but radiation data are often not readily available.
reduce the error in snowpack simulations and
The use of satellite data to determine snow-covered
forecasting.
area can greatly augment simple snow models. In
Current ability to forecast the weather con-
fact, in an operational system with (1) good sat-
strains our ability to forecast snowmelt runoff.
ellite snow covered area data, (2) frequent model
Future methods of estimating and forecasting
updates to match runoff conditions, and (3) expe-
snowmelt runoff for large river basins are likely
rienced forecasters, the detail in snowmelt algo-
to be driven by mesoscale meteorological forecast
rithms is a relatively less important aspect of the
models. Tying snowmelt models with mesoscale
model system. In this context, the key to improved
meteorological model forecasts could be the key
temperature-index modeling would be (1) fuller
to improved short term snowmelt runoff forecasts
use of satellite remote sensing data, and (2) the
within a decade. These meteorological forecasts
integration of terrain data to better define melt
are spatially continuous over a landscape at grid
patterns. This type of method would be more suit-
resolutions (10100 km) compatible with distrib-
able for regions that are not frequently overcast,
uted hydrologic modeling approaches and pro-
since this would interfere with the remote map-
vide the parameters needed to drive an energy
ping of snow cover. The third group is models that
balance model of snowpack accounting.
require additional predictive data including, rela-
In the interim, if there are not adequate data to
tive humidity, wind speed, and either radiation
run an energy balance model with more detailed
or cloud cover data. These full energy balance
snowpack accounting over an entire watershed,
algorithms may be drivable with mesoscale mete-
one could be run at a few key sites where data are
orological model data forecasts, a data source
available. Information on snowpack ripeness and
that is likely to figure heavily in future forecast-
onset of meltwater outflow provided by SNTHERM
ing approaches. Simplified energy balances based
or SNAP could greatly aid forecasters. This,
on temperature and precipitation could also be
however, implies that good knowledge of what
configured to drive the more detailed snowpack
is going on at a handful of sites in a basin has
process models, SNTHERM and SNAP.
strong implications for what is occurring on the
basin as a whole. This leads to the idea of distrib-
uting point mass and energy models across drain-
DISCUSSION
age basins, large or small. The configuring and
The more detailed snow models, SNTHERM
evaluation of distributed mass and energy balance
and SNAP, are candidates for today's distributed
models such as SNTHERM are the topics of cur-
and operational snowmelt models. Data handling
rent applied research at CRREL. The use of mesos-
is much less difficult today, and computers are
cale meteorology models to drive snowmelt mod-
much speedier than during the 60s and 70s when
els is being investigated (Melloh et al., in prep), as
the existing operational models were developed.
well as optimal methods of segmenting basins into
Though we have the electronic capabilities to take
hydrologic response units for snowmelt (Melloh
and Jordan, in prep., Davis pers. comm.∗ ).
advantage of more detailed models and handle
more meteorological data, the type and number of
meteorological sensors accessible electronically
CONCLUSIONS
may not have increased. Meteorology at high ele-
vations in drainage basins has historically been
Data availability and time constraints will con-
underrepresented, and this is likely still the case.
tinue to drive the choice of surface energy balance
Current operational methods and models did,
models for a particular application; thus both tem-
however, show significant shortcomings in simu-
perature index and full energy balance methods
lating snowpack conditions during the upper Mid-
are needed. Slope, aspect, and forest cover are
west (Red River at the north, and Missouri basins)
extremely important to snowmelt; one need only
flooding in 1997. Without the ability to incorpo-
look outside to see this demonstrated in real life.
rate full energy balance methods, or satellite snow-
In the past only temperature and precipitation
covered area maps, it may not be possible to fore-
data were available to operational forecasters; to-
cast conditions beyond the range of "normal"
day we also have digital terrain and forest cover
meteorological conditions, as was suggested by
data and our methods should take advantage of
Anderson (1976). This improved predictive capa-
bility of full surface energy budget approaches and
more detailed internal process accounting should
* Personal communication, R. Davis, CRREL, 1998.
14
Return to contents
Previous Page
