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

1. 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

2. Overview Success in several NextGen operational concept tests has led to accelerating the pace – referred to as NowGen Better a/cmeasurements 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 alreadyexists 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

3. 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 substantial impact in fuel costs in past Further progress essentialfor NextGen! Will show specific examples of how a/c sensors help in NextGen operations of: -- “flow corridors” -- “choreographed ascent / descent”

4. Definitions: ADS – B, 4DT, Update rate, RNP Automatic Dependent Surveillance – Broadcast(ADS-B): a/c determine own position using GPS and broadcast it (with other 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 infois 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/c within 0.3 nm of intended horizontal position 95 % of the time

5. One more definition and interactions between them 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 (atmosphere is a chaotic system) Futurea/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 aviation is primarily a function of accurate environmental data provided by commercial aircraft

6. Position error for a given aircraft ΔP 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]

7. Model wind errors from a regional forecast model Forecast wind errors as a function of pressure (from NOAA/FSL)

8. 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 with accurate 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 TATT bias, 1.5 m/s wind error):RNP = 0.26 nm Sudden corridor change [rapidly growth in enroute convection or 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!

9. How does one predict convection and support aviation with this skill? Forecast errors of RH as a function of pressure (from NOAA/FSL)

10. Choreographed Arrival/Departure (future plans) Airport terminals will have 4-D picture of terminal area (out to ~ 150 nm radius) from forecast model with 5 min refresh rate 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 accuratewind, temp (T), andwater vapor (q) from commercial aircraft in the greater (150 nm) terminal area for analysis and projection of convection: turbulence, thunderstorms, microbursts, and wake vortices

11. 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!

12. 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 dissipation at very small scales – when it exists, it is continuous Turbulence requires a continuous supply of energy (for such viscous losses). If no energy is supplied, turbulence decays rapidly – thus it can come and go rapidly

13. We have a passive method of determining when turbulence exists along this GPS ray path … But just where is italong this ray path ? -- for aviation the concern could be from anywhere near the surface to 40,000 feet GPS receiver

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

15. 3-D Map GPS Surface Site Turbulence Processing Center

16. Patent claims (already granted) cover the complete system 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

17. Summary of GPS errors removed GPS signals travel at speed of light through a vacuum, butthe rays are slowed (Δ time delay) by effects oftemperature (T) & water vapor(q)These averaged delays need not be removedfor 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 ~ laminar fluid 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

18. 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)

19. 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 themean of the sample coming from the T and q presentin the fluid (atmosphere); the variance comes from the fluid flow(turbulence) disturbing the flow Variance values for each GPS slant path will be computed by the GPS receiver software – then sent to the TPC

20. 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 satellites and 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 large over-determined matrix equation A X = B

21. 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 Ngrid pointsX = [N x 1] B: measured vector variances for the R rays B = [ R x1] [ 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 as satellites and a/c move so A’s and B’s change

22. System Design Concepts Need more GPS rays than grid points! Useful national map would have approximately 250,000 grid points (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 active aircraft yields ~ 300,000 rays – providing a Ratio ~ 1.2

23. 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)

24. 3.5 years of extensive simulations performed Turbulence over 4 vertical layers with 4 intensities (LGT, MOD, SVR, & mixed) 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 Can test over an abbreviated horizontal grid in 2-D with full vertical scale like 225 grid points shown here,or achieve sameresults with extended grid (2525 grid points and 3720 rays performed on laptop pc)

25. Many simulations performed – always more rays than grid pts Turbulence (simple to complex) reproduced using various 2 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 A matrix 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

26. 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 have error 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 grid points near the surface and grid edge get largest errors. Part of the TPC quality control (QC) procedures are to set all computed values less than ~60 to zero Other QC procedures: using pilot reports and automatic turbulence reports from aircraft for feedback; inter comparison and integration of significant radar echoes; etc.

27. 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 provide ample 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

28. 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. in the 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 whereC is - (fraction exponentially decreasing with height)

29. 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 less likely to be isotropic (same variability in all directions), but perhaps 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 precisely quantifiable

30. 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 emissions Evolving benefits for NowGen and NextGen operations Reduce 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

31. 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 carriers with cost added to ticket price

32. Expenses: Capital costs and recurring costs Capital costs 1000 aircraft equipped cost = $10 K each = $ 10.0 M 7500 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 costs Aircraft (1000) communication costs $ 4.8 M / year Ground communication (7500) costs $ 4.5 M / year TPC personnel costs (salary & benefits) $ 6.0 M / year Computer, facilities, overhead, travel, etc. $ 6.0 M / year Receiver maintenance and replacement $ 1.1 M / year (after first 2 years) at 5% of capital costs TOTAL $ 22.4 M / year

33. 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 but assume 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 operations in 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!

34. 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 code Variance processing (5 rays each 6 seconds apart = 30 seconds) Duty cycle 3 minutes Up to 30 second delay thru avionics units (GPS receiver thru ARINC to TPC) Processing time in the TPC for sub-map and /or final map Up to 30 second delay for communication of map/products to a/c and other users Product is approximately (10 – 3.5) = 6.5 minutes old

35. 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 code Variance processing (5 rays each 6 seconds apart = 30 seconds) Duty cycle 5 minutes Up to 30 second delay thru avionics units (GPS receiver thru ARINC to TPC) Processing time in the TPC for sub-map and /or final map Up to 30 second delay for communication of map/products to a/c and other users Product is approximately (15 – 5.5) = 9.5 minutes old

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