Академический Документы
Профессиональный Документы
Культура Документы
120
ical controls on forest transpiration, require an
accurate estimate of snowpack amount and loss in
order to accurately time soil water depletion and
the onset and severity of summer water stress
(Giles et al. 1985; Running 1984; Black and Spittelhouse 1981; Troendle and Leaf 1980).
Forest ecosystem process models predominately
use daily to monthly time steps to model the forest
hydrologic budget and its effect on canopy processes like net photosynthesis and transpiration
(McMurtrie 1992; Ryan 1993; Friend et al. 1993).
One general forest ecosystem model, FORESTBGC (Running and Coughlan 1988; Running and
Gower 1991) has demonstrated how LAI and daily
climate data can explain differences in forest carbon and water cycles without site specific model
calibration.
Process models represent snowpack as a state
variable contains a mass of water (M3/HA) which
accumulates during the winter months and releases
during the spring months. The melt water can be a
significant source of water for generating spring
runoff and resaturating the soil to provide water for
transpiration and growth (Running and Coughlan
1988). For site water balance modeling, simulations can use data describing spring conditions
such as an accumulated and isothermal snowpack
(McLeod and Running 1988). If initialized with an
isothermal snowpack, a process model like FOREST-BGC can employ a simple air temperature
index melt model. There are critical limitations,
however, to an isothermal snowpack model including obtaining the initial amount of snow and estimating when it becomes isothermal. This data can
be measured for a single site, but for landscape level simulations in mountainous terrain data are often
difficult to obtain and are highly spatially variable.
One approach to modeling regional hydrologic
budgets is to divide a heterogeneous study area into
smaller, more homogeneous sub-sites, and execute
individual model runs on each sub-site (Nemani et
al. 1993; Paniconi and Wood 1993). For mountainous terrain, the amount of snow accumulation and
the rate of snowmelt are partially dependent on
landscape features that tend to be highly variable
over short distances (U.S. Army Corps 1956). For
parameterizing a temperature index snow model
for remote sites or regions, snow amounts and the
onset of isothermal snowpack conditions must be
estimated remotely. Direct measurements of
121
Coughlan 1988) and the climate simulator model
MTCLIM (Running et al. 1987). The RHESSys
Snow Model (RSM) was built to compute a spatially varying snowpack for regional applications of
FOREST-BGC in RHESSys. This section describes
RSM and the ecosystem process model FORESTBGC, and the climate simulator, MTCLIM, as they
pertain to the RSM.
122
T min + T day
Tnight = __________
2
(1)
important feature of both models is using a definition of forest structure (leaf, stem and root biomass) obtainable from remotely sensed forest LAI.
Previous stand level models (Leaf and Brink
1973a; Solomon et al. 1976) used canopy closure
or cover density from aerial photographs to sample
catchment forest cover. In the RSM, a catchment
can be parameterized using satellite derived LAI
and topography derived from digital elevation
models. Another change from previous snow models is having a separate climatic simulator,
MTCLIM, which can explicitly estimate daily S
from integration of potential inputs at hourly intervals mitigated by daily transmissivity estimated
from air temperature amplitude.
(2)
Cint
____
qsub = MIN LAI .
_____
,
+
(3)
123
interception coefficient (1.0 mm LAI1). Although
water losses from canopy interception of snow can
be greater than of rain (Waring and Schlesinger
1985), FOREST-BGC's rain interception coefficient also includes interception losses from the litter layer. Sublimation driven by turbulent transfer
is not incorporated in this formulation.
(4)
where is the current temperature sum for determining isothermal snow conditions (C), 1 is
the temperature sum of the previous day (C), is
the temperature increment parameter (C), and T24
is the weighted sum of the daily maximum and
minimum air temperatures (C) (Equation 5). is
initialized to 0.0C with a new snowpack and is
defined for all days with a snowpack. Snowfall
events do not change nor is the calculation
dependent on the mass of water in the snowpack.
(5)
124
around, a seasonally adjusted decay rate is calibrated empirically (Leaf and Brink 1973) to reflect the
changing albedo-time relationship. A temperature
based albedo decay need not be seasonally adjusted
because by definition it will have a slower decay
rate in the cold winter month and correspondingly
an accelerated decay rate during the warm melt
phase (U.S. Army Corps 1956; Baker et al. 1990).
Snow surface albedo is calculated as
t = MIN (t1), max . ___ (6)
LAI
C .___
ext 2.2
Sf = S . e .
(7)
Sf
qmelt = ___
(8)
S
f
qsub = ___
+
(9)
87,88,89
87,88,89
85,86,87
84,85,86
86,88,89
87,88,89
88,89,90
85,86,87
88,89,90
88,89,90
30
101
30
30
30
30
30
30
30
30
0.52
1.0
1.0
1.0
1.0
1.0
0.52
1.0
0.52
1.0
1.0
1.0
1.3
1.3
1.4
1.5
1.0
1.0
1.3
1.0
qmelt
Snow ppt.
(C) coeff.
Multiplier
(mm/C)
31
39
70
59
29
40
78
29
27
25
23
21
n/a
n/a
n/a
n/a
68
n/a
22
n/a
274
391
1112
853
279
789
523
170
208
358
94
171
463
336
125
359
206
99
76
143
11
10
6
7
10
5
15
17
13
7
33
23
15
18
23
11
38
29
36
17
24
12
n/a
n/a
n/a
n/a
27
n/a
29
n/a
Outputs
1Generic
value of 30 () was used in generic results; 2Generic value of 1.0 (mm/C) was used in generic results; 3 = | observed - predicted | / n, where n = number of days
observed or predicted snow > 0; 4Max Snow is maximum observed SWE over 3 water years; 5Mean Snow is average observed SWE for days with snowpack for the 3 water
years; 6Max is as a % of Max Snow; 7Mean is as % of Mean Snow.
34.3
34.2
38.2
38.1
40.1
40.3
44.2
42.2
46.5
46.5
2225
2417
2682
2896
2658
3261
1067
1756
1426
1905
AZ1
AZ2
CA1
CA2
CO1
CO2
OR1
OR2
MT1
MT2
Baker Butte
Promontory
Sonora Pass
Virginia Lakes Ridge
Stillwater Creek
Lake Irene
Jump Off Joe
Quartz Mountain
Lubrecht Flume
NFK Elk Creek
Site data
Elevation Latitude Water years
(m)
()
OctSept
SNOTEL
Abbrev. Station ID name
Inputs
Table 1. SNOTEL Site descriptions, summary input data and RESSys Snow Model (RSM) parameters are listed for generic and adjusted simulations. Output results are summarized with predicted (snow water equivalent) SWE errors () which average results from 3 water year simulations. Adjusted simulations were run for sites with high and listed,
sites without adjusted results were filled with n/a.
125
126
Equation (8) with the exception of using the sum of
and (MJ mm1 M2). Total sublimation is the
sum of sublimation from intercepted snow held in
the canopy (Equation 3) and from snowpack on the
forest floor (Equation 9).
3. Methods
3.1. Study sites
10 study sites were used with site data collected
during 3 water years (defined as October to September). Two sites, each pair from five western
states: Arizona, California, Colorado, Montana and
Oregon, were chosen (Fig. 1) to represent some of
the more diverse snow regimes found in western
North America as described by Baker (1944). The
site data were collected using the Central Forecast
System's central computer database administered
by the Soil Conservation Service (SCS) (Soil Conservation Service 1985). The selection criteria for
the 10 sites were: (1) Station pairs located within
the same state, and as close as possible to one
another so that they have similar mesoscale climate
influences, (2) each state has a high and low elevation station, and (3) each station has 3 water years
data collected during the snow season.
All SNOTEL stations are located in mountainous areas in forest clearings of variying size and
were simulated with an LAI of 0.0. The one exception is the Baker Butte AZ SNOTEL station which
is located along an abandoned and very narrow
road that runs north to south, and the station is
heavily shaded to the east and west. An LAI of 1.0
was used to represent the snow pillow shading
(USGS SNOTEL site description).
127
3.4. Forest cover and topography
Sensitivity tests were run for an Oregon and Colorado station and both Montana stations using
combinations of aspect, elevation, and LAI. Two
stations were selected that represent the most
extreme conditions: Jump Off Joe, Oregon (1067
m), a warm maritime climate with the lowest elevation of all ten stations and Lake Irene, Colorado
(3261 m), a cold continental climate and the highest elevation station (Baker 1944). Both Montana
SNOTEL stations represent an elevation gradient
since they are located in the same small catchment
(17 km2) with a 500 m elevation difference.
For each of the 3 sites, a set of simulations were
made for LAI's of 0 and 6, to represent a forest
clearing and a typical forest stand. An LAI of 6 is
half the maximum LAI observed for lodgepole
stands in the Central Rocky Mountains (Kaufmann
et al. 1982) and is the LAI value used to represent
an average forest stand. In addition to LAI, three
aspects; facing north and south with 25 slope and
a flat surface, were chosen to show the effect of
forest cover and aspect combinations. Elevation
sensitivity tests were replicated in Montana on a
dry (1990) and wet (1991) water year.
Motoyama (1990).
Summaries of daily SWE predictions are in
Table 1 where indicates how well predicted SWE
tracked observed SWE for the three snow seasons.
was less for higher elevation and colder stations
while warmer and lower elevation stations, like
Arizona and Oregon, experienced intermittent winter melt episodes and had a greater . In general, it
is easier to accumulate and melt snow when there
is a well defined accumulation and ablation season,
as is found in the higher elevation study sites.
Precipitation gauge corrections (Table 1) reconciled differences between snow pillow SWE and
SWE estimated from gauged precipitation for
SNOTEL stations in Colorado, California and
Montana. When a precipitation gauge clogs, precipitation can be approximated by accumulated
snow increments. To correct for the errors, a precipitation correction coefficient adjusts gauge measurements to match snow accumulation (U.S. Army
Corps 1956; Motoyama 1990; Hinzman and Kane
1991). The alternative explanation for discrepancies between pillow and precipitation measurements at the same location is snow redistribution
from the canopy to the snow pillow. Snow redistribution is well documented for coniferous forests
ranging from Colorado (Gray and Troendle 1967)
to Oregon (Berris and Harr 1987) to Alberta (Golding and Swanson 1978). This study used gauge corrections to SWE deficiencies because the coefficients were approximated proportional to elevation
and snowfall intensity which indicates gauge clogging (Hinzman and Kane 1991).
4.1.2. Snowpack depletion dates
A scatter plot (Fig. 3) of observed to predicted
snow depletion dates was made for all 30 simulations using generic RSM parameter values of =
30 and qmelt = 1.0 mm/1C (Table 1). Each station
is plotted three times, once for each water year, and
is identified by its state abbreviation and station
number given in Table 1. Located at the upper right
of the scatter plot (Fig. 3) are the snowy, high elevation sites like Colorado and California, while the
lower left are warmer and lower elevation sites of
Oregon and Arizona. A two week interval drawn as
dotted parallel lines in the scatter plot (Fig. 3) contained 18 of the 30 predicted dates. The average
error between observed and predicted dates is 6.2
128
Fig. 3. Scatter plot and regression line for observed and predicted snow depletion dates for each SNOTEL station and
water year. Stations are identified by state abbreviation and station numbers from Table 1, and the abbreviation's center letter
marks the exact date. Lines defining an error interval of one
week contain 18 of the 30 simulations.
4.2. Albedo
Both snow depletion and daily SWE predictions
require an accurate estimate of S and snow albedo
(Aguado 1985). Snow surface albedo is usually
calculated as a function of time because time is
both a good predictor and easy to measure (Baker
et al. 1990). Models that use time usually divide
the snow season into a pre- and post-ablation
phase. In the pre-ablation phase albedo decay is
slow while in the post phase albedo decay is accelerated (Gray and Landine 1987; Loth and Graf
1993). Baker et al. (1990) demonstrated that an
albedo calculation with temperature index, , can
also predict albedo. Air temperatures change from
the winter to spring so a calculation need not
define a pre and post ablation phase. directly
relates albedo decay to air temperature which
129
makes the albedo calculation more adaptive
through the snow season and also between sites.
130
Fig. 6. Simulated snow loss attributed to sublimation, temperature melt and radiation melt for Jump Off Joe, Oregon (1067
m) and Lake Irene, Colorado (3261 m) under a forest canopy
(LAI = 6) and in the open (LAI = 0).
131
isothermal conditions as . Therefore, the thermal
quality of the snow that drops off the canopy is, for
purposes of model simplicity, set to the snowpack's
condition, , although snow drop may really occur
as isothermal clumps or as melt water. A second
compensating error is in sublimation. Canopy snow
is sublimated with S and snow cover is sublimated
with S during sub-freezing conditions. This design
may generate an overestimate of sublimation from
S. On the opposing side, no turbulent transfer driven sublimation process is in RSM. Forest-BGC
and RHESSys does not have a dynamic windspeed
variable so the inclusion of a wind driven melt
process such as turbulent transfer would add to the
model's data needs and violate a design requirement. Further model verification is necessary to
determine if wind speed, a difficult variable to
obtain and highly variable in mountainous terrain,
is needed to improve model results.
Field data from Berris and Harr (1987) documented higher daylight air temperatures and melt
rates in the Oregon forest clearings than under the
adjacent canopy. Their reported air temperature differences were highest when S was prominent
showing a strong link between air temperature and
incoming S. Their microclimate and snowmelt data
indicate that a canopy reduces both thermal and
solar components of melt, as is predicted by RSM.
S melt dominates the temperature melt process in
the open simply because S is much greater in the
open. These data indicate that the major melt
process driving RSM are tracking observed microclimate conditions in the open and forest but the
data do not indicate if RSM snowmelt components
are correct except in relative magnitudes.
4.4.2. Topography and LAI
Aspect and elevation sensitivity were tested using
the same two Oregon and Colorado stations and
both Montana stations which are located in the
same small basin and also represent on an elevation
gradient of 500 m.
Aspect
Although their climates were different, the relative
melt rates between the three surfaces; N, S and flat,
were similar in both Oregon and Colorado climates. Colorado was more sensitive to aspect than
Oregon, which is a result of Colorado's clearer
132
Table 2. Observed and simulated snowpack depletion dates showing snow depletion sensitivity to elevation, LAI, and aspect. Elevation
sensitivity was replicated at the MT stations and entered as year 2 while all single year observations were placed into year 1. Locations
are all in the inter-mountain states of western U.S.: CO, MT, and NM. Observed depletion dates are from the literature and simulated
dates are made with SNOTEL station data, no climatic data were available for simulating snow depletion at literature sites. The SNOTEL depletion data is listed for observed, flat, LAI = 0 entries in CO and MT and simulation results were made for those exact sites.
Measurement Year Elevation
Observed
High
Low
2
Simulated
High
Low
High 1
Low 1
High 1
Low
Aspect
Location
CO
Flat
MT4
LAI 0
09-May
LAI 1.53.0 20-May1
LAI 56
02-Jun1
LAI 9
LAI 0
20-Apr
LAI 1.53.0
LAI 56
LAI 0
10-May
LAI 0
19-Apr
LAI 0
13-May
LAI 56
06-Jun
LAI 0
12-Apr
LAI 56
22-Apr
LAI 0
18-May
LAI 56
09-Jun
LAI 0
13-Apr
LAI 56
22-Apr
CO
North
NM3
MT4
CO
South
NM
06-May3
06-May3
17-Jun3
10-Jun3
MT4
13-May2
08-Apr2
09-Jun
26-Jun
16-May
07-Jun
18-Apr
26-Apr
23-May
10-Jun
19-Apr
25-Apr
05-Jun
23-Jun
12-May
05-Jun
10-Apr
21-May
17-May
08-Jun
10-Apr
21-Apr
1Observed LAI is 1.52.5 and 5.0 both at 2680 m in Colorado (CO) (Gray and Troendle 1982); 2N slope with LAI 5 and S slope with
LAI 3 at 2750 m site in New Mexico (NM) (Gray and Coltharp 1967); 3N and S hillslope pairs with and without LAI at 3400 m with 7
day sampling interval (Gray and Coltharp 1967); 4All MT data is from SNOTEL stations MT1 and MT2 in Table 1 and for water years
'89 (1) and '90 (2).
133
which allows comparisons with simulated and
observed data. SNOTEL data was readily available
which allowed replication over two new seasons:
WY 1990 (as year 1) and 1991 (as year 2). All other observed and simulated comparisons were from
varying water years and merged (as year 1) in
Table 2.
At high elevations, both observed and simulated
hillslopes had small differences between opposing
N-S slopes (Table 2). The resolution of the
observed data is 7 days; hence the 7 day difference
for an LAI of 9 may not be 7 days. The simulated
differences for Colorado N-S slopes with LAI of 0
were 4 days apart and with LAI of 6, 3 days apart.
With a weekly sampling interval the difference
would also be 7 days.
Larger melt differences occurred between opposing LAI treatments (forested and bare) than opposing site aspect (N-S) but differences were greatest
between forested north slopes and bare south
slopes. With an observed LAI sensitivity test on a
south facing slope, a high elevation site with an
LAI of 9 prolonged the observed snowpack by 35
days (NM). A simulated LAI of 6 prolonged site
depletion by 18 days (CO). The largest melt differences between various LAI and aspect conditions
occurred between forested north and bare south
surfaces. When aspect was set to opposing N-S
conditions the melt differences widened from 35 to
42 and from 18 to 21 days respectively for LAI of
9 on a N (NM observed) and LAI of 6 (CO simulated). For flat surface, observed depletion in CO
had a 13 day difference between high elevation
sites with a LAI difference of 2.04.5 (1.53.0 versus 5.06.0).
Simulation results from both Montana SNOTEL
stations indicate that the effect of site orientation is
codependent on elevation (Table 2). The time lags
between depletion dates for bare N-S aspects
increased from 4 days at the high station to 8 days
at the low station. With no elevation effect the two
sites should have melted with the same time difference. The 4 day difference (84) is a doubling of
time for a 500 m elevation drop. Simulations made
again for the next water year also had a doubling
between the two elevations. The high site N-S difference was 5 days and the low site was 9.
To ascertain if radiation was responsible for the
melt differences between elevations, a simulated
134
4.4.4. Aspect, elevation, and yearday
The differences in melt can be quantified and
explained by comparing S on different slopes. High
net radiation gradients found between N-S hillslopes decrease as the sun's azimuth increases in
the late spring (Buffo et al. 1976) causing a
decreasing effect of aspect on snow depletion. On
March 15 the simulated input is 15,000 MJ M2
Day1 on a north slope and 30,000 MJ M2 Day1
on a south slope. The respective S inputs on June
21st are 30,000 MJ M2 Day1 and 33,000 MJ M2
Day1 for N-S slopes. Lower elevation snowpacks
become isothermal early in the spring and experience greater net radiation gradients between hillslopes, and melt with increased sensitivity to
aspect. High elevation snowpacks are cold and
become isothermal in late spring when radiation
differences between hillslopes is less and total
incoming S is high. Both N-S hillslopes receive
near equal portions of S in the late spring therefore
aspect differences are less than earlier in the year
(Aguado 1985). Adding LAI shades the snowpack
from S and reduces the aspect effect in the early
part of the spring when they can be large, as seen in
Table 2 and Figure 8.
RSM results extrapolated to a forested watershed
would predict an earlier and higher peak discharge
in a harvested watershed due to the reduction in
LAI causing 1) faster melt in the forest clearings
and 2) overall increase in meltwater due to reductions in canopy sublimation losses. Sublimation
losses are difficult to accurately measure in the
field, however Golding and Swanson (1986)
reported 33 mm losses on the north slope and 58
mm on the south slope for a partially harvested
basin in Alberta, Canada with equal harvest treatments. The increased losses in water yield on the
south slope suggest that S contributes to the difference by means of increased sublimation. Troendle
and King (1985) observed annual discharge patterns for 30 years in experimental watersheds with
various harvest treatments and found runoff patterns supportive of RSM predictions. They found
large and early peak stream discharge after forest
cutting treatments and a gradual returning to predisturbance discharge patterns which are less sudden and occur later in the year as the forest grows
back and LAI increases.
5. Summary
Snow accumulation and melt can be simulated for
a wide range of conditions found in the western
United States using a simple snow model and a
generic set of model parameters. RSM predicted
snow depletion with a R2 of 0.91 and an error of
6.2 days from observed depletion dates for 30 simulations at 10 diverse sites in western North America. For the purpose of calculating site water balance and plant water stress, the snow depletion date
is important since melt water is a significant source
of site water and its disappearance marks the
beginning of soil water depletion (Knight et al.
1985). Tracking daily SWE accumulation and loss
is important for validation to SNOTEL data and for
estimating spring runoff. Strangely, a simulation
with a low SWE error did not necessarily indicate
an accurate depletion date prediction and vice
versa. This independence of the two properties is
significant when considering that validation of
regional snowmelt models can be made with
remotely sensed observations of depletion or by
streamflow measurements with SNOTEL point
estimates. For validation, regional simulated snow
cover should be compared to both remotely sensed
snow cover and stream flow in gauged watersheds.
RSM accumulated smaller snowpacks under a
forest canopy than in clearings and had greater net
water losses to sublimation. Snowpacks under a
canopy however, melted slower than in the open
because of reduced melt energy at the snow surface
which caused the melt to occur later into the season. LAI is the most important variable determining snow accumulation and melt for high elevation
sites. As elevation decreases, the model shows that
site aspect becomes increasingly important in
determining snow depletion dates, although still a
secondary variable.
Extrapolating RSM results to a landscape, a partially harvested watershed will have an earlier and
higher peak runoff due to faster melting in the
openings and less sublimation losses than a forested watershed. With RSM's simple parameterization
needs, snow cover simulations on a basin scale are
possible with satellite derived LAI and digital
topography from DEMs. A computer version of the
RHESSys Snow Model is available from the First
Author upon written request.
135
Acknowledgements
This research was guided by Steve W. Running,
and conversations with L.E. Band, D.L. Potts, R.R.
Nemani, J. Dungan. Assistance in understanding
the conditions at each SNOTEL site came from
many USGS personnel located in the States of AZ,
CA, CO, MT, and OR. Thanks to S. Alexander and
J. Dungan for helpful comments on this manuscript. This research was funded by a NASA grant.
References
Aguado, E. 1985. Radiation balances of melting snow covers at
an open site in the Central Sierra Nevada, California. Water
Resources Research 21(11): 10491034.
Anderson, E.A. 1968. Development and testing of snowpack
energy balance equations. Water Resources Research 4(1):
1936.
Anderson, E.A. 1973. SNOW-17 a model based on temperature
and precipitation. NOAA, publication SW-102.
Baker, D.G., D.L. Ruschy and D.B. Wall. 1990. The albedo
decay of prairie snows. American Meteorological Society
29: 179187.
Baker, F.S. 1944. Mountain climates of the Western United
States, Ecological Monographs 14(2): 223254.
Band, L.E., D.L. Peterson, S.W. Running, J. Coughlan, R. Lammers, J. Dungan and R. Nemani. 1991. Forest ecosystem
processes at watershed scale: Basis for distributed model.
Ecological Modeling 56: 171196.
Berris, S.N. and R.D. Harr. 1987. Comparative snow accumulation and melt during rainfall in forested and clear-cut plots
in the Western Cascades of Oregon. Water Resources
Research 23(1): 135142.
Black, T.A. and D.L. Spittelhouse. 1981. Modeling the water
balance for watershed management. Interior West watershed
management. Edited by D.M. Baumgartner. WSU. Pullman,
WA.
Bristol, K.L. and G.S. Campbell. 1984. On the relationship
between incoming solar radiation and daily maximum and
minimum temperature. Agriculture and Forest Meteorology
31: 159166.
Buffo, J., L. Fritschen and J. Murphy. 1972. Direct solar radiation on various slopes from 0 to 60 north latitude. Research
Paper PNW-142, USFS. Portland, OR.
Coughlan, J.C. and J.L. Dungan. 1996. Combining remote
sensing and forest ecosystem modeling: An example using
the Regional Hydrological Simulation System (RHESSys).
In: Gholz, H.L., Nakane, K. and Shinoda, H. (eds). The use
of remote sensing in the modeling of forest productivity at
scales from the stand to the globe. Kluwer Academic Publishers, Dordrecht pp. 135158.
Friend, A.D., H.H. Schugart and S.W. Running. 1993. A physiology-based gap model of forest dynamics. Ecology 74(3):
792798.
Gary, H.L. and G.B. Coltharp. 1967. Snow accumulation and
136
Marks, D. and J. Dozier. 1992. Climate and energy exchange at
the snow surface in the alpine region of the Sierra Nevada 2.
Snow cover energy balance. Water Resources Research
28(11): 30433054.
Motoyama, H. 1990. Simulation of seasonal snowcover based
on air temperature and precipitation. Journal of Applied
Meteorology 29: 11041110.
Nemani, R.R., L.L. Pierce, L.E. Band and S.W. Running. 1993.
Forest ecosystem processes at watershed scale III: Sensitivity to remotely sensed LAI estimates. International Journal of
Remote Sensing 14(13): 25192534.
Nemani, N., S.W. Running, L.E. Band and D.L. Peterson. 1993.
Regional Hydrological Simulation System: An illustration
of the integration of ecosystem models in a GIS. In Environmental Modeling with GIS. pp. 296304. Edited by M.F.
Goodchild, B.O. Parks and L.T. Steyaert. Oxford University
Press, Oxford.
Paniconi, C. and E.F. Wood. 1993. A detailed model for simulation of catchment scale substrate hydrologic processes.
Water Resources Research 29(6): 16011620.
Price, A.G. 1988. Prediction of snowmelt rates in a deciduous
forest. Journal of Hydrology 101: 145157.
Prevost, M., R. Barry, J. Stein and A.P. Plamondon. 1991.
Snowmelt modeling in a balsam fir forest: comparisons
between an energy balance model and other simplified models. Canadian Journal of Forest Research 21: 110.
Running, S.W. and S.T. Gower. 1991. FOREST-BGC, a general
model of forest ecosystem processes for regional applications II. Dynamic Carbon allocation and nitrogen budgets.
Tree Physiology 9: 147160.
Running, S.W. and R.R. Nemani. 1991. Regional hydrologic
and carbon balance responses for forests resulting from
potential climate change. Climatic Change 19: 349368.
Running, S.W. 1984. Microclimate control of forest productivity: analysis by computer simulation of annual photosynthesis/transpiration balance in differing environments. Agricultural and Forest Meteorology 23: 267288.
Running, S.W. and J.C. Coughlan. 1988. A general model of
forest ecosystem processes for regional applications I.
Hydrologic balance, canopy gas exchange and primary production processes. Ecological Modeling 42: 125154.
Running, S.W., R.R. Nemani and R.D. Hungerford. 1987.
Extrapolating of synoptic meteorological data in mountainous terrain and its use for simulating forest evapotranspiration and photosynthesis. Canadian Journal of Forest
Research 19(6): 472483.
Ryan, M.G., E.R. Hunt, Jr., R.E. McMurtrie, G.I. Agren, J.D.
Aber, A.D. Friend, E.B. Rastetter, W.J. Pulliam, R.J. Raison
and S. Linder. (in press). Comparing models of ecosystem
function for coniferous forests. I. Model description and val-
Cext
Cint
LAI
qmelt
qsub
S
Sf
t24
tday
tmax
tnight
albedo (%)
temperature index increment (C)
canopy light extinction coefficient (C)
canopy rain interception coefficient (mm/LAI)
latent heat of fusion (MJ/kg)
latent heat of vaporization (MJ/kg)
leaf area index
albedo temperature sum (C)
isothermal temperature sum (C)
radiation snowmelt (mm/day)
radiation sublimation (mm/day)
incoming shortwave radiation (MJ/M2/Day)
incoming shortwave radiation on the snow surface
(MJ/M2/Day)
24 hour weighted air temperature average (C)
daylight air temperature (C)
maximum 24 hour air temperature (C)
night air temperature (C)