Low Level WInd Shear Probabilities

We are evaluating two ensemble forecasts of low level wind shear (LLWS). LLWS is one of the hazards that are forecast as part of our turbulence package.  Both the AFWA and SREF have been post-processed to produce a probabilistic forecast of low level (0-2000 feet AGL) greater than 30kts.

09Z SREF run on 19 Feb, LLWS probabilities between 21Z Feb 19 - 09Z Feb 21.

12Z AFWA run on 19 Feb, LLWS probabilities between 21Z Feb 19 - 09Z Feb 21.

Icing prediction using Mixing Ratios

Several of the products that we are looking at for this experiment use mixing ratios to predict regions of icing.  Both the AFWA and SREF ensembles use mixing ratio from each of the members to predict where icing will occur.

The AFWA takes a worst-case approach for explicit prediction of super-cooled liquid.  Shown below is the ensemble maximum SLD mixing ratio (shaded) with the lowest freezing level from the ensemble contoured over the top.  PIREPs for lgt-mod icing or greater have been overlaid on this forecast.'

11 Feb 00Z AFWA forecast of worst case of SLD mixing ratio and lowest freezing level height.  PIREPs are for lgt-mod icing, or greater.

Post-processing on the SREF takes a slightly different approach, looking at cloud water mixing ratio between 0 and -25 C in each member, and produces a probability that icing will occur.  PIREPs of lgt-mod icing or greater are also shown in this loop.

11 Feb 09Z SREF probability of icing.  PIREPs are for lgt-mod icing, or greater.

IFR Verification

As a follow-up to yesterday's post concerning probabilistic flight rule prediction, here are some quick verification images.  For the C&V prediction, the forecast period is 21Z - 09Z.

Each image contains the METAR observations of flight category overlaid on the gridded data set.  The first gridded data set is an analysis of observed ceiling and visibility conditions, the National Ceiling and Visibility Analysis (NCVA).  Hourly data is show along with the METARs that went into this analysis.

NCVA Field and METAR flight category for 21Z 11 Feb - 09Z 12 Feb.

The primary forecast concern in this area was the lifting of the ceilings, bringing terminals out of IFR conditions.  The dataset shown below is the GFS-LAMP IFR conditions.  Regions shaded in green are where the 18Z GFS-LAMP run from 11 Feb is predicting IFR conditions.  It clears the IFR conditions fairly well in PA/NJ/NY/CT, but keeps DL/MD/VA in IFR conditions too long.

11 Feb 18Z GFS-LAMP forecast valid same period as NCVA.

Additionally, the SREF and AFWA probability of flight category as highlighted in yesterday's post is show below.  The SREF is the 09Z run from Feb 11, and the AFWA is the 00Z run from the same day.  Only the MVFR and IFR conditions are shown in each of these products. The SREF forecast does not seem to discriminate well between MVFR and IFR conditions (all or nothing) and clears the IFR conditions along the northeast coast fairly well.  The AFWA forecast tends to clear the IFR conditions to quickly along the northeast coast.

Another feature of both of these forecast is the forecast areas of IFR around the Great Lakes region that did not verify.  MVFR conditions were observed in MI and NWRN OH.  In the SREF this area is especially overdone and we are again seeing the issues that the SREF has discriminating between MFVR and IFR conditions.  The are in the AFWA is also overdone, with IFR conditions forecast in PA where only MVFR was observed.

11 Feb 09Z SREF forecast valid same period as NCVA.

11 Feb 00Z AFWA forecast valid same period as NCVA.

Ensemble Prediction of Flight Rules

One of the things being explored in the Winter Weather Experiment is the use of mesoscale and high-resolution ensembles for prediction of ceiling and visibility.  The two primary ensembles being explored for this purpose are the Short Range Ensemble Forecasting (SREF) System and the Air Force Weather Agency (AFWA) mesoscale ensemble.

The SREF is composed of 21 ensemble members with post-processed output on a ~16km horizontal resolution, while the AFWA ensemble has 10 members run at a ~4km horizontal resolution.  Each individual member is obtained by the Aviation Weather Center where we perform additional analysis that is not done in the basic ensemble post-processing.

Each of the individual member solutions contains a prediction of ceiling and visibility, for which ensemble statistics are computed.  Additionally, the flight rule category is computed for each ensemble member, from which the probability of the flight category can be determined.

These probabilities for the time period between 21Z Feb 11 - 09Z Feb 12 are shown for the AFWA and the SREF below.  Note that these probabilities are not calibrated and weight each of the ensemble members equally.  Both the weighting of the members and the calibration will need to be explored in order to improve the reliability of these forecasts (see: Reliability and Resolution). 

Probability of Flight Rules as predicted by the 00Z AFWA from 11 Feb 2013.  Valid 21Z 11 Feb - 09Z Feb 12.  MVFR, IFR, and LIFR conditions are shaded in blue, red, and yellow, respectively, shaded starting at 40%.

Probability of Flight Rules as predicted by the 03Z SREF from 11 Feb 2013.  Valid 21Z 11 Feb - 09Z Feb 12.  MVFR, IFR, and LIFR conditions are shaded in blue, red, and yellow, respectively, shaded starting at 40%.

Capacity Reduction Fields

One of the tools that has been used to diagnose the potential impacts to air traffic is the probability of airspace capacity reduction fields produced by the NCAR Research Applications Lab.  This tool uses members of the Air Force Weather Agency ensemble to diagnose the potential reduction of airspace due to weather.  This product can be thought of as a geometry problem, with each member analyzed to determine the extent to which flow into and out of a 100-km grid box will be obstructed using a mincut technique.  The capacity reduction for each of the members are then combined to produce a field that depicts the probability of a certain amount of airspace reduction, at a certain level, and in a certain direction of flow.

Note that because this problem simply uses the geometry of the grid to determine capacity reduction, it does not take into account planned or actual traffic.
Thus, using a simple example, if the number of planes that can fly through a particular portion of airspace over an hour is 80, and the demand for this hour is only 25 aircraft, a 50% reduction of airspace capacity (in the geometric sense, as is diagnosed in these products) will not impact aviation operations through this region.

Shown here are fields of 50% reduction in airspace capacity at FL300 (30,000 ft).  This shows the proportion of AFWA members that show a blockage of at least 50% for that particular 100-km grid box.  These are computed for two directions of flow: east-to-west/west-to-east, and north-to-south/south-to-north.  To get an idea of how these fields are derived, plots of the individual members are shown below the capacity reduction fields.

The first set of images is valid at 1700, when the AFWA members are capturing the system off the Atlantic coast.  The first frame depicts the reduction in the north-south direction, with the second frame depicting the east-west capacity reduction.  Because the line is oriented southwest-to-northeast fashion, the difference between the two directions is not as pronounced.  Another contribution to this similarity is the scale at which these fields are produced.

Probability of 50% reduction of FL300 air capacity in North-South direction at 1700 UTC.

Probability of 50% reduction of FL300 air capacity in East-West direction at 1700 UTC.

AFWA ensemble members showing simulated composite reflectivity >= 40 dBZ at 1700 UTC.

The next three images are valid at 2000 UTC, after several members of the AFWA are developing convection along the coastal sea breeze.  The difference between the two orientations is a bit more pronounced in this case, especially along the South Carolina coast.  The capacity reduction product indicates that the flow will face a greater (geometric) constraint in the north-south direction than in the east-west direction, which can be seen by looking at the individual solutions in the third image.   Along the Florida and Georgia Atlantic coast there is more reduction in the east-west direction due to the orientation of the convection in this region.

Probability of 50% reduction of FL300 air capacity in North-South direction at 2000 UTC.

Probability of 50% reduction of FL300 air capacity in East-West direction at 2000 UTC.

AFWA ensemble members showing simulated composite reflectivity >= 40 dBZ at 2000 UTC.

Convection in the Northeast.

The first week of this year's summer experiment is almost completed.  The week has certainly gone by quickly.  Today the high-resolution model evaluation and verification desk had a bit more time to put together a quick post. 

We were focused primarily on the timing and location of convection impacting the northeast.  One of the products used for this diagnostic was the probability of a lightning strike. The image below is a loop from 12Z-06Z centered over the northeast.  The field is the probability of a lightning strike within 20 nautical miles of a grid point, as calculated from the AFWA mesoscale ensemble prediction system.  This product indicated a significant lightning threat moving through Nova Scotia and Maine during this time.

AFWA probability of a lightning strike, 12Z-06Z.

Similar to the previous post, we took a look at spaghetti diagrams of the Air Force Weather Agency ensemble and the Storm Scale Ensemble of opportunity. Both ensembles indicated that the threat would primarily be cellular in nature.  The AFWA ensemble progressed the system more quickly and further east than did the SSEO.  Upon investigating the initial conditions for these ensembles, it was determined that this was likely due to the fact that all of the SSEO members were initialized off the 00Z operational NAM, while the AFWA members were initialized from 12Z and 18Z global model runs from the previous day which had the surface trough further east in earlier model runs.

The AFWA ensemble also developed more cellular convection southward, stretching across Pennsylvania. To get an idea of the potential impacts these solutions have on aviation operations, the top 25 jet routes have been plotted in each of these figures.  The AFWA solution then indicates more constraints to aviation traffic flow.

AFWA paintball plot of areas of simulated composite reflectivity >= 40 dBZ

SSEO paintball plot of areas of simulated composite reflectivity >= 40 dBZ

Convective Timing

This morning's analysis of the high-resolution ensemble models focused on the ongoing convection in the southeastern US.  One of the questions asked was when would convection clear out of the Atlanta, GA area.  One of the tools used were spaghetti/paintball plots showing the location of each of the individual members of the ensembles.  These graphics allow a quick visual of the solutions presented by the various ensemble members.

Spaghetti plot of AFWA ensemble members with composite reflectivity >= 40 dBZ, valid 17Z.

Spaghetti plot of SSEO ensemble members with composite reflectivity >= 40 dBZ, valid 17Z.

In general, more spread was observed in the AFWA ensemble than in the SSEO.  Timing was fairly good, with the location of leading edge of convection depicted in the same general area in southeast Georgia.

Radar and aircraft observations valid at 17Z.