Environmental Products for Aviation: A/C Sensors and a Breakthrough Process for Turbulence

Dr. Rex J. Fleming

Global Aerospace, LLC

(also representing SpectraSensors, Inc.)

2009 ACM / AEEC General Session

March 30 – April 1, 2009 Minneapolis, Minnesota

Overview Success in several NextGen operational concept tests has led to accelerating the pace – referred to as NowGen

Better a/c measurements and products are required for both: -- to help improve the required navigation performance -- to improve performance in new ATM operations

Michael Lewis (Director of Business Development for Boeing Advanced ATM) has this quote concerning NextGen:“required innovation already exists as a technology, weather prediction is currently not as good as it needs to be”

Show impact of improved a/c sensors on two NextGen ops

Reveal new method for identifying turbulence for aviation

Motivation for weather improvements

Giovanni Bisignani (CEO of IATA) claims: “ shaving one minute off each commercial flight would save 5.0 million tons of CO2 emissions and $3.8 billion in fuel costs each year ”

Carbon taxes may begin in January 2010!

Weather product improvements have made substantialimpact in fuel costs in past

Further progress essential for NextGen! Will show specific examples of how a/c sensors help in NextGen operations of: -- “flow corridors” -- “choreographed ascent / descent”

Definitions: ADS – B, 4DT, Update rate, RNP

Automatic Dependent Surveillance – Broadcast (ADS-B): a/c determine own position using GPS and broadcast it (withother relevant info) to ground stations and other a/c

4-dimensional trajectory (4DT) [each a/c ‘s current & future path in space/time] common knowledge of each 4DT by all hands enables major increases in air traffic throughput.

The Update rate is the frequency with which 4DT and other info is simultaneously upgraded across the network to all players; this is more than ADS-B -- still being worked on

Required Navigational Performance: RNP = 0.3 means all a/cwithin 0.3 nm of intended horizontal position 95 % of the time

One more definition and interactions between them

Future a/c ground speeds will depend upon sum of the “true air speed (TAS)” (TAT probe error and / or predicted temperature error) and the model wind prediction error

RNP optimized for safety, but position uncertainty will grow until the next update rate of 4DT for all aircraft; A “delta Δ position” in the future depends upon (Δ velocity) x (time)

A successful and efficient model refresh rate for aviationis primarily a function of accurate environmental dataprovided by commercial aircraft

Model Refresh Rate: 4DT requires future positions and velocities – from prediction models with an analysis/ forecast cycle or refresh rate; model errors grow in time (atmosphereis a chaotic system)

ΔP = ΔP (GPS)

+ { ΔV (TAS)Ran + ΔV (TAS)Bias + ΔV (OBS) + ΔV (ATM) } [Time]---------------------------------------------------------------------------------------

2nd largest error from [ΔV (TAS)Bias ] [Time]

(TAS)Bias = (M) ([Speed of sound) = (M) [(401.87)(ΔTS) ]1/2

M = Mach number (from pitot tube) TS from TAT: TT = TS ( 1 + 0.2 M 2 )-------------------------------------------------------------------------------------------------------

Largest error from [ΔV (model predicted wind error)] [Time]

Position error for a given aircraft ΔP

Forecast wind errors as a function of pressure (from NOAA/FSL)

Model wind errors from a regional forecast model

Flow Corridors: long tubes or “bundles of near parallel 4DT assignments which achieve a very high traffic throughput

Corridors planned in advance; adjusted every 2 hours withaccurate wx forecasts for efficiency – use of jet stream winds,avoiding turbulence, convection [temp (T) & water vapor [q])

Results (update rate=20 sec, 3 m/s wind error): RNP = 0.49 nm(eliminating TAT T bias, 1.5 m/s wind error): RNP = 0.26 nm

Sudden corridor change [rapidly growth in enroute convectionor moderate to severe turbulence, loss of key terminal] could lead to massive delays, safety issues, etc.

Unlike havoc on highway where autos can come to a stop -- high density a/c traffic needs to be shuffled somewhere!

Forecast errors of RH as a function of pressure (from NOAA/FSL)

How does one predict convection and support aviation with this skill?

Choreographed Arrival/Departure (future plans)

Results (update rate=16 sec, 3 m/s wind error): RNP = 0.31 nm eliminating TAT T bias, 1.5 m/s wind error): RNP = 0.16 nm

Continuous descent approach (CDA) can occur in all weather conditions, with mixed fleets, and even at busy terminals!

However, need accurate wind, temp (T), and water vapor (q) from commercial aircraft in the greater (150 nm) terminal area for analysis and projection of convection: turbulence,thunderstorms, microbursts, and wake vortices

Airport terminals will have 4-D picture of terminal area (out to~ 150 nm radius) from forecast model with 5 min refresh rate

Let’s stop flying into turbulence to find it!

Let’s start avoiding turbulence!

Stop wine and coffee spills in our laps!

Stop bouncing flight attendants off cabin ceilings!

Atmospheric Turbulence is extremely complex!

No mathematical solution nor adequate model exists

Turbulence pulls energy from the mean flow at large scales-- this gain approximately balanced by viscous dissipationat very small scales – when it exists, it is continuous

Turbulence requires a continuous supply of energy (for suchviscous losses). If no energy is supplied, turbulence decays rapidly – thus it can come and go rapidly

GPS receiver

We have a passive method of determiningwhen turbulence exists along thisGPS ray path …

But just where is it along this raypath ? -- for aviation the concerncould be from anywhere near the surface to 40,000 feet

Solution: provide more GPS rays than grid points using both airborne and ground-based GPS receivers!!

GPS Surface Site Turbulence Processing Center

3-D Map

GPS receivers on commercial aircraft and at ground sites

Communication of variance (turbulence) information from each GPS receiver to the Turbulence Processing Center (TPC)

TPC computes geometry of GPS rays

TPC uses mathematical algorithms and special techniques to produce national or regional maps around the Earth

Communication of tailored products to Air Navigation Service Providers (ANSP), pilots, dispatch, and others with need to know

Patent claims (already granted) cover the complete system

Summary of GPS errors removed

GPS signals travel at speed of light through a vacuum, but the rays are slowed (Δ time delay) by effects of temperature(T) & water vapor (q) These averaged delays need not be removed for the turbulence product

A variety of processes remove other errors: ionospheric delay, receiver clock errors, satellite clock errors and ephemeris errors (WAAS and LAAS)

The mean GPS signal delay is a function of the ~ laminarfluid flow -- the (nearly horizontally uniform) temperature (T)and water vapor (q) fields – the fluid properties which get colder and drier as one goes higher in the atmosphere

Why GPS variance reveals atmospheric turbulence

Mean GPS signals are typically slowly varying: given by an equation involving T and q. Scientists have assumed values of T and approximated q or vice versa. We do not need to separate effects of T and q in the turbulence calculation

Turbulence is a function of the fluid flow -- not the fluid itself which upsets the laminar flow conditions in T and q when the turbulence is present -- producing the high frequency variability (variance of the GPS signal delay about the mean)

Variance data about the geophysical mean signal computed by the GPS receivers (e.g. data at 5 Hertz for 6 s = 30 samples)

Given 30 consecutive samples X (i) of excess delay from a single ray, the variance is:

Variance = { Σ (X (i) – u)2 } / 30 where i = 1,30

and where u is the mean of the sample coming from the T and q present in the fluid (atmosphere); the variance comes from the fluid flow (turbulence) disturbing the flow

Variance values for each GPS slant path will be computedby the GPS receiver software – then sent to the TPC

National Turbulence Map from satellites in view (7 of 8 used)

Variance information computed in GPS receiver from 5 rays (each 6 seconds apart) simultaneously from 7 visible satellitesand communicated as single packet of info every 30 seconds

Repeat above every 3 minutes (duty cycle 30 s / 3 m = 1 / 6) so TPC computes national sub-map every 3 minutes

Produce national map and products (based upon 3 - 4 sub-maps) every 15 min; disseminate map and product

Turbulence map produced from solving largeover-determined matrix equation A X = B

A X = B X is vector of unknown turbulence at (N) points

A: calculated weight matrix (# of rays (R) by # of points (N)

A = [ R x N ] is large matrix with many zero values

X: vector of turbulence values at N grid points X = [N x 1]

B: measured vector variances for the R rays B = [ R x 1]

[ R x N ] [ N x 1 ] = [ R x 1 ] R >N thus over-determined (SVD)

All values change with each map. Grid is stationary, but X =turbulence values change with time; rays change position assatellites and a/c move so A’s and B’s change

System Design Concepts

Need more GPS rays than grid points!

Useful national map would have approximately 250,000 gridpoints (100 x 100 x 25) extending over ocean, into Canada &Mexico, up to ~ 40,000 feet (12 km with 0.5 km resolution)

Excellent results with Ratio of (# of rays) / (# of points) = 1.16

Typical arrangement of 7500 surface sites and 1000 activeaircraft yields ~ 300,000 rays – providing a Ratio ~ 1.2

7500 ground sites from: map of 5588 largest airports of 14,858 nationally.

1000 potential aircraft from 4552 commercial aircraft in US fleet (ATA report 2007)

12.0 km 11.5 km 11.0 km 45 45 45 45 45

10.5 km 90 90 9090 120 10.0 km 120 90 9.5 km 105 105 105 105 9.0 km 8.5 km 8.0 km 7.5 km 7.0 km 6.5 km 6.0 km 5.5 km 5.0 km 4.5 km 4.0 km 3.5 km 3.0 km 2.5 km 2.0 km 1.5 km 1.0 km 0.5 km surface J = 1 J = 2 J = 3 J = 4 J = 5 J = 6 J = 7 J = 8 J = 9

Ratio of rays to grid points was 1.16, this gave perfect solution

All rays must exit top of grid; low elevation angle rays not used

3.5 years of extensive simulations performedTurbulence over 4 vertical layers with 4 intensities (LGT, MOD, SVR, & mixed)

Can test over an abbreviated horizontal grid in 2-D with full vertical scale like 225 grid points shown here, or achieve same results with extended grid (2525 grid points and 3720 rays performed on laptop pc)

Many simulations performed – always more rays than grid pts

Turbulence (simple to complex) reproduced using various2 and 3-D grid resolutions; satellites and aircraft moved; and random errors added to ray values to find possible failure modes – none found!

But an important processing change had to be made – the Amatrix is geometrically sensitive to satellite, aircraft, and ground receiver positions with random errors added.

The A – matrix is primarily zeros (a single ray influences at most 0.8% of the available grid points)

The scaled variance values (0, 2, 3, and 4) for (no, light,moderate, and severe turbulence) communicated from the GPS receivers are rescaled to (0, 60, 90, and 120) in the TPC

With random errors added to all the B –vector variance values (both turbulent and non-turbulent) the turbulent cells are still accurate to 95-99%, but the non-turbulent cells haveerror values ranging from extremely small (10 -4 ) to quite large (up to 40% of the “light turbulence” values of [60]) Via the A matrix, the SVD method disperses the largest errors to grid points with the least ray influence – thus gridpoints near the surface and grid edge get largest errors.

Part of the TPC quality control (QC) procedures are to set allcomputed values less than ~ 60 to zero

Other QC procedures: using pilot reports and automaticturbulence reports from aircraft for feedback; intercomparison and integration of significant radar echoes; etc.

This product is less important in lowest 1 kilometer (~ 3000 ft) -- typical height of the planetary boundary layer (PBL)

Often PBL is quiet, but usually small scale turbulence due to breaking gravity waves in an unstable PBL

The 22 million arrivals and departures in the USA provideample information on turbulence in the PBL as pilots and controllers readily communicate significant information

As more aircraft provide real time information on winds, temp,and water vapor during ascent/ descent, future models can predict changes in the PBL stability (if not turbulence itself)

The lowest levels over the ocean are not currently covered by GPS rays – only low elevation angles from coastal ground sites and rays from over ocean aircraft are available

Turbulence usually occurs in thin layers, but can cover multiple layers, in either case it is continuous

Is variance signal passing through turbulence in upper layers additive with a turbulence layer in a lower layer (e.g. inthe boundary layer) – a case where 1 + 1 = 2.1 (not 2) ?

We have solved this when A + B = (constant fraction) (A + B); even when the fraction is changing as a function of altitude!

Fraction is exp (C x LK) where LK is the base for interpolation

where C is + (fraction exponentially increasing with height) or where C is - (fraction exponentially decreasing with height)

Most rays will not see any turbulence – not that ubiquitous

Identifying small thin layers of turbulence is relatively easy(turbulence is homogeneous and isotopic on small scales)

Moderate to severe turbulence on large scale is lesslikely to be isotropic (same variability in all directions), butperhaps nearly so from a GPS variance perspective

The re-scaling of variance values to (60, 90, and 120) in the TPC for turbulence of (light, moderate, and severe)will help identify a blending of information

The evaluation over and between sub-maps will help identify this blending – the important thing is that such turbulent areas will be easily identifiable – if not preciselyquantifiable

Time for action on two major activities

Accelerate accurate a/c measurements of winds, T, and q

Implement passive turbulence system; GPS receivers abound,but must be modified; need corporate partner to help initiate

Immediate benefit of less fuel use and less carbon emissionsEvolving benefits for NowGen and NextGen operationsReduce potential passenger’s personal anxiety about flying

These basic infrastructure improvements serve all of society(beyond aviation) and will help stimulate all economies (in the USA weather and climate impact 1/7 of GDP)

Carbon taxes could help fund all the above requirements

Expected Revenue

Facts from 2007 Air Transport Association (ATA) Annual Report

-- 11,365,000 aircraft departures from US airlines

-- 769.2 M passengers per this year

-- charge $15 per aircraft departure for total turbulence (24 by 7) coverage this is $ 170.5 M per year

-- charge $0.22 per passenger for same total coverage this is $ 169.2 M per year

Revenue funds paid by Federal Government Agency and/or by air carrierswith cost added to ticket price

Capital costs1000 aircraft equipped cost = $10 K each = $ 10.0 M7500 GPS surface sites cost = $1.4 K each = $ 10.5 M Initial installation costs $10 M (a/c) $10 M (sfc) = $ 20.0 M

Recurring costsAircraft (1000) communication costs $ 4.8 M / yearGround communication (7500) costs $ 4.5 M / yearTPC personnel costs (salary & benefits) $ 6.0 M / yearComputer, facilities, overhead, travel, etc. $ 6.0 M / yearReceiver maintenance and replacement $ 1.1 M / year (after first 2 years) at 5% of capital costsTOTAL $ 22.4 M / year

Expenses: Capital costs and recurring costs

Return on Investment (ROI)

Profit per year = ( $170.5 M - $22.4 M) = $ 148.1 M per year

Full accounting with depreciation of assets, etc., later butassume effective tax rate of 15% gives PAT = $125.9M

Knowledge of where turbulence exits in advance reduces personnel anxiety, saves fuel use, reduces CO2 emissions, and will be essential in safe, optimal use of new operationsin NowGen and NextGen

ROI achieved one year after full operation; opportunity for profitable investment with long term cash cow; help make the world a better place to fly!

Timing options for Passive GPS Turbulence Monitoring System (10 minute map: TPC processing in 2.0 min.) Time in Minutes 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Color codeVariance processing (5 rays each 6 seconds apart = 30 seconds) Duty cycle 3

minutesUp to 30 second delay thru avionics units (GPS receiver thru ARINC to TPC)Processing time in the TPC for sub-map and /or final mapUp to 30 second delay for communication of map/products to a/c and other

usersProduct is approximately (10 – 3.5) = 6.5 minutes old

Timing options for Passive GPS Turbulence Monitoring System (15 minute map: TPC processing in 3.0 min. – 1 computer) Time in Minutes 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Color codeVariance processing (5 rays each 6 seconds apart = 30 seconds) Duty cycle 5

minutesUp to 30 second delay thru avionics units (GPS receiver thru ARINC to TPC)

Processing time in the TPC for sub-map and /or final mapUp to 30 second delay for communication of map/products to a/c and other

usersProduct is approximately (15 – 5.5) = 9.5 minutes old

Transmission time

Computer speed 1 minute

Computer speed 2 minutes

Computer speed 3 minutes

Computer speed 4 minutes

3 Minutes 5.5 ( 9 ) 6.5 ( 10 ) 8.5 ( 12 ) 10.5 ( 14 )

4 Minutes 7.5 ( 12 ) 8.5 ( 13 ) 9.5 ( 14 ) 10.5 ( 15 )

5 Minutes 7.5 ( 13 ) 8.5 ( 14 ) 9.5 ( 15 )

10.5 ( 16 )

Latency time (average age of input data in minutes)Frequency of issuance of national turbulence map (in minutes)