Upper Colorado River Basin Water Forum November 7, 2018 Communicating Uncertainty and Risk in Water Resources: Using Innovative Displays of Hydrologic Ensemble Traces Richard Koehler, Ph.D. NOAA/NWS, Boulder, CO
Define terms Ensemble traces A set of future flow simulations at a location. Risk The probability of an adverse event. Perception of risk is subjective. Uncertainty : Technical Inability to precisely determine the true value or form of model states and variables. Uncertainty : Law That which is unknown or vague. Uncertainty : General audience Not exactly known, such as the future. Used for this talk
Ensemble overview Each trace is a simulation of future flow Now Past Future Future meteorological conditions (forcings) 1 5 days River Forecast Centers and Weather Prediction Center* guidance observed forecast 1 15 days Environmental Modeling Center* Global Forecast System (type of NWP) Flow 16 270 days Climate Prediction Center* Seasonal climate forecast Saved model states (current conditions) Ensemble traces Time 271 days and longer Historic climatology * National Centers for Environmental Prediction NCEP, part of NOAA s National Weather Service
Ensemble analysis Each trace is a simulation of future flow Flow Now Past Future observed forecast Probability calculations based on multiple traces Exceedance probability Large flows Low chance of being exceeded Median flows Medium chance of being exceeded Saved model states (current conditions) Ensemble traces Time Small flows High chance of being exceeded
Study site Colorado
Ensemble plot evolution Water supply forecast, April 1 to July 31 Colorado River near Cameo, CO Daily flow vol (kaf/day) Current assumption: Lines in a single plane Spaghetti plot Issued 4/1/2018 Source: CBRFC New assumption: Profiles in multiple planes Aerial view Raster-based plot Day of Year Ground view Spaghetti plot
Time map raster framework X (day) Y (year) Z (value) Short-term coordinate Long-term coordinate Raster cell color Daily flow volume (kaf/day) 5 10 15 20 25 30 35 40 45 50 55
Ensemble raster hydrograph Raster hydrograph View all days, all traces Easy comparisons Within years Between years Flow properties Magnitude, frequency Duration, timing, flow change Patterns across timescales Daily, weekly, monthly Seasonal, annual
Assess simulation variability Wet Dry April May June July Wet Rank Dry 5 10 15 20 25 30 2 12 1981 25 1982 1983 1984 3 14 18 1985 1986 19 15 27 30 1988 31 1989 32 1987 9 1991 1990 1 1993 1995 6 1996 23 24 26 1994 1992 Rank vs Year April 1 to July 31, 2018 water supply volume Colorado River at Cameo, CO April 1, 2018 forecast 5 1999 1997 21 2001 22 2000 1998 1985 1990 1995 2000 2005 2010 35 Simulation Year (Rank: 1 = Largest volume, 35 = smallest volume) 11 2003 17 2004 28 2002 16 2005 33 29 13 2006 8 2007 4 2008 2009 2010 34 10 2011 7 2013 20 2014 2012 201
Ranked raster hydrograph Identify flow volume patterns Condition Rank numbers Wet 1 12 Late peak, extended higher flow Higher daily flow starts end of May Wet Median 13 24 Most volume mid-may to mid-june Dry 25 35 Low volumes mid-may to mid-june July typically has very low daily flow Very low daily flow can start in mid-june Dry
Custom assessment plots Threshold contour 20 kaf/day Rank 2 6 10 14 18 22 26 30 34 Seasonal maximum and timing April May June July 1989 2001 2007 1996 1992 1987 2006 2012 1995 1984 2011 1999 1997 2015 2010 1993 2013 2003 1981 2009 1985 2008 1991 2005 1986 1990 2014 2000 1994 1988 1998 2002 2004 2018 90 100 110 120 130 140 150 160 170 180 190 200 210 19.6 to 21.6 21.6 to 23.6 23.6 to 26.8 26.8 to 27.8 27.8 to 30.6 30.6 to 34.7 34.7 to 46.0 46.0 to 59.5 Day of Year Daily maximum (kaf/day) 1982 1983 Observed
Exceedance probabilities Daily value base on multiple traces Integrate into raster hydrographs Weibull plotting positions, exceedance prob, P = m / (n+1) (m is rank, n is # traces) kaf/day Day 150 kaf/day 60 Day 150 analysis 50 P = 1/(35+1) = 0.0278 = 2.78% 40 30 20 Day of Year 10 0 larger flows 50% smaller flows 1 2 5 10 20 30 40 50 60 70 80 90 95 98 99% lower Exceedance probability (%) higher chance chance
Probability time map plot Wet Summary April and May Ensembles do not foretell forecast seasonal volume June and July Probabilities stratify by wet/dry conditions Dry Exceedance prob 50% Filled area, lower flows
Combination plot with multiple overlays Wet Max flow volume (kaf/day) 19.6 to 21.6 21.6 to 23.6 23.6 to 26.8 26.8 to 27.8 27.8 to 30.6 30.6 to 34.7 34.7 to 46.0 46.0 to 59.5 Exceedance prob 50% Filled area, lower flows Dry
GIS-type conditional analysis Criteria: ex_prob 15% and Q vol 30 kaf/day Wet Summary: 13 of 35 traces meet criteria of at least 1 day April May June July Wet: 11 of 12 Med: 2 of 12 Dry: 0 of 12 Dry
Summary Communication requires defining terms Communicate visually - Change assumptions if needed - Ensemble order (wet to dry), view conditions of interest Raster plot versatility - View multiple timescales, overlay multiple layers - Put flow, probability, uncertainty into context Improve forecast insight - Conditional analysis, more what-ifs - Enhanced and defendable decision support
Thank you Questions?