243
Modeling Snowmelt Runoff under Climate Change Scenarios
In addition to changes in the annual volume of flows, there is a change in the timing of when the snowpack melts and becomes stream flow, as is evident in Figures
10.13 and 10.14. The timing of the historical simulated values closely corresponds
to the timing of the measured flows. However, all the climate change scenarios produce earlier spring runoff. This is consistent with what others have found for the
Pacific Northwest, both looking at historical trends from the 1950s to the present
and using climate forecasts (e.g., Stewart et al. 2004). Table 10.5 lists the number of
days that the hydrograph peak advances earlier in the spring relative to the time of
occurrence of the peak for the simulated historical flows. These ranges between 5
and 8 days, with 2080 climate scenario value peaks occurring earlier than those of
the 2030 scenarios. Another way of looking at this is to consider the relative timing
of when 50% of the total annual volume of runoff has occurred, as shown in the last
column of Table 10.5. For the 2030 scenarios, this ranges between 3 and 5 days, but
for the 2080—Wet and 2080—Dry scenarios, this occurs 13 or 14 days earlier than
the historical case. Corresponding to a shift toward earlier spring runoff, there is a
reduction in summer runoff that manifests itself in several of the changed-climate
scenarios. These shifts in timing are partially the result of increased winter time precipitation, especially in the two Wet scenarios, but are mostly driven by the modeled
temperature increases, which appear in every month of every scenario.
10.5 CONCLUSIONS
In this chapter, we have reviewed and discussed the two major types of snowmelt
models based on a temperature index (degree-day) approach and a surface energy
balance approach, respectively. Although the energy balance approach is more physically accurate, the lack of availability of the data required to run these models and
the complexity of assembling data sets over a spatially distributed grid reduce the
usability of these models. The data required by energy balance models are not usually provided by climate models. Temperature index models, such as SRM, on the
other hand, have much simpler data requirements. This makes them more suitable
both for historical and operational studies and makes them even more desirable for
climate scenario studies, since temperature and precipitation are the most common
and readily available outputs from climate models.
In addition to temperature and precipitation, SRM requires time series of SCA as
a data input that is assumed to be provided by means of remote sensing. At present,
the snow cover product from the MODIS sensor borne by the Terra and Aqua satellites is considered to be the best source of data owing to their short repeat cycle and
good spatial resolution. The downside of this sensor is its inability to penetrate cloud
cover. Various spatial and temporal filters and combinations of data from different satellite sensors and platforms including passive microwave are currently being
developed to minimize the cloud obscuration problem. In the work presented here,
we used an 8-day composite temporal filtered MODIS product for our SCA.
No satellite data existed in many years from which we cannot derive the essential S(t) curves. In order to synthesize SCA time series, or S(t) curves, and overcome this issue of lacking satellite data, we developed and presented a method of
merging MODIS snow cover products with ground-based SNOTEL data. SRM can
Modeling Snowmelt Runoff under Climate Change Scenarios
In addition to changes in the annual volume of flows, there is a change in the timing of when the snowpack melts and becomes stream flow, as is evident in Figures
10.13 and 10.14. The timing of the historical simulated values closely corresponds
to the timing of the measured flows. However, all the climate change scenarios produce earlier spring runoff. This is consistent with what others have found for the
Pacific Northwest, both looking at historical trends from the 1950s to the present
and using climate forecasts (e.g., Stewart et al. 2004). Table 10.5 lists the number of
days that the hydrograph peak advances earlier in the spring relative to the time of
occurrence of the peak for the simulated historical flows. These ranges between 5
and 8 days, with 2080 climate scenario value peaks occurring earlier than those of
the 2030 scenarios. Another way of looking at this is to consider the relative timing
of when 50% of the total annual volume of runoff has occurred, as shown in the last
column of Table 10.5. For the 2030 scenarios, this ranges between 3 and 5 days, but
for the 2080—Wet and 2080—Dry scenarios, this occurs 13 or 14 days earlier than
the historical case. Corresponding to a shift toward earlier spring runoff, there is a
reduction in summer runoff that manifests itself in several of the changed-climate
scenarios. These shifts in timing are partially the result of increased winter time precipitation, especially in the two Wet scenarios, but are mostly driven by the modeled
temperature increases, which appear in every month of every scenario.
10.5 CONCLUSIONS
In this chapter, we have reviewed and discussed the two major types of snowmelt
models based on a temperature index (degree-day) approach and a surface energy
balance approach, respectively. Although the energy balance approach is more physically accurate, the lack of availability of the data required to run these models and
the complexity of assembling data sets over a spatially distributed grid reduce the
usability of these models. The data required by energy balance models are not usually provided by climate models. Temperature index models, such as SRM, on the
other hand, have much simpler data requirements. This makes them more suitable
both for historical and operational studies and makes them even more desirable for
climate scenario studies, since temperature and precipitation are the most common
and readily available outputs from climate models.
In addition to temperature and precipitation, SRM requires time series of SCA as
a data input that is assumed to be provided by means of remote sensing. At present,
the snow cover product from the MODIS sensor borne by the Terra and Aqua satellites is considered to be the best source of data owing to their short repeat cycle and
good spatial resolution. The downside of this sensor is its inability to penetrate cloud
cover. Various spatial and temporal filters and combinations of data from different satellite sensors and platforms including passive microwave are currently being
developed to minimize the cloud obscuration problem. In the work presented here,
we used an 8-day composite temporal filtered MODIS product for our SCA.
No satellite data existed in many years from which we cannot derive the essential S(t) curves. In order to synthesize SCA time series, or S(t) curves, and overcome this issue of lacking satellite data, we developed and presented a method of
merging MODIS snow cover products with ground-based SNOTEL data. SRM can