Steps towards inges/ng radar retrieved refrac/vity into numerical models Ya-Chien Feng* and Frédéric Fabry 639 *[email protected] Acknowledgement: The authors thank Prof. Daniel Kirshbaum for providing landuse and soil moisture files produced by the WRF iniFalizaFon and Dr. Dominik Jacques for preparing the ensemble forecasFng data and discussion. This project was undertaken with the financial support of the Government of Canada provided through the Department of the Environment. The views expressed herein are solely those of McGill University. What is radar retrieved refrac/vity (N)? φ i (r ) = 2 π f 2 r i n c f: radar transmiQer frequency c: speed of light in vacuum n air_near_sfc ~ 1.0003 Radar Fixed ground target N = ( n − 1) × 10 6 = 77.6 P T + 3.73 × 10 5 e T 2 r: range of radar beam path P: pressure (mb) T: temperature (K) e: water vapor pressure (mb) The Fme for waves to travel back and forth between the radar and a fixed target. • The phase ϕ of a fixed ground target changes with the refracFvity N of air, a proxy of low-level moisture (Fabry et al., 1997). See image at right. • Radar refrac/vity is sensi/ve to the moisture varia/on: T = 18°C, ΔT = 1°C à ΔN =1 Δq = 0.2g/kg à ΔN =1 • The characteris/cs of errors are cri/cal for data assimila/on, or radar refrac/vity networks implementa/on. • Radar N fields have been assimilated to improve the iniFal condiFon of low-level moisture for a beQer quanFtaFve precipitaFon forecasFng result, but with assumed observaFonal errors and representaFve height. Quan/fying the observa/on errors for varia/onal methods • Radar refracFvity provides valuable high-resoluFon low-level thermodynamic fields to study the land-surface atmospheric interacFon, and help model evaluaFon and data assimilaFon in mesoscale and regional numerical weather predicFon models. • A beQer understanding of the biases and quanFficaFon of errors provides solid data characterisFcs for variaFonal method, such as data assimilaFon and synergizing with other boundary layer instruments. • Radars and ground targets are there in the naFonal radar network. Why not make refrac4vity opera4onal? Ground target reliability Good Bad Refrac/vity map to study land-surface atmospheric interac/on • Even though the target heights can be esFmated from terrain, there are sFll some uncertainFes. • The σ(N) is larger for steeper terrain slopes under the most extreme dN/dz condiFons; in Montreal, such condiFons are rare. • The radar refracFvity retrieval provides insights on high-resolu/on near-surface moisture, which is important to understanding convecFon iniFaFon and evoluFon as well as boundary layer processes. • The radar retrieved N shows consistent trends with the N of the surface sta/ons. - Radar N at ~20 m above terrain, close the lowest level of model. - The difference in N is largely due to the refracFvity profiles (dN/dz) and height difference. Terrain (m) • σ 2 (ϕ) can be esFmated by ground target reliability. • σ(N) is NOT randomly distributed. N is reliable in the dense good targets area. σ (N) is within ±1 N-unit. • Larger observaFonal errors in the less dense target area is quanFfied. σ(N) - The error variance (σ 2 ) of N is affected by the measured phase (ϕ) of a single target and retrieval method: σ 2 (N)=σ 2 [ retrieval method (σ 2 (ϕ))] σ 2 (ϕ) = σ 2 (Radar transmiQer frequency, hardware) + σ 2 (Antenna phase paQern, refracFvity profile dN/dz) + σ 2 (RefracFvity profile dN/dz, target heights) + σ 2 (Atmospheric scinFllaFon) + σ 2 (Target movement) + σ 2 (complexity of targets: point/extended) Error: Uncertainty. Can not be corrected, but can be simulated Bias: SystemaFc. Can be corrected. (Feng et al. 2016) σ(N): uncertain height • The N map shows the near- surface air thermodynamic spa/al contrast between urban and suburban areas in Montreal (Canada). - N urban < N rural or suburban - The local variaFon of water vapor is affected both by horizontal moisture advecFon and by evaporaFon from the land surface. Refrac/vity in the opera/onal radar network for a regional model • Can radar refracFvity from the naFonal radar network provide a low- level thermodynamic constraint for the model at regional scales? - Horizontal auto-correlaFon of refracFvity field at the center of each square grid at the lowest model level suggests N informaFon would propagate. (Environment Canada regional model, 15 km resoluFon, forecasFng 12hr, 20 ensembles though more are needed) • We believe the average refracFvity over a small region (~30 km radius) can be very precise and would be more representaFve than point observaFons. Dry, warm Moist, cool N average in three clear days urban Summer volumetric soil moisture N diurnal variaFon in urban/suburban • The N map shows consistency with the climatology of summer volumetric soil moisture (m 3 m -3 ) from U.S. Geological Survey land- use database. - High spaFal-temporal resoluFon of N might help study land-surface atmospheric interacFon over a heterogeneous land surface. Summary: What can radar refrac/vity do for you? Horizontal auto-correlaFon of N
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Stepstowardsinges/ngradarretrievedrefrac/vityintonumericalmodelsYa-ChienFeng*andFrédéricFabry
639
Acknowledgement:TheauthorsthankProf.DanielKirshbaumforprovidinglanduseandsoilmoisturefilesproducedbytheWRFiniFalizaFonandDr.DominikJacquesforpreparingtheensembleforecasFngdataanddiscussion.ThisprojectwasundertakenwiththefinancialsupportoftheGovernmentofCanadaprovidedthroughtheDepartmentoftheEnvironment.TheviewsexpressedhereinaresolelythoseofMcGillUniversity.
Whatisradarretrievedrefrac/vity(N)?φi (r) = 2π f
2ri nc
f:radartransmiQerfrequencyc:speedoflightinvacuumnair_near_sfc~1.0003
Radar Fixedgroundtarget
N = (n −1)×106 = 77.6 PT+ 3.73×105 e
T 2
r:rangeofradarbeampathP:pressure(mb)T:temperature(K)e:watervaporpressure(mb)
TheFmeforwavestotravelbackandforthbetweentheradarandafixedtarget.
• Thephaseϕofafixedgroundtargetchangeswith therefracFvityNofair,aproxyoflow-levelmoisture(Fabryetal.,1997).Seeimageatright.
• Radarrefrac/vityissensi/vetothemoisturevaria/on:T=18°C,ΔT=1°CàΔN=1
Δq=0.2g/kgàΔN=1
• Thecharacteris/csoferrorsarecri/calfordataassimila/on,orradarrefrac/vitynetworksimplementa/on.• Radar N fields have been assimilated to improve the iniFal condiFon of low-level moisture for a beQerquanFtaFveprecipitaFonforecasFngresult,butwithassumedobservaFonalerrorsandrepresentaFveheight.
Quan/fyingtheobserva/onerrorsforvaria/onalmethods
• Radar refracFvity provides valuable high-resoluFon low-level thermodynamic fields to study the land-surfaceatmospheric interacFon, and help model evaluaFon and data assimilaFon in mesoscale and regional numericalweatherpredicFonmodels.
• AbeQerunderstandingof thebiasesandquanFficaFonoferrorsprovidessoliddatacharacterisFcs forvariaFonalmethod,suchasdataassimilaFonandsynergizingwithotherboundarylayerinstruments.
• RadarsandgroundtargetsarethereinthenaFonalradarnetwork.Whynotmakerefrac4vityopera4onal?
Groundtargetreliability Good
Bad
Refrac/vitymaptostudyland-surfaceatmosphericinterac/on
• EventhoughthetargetheightscanbeesFmatedfromterrain,therearesFllsomeuncertainFes.
• Theσ(N)islargerforsteeperterrainslopesunderthemostextremedN/dzcondiFons;inMontreal,suchcondiFonsarerare.
• TheradarrefracFvityretrievalprovidesinsightsonhigh-resolu/onnear-surfacemoisture,whichisimportanttounderstandingconvecFoniniFaFonandevoluFonaswellasboundarylayerprocesses.
• The radar retrieved N showsconsistent trends with the N ofthesurfacesta/ons.-RadarNat~20maboveterrain,closethelowestlevelofmodel.-ThedifferenceinNislargelydueto the refracFvity profiles (dN/dz)andheightdifference.
Terrain(m)
• σ2(ϕ) can be esFmated by groundtargetreliability.
• σ(N)isNOTrandomlydistributed.N is reliable in thedensegoodtargetsarea.σ(N)iswithin±1N-unit.
• Larger observaFonal errors in the lessdensetargetareaisquanFfied.
σ(N)
- Theerrorvariance(σ2)ofNisaffectedbythemeasuredphase(ϕ)ofasingletargetandretrievalmethod:σ2(N)=σ2[retrievalmethod(σ2(ϕ))]σ2(ϕ)=σ2(RadartransmiQerfrequency,hardware)
+σ2(AntennaphasepaQern,refracFvityprofiledN/dz) +σ2(RefracFvityprofiledN/dz,targetheights)+σ2(AtmosphericscinFllaFon) +σ2(Targetmovement)+σ2(complexityoftargets:point/extended) Error:Uncertainty.
Cannotbecorrected,butcanbesimulated
Bias:SystemaFc.Canbecorrected.(Fengetal.2016)
σ(N):uncertainheight
• The N map shows the near-surfaceairthermodynamicspa/alcontrast between urban andsuburban areas in Montreal(Canada).-Nurban<Nruralorsuburban-ThelocalvariaFonofwatervaporis affected both by horizontalmoisture advecFon and byevaporaFonfromthelandsurface.
Refrac/vityintheopera/onalradarnetworkforaregionalmodel• CanradarrefracFvity fromthenaFonal radarnetworkprovidea low-
levelthermodynamicconstraintforthemodelatregionalscales?-Horizontalauto-correlaFonofrefracFvityfieldat thecenterofeachsquare grid at the lowestmodel level suggestsN informaFonwouldpropagate. (Environment Canada regionalmodel, 15 km resoluFon,forecasFng12hr,20ensemblesthoughmoreareneeded)
• WebelievetheaveragerefracFvityoverasmallregion(~30kmradius)can be very precise and would be more representaFve than pointobservaFons.
Dry,warm
Moist,coolNaverageinthreecleardays
urban
Summervolumetricsoilmoisture NdiurnalvariaFoninurban/suburban
• The N map shows consistencywith the climatology of summervolumetric soil moisture (m3m-3)from U.S. Geological Survey land-usedatabase.- High spaFal-temporal resoluFonofNmighthelpstudyland-surfaceatmospheric interacFon over aheterogeneouslandsurface.
Summary:Whatcanradarrefrac/vitydoforyou?
Horizontalauto-correlaFonofN
Related Documents
