Validation ExerciseGDI-BT r^2First: Where is the GDI expected to work?

Where is the GDI expected to work best?Recall driving processes: ECI: High-THTE column in the low-mid troposphere(2) MWI: Mid-level ridge stabilizing(3) II: Subsidence inversions causing low-level stabilization + drying

Is this what we found?

GDI vs Brightness Temperature: Index values from 1x1 deg GFS00 compared against IR4 Brightness Temperatures from GOES. Each value represents a 12-hour average (GFS00 F00-F12, GOES 00z-12z)Timeframe: July-November 2013 (Wet season)

Validation ExerciseNorth AmericaGDI vs OLR: Index values from 1x1 deg GFS compared against OLR. Each value represents a 24-hour average (GFS00 F00-F24, OLR 00z-00z_day+1).

GDI vs Rain: Index values from 1x1 deg GFS compared against 24-hr rainfall analyses from CPC. Each value represents a 24-hour average (GFS00 F12-F36, OLR 12z-12z_day+1)Timeframe: October 2013-February 2014 (Wet season)

South AmericaBoth continents in the respective rainy seasonsSouthamerica had more thorough analysis thanks to Silvia and the rest of the team.

CPC Rain-Stations density very low in some places -Convective rainfall can be very localized and not being measured.Dataset limitationsOLR and-Assumption that temperatures/OLR Brightness measurements are representative.Temperature-Cirrus contamination, especially under subtropical jets and around MCSsIndex-Dependent on model output. If the model Values is wrong then all indices are.

Validation PhilosophyValidation question: How well does the GDI capture area coverage and type of convection (shallow vs deep)?-We are interested more in this than rainfall amounts. This also reflects better when validating with BT or OLR.

This means that in areas such as the Gulf of Honduras the GDI can diagnose 50% of the variance of the convective regime. Scatterplots coming laterResults in the Caribbean!GDI correlates well with brightness temperature for the most part.

Comparison against other stability indicesRed means the GDI did better than the other index r2 of the GDI were compared to those of other indices. Question: How much better does the GDI resolve the potential for convection?

Comparison against Precipitable Water (PWAT)r2 GDI - PWATPWAT is not a stability index but is a good indicator of the amount of water vapor available for precipitation.

Why does the GDI compete with PWAT?Lets revisit the large scale transport of heat and moisture.

PWAT reflects the deep-layer moisture transport. Is the vapor integrated throughout the troposphere.

The GDI also captures this transport (for the most part). Yet since it considers inversions and troughs/ridges, the GDI does a better job defining more discrete regions with convective instability. PWAT, as the K does, has problems differentiating between shallow convection regimes, overestimating the potential.

It is easier to identify regions with the potential for convection with the GDI than with PWAT.

Scatterplots! Brightness T vs Index Value

Two useful applications2) Find and track perturbations in the trades Great to identify potential for thunderstorms (aviation) as it resolves finer structures. Captures convective instability associated with Tropical, Easterly and TUTT-induced waves. Also captures weaker perturbations that produce enhanced shallow convection.Forecasting of type of convective regime expected especially in tropical and subtropical regions. To some extent, expected rainfall amounts can be inferred from convection type. Independent to some extent of convective parameterization.

Tracking waves in the trades with the GDI

Tracking waves in the trades with the GDI

Summary +Here, important processes aside from heat and moisture availability are: -Subsidence Inversions -Mid-level troughs and ridges. (2) Beats the K-index across most of the Caribbean & South America. (3) Only competes with PWAT but captures much better the structure of convection. (6) Wide range of significant applications: Aviation (potential for t-storms); general forecasting of tropical convection; monitor trade wind perturbations. (1) Up to now the GDI seems to be the best available tool to evaluate the potential for tropical convection. Works best in trade wind regimes and in areas downwind. (5) Best way to use the GDI: Look at the GDI, PWAT and flow. Dynamics are very important. (4) As forecast time increases, the GDI can beat model rainfall as a predictor since it is to some extent independent of the convective parameterization.

Questions?More information athttp://www.wpc.ncep.noaa.gov/international/gdi/Realtime 7-day GDI forecasts using GFS model output.More information, including documents and package to install the GDI if you are a Wingridds user.Contacting usjose.galvez@noaa.gov

GDI values. General assessment for convective regimes. These values will vary slightly upon location and time of the year.*The GDI works best in areas where the ECI is high. This is true for most of the tropics. For some regions in mid-latitudes this is true for the summer months. The same is true for equatorial tiers of cold-current environments.GDI also works where II varies. This means that when it meanders between 0 and negative values. This occurs in trade-wind-type inversions. In other places such as cold current the inversions may be too low. Furthermore in inland locations in continents trade wind inversions become modified by the land and lose their properties. Thus it works best in eastern coasts.Finally the GDI is sensitive to mid-level temperatures. It should work best where mid-levels alternate between warm ridges and cool troughs. The combination of all these factors is reached at its best in the subtropics. Some tropical locations also get benefits from this variability. One east the eastern and northern coast of Brasil where the Cavado do Nordeste forms and evolves. Similarly, mid-level troughs amplify in the eastern Caribbean into northeastern South America.*Yes, this is what we found. These are the validation correlation maps (refer to next presentation). It shows the r^2 or determination coefficient calculated by comparing the GDI against brightness temperature in the Caribbean, and against OLR in South America. Note that the scales are slightly different. R^2 exceed .5 across most of the northern Caribbean, especially across Mexico and Gulf of Honduras. This means tha the GDI alone was able to explain more than 50% of the brightness temperature variance.

Across South America, correlations with OLR exceed 0.7 (or 70% of the variance) in southeastern Brasil. Yellows are o.5, so note the large area in which the GDI explained over half of the OLR variance. In general, these areas concide with the areas where the GDI was expected to perform. So the results make sense and the r^2 were quite high in several regions.

*Errors of validation:Model error, affects all indicesData error (OLR or Rainfall Analyses are not representative)(3)Cirrus error on OLR*Errors of validation:Model error, affects all indicesData error (OLR or Rainfall Analyses are not representative)(3)Cirrus error on OLR*Errors of validation:Model error, affects all indicesData error (OLR or Rainfall Analyses are not representative)(3)Cirrus error on OLR*-In places like the Bahamas, dynamic forcing can play an important role in convection, thus scatter can be large in intermediate ranges such as the GDI 0-10 values.*