Ed Maurer Partitioning precipitation into rain and snow for event-based hydrologic modeling in the Pacific Northwest U.S. Edwin Maurer Civil Engineering Department Santa Clara University H51G-04
Ed Maurer Motivation Precipitation type can drive flood simulations Determination of type in hydrology models is dubious Unique data presents opportunities to improve precipitation type determination with radar Potential for transferability
Ed Maurer Primary Questions Do surface temperature-based methods work adequately for determining whether precipitation is falling as rain, snow, or a mixture? Can using a reflectivity from a vertically- pointing radar be used to improve this, and ultimately streamflow simulations? Can information on derived rain-snow partitioning be transferred to neighboring watersheds?
Ed Maurer Area of Focus Improvement of Microphysical PaRameterization through Observational Verification Experiment (IMPROVE-2). Intensive field observation campaign: 26 Nov- 22 Dec 2001 IMPROVE-2 domain overlaps with South Santiam River basin: total basin area of 1,440 km 2.
Ed Maurer South Santiam River Basin High orographic influence Winter storms include mix of rain and snow Ground-based Meteorological Observations: Hourly P Co-op Stations SNOTEL IMPROVE P USGS Radar
Ed Maurer Surface Air Temperature for Rain-Snow Determination Accumulation Melt T is not a good indicator of accumulation or melt Probably not good indicator of P type JUMP OFF JOELITTLE MEADOWS Each 6-hourly observation where P>0 1.determine change in swe 2.find P, Tavg 3.Plot d(swe)/d(P) vs. T 11/25 12/01 12/07 12/13 12/19
Ed Maurer Scenarios for Precipitation Type Determination Three scenarios: 1.Base Case – published T thresholds (0.0 °C and 0.7 °C) 2.Alternative 1 – 0°C level from Radar Data 3.Alternative 2 – Radar-derived T thresholds
Ed Maurer Vertically Pointing Radar Data – Reflectivity Data NOAA/ETL S-band vertically pointing radar Sample from 2215 UTC 13 Dec UTC 14 Dec 2001 Bright band in red, the top is associated with 0°C temperatures. Approx. 300 meter thickness 11/25 12/01 12/07 12/13 12/19 Observed 0° Level Based on Bright Band Identification
Ed Maurer Alternative 1: Using Radar Detected Melting Layer in Hydrologic Model Radar-detected bright band 0°C level – Melting begins Snow at land surface Rain below bright band Melt complete
Ed Maurer Alternative 2: Radar-derived surface air temperature index Radar-detected bright band Surface air temperature at pixels set to T min(rain) Surface air temperature at pixels set to T max(snow)
Ed Maurer Alternative 2: Using radar to set air temperature thresholds MinimumMaximumAverage T min (Rain) T max (Snow) Average over period Basin average surface air temperatures for snow/rain inferred from radar 0°C level Dynamic variability of radar-derived T max(snow) and T min(rain) Average over basin and time period shows values outside published range
Ed Maurer Stream Flow Simulation Gauge elev. 320 m Gauge elev. 230 m Gauges selected based on: observed data for period no effects from dams DHSVM implemented with: 150 m spatial resolution 3-hour time step Gridded observed meteorology
Ed Maurer Improvement in simulated hydrographs Gauge Gauge Base Case3846 Alternative Alternative In all cases, improvement is seen over the base case, esp. peaks 3, 4, 5. 26% reduction in RMSE for gauge in higher elevation basin Temperature index derived from radar data achieves most of improvement seen in direct use of radar freezing level Base Case Alt. 1 Alt. 2 RMSE for flows over 50 m 3 /s 11/25 12/01 12/07 12/13 12/19
Ed Maurer Snow Simulations at SNOTEL site Simulated SWE at Little Meadows SNOTEL site, upstream of Gauges Alt. 1 shows dramatic improvement over base case Alt. 2, while better than Base Case later, substantially overestimates melt in intermediate period Base Case Alt. 1 Alt. 2 11/25 12/01 12/07 12/13 12/19
Ed Maurer Transferring methods to neighboring watershed Gauge Gauge Gauge elev. 200 m Gauge elev. 485 m
Ed Maurer Changes at transferred sites Gauge Gauge Base Case4464 Alternative Alternative Higher elevation basin sees minor benefit using radar-detected 0 o level Increasing from ~45 to ~80 km appears beyond the transfer range for “calibrated” temperature index for T max(snow) and T min(rain) RMSE for flows over 50 m 3 /s ( ) and 40 m 3 /s ( ) Base Case Alt. 1 Alt. 2 11/25 12/01 12/07 12/13 12/19
Ed Maurer Radar as a calibration tool Apply to same period of previous year: 11/25/ /19/2000 shown as shaded region Gauge Gauge Base Case1812 Alternative Radar-derived T max(snow) and T min(rain) derived using December Decrease RMSE for same period in 2000 by 20% at higher elevation gauge RMSE for flows over 10 m 3 /s Alt. 2 Base Case
Ed Maurer Conclusions Surface air temperature is not a good indicator of precipitation type Radar-detected freezing levels can improve P partitioning into rain/snow in hydrologic simulations T max(snow) and T min(rain) derived from radar-detected 0°C levels achieve much of the benefit of direct use of freezing levels for concurrent period Benefits are not realized when transferring to other basins Derived T max(snow) and T min(rain) show some promise in transferring to same period and basin in previous year