. Rev. 129 (4): 17 569-585. 18
Christensen, J.H., B. Hewitson, A. Busuioc, A. Chen, X. Gao, I. Held, R. Jones, R.K. Kolli, W.-T. Kwon, 19 R. Laprise, V. Magaña Rueda, L. Mearns, C.G. Menéndez, J. Räisänen, A. Rinke, A. Sarr and P. 20 Whetton (2007), Regional Climate Projections. In: Climate Change 2007: The Physical Science Basis. 21 Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on 22 Climate Change [Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor 23 and H.L. Miller (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, 24 NY, USA. 25
Chylek, P., V. Srivastava, L. Cahenzli, R. G. Pinnick, R. L. Dod, and T. Novakov (1987), Aerosol and 26 graphitic carbon content of snow, J. Geophys. Res., 92(D8), 9801–9809. 27
Clarke, A., and K. Noone (1985), Soot in the Arctic: A cause for perturbation in radiative transfer, Atmos. 28 Environ., 19(12), 2045–2053. 29
Colle, B.A., C.F. Mass, and K.J. Westrick (2000), MM5 precipitation verification over the Pacific 30 Northwest during the 1997-99 cool seasons. Wea. and Forecasting, 15, 730-744. 31
39
Conway, H, A. Gades, C. F. Raymond (1996), Albedo of dirty snow during conditions of melt, Water 1 Resour. Res. 32 (6): 1713-1718. 2
Cooke, W. F., C. Liousse, H. Cachier, and J. Feichter (1999), Construction of a 1x1 fossil fuel emission 3 data set for carbonaceous aerosol and implementation and radiative impact in the ECHAM4 model, J. 4 Geophys. Res., 104(D18), 2137–2162. 5
Davidson, C. I., S. Santhanam, R. C. Fortmann (1985). Atmospheric Transport and Deposition of Trace-6 elements onto the Greenland ice-sheet. Atmospheric Environment. 19 (12): 2065-2081. 7
Easter, R. C., S. J. Ghan, Y. Zhang, R. D. Saylor, E. G. Chapman, N. S. Laulainen, H. Abdul-Razzak, L. 8 R. Leung, X. D. Bian, and R. A. Zaveri (2004), MIRAGE: Model description and evaluation of 9 aerosols and trace gases, J. Geophys. Res., 109, D20210, doi:10.1029/2004JD004571. 10
Essery, R., J. Pomeroy, J. Parviainen, and P. Storck (2003), Sublimation of snow from 11
coniferous forests in a climate model, J. Climate, 16, 1855-1864. 12
Fahey, K. M., and S. N. Pandis (2001), Optimizing model performance: variable size resolution in cloud 13 chemistry modeling, Atmos. Environ., 35, 4471-4478. 14
Fast, J. D., et al. (2006), Evolution of ozone, particulates, and aerosol direct radiative forcing in the 15 vicinity of Houston using a fully coupled meteorology-chemistry-aerosol model, J. Geophys. Res., 16 111, D21305, doi:10.1029/2005JD006721. 17
Fekete, B. M., C. J. Vorosmarty, and W. Grabs (2000), Global composite runoff fields based on observed 18 discharge and simulated water balance, Rep. 22, Global Runoff Data Cent., Koblenz, Germany. 19 (Available at http://www.grdc.sr.unh.edu/html/paper/ReportUS.pdf). 20
Feng, S., and Q. Hu (2007), Changes in winter snowfall/precipitation ratio in the contiguous United States. 21 J. Geophys. Res., 112, D15109, doi:10.1029/2007JD008397. 22
Flanner, M. G., and C. S. Zender (2006), Linking snowpack microphysics and albedo evolution, J. 23 Geophys. Res., 111, D12208, doi:10.1029/2005JD006834. 24
Flanner, M. G., C. S. Zender, J. T. Randerson, and P. J. Rasch (2007), Present-day climate forcing and 25 response from black carbon in snow, J. Geophys. Res., 112, D11202, doi:10.1029/2006JD008003. 26
Ghan, S., R. Easter, J. Hudson, and F. M. Breon (2001), Evaluation of aerosol indirect radiative forcing in 27 MIRAGE, J. Geophys. Res., 106, 5317-5334. 28
Ghan, S. J., T. Shippert (2006), Physically based global downscaling: Climate change projections for a 29 full century, J. Clim. 19 (9): 1589-1604. 30
Grell, G. A., S. E. Peckham, R. Schmitz, S. A. McKeen, G. Frost, W. C. Skamarock, and B. Eder (2005), 31 Fully coupled "online" chemistry within the WRF model, Atmos. Environ., 39, 6957-6975. 32
Grenfell, T. C., D. K. Perovich, J. A. Ogren (1981), Spectral Albedos of an Alpine Snowpck. Cold 33 Regions Science and Technology. 4 (2): 121-127. 34
40
Gustafson, W. I., et al. (2007), Impact on modeled cloud characteristics due to simplified treatment of 1 uniform cloud condensation nuclei during NEAQS 2004, Geophys. Res. Lett., 34, L19809, 2 doi:10.1029/2007GL0300321. 3
Hansen J, and L Nazarenko. 2004. “Soot climate forcing via snow and ice albedos.” Proc. Natl. Acad. Sci. 4 101: 423-428. 5
Hansen, J., et al., 2005: Efficacy of climate forcings. J. Geophys. Res., 110, D18104, 6 doi:10.1029/2005JD005776. 7
Hartman, R. K., A. A. Rost, and D. M. Anderson (1995), Operational processing of multi-source snow 8 data. Proc. Western Snow Conf., Sparks, Nevada, 147–151. 9
Hong, S.-Y., Y. Noh, and J. Dudhia (2006), A new vertical diffusion package with an explicit treatment 10 of entrainment processes. Mon. Wea. Rev., 134(9), 2318–2341. 11
IPCC, 2007: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the 12 Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, 13 M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.)]. Cambridge 14 University Press, Cambridge, United Kingdom and New York, NY, USA, 996 pp. 15
Jacobson, M. (2001), Strong radiative heating due to the mixing state of black carbon, Nature, 409, 695–16 697. 17
Jacobson MZ. (2004), Climate response of fossil fuel and biofuel soot, accounting for soot's feedback to 18 snow and sea ice albedo and emissivity. J. Geophys. Res. Art. No. D21201. 19
Kanamitsu, M., W., Ebisuzaki, J. Woollen, S-K Yang, J.J. Hnilo, M. Fiorino, and G. L. Potter (2002), 20 NCEP-DOE AMIP-II Reanalysis (R-2). Bul. of the Amer. Met. Soc. 1631-1643. 21
Leung, L. R., Y. Qian Y (2003), The sensitivity of precipitation and snowpack simulations to model 22 resolution via nesting in regions of complex terrain. J. Hydrometeorol. (6): 1025-1043. 23
Leung, L. R., Y. Qian, J. Han, J. O. Roads (2003), Intercomparison of global reanalyses and regional 24 simulations of cold season water budgets in the western United States, J. Hydrometeorol. 4 (6): 1067-25 1087. 26
Leung, L. R., Y. Qian, X. D. Bian, W. M. Washington, J. G. Han, J. O. Roads (2004), 27 Mid-century ensemble regional climate change scenarios for the western United States. 28 Climatic Change 62 (1-3): 75-113. 29
Leung, L.R., J. Done, J. Dudhia, T. Henderson, M. Vertenstein, and B. Kuo (2005), Preliminary results of 30 WRF for regional climate simulations. Workshop on Research Needs and Directions of Regional 31 Climate Modeling Using WRF and CCSM, Boulder, CO. [Available online at 32 http://box.mmm.ucar.edu/events/rcm05/presentations/leung_rcm_workshop.pdf.]. 33
41
Lynch, A., D.L. McGinnis and D.A. Bailey (1998), The seasonal cycle and snow melt in a regional 1 climate system model: Influence of land surface model, J. Geophys. Res., 103, 29037-29049. 2
Marshall S, and R. J. Oglesby (1994), An Improved Snow Hydrology For Gcms .1. Snow Cover Fraction, 3 Albedo, Grain-Size, And Age. Climate Dynamics. 10 (1-2): 21-37. 4
Marshall, S, R. J. Oglesby, A. W. Nolin (2003), The predictability of winter snow cover over the western 5 United States. J. Clim. 16 (7): 1062-1073. 6
Maurer, E. P., G. M. O’Donnell, D. P. Lettenmaier, and J. O. Roads (2001), Evaluation of the land 7 surface water budget in NCEP/NCAR and NCEP/DOE Reanalyses using an off-line hydrologic 8 model. J. Geophys. Res., 106 (D16), 17 841–17 862. 9
Mote, P., Hamlet A. F., Clark M. P., and Lettenmaier D. P. (2005), Declining mountain snowpack in 10 western North America. Bul. of the Amer. Met. Soc. 86(1): 39-49. 11
Niu, G.-Y., and Z.-L. Yang (2007), An observation-based formulation of snow cover fraction and its 12 evaluation over large North American river basins, J. Geophys. Res., 112, D21101, 13 doi:10.1029/2007JD008674. 14
Pielke, RA (1984), Mesoscale Meteorological Modelling. Academic Press, Inc., London. 15 Qu, X., and A. Hall (2006), Assessing snow albedo feedback in simulated climate change, J. Clim., 19, 16
2617–2630. 17 Randall, D.A., R.D. Cess, et al. (1994), Analysis of snow feedbacks in 14 general circulation models, J. 18
Geophys. Res., 99, 20757-20771. 19 Robinson, D. A., and G. Kukla (1985), Maximum surface albedo of seasonally snow-covered lands in the 20
Northern Hemisphere, J. Clim. Appl. Meteorol., 24, 402– 411. 21 Schlosser, C. A., A. G. Slater, A. Robock (2000), Simulations of a boreal grassland hydrology at Valdai, 22
Russia: PILPS phase 2(D), Mon. Wea. Rev. 128 (2): 301-321. 23 Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang, J. G. Powers. (2005), A 24
Description of the Advanced Research WRF Version 2. NCAR/TN–468+STR NCAR TECHNICAL 25 NOTE. National Center for Atmospheric Research, Boulder, Colorado, USA. 26
Thomas, G., and P. R. Rowntree (1992), The boreal forests and climate, Q. J. R. Meteorol. Soc., 118, 27 469–497. 28
Taylor, K.E., D. Williamson, and F. Zwiers (2000), The sea surface temperature and sea-ice concentration 29 boundary conditions for AMIP II simulations, PCMDI Report No. 60, Program for Climate Model 30 Diagnosis and Intercomparison, Lawrence Livermore National Laboratory, Livermore, California, 25 31 pp . 32
Walland, D. J., I. Simmonds (1996), Modelled atmospheric response to changes in Northern Hemisphere 33 snow over, Climate Dynamics, 13 (1): 25-34. 34
42
Warren, S. G., and W. J. Wiscombe. (1980), A model for the spectral albedo of snow. II. Snow containing 1 atmospheric aerosols. J. Atmos. Sci., 37, 2734-2745. 2
Warren, S. G., and W. J. Wiscombe (1985), Dirty snow after nuclear war. Nature, 313, 467-470. 3 Westrick, K.J., and C.F. Mass (2001), An evaluation of a high-resolution hydrometeorological modeling 4
system for prediction of a cool-season flood event in a coastal mountainous watershed. J. 5 Hydrometeorol., 2, 161-180. 6
Wild, O., X. Zhu, and M. J. Prather (2000), Fast-J: Accurate simulation of in- and below cloud photolysis 7 in tropospheric chemical models, J. Atmos. Chem., 37, 245 –282. 8
Xu, K.M., and D.A. Randall (1996), Explicit simulation of cumulus ensembles with the GATE Phase III 9 data: Comparison with observations, J. Atmos. Sci., 53, 3710–3736. 10
Yang, Z.-L. and G.Y. Niu (2000), Snow-climate interaction in NCAR CCM3, Preprints, 15th Conference 11 on Hydrology, American Meteorological Society Meeting, Long Beach, CA, USA, January 2000, pp. 12 24-27. 13
Zaveri, R. A., and L. K. Peters (1999), A new lumped structure photochemical mechanism for large-scale 14 applications, J. Geophys. Res., 104, 30,387– 30,415. 15
Zaveri, R. A., R. C. Easter, and A. S. Wexler (2005a), A new method for multicomponent activity 16 coefficients of electrolytes in aqueous atmospheric aerosols, J. Geophys. Res., 110, D02201, 17 doi:10.1029/2004JD004681. 18
Zaveri, R. A., R. C. Easter, and L. K. Peters (2005b), A computationally efficient Multicomponent 19 Equilibrium Solver for Aerosols (MESA), J. Geophys. Res., 110, D24203, 20 doi:10.1029/2004JD005618. 21
Zaveri, R. A., R. C. Easter, J. D. Fast, and L. K. Peters (2008), Model for Simulating Aerosol Interactions 22 and Chemistry (MOSAIC), J. Geophys. Res., in press. 23
24
25
43
Figure Captions: 1 2
1. Model domain and subregions (a), elevation (b) and daily Black Carbon emission (c) for 3
particle diameter <10µm (unit: mg m-2 day-1). 4
2. Spatial distribution of WRF-Chem simulated and IMPROVE observed BC concentration 5
(particle diameter < 10µm) at near surface for (a) MAM, (b) JJA, (c) SON and (d) DJF. 6
3. Monthly averaged BC concentrations for the IMPROVE stations within the domain 7
along with the coincident WRF-Chem grid point averages. The minimum and maximum 8
value is given by the narrow bar and 25th percentile and 75th percentile is given by the 9
wider bar. 10
4. Accumulated BC deposition (dry + wet) for particles less than 10µm in diameter for the 11
period of (a) DJF and (b) MAM. Depositions units are µg m-2. Areas in WRF-Chem with 12
seasonal mean snow cover greater than 1 cm in depth are overlaid with white hatching. 13
5. DJF averaged (a) BC-SNOW mixing ratio (ng/g) and (b) corresponding snow albedo 14
perturbation. 15
6. The spatial distribution of (a) maximum snow albedo (SNOALB) used by the standard 16
WRF Noah Land Surface Model (LSM) and (b) mean albedo simulated for March in the 17
WRF-RCM control simulation. 18
7. Spatial distribution of simulated (left) and observed (right) precipitation in mm day-1 (top) 19
and surface air temperature in oC (bottom) for DJF. 20
8. Monthly mean time series of simulated and observed precipitation and surface air 21
temperature for CRB and SSJ basin. 22
9. Spatial distribution of simulated (a) and three observed (b:CMC; c:NOHRSC; d: 23
SNOTEL) snow water equivalent datasets (SWE, unit: mm) for DJF and simulated (e) 24
and CMC (f) SWE for MAM. 25
10. Spatial distribution of simulated (a) and observed (b) runoff (GRDC) for MAM, unit: mm 26
day-1. 27
11. The spatial distribution of change in March for (a) net shortwave radiation flux at the 28
surface (NSW, W m-2), (b) surface (2 meter) air temperature (oC), (c) skin temperature 29
(oC), and (d) surface albedo. 30
44
12. Changes in monthly mean Snow Water Equivalent (SWE, unit: 10mm), albedo (*0.1), 1
surface Net Shortwave radiation flux (NSW, *5 W m-2), 2-meter air temperature (oC) and 2
skin temperature (oC) averaged over the CRB and SSJ basin. 3
13. The spatial distribution of change in mean Snow Water Equivalent (SWE, mm) from 4
January to April. 5
14. The spatial distribution of change in runoff (mm/day) from February to May. 6
15. Changes in monthly mean surface water budget averaged over the CRB and SSJ basin. 7
Shown in the figure are changes in precipitation (P), snowpack accumulation rate (SP), 8
runoff (R), soil moisture accumulation rate (SM), and evapotranspiration (ET) in mm/day. 9
16. Diagram for feedbacks between albedo, solar radiation, temperature, and snowpack.10
45
1
Table 1. Comparison of measured and simulated soot mixing ratio in snow and corresponding snow albedo perturbation Measured, Cascade, Washington 22-59 ng/g (melting or wet snow) Grenfell et al., 1981. Measured, New Mexico, West Texas (32N, 106W)
4.9 – 15.9 ug/L Chylek et al., 1987 Measured, Hurricane Hill (48N, 123.5W)
10.1-18.5 (ng/g) Clarke and Noone, 1985 GCM Simulated, NH Land
20-60 ng/g (1.5-14%) Hansen and Nazarenko, 2004 GCM Simulated, western US
10-46 ng/g Flanner et al., 2007 GCM Simulated, New Mexico, West Texas, Washington
14-32 ng/g Jacobson, 2004. GCM Simulated, Global average
5 ng/g 0.4% (1.0%, NH) Jacobson, 2004. This study, Western US
10-120 ng/g (1 – 10%)
Table 2 Mean changes in March in maximum snow albedo (maxsnoalb), surface albedo (albedo), surface downward solar radiation flux (downsw, W/m2) and net solar radiation flux (nsw, W/m2), surface air temperature (t2, C) and skin temperature (ts, C), precipitation 9mm/day), evaporation (mm/day), snow water equivalent (swe, mm), soil water content (mm), and runoff (mm/day) averaged for CRB and SSJ basins as well as CR and SN regions (see Figure 6). Change in percentage is in bracket. CRB SSJ CR SN
maxsnoalb -0.026 (-5.52%) -0.008 (-1.45%) -0.031 (-5.43%) -0.015 (-2.81%)
albedo -0.019 (-6.51%) -0.004 (-1.57%) -0.033 (-9.07%) -0.010 (-2.99%)
downsw -0.406 (-0.23%) 0.054 (0.025%) -1.112 (-0.53%) -0.388 (-0.17%)
nsw 3.277 ( 2.58%) 0.833 ( 0.51%) 6.360 ( 4.77%) 2.066 ( 1.35%)
t2 0.319 0.071 0.557 0.106
ts 0.398 0.086 0.721 0.156
prcp 0.006 ( 0.19%) 0.009 ( 0.18%) 0.021 ( 1.02%) 0.019 ( 0.29%)
evap 0.025 ( 2.23%) 0.010 ( 0.54%) 0.041 ( 4.04%) 0.020 ( 1.30%)
swe -8.800 (-6.22%) -2.689 (-3.15%) -10.363 (-11.70%) -8.135 (-3.06%)
soilm 1.607 ( 0.25%) -0.154 (-0.025%) 3.788 ( 0.69%) 0.546 (0.094%)
runoff 0.015 ( 0.50%) 0.025 ( 0.58%) 0.027 ( 1.57%) 0.083 ( 1.35%)
1: Central Rockies (CR); 2: Sierra Nevada (SN)
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11
Figure 12
Figure 13
Figure 14
Figure 15
Surf
ace
Alb
edo
Skin
and
Air
War
ming
Snow
pack
an
d Sn
owFr
action
Surf
ace
Net
Solar
Radiat
ion
Max
imum
Albed
oov
er
Dee
p Sn
owSo
ot o
n Sn
ow
Figu
re 1
6