The demand for physically-based distributed hydrological models to capture climatological patterns as well as extreme weather events has necessitated the need for accurate meteorological driving data (temperature, precipitation, solar radiation etc.). Due to the current limited meteorological observations over the majority of western U.S. basins, missing observations are commonly estimated from another source (i.e. solar radiation empirically estimated from temperature and relative humidity or simulated precipitation from a mesoscale model). However, the best choice of what source to use for a given variable is not always straightforward and the user’s choice may significantly impact the quality of simulated snowpack and streamflow. Here, we address this known issue by addressing the following questions:
Here, we address this known issue by addressing the following questions:
- How do simulated meteorological variables compare with in-situ observations, particularly in high elevation areas where stations are typically unavailable?
- How and where does the source of forcing data impact simulations of snowpack?
- How does the choice of forcing data impact streamflow simulations in basins draining different elevation ranges?
We found that problems with estimating precipitation comprise the largest source of forcing data errors in mountain hydrology (Lundquist et al. 2010; Wayand et al. 2013; Raleigh et al. in prep). A mesoscale atmospheric model (Weather Research and Forecasting, WRF) was able to reproduce basin-wide precipitation as well as more conventional means (extrapolating from a lower-elevation station assuming fixed climatological patterns with elevation), but even in a well-instrumented basin, we did not have enough long-term distributed observations to determine a clearly preferable method (Wayand et al. 2013). After precipitation, improper representation of the change in temperature with elevation led to significant hydrologic errors, particularly with regards to the rain-snow transition elevation (Minder et al. 2010). Fortunately, WRF represents the change in temperature with elevation well (Wayand et al. 2013; Minder et al. 2010), and many satisfactory methods exist for patching missing temperature observations in complex terrain (Henn et al. 2013). Errors in net radiation, both long-wave and short-wave, have the greatest impact on snowmelt rates (Wayand et al. 2013, Raleigh et al. in prep). Estimation of incoming long-wave irradiance depends on relative humidity (RH), and WRF estimated distributions of RH in complex terrain better than any current empirical methods (Feld et al. 2013). The relative skill of WRF vs. empirical methods at directly providing incoming radiation forcing could not be conclusively determined from the available observations (Wayand et al. 2013), particularly considering problems with in situ radiation measurements (Raleigh et al. in prep). However, snow surface temperature clearly reflects biases in incoming long-wave irradiance, and dew point temperatures provide a practical option for estimating snow surface temperatures (Raleigh et al. submitted). Dew point temperature can be obtained from an inexpensive, self-logging sensor with reasonable accuracy (Feld et al. 2013).
In summary, WRF, when run at 6-km resolution with Reanalysis boundary conditions, was able to match observations as well as, and sometimes better than, more conventional empirical techniques, so we would recommend WRF as a primary source of driving data for hydrologic models in mountainous terrain.