QPF & WinQPF/HAS-QPF (with apologies to Dave Ondrejik and thanks to Pete Manousos, although he may not know it) Presented during COMET Hydrometeorology Course Your Operator Chris Bovitz, Hydrologic Forecaster West Gulf River Forecast Center, Fort Worth, Texas Safe - Reliable - Courteous
WinQPF/HAS-QPF n n Goal of presentation – –History of QPF project – –Origins of HAS unit/function – –Origins of WinQPF and HAS-QPF – –QPF Verification – –Probabilistic QPF – –A forecasting technique
Intro: Terminology n n Mean Areal Precipitation (MAP) – –Precipitation accumulation spatially and temporally averaged over a defined region, usually a drainage basin n n Future Mean Areal Precipitation (FMAP) – –(a.k.a. QPF) Forecast precipitation accumulations spatially and temporally averaged over a defined area. n n Future Mean Areal Temperature (FMAT) – –Forecast temperatures spatially & temporally averaged over an area.
Goal of RFC QPF Program n n To accurately produce GRIDDED FMAP/FMAT for each subbasin for use in river forecast model
History of QPF n n Hydrologic forecasts without QPF
Hydrologic Forecasts Prior to QPF n n Used only observed amounts through 12Z n n Ignored that it could still be raining, or that rain was certain n n Led to stair-stepped river forecasts during sustained rain events
Example of “Stair Stepping” There must be a better way
The Origin of QPF n n QPF initiated in the 1970s n n Event-driven – –Composed only during significant events (but what’s significant?) n n One QPF value supplied per basin – –In 6-hour increments (up to 72 hours) n n Matches time step of RFC river model – –QPF manually entered into RFC model in tabular form n n Resulted in 2 forecasts produced by RFC – –Official (without QPF) and Contingency (with QPF) n n Gave forecasters “what-if” scenarios n n But which forecast is the official forecast?
QPF Begins to Evolve n n RFCs wanted WSFO input – –Wanted opinion of “local experts,” i. e., WSFO meteorologists with expertise in forecasting their areas – –At this time, RFCs were still staffed mainly with hydrologists (no designated meteorologists) n n Critical Flood Support Office (CFSO) were created to coordinate QPF among WSFOs
Dawn of Daily QPF n n 1988: QPF supplied daily – –Even during dry weather – –CRW and PIT first offices to supply daily QPF (to OHRFC) n GUI created –Increased efficiency in entering and sending forecast to RFC –Developed by Matt Peroutka (CLE) –Used smaller QPF regions n Still was used at PIT as recently as 1998 n Gradually, other WFOs produced daily QPFs –Eventually became mandatory
HAS Unit & Function is Formed n n HAS - Hydrometeorological Analysis & Support n n Meteorologist or Met-trained hydrologist – –Need to be knowledgeable in Met to best coordinate with WFO meteorologists – –Need to be familiar with Hydro to brief fellow hydrologists
HAS Forecaster Duties n n Produce best possible precipitation data set for use in hydrologic models n n Produce daily FMAP (QPF) – –With updates if needed or requested by hydrologists or WFO n n Coordinate with RFC Senior Duty Hydrologist (SDH) – –Primarily via briefings n n Produce daily HMD (Hydrometeorological Discussion) – –Allows other WFOs and public to know about QPF used in river forecasts and the current state of the rivers in RFC’s area of responsibility n n Other duties – –Coordinate with WFOs via telephone or hydrometeorological coordination message (HCM) – –Act as liaison between RFC hydrologists and WFO meteorologists – –Produce extended QPF – –Produce FMATs for snowmelt runoff forecasts – –Work on research projects, other RFC-WFO coordination
A need existed to enhance QPF entry and modification n n 1992: Risk reduction project began n n Needed ability to display multiple WFOs’ QPFs side-by-side n n Needed easy editability to facilitate HAS adjustments
WinQPF Developed n n Developed by Mark Fenbers (Senior HAS, OHRFC) n n QPF entry via a Windows GUI – –Configurable for each WFO n n Allows user defined preferences – –Colors, overlays, etc. n n Computes FMAP in gridded format n n Extended QPF capability n n FMAT capability n n Transmit and archive capability n n Good stepping stone for probabilistic QPF (future)
HAS-QPF Developed ( ) n n Similar look and feel to WinQPF n n Compiled for Unix operating system on Hewlett-Packard, IBM workstations n n Used by HAS at most RFCs – –Western offices use Mountain Mapper – –Will not not be incorporated into AWIPS – –But stand-alone program does work – –New software is under development
QPF Verification n Key to self-calibration n Compares FMAP values to MAP values from WFOs (and RFCs) n Daily MAP sent to WFOs n Scores depend on quality of MAP values –Rain gages not always maintained –Inaccurate location (lat/lon) –Majority of reports are 24hr - leads to disaggregation problems –Algorithms make assumptions n Precipitation evenly spread over drainage basin in space and time
QPF Verification - How to Score? n n How do we come up with a single value which incorporates timing, amount, and areal coverage of precipitation? n n Just use ye olde POD, FAR, CSI? – –What are good POD/FAR/CSI scores during synoptic events? Landfalling tropical cyclone events? Stratiform events? Convective events? n n Difficult to draw useful conclusions from raw verification scores of individual basins n n Bayesian Correlation Score (BCS) (a.k.a Bayesian Informativeness Score - BIS) –Krzysztofowicz, Mon. Wea. Rev., 1992 –From Krzysztofowicz, Mon. Wea. Rev., 1992 –Attempts to reduce a three-dimensional problem (Areal coverage, timing, amount) into a value which can be easily compared –1=perfect forecast 0=no skill –Major contributor to BIS is standard deviation of precip climatology n n Attempts to normalize variability between different locations
QPF Verification - Results n n On average, how do we do? – –Major events are underforecasted – –Minor events are overforecasted n n Why? – –Underestimate scattered nature of precipitation – –“Trained” to seek the highest value – –Hard to forecast extreme events – –Point values versus mean areal values – –Personal wet/dry biases
QPF Verification n n Verification not only helps self-calibration, but also can answer other questions – –Are we better off with or without QPF? – –Does RFC HAS improve the WFO forecast? – –Does RFC HAS improve over NCEP? – –Do we have the skill to produce 48- or 72-hour QPF?
Probabilistic QPF (PQPF) n n Similar to “contingency” concept, but more specific n n Gives forecasts with different probabilities of occurring – –By exceedence fractiles (what?) n n ” n n ” n n ” – –River forecasts will be generated the same way n n Puts the “decision-making” power in the hands of the decision makers
WinQPF/HAS-QPF A FORECASTING TECHNIQUE A FORECASTING TECHNIQUE