Improving UpperLevel Performance in AMPS Longwave Radiation Jordan

Improving Upper-Level Performance in AMPS: Longwave Radiation Jordan G. Powers, Steven M. Cavallo, and Kevin W. Manning

• Motivation for AMPS Investigation – Examination of WRF simulations of Atlantic basin hurricanes: T biases at upper levels found – Model top cooling from longwave (LW) processes (RRTM LW scheme) significantly higher than observation • AMPS Testing – Analysis of summer and winter periods to assess extent of problem – Test simulations with RRTM LW scheme modifications performed

Upper-level T Biases: WRF 2009 Atlantic Basin Hurricane Forecasts (WRF) (v. Time) WRF–GFS Analysis (v. Time) -10 K max Upper-level cooling over time Output from fcst hr 6

RRTM LW Scheme Modification— Atlantic Basin Experiments Heating Rate Differences Bias reductions from mods (Modified – Unmodified RRTM) W/o H 2 O adj: Refined buffer layer and T profile Full mods: H 2 O adjustment (std profile) in buffer layer (to avoid excessive – 1 week period / Fcsts every 12 hrs / 6 -hr fcsts – MLS= Mid-Latitude Summer / TROP= Tropical MT moisture) Note: SLP RMSEs also decrease with modified scheme.

Configuration of AMPS for Investigation / Testing Domains: 45 -km / 15 -km 45 km Test fcsts: 6 -hr IC/BCs: GFS Test periods: Summer January 1 -7, 2010 Winter July 1 -7, 2009 15 km

RRTM LW Scheme: Original Model Top Treatment • Buffer layer from model top (MT) to top of atmosphere (TOA) – Extra computational level in LW scheme only: No new model η-level • Layer properties – T isothermal: MT value – qv constant: MT value – O 3 set to. 6 O 3 MT value

RRTM LW Modifications • Computational layer refined: Multiple levels to TOA added – p= 2. 5 mb – Extra levels in scheme, not η-levels (no significant extra run time) • Improved T representation – Temps at new levels related to average T profile (using T at MT) • Excessive moisture prevented: Layer H 2 O= 5 ppmv • O 3 interpolated from table

WRF Water Vapor Issue • Potential for Excessive Moisture at High Levels: Affects LW Flux Calculations – <Jan 2010: No H 2 O vapor fields above 100 mb in GFS files – WPS assumption (where nec’y): 5%≥ RH ≥ 1% for 300– 50 mb – Problem: Too moist in stratosphere • Standard profile value: 5 ppmv • WRF-Var minimum qv: qv= 1 e-6 kg/kg (o(5 ppmv)) (if qv < 1 e-6 kg/kg) WRF: Atlantic Basin Tests

AMPS Upper-Level Water Vapor Winter Testing Summer Testing Domain avg qv Top η 1/2 Level Sounding maxima ( 12 mb) WRF-Var min qv 1 e-6 SAW= Sub-Arctic Winter SAS= Sub-Arctic Summer

Analysis of AMPS Heating Rates: Original RRTM LW Winter Summer Excessive LW Heating Rates SAS LW SAS SW Net= ∂ /∂t LW + ∂ /∂t SW AMPS–SAS AMPS–SAW Cooling bias Heating Rate Bias SAW= Sub-Arctic Winter MLW= Mid-Latitude Winter SAS= Sub-Arctic Summer MLS= Mid-Latitude Summer

AMPS Differences from Standard Profiles and Single-Column (SC) Tests Winter : Extrapolated SC AMPS’s lesser cooling rate may reflect colder Antarctic stratosphere for SC model Summer SC top value: Artifact of extra level : Extrapolated SC AMPS: Cooling bias SC SAS: Projected cooling bias at MT (excl. artifact) for SC model Summer SC SAS test: Problem in RRTM LW scheme SAW Temps/SAS Temps: SC model run w/given temp profiles Single column: SC version of RRTM (run from domain-avgd profile of T)

Analysis of AMPS Heating Rates: Modified RRTM LW Winter Summer Heating Rates Max ~1. 8 K/d Modified – Control= Original RRTM LW MT T 5 days: ~9 K Experiment= Modified RRTM LW

Model Top Improvement: Summer ∂ /∂t (LW) Control ∂ /∂t (LW) New Mods reduce cooling and eliminate excess qv impacts ∂ /∂t (LW) New–Control _6 h (Total) New–Control hr 6 – hr 0 Mods reduce cooling bias ∂ /∂t (LW)= Instantaneous heating rates avg’d/fcst hr 6 ∂ /∂t (Total)= hr 6 – hr 0 Level = η 1/2

Summary • WRF MT cooling bias seen in Antarctic/AMPS application – Summer signal – Moderate compared to non-polar WRF applications • AMPS upper-level H 2 O vapor – Localized high qv biases near MT from soundings – Large vapor amounts can influence LW calculations • RRTM LW Mods: Decreased MT cooling & T errors in AMPS – Mods reduce LW flux errors and excessive cooling – Mods avoid LW errors due to areas of excessive qv at MT