polar wrf forecasts on the arctic system reanalysis domain: atmospheric hydrologic cycle workshop on...
Post on 21-Dec-2015
219 views
TRANSCRIPT
Polar WRF Forecasts on the Arctic System Reanalysis Domain:
Atmospheric Hydrologic Cycle
Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model
Aaron B. Wilson* , David H. Bromwich, and Keith M. Hines
Polar Meteorology GroupByrd Polar Research CenterThe Ohio State University
Outline
• Motivation• Model Configuration and Domain• Precipitation
– Annual, Seasonal, Monthly– Major Arctic River Basins
• Clouds• Shortwave and Longwave Radiation• Summary and Conclusions
Motivation
• The Arctic System Reanalysis– Expansion of Polar WRF Development – Optimizing performance over the Arctic w/o penalty to
other areas
• Polar Frontier Project and outreach applications• Evaluation of Polar WRF short-term weather
forecasts of the atmospheric hydrologic cycle as a compliment to the surface/upper air analysis – Wilson et al. 2011: JGR doi:10.1029/2010JD015013
Model Configuration• Polar WRF version 3.1.1 with 39 levels in the vertical
• Top set at 10 hPa and lowest level centered at 8 m AGL• Physics
• WRF Single moment 6-Class*• Grell-Devenyi 3-D Ensemble• RRTM Longwave*• Goddard Shortwave*• MYJ PBL• Noah Land Surface Model with Eta Similarity
• Lower Boundary Conditions• SST: NCEP 0.5º RTG_SST Analysis• Sea Ice: Fractional Sea Ice
• Bootstrap SSM/I 25 km from NSIDC• Seasonal transition of sea ice albedo
• Fixed albedo winter and spring 0.82• June: Linear decrease in albedo 0.5 by the end of the month representing a mix
of bare ice and melt ponds• July: As ponds deepen and become less reflective albedo increases to 0.65
(representing bare ice only) • August 15: 21 day linear increase of albedo to 0.82
Model Domain and Input Data
• 2-Way Nested Domain• Outer: 180km
• Includes most of the NH
• Inner: 60km• Includes major river
basins flowing through the Arctic
• Broad Scale Evaluation • Tractable
• NCEP/NCAR 1 x 1 Final Analysis (6-hr)
• Forecast mode: 48-hr simulations with hours 24-45 retained for analysis
• 240 second time step with 3-hr output
• Defined Polar and Mid-Latitude Sub Domains 60º N
• Precipitation Data: GHCN, AHCCD, ERA-Interim
• Clouds: NCDC, CloudSat/Calipso, MODIS
• Radiation: BSRN, ARM, SPRS, FMI
Annual Precipitation: Spatial Comparison
• Spatially consistent with ERA-Interim Reanalysis– Highest Precipitation totals located throughout the
mid-latitudes and sub-polar storm track regions– Dry throughout the Canadian Archipelago
• Slightly higher totals in Pacific NW N. America– Both PWRF and ERA-Interim are higher than GPCP
here• Polar WRF shows great detail in higher terrain
Polar WRF ERA-Interim GPCP
Annual Precipitation• Mid-Latitudes (305): +37.3 mm
(+4.6%)– 62% within ±50% (35% within 25%)– NA, Europe, Asian regions: similar
results• Polar (78): -58.8 mm (-9.4%)
– 69% within ±50% (44% within 25%)• Few stations within 5%• NOT spatially homogenous• Horizontal resolution: a difficult
obstacle in grid-point analysis– Large +/- biases in areas of complex
terrain• Smooth effects of small scale
circulations in area of complex terrain• The fjords of Norway
• Canadian Archipelago– Dry throughout the entire year– Mean equatorward water-vapor
transport (Serreze et al. 1995)– Winds are inconsistent (southerly)
Monthly Precipitation
• Mid-Latitudes warm/cool season discrepancy– Cool months: Negative
biases when precipitation synoptically driven
– Large (+) Biases in Spring and Summer (Jun: 35.2%, Jul: 16.2%, Aug: 15.2%)
– Warm months tied to convection
• Polar – Negative throughout the
year (-5% to -20%)– Positive bias only in July
due to convection near stations in the southern part of boundary
Evaporation
• Annual mean 2 m dew point temperature biases in the mid-latitudes led to an investigation of evaporation
• ERA-Interim 2 m dew point biases are smaller compared to observations than Polar WRF
• Total evaporation on land shows Polar WRF overpredicts evaporation for July compared to ERA-Interim especially for mid-latitudes
• Some regions of the Arctic underpredicted and may help explain negative precipitation biases
Polar WRF
ERA-Interim
Convective Precipitation• 3 Sensitivity Simulations
(WRF6C, Morrison, Kain-Fritsch)– Little change in the overall
total and convective precipitation (WRF6C, Morrison, Kain-Fritsch)
• Grid-nudging of specific humidity towards a drier state in the lower atmosphere – yields negative
precipitation bias (25% decrease) and ~1/2 convection
• Other areas to investigate include soil moisture and interaction with PBL scheme
Sensitivity simulations for July 2007
Arctic River Basins
• Important for the fresh water supply to the Arctic Ocean
• Headwaters begin as far south as 45ºN representing a strong link between mid-latitude atmospheric processes and effects in the Arctic.
• Arctic climate system and global ocean circulation
ObYenisei
LenaMackenzie
Precipitation: Arctic River Basins
Convection in spring and summer for the Russian rivers lead to large summer biases
Mackenzie River biases related to smoothed terrain effects Negative overall precipitation especially in late spring Southern extent of the region influence by Gulf of Mexico through
Great Plains low level jet and cyclogenesis Single events of > 100 mm are observed (MAGS)
8 18
5 13
+80.9 mm (15%)
+204.4 mm (57.5%)
+93.4 mm (24.2% -29.5 mm (-0.6%)
Diurnal 2 m Temperature Cycle
• Larger model diurnal 2 m temperature range (i.e. warm day, cool night) suggests too little cloud cover / too thin
• Affects other state variables… and must be seen in the cloud fractions and radiation
Cloud Fraction Biases• Estimated cloud fraction based on cloud liquid water and ice (Fogt and Bromwich 2008)– Converted 3-hr
observed NCDC cloud categories to decimal value
• January:– Positive biases in
western Europe and NA associated with storm tracks
• July:– Majority of
stations reflect negative CF biases
– Many stations have < -25% CF compared to observed
Clouds Continued• Model shows (+)
CF associated with higher terrain perhaps too strongly
• Storm tracks in N. Pacific and N. Atlantic depicted well
• North Slope CF reasonably well matched with MODIS and CloudSat/Calipso– No increase in
cloudiness adjacent to the coast
• Only conservative method yields results that approach observed
Radiation Sites
Longwave and Shortwave Radiation Biases
July 2007
Longwave W m--2 Shortwave W m-2
NAME OBS BIAS RMSD CORR NUM OBS BIAS RMSD CORR NUM
ABS 330.0 -54.1 59.1 0.46 247 198.2 14.5 143.5 0.78 248
ATQ 321.7 -39.1 48.5 0.10 248 263.2 74.0 124.4 0.91 248
BAR 301.1 -40.2 49.1 0.08 248 244.7 96.7 141.3 0.90 248
CAB 360.1 -42.1 48.2 0.60 225 185.0 95.0 211.8 0.80 225
FPE 373.1 -15.9 25.3 0.81 248 275.1 65.0 153.6 0.92 248
PAY 348.2 -47.0 55.9 0.40 248 240.0 12.9 176.9 0.84 248
SOD NA NA NA NA NA 166.8 59.7 155.0 0.78 248
SXF 367.4 -1.0 17.1 0.85 248 294.7 70.4 142.3 0.94 248
TAT 410.4 -26.1 30.9 0.57 248 146.3 110.4 228.5 0.82 248
XIA 396.6 -4.4 23.1 0.65 248 212.0 123.8 210.5 0.92 248
Expands previous studies with two additional sites (Abisko and Sodankylä)
Compared with middle months of 4 seasons with July shown hereNegative Longwave Radiation Biases…most significant at 99%.Positive Shortwave Radiation Biases… most significant at 99%.Poor longwave correlations/ Good shortwave correlations.
July 2007: Mid-Latitudes
FPE 373.1 -15.9 25.3 0.81 248 275.1 65.0 153.6 0.92 248
Longwave W m-2 Shortwave W m-2
• Fort Peck, Montana– Grassy flat location in
the northern Great Plains
– SW radiation often overpredicted by the model
– LW radiation generally underpredicted throughout the month
– Area shows 3ºC warm biases consistent with too much SW reaching the surface and a lack of radiative clouds for LW
– Note the difference on 1st, 7th, 11th, and 25th
July 2007: Polar
Longwave W m-2 Shortwave W m-2
ATQ 321.7 -39.1 48.5 0.10 248 263.2 74.0 124.4 0.91 248
• Atqasuk, Alaska– Flat tundra on the
North Slope of Alaska (~70 km away from the coast)
– SW radiation also overpredicted by the model
– LW radiation greatly underpredicted throughout the month
– Note large differences around the 10th and 25th but LW also not as tied to SW
July 2007: Polar
Longwave W m-2 Shortwave W m-2
ABS 330.0 -54.1 59.1 0.46 247 198.2 14.5 143.5 0.78 248
• Abisko, Sweden– Slightly sloping tundra
on the south shore of Lake Abiskojaure with terrain SW increasing rapidly
– SW radiation also overpredicted by the model but seems offset slightly by equal/opposite errors
– Abisko experiences less cloudy conditions due to down-sloping effect from the higher terrain SW not well represented by model (14th, 17th, and 19th)
Cloud Water/ Cloud Ice
• Scatter plots of Model LW vs. Observed Longwave for various model cloud species– (a) Cloud water and/or cloud ice available– (b) No Cloud water or ice– (c) Cloud water regardless of cloud ice– (d) Cloud ice only
• Mid-latitudes– LW correlations are strong for all 4 cases – When cloud water or ice is available,
model biases are negative – “Model Clear Sky”: Correlations increase
and model agrees better with observations
– Again, when cloud water is present (c) the model performs worse (Cloud ice has a zero effect on switch in RRTM scheme)
• Polar Region– Model LW suffers greatly compared to
observations– Apparent insensitivity between cloud
water/cloud ice conditions and “clear sky”
Shortwave and Longwave BiasesRevisited…ASR style.
• LESS Negative Longwave Radiation Bias: Many still significant• LESS Positive Shortwave Radiation: Fewer significant
differences• Improved Longwave correlations/ Mixed Shortwave
correlations.
July 2007
Longwave W m--2 Shortwave W m--2
NAME OBS PWRF ASR PWRF ASR OBS PWRF ASR PWRF ASR
ABS 330.0 -54.1 -2.7 0.46 0.41 198.2 14.5 -12.7 0.78 0.67
ATQ 321.7 -39.1 -19.1 0.10 0.22 263.2 74.0 32.0 0.91 0.90
BAR 301.1 -40.2 -12.4 0.08 0.24 244.7 96.7 27.2 0.90 0.86
CAB 360.1 -42.1 -11.0 0.60 0.71 185.0 95.0 76.0 0.80 0.81
FPE 373.1 -15.9 -8.7 0.81 0.82 275.1 65.0 49.4 0.92 0.93
PAY 348.2 -47.0 -17.3 0.40 0.63 240.0 12.9 47.6 0.84 0.92
SOD NA NA NA NA NA 166.8 59.7 11.0 0.78 0.79
SXF 367.4 -1.0 -0.1 0.85 0.88 294.7 70.4 36.6 0.94 0.96
TAT 410.4 -26.1 -9.6 0.57 0.65 146.3 110.4 109.3 0.82 0.85
XIA 396.6 -4.4 4.8 0.65 0.68 212.0 123.8 108.6 0.92 0.93
July 2007: Polar
Longwave W m-2 Shortwave W m-2
ATQ 321.7 -19.1 36.0 0.22 248 263.2 32.0 105.6 0.90 248
Summary and Future Work• Model
– Precipitation• Spatially consistent with ERA-Interim Reanalysis and GPCP• Small Annual (+) Biases in Mid-Latitude, Larger Annual (-) Polar Biases• Large (+) spring and summer biases tied to convection including
Russian sector rivers• Related to high evaporation and a moist lower boundary layer
– Cloud Fraction• Appears too low based on cloud water and cloud ice calculation• Cloud frequency technique compares better to MODIS and
CloudSat/Calipso discrepancies still exist– Radiation
• Significant (+) SW Down Biases and (-) LW Biases• Despite cloud water in the model, LW biases are still negative• Insensitivity to cloud water/cloud ice in the polar region (Perhaps
biggest concern needed to address in the future)• ASR
– Precipitation• Grid Nudging specific humidity decreases convection
– Better constrained moisture field in the boundary layer should improve performance
• ASR Precipitation needs to be analyzed– Radiation
• Improvements in biases for ASR in both SW and LW radiation
Thank You!NSF IPY Grant ARC-0733023
Columbus Zoo and Aquarium