• Cardinali, C., , Isaksen L. , , and Andersson E. , 2003: Use and impact of automated aircraft data in a global 4DVAR data assimilation system. Mon. Wea. Rev., 131, 18651877, doi:10.1175//2569.1.

    • Search Google Scholar
    • Export Citation
  • Collinson, R. P. G., 2011: Introduction to Avionics Systems. 3rd ed. Springer, 530 pp., doi:10.1007/978-94-007-0708-5.

  • Cress, A., , and Wergen W. , 2001: Impact of profile observations on the German Weather Service’s NWP system. Meteor. Z., 10, 91101, doi:10.1127/0941-2948/2001/0010-0091.

    • Search Google Scholar
    • Export Citation
  • de Haan, S., 2010: Quality assessment of high resolution wind and temperature observations from ModeS. Rev. ed. KNMI Scientific Rep. WR 2009-07, 31 pp.

  • de Haan, S., 2011: High-resolution wind and temperature observations from aircraft tracked by Mode-S air traffic control radar. J. Geophys. Res., 116, D10111, doi:10.1029/2010JD015264.

    • Search Google Scholar
    • Export Citation
  • de Haan, S., 2013: An improved correction method for high quality wind and temperature observations derived from Mode-S EHS. KNMI Tech. Rep. TR-338, 43 pp.

  • de Haan, S., , and Stoffelen A. , 2012: Assimilation of high-resolution Mode-S wind and temperature observations in a regional NWP model for nowcasting applications. Wea. Forecasting, 27, 918937, doi:10.1175/WAF-D-11-00088.1.

    • Search Google Scholar
    • Export Citation
  • de Haan, S., , de Haij M. , , and Sondij J. , 2013: The use of a commercial ADS-B receiver to derive upper air wind and temperature observations from Mode-S EHS information in The Netherlands. KNMI Tech. Rep. TR-336, 44 pp.

  • de Leege, A., , van Paassen M. , , and Mulder M. , 2013: Using automatic dependent surveillance-broadcast for meteorological monitoring. J. Aircr., 50, 249261, doi:10.2514/1.C031901.

    • Search Google Scholar
    • Export Citation
  • EUMETNET, 2015: E-AMDAR. Accessed 7 July 2015. [Available online at http://www.eumetnet.eu/e-amdar.]

  • Eyre, J., , and Reid R. , 2014: Cost-benefit studies for observing systems. Met Office Forecasting Research Tech. Rep. 593, 11 pp.

  • Finlay, C. C., and et al. , 2010: International geomagnetic reference field: The eleventh generation. Geophys. J. Int., 183, 12161230, doi:10.1111/j.1365-246X.2010.04804.x.

    • Search Google Scholar
    • Export Citation
  • Graham, R. J., , Anderson S. R. , , and Bader M. J. , 2000: The relative utility of current observation systems to global-scale NWP forecasts. Quart. J. Roy. Meteor. Soc., 126, 24352460, doi:10.1002/qj.49712656805.

    • Search Google Scholar
    • Export Citation
  • ICAO, 2012: Technical provisions for Mode S services and extended squitter. 2nd ed. International Civil Aviation Organization Doc. 9871, AN/460, 326 pp.

  • Jacobs, N. A., , Mulally D. J. , , and Anderson A. K. , 2014: Correction of flux valve–based heading for improvement of aircraft wind observations. J. Atmos. Oceanic Technol., 31, 17331747, doi:10.1175/JTECH-D-13-00175.1.

    • Search Google Scholar
    • Export Citation
  • KNMI, 2014: QEvC info. Accessed 10 June 2015. [Available online at http://www.knmi.nl/samenw/geoss/eumetnet/E-Amdar/QEvC/.]

  • Köllner, G., 2013: Mode-S Beast. Accessed 8 July 2014. [Available online at http://www.modesbeast.com/.]

  • Lorenc, A. C., , and Marriott R. T. , 2014: Forecast sensitivity to observations in the Met Office global numerical weather prediction system. Quart. J. Roy. Meteor. Soc., 140, 209224, doi:10.1002/qj.2122.

    • Search Google Scholar
    • Export Citation
  • Maus, S., , Macmillan S. , , McLean S. , , Hamilton B. , , Thomson A. , , Nair M. , , and Rollins C. , 2010: The US/UK World Magnetic Model for 2010-2015. British Geological Survey, NOAA Tech. Rep. NESDIS/NGDC, 98 pp.

  • Petersen, R. A., 2016: On the impact and benefits of AMDAR observations in operational forecasting. Part I: A review of the impact of automated aircraft wind and temperature reports. Bull. Amer. Meteor. Soc., doi:10.1175/BAMS-D-14-00055.1, in press.

  • Stone, E. K., , and Kitchen M. , 2015: Introducing an approach for extracting temperature from aircraft GNSS and pressure altitude reports in ADS-B messages. J. Atmos. Oceanic Technol., 32, 736743, doi:10.1175/JTECH-D-14-00192.1.

    • Search Google Scholar
    • Export Citation
  • Strajnar, B., 2012: Validation of Mode-S Meteorological Routine Air Report aircraft observations. J. Geophys. Res., 117, D23110, doi:10.1029/2012JD018315.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Two maps of the United Kingdom. (a) Map has 0.025° longitude × latitude squares, where the colors indicate the lowest observation GNSS altitude within that box between 21 and 31 May 2015, inclusive. The bright red boxes indicate an altitude between 0 and 100 m, suggesting that there are observations very close to the surface. (b) Map has nonopaque circles for each observation between 0600 and 0700 28 May 2015. The colors represent which receiver was used; during this time 310 000 observations were derived.

  • View in gallery

    (a) Number of observations derived per day from aircraft Mode-S messages received using the five Met Office receivers. (b) Average number of observations per hour derived from aircraft Mode-S messages received using the five Met Office receivers using the same data as in (a).

  • View in gallery

    The o−b mean values (gray lines) and RMS (black lines) for 7 days of data from 9 to 15 Aug 2013 in 250-m altitude bins. The solid lines show the uncorrected data; the dashed lines show the uncorrected data after a heading correction has been applied. The 7 days of 1–7 Aug 2013 data have been used to create the corrections. (a) The u wind component, (b) the υ wind component, (c) the number of observations per 250-m high altitude bin, and (d) the temperature.

  • View in gallery

    A histogram of the heading corrections generated from Mode-S EHS–derived data generated between 1 and 7 Aug 2013, and applied to the data received for 9–15 Aug 2013.

  • View in gallery

    The o−b statistics for Mode-S EHS–derived (solid black circles) and AMDAR (solid gray triangles) u components using the UKV model. (a) The mean o−b for each model run, (b) the RMS o−b for each model run, (c) the number of Mode-S o−b values, and (d) the number of AMDAR o−b values. Heading corrections were introduced on 28 Aug 2014.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 294 294 31
PDF Downloads 203 203 23

A Network of Mode-S Receivers for Routine Acquisition of Aircraft-Derived Meteorological Data

View More View Less
  • 1 Met Office, Exeter, United Kingdom
© Get Permissions
Full access

Abstract

Aircraft are an important source of meteorological observations for both numerical weather models and aviation weather forecasting. There has been increasing interest in novel methods for gathering aircraft-based observations, especially mode-selective (Mode-S) enhanced surveillance (EHS)-derived data. This paper reports on the progress made in the United Kingdom at the Met Office on receiving and processing these data. Five receivers have been deployed, providing up to 5.7 million observations of horizontal wind and temperatures per day over the United Kingdom. The receivers are relatively low cost and deploying them at existing operational weather radar sites has been shown to be an ideal choice. Heading corrections are required to improve the quality of the wind observations. When corrected the Mode-S EHS wind data have similar observations-minus-background (o−b) statistics as Aircraft Meteorological Data Relay [AMDAR; using a Met Office version of the Unified Model over the United Kingdom (UKV)]. For the u wind component, the average per model run o−b root-mean-square value for the Mode-S EHS–derived data was 2.45 and 2.12 m s−1 for AMDAR. The AMDAR data are assimilated into the model.

Corresponding author address: Ed Stone, Met Office, FitzRoy Road, Exeter, Devon EX1 3PB, United Kingdom. E-mail: ed.stone@metoffice.gov.uk

Abstract

Aircraft are an important source of meteorological observations for both numerical weather models and aviation weather forecasting. There has been increasing interest in novel methods for gathering aircraft-based observations, especially mode-selective (Mode-S) enhanced surveillance (EHS)-derived data. This paper reports on the progress made in the United Kingdom at the Met Office on receiving and processing these data. Five receivers have been deployed, providing up to 5.7 million observations of horizontal wind and temperatures per day over the United Kingdom. The receivers are relatively low cost and deploying them at existing operational weather radar sites has been shown to be an ideal choice. Heading corrections are required to improve the quality of the wind observations. When corrected the Mode-S EHS wind data have similar observations-minus-background (o−b) statistics as Aircraft Meteorological Data Relay [AMDAR; using a Met Office version of the Unified Model over the United Kingdom (UKV)]. For the u wind component, the average per model run o−b root-mean-square value for the Mode-S EHS–derived data was 2.45 and 2.12 m s−1 for AMDAR. The AMDAR data are assimilated into the model.

Corresponding author address: Ed Stone, Met Office, FitzRoy Road, Exeter, Devon EX1 3PB, United Kingdom. E-mail: ed.stone@metoffice.gov.uk

1. Introduction

Aircraft-based observations (ABO) are often shown to be among the highest impact observations used in numerical weather prediction (NWP) models (Lorenc and Marriott 2014; Cardinali et al. 2003; Graham et al. 2000; Eyre and Reid 2014). At the Met Office, the most numerous ABO type of data are Aircraft Meteorological Data Relay (AMDAR), where meteorological information from commercial aircraft is automatically sent to national weather services using either satellites or ground stations. In the United Kingdom, AMDAR profiles (ascents or descents) are collected as part of the U.K. AMDAR and EUMETNET AMDAR (E-AMDAR) programs (EUMETNET 2015). The number of observations is limited, as only a subset of aircraft from some airlines carries the software necessary to report the meteorological data, and this limits the number of airports from which profiles are available. The number of airlines and observations is increasing with time but still only a small fraction of commercial aircraft report the data. Recently, there have been investigations into alternative methods of gathering meteorological data from aircraft that are now being referred to as aircraft-derived data (ADD) (de Haan 2011; de Haan and Stoffelen 2012; Strajnar 2012; Stone and Kitchen 2015; de Leege et al. 2013).

Mode-selective enhanced surveillance (Mode-S EHS) is a messaging regime used by aircraft and air traffic management (ATM) for the communication of an aircraft’s current situation (position, movement, intention, etc.); the standards are defined by the International Civil Aviation Organization in ICAO (2012). Mode-S EHS data can be used to derive both wind and temperature information at the aircraft’s location. The quality of the wind data has been shown to be comparable to that of AMDAR; temperature observations are of significantly lower quality (de Haan 2011). Further work by de Haan and Stoffelen (2012) has shown a positive impact on the High Resolution Limited Area Model (HIRLAM) NWP model. Mode-S EHS data require interrogation of aircraft by ATM secondary surveillance radar (SSR). Mode-S meteorological routine air reports (MRAR) are directly reported meteorological data from aircraft. The data have been shown to be of similar quality to AMDAR (Strajnar 2012). MRAR messages directly report the temperature and wind as available on the aircraft in response to interrogation by SSR. MRAR data have not been considered further in this paper, as they are not routinely requested in U.K. airspace and very few aircraft are capable of reporting the data (Strajnar 2012). Automated Dependent Surveillance Broadcast (ADS-B) extended squitter (ADS-B ES) data have also been used to derive layer temperatures from pressure and Global Navigation Satellite System (GNSS) altitudes, which may prove useful (Stone and Kitchen 2015; de Leege et al. 2013). The ADS-B ES messages are routinely broadcast from aircraft and are a subset of the broader Mode-S standard, although they do not require SSR to interrogate the aircraft (ICAO 2012).

In most of the previously published work, the data have been collected by ATM organizations and distributed to national weather services for processing, although de Haan et al. (2013) showed it was possible to collect the data using a commercial receiver. It is known that Mode-S EHS is currently being processed operationally at the Royal Netherlands Meteorological Institute (KNMI) using data provided by the European Organisation for the Safety of Air Navigation (EUROCONTROL) Maastricht Upper Area Control Centre1, which covers Belgium, Germany, Luxembourg, and the Netherlands. The Slovenian Environment Agency is operationally processing Mode-S EHS and Mode-S MRAR data provided by their national ATM. In the United Kingdom it is not feasible to collect the data from the national ATM, as they do not have the infrastructure required to collate and provide the data, and there are also a number of independent airports operating Mode-S SSR. The Met Office has therefore been trialing receiver equipment for Mode-S EHS and ADS-B ES data since mid-2011 with semicontinuous collection and processing for a slowly increasing number of sites since June 2013. This paper reports the progress that has been made at the Met Office toward deploying an operational network of Mode-S receivers in the United Kingdom and the processing of the received data. The data presented in this paper use the Mode-S EHS data, requiring an ATM SSR infrastructure to interrogate the aircraft and to initiate the data broadcast. The same network is used to collect only ADS-B ES messages, and the conclusions drawn below can be extrapolated to this data type. ADS-B ES messages are used to provide additional parameters and the locations for the Mode-S EHS–derived data.

2. Data derivation

The method described in this section to derive wind and temperature observations is the same as that introduced by de Haan (2011), although slightly reformatted. It is important to note that unlike most of the existing published work, aircraft locations are derived from broadcast messages rather than from the primary radar signature.

Two vectors are defined for an aircraft in flight—the ground movement vector [, where and are the components of the ground movement (knots; 1 kt = 0.51 m s−1) in the north and east directions, respectively] and the air movement vector [, where is the true airspeed (TAS) and is the aircraft’s heading relative to true north]. Note that θ can be converted to the true heading from the reported magnetic heading using the latest International Geomagnetic Reference Field (Finlay et al. 2010) or the World Magnetic Model (Maus et al. 2010). The difference between the two vectors is due to the wind acting on the aircraft such that
e1
where is the wind vector. All of the parameters used are nominally instantaneous values as reported by the aircraft; there is likely to be some unknown and variable averaging and quality control (QC) process within the avionics, sensors, and software used by the world’s airlines and aircraft manufacturers.
The Mach number M is measured by the pitot tube and is used in the derivation of the true airspeed with temperature; the Mach number and true airspeed are broadcast from the aircraft, allowing the temperature to be calculated using
e2
where T is the temperature (K), is the true airspeed (kt), and 38.975 kt K−1/2 is a conversion factor incorporating the ratio of specific heats and converting from knots to meters per second (Stone and Kitchen 2015).

The temperature and wind observations (referred to as Mode-S EHS–derived data) is a near-instantaneous measurement at the location of the aircraft. The error in the Mode-S EHS–derived temperature (T) is estimated to be ±3 − 5 K (these estimates have been done using comparisons to AMDAR, radiosondes, NWP, and other in situ instruments) (de Haan 2011; A. K. Mirza et al. 2016, unpublished manuscript).

ICAO (2012) provides a full list of the messages and their contents, and they can therefore be used to identify the messages required for the generation of meteorological data. There are four messages required to derive a wind and temperature observation. Two messages are downlink format 17 (DF-17)-type messages; these are continuously broadcast by capable aircraft (fitted with Mode-S/ADS-B transponders) and are often referred to as ADS-B ES messages. These two messages provide information on the location and ground movement vector of the aircraft. The other two messages are subtypes (Comm-B data selector) of either DF-21 or DF-22 messages (DF-2x) and must be requested by an SSR. The DF-2X message format has a 56-bit data section that can carry any one of many predefined subsets of aircraft data; these are known as binary data store (BDS). The two BDS datasets that are required are BDS 5.0 and BDS 6.0, both of which are requested as part of the Mode-S EHS standard defined by EUROCONTROL and can be transmitted in the data section of either the DF-21 or DF-22. These messages provide information on the air movement vector of the aircraft and additional maneuvering parameters, which are useful for quality control of the data. The collection of DF-17 messages is common by aircraft enthusiasts and there are several websites devoted to the collection and mapping of these messages, allowing commercial aircraft to be tracked across large areas of the world. The DF-2x messages are part of the same standard as DF-17 and can be collected using the same equipment; DF-2x messages do not explicitly contain their subtype; therefore, it must be inferred from the format of the bits in the data. A summary of what the different messages are used for can be found in Table 1.

Table 1.

A brief summary of the message types required to calculate a wind and temperature observation.

Table 1.

3. Met Office network and processing

The Met Office Mode-S network currently comprises five receivers, four of which are collocated with operational weather radars. Mode-S/ADS-B operates at 1090 MHz (≈1 GHz), which requires line of sight for communication. Radar sites offer excellent horizons and existing communication infrastructure. A Met Office receiver comprises a commercial USB Mode-S receiver and decoder2, utilizing a field programmable gate array processor to allow the processing of a 1-MB s−1 datastream (Köllner 2013). The decoder is connected to a standard PC running bespoke software logging all DF-17 and DF-2x messages. A vertical high-gain (≈7.5 dBd) collinear antenna with a center frequency tuned to 1090 MHz combined with three 17-cm ground plane radials is attached to the receiver via military specification low-loss coaxial cable. The antenna is mounted as high as possible at the sites. Because of collocation with radar, the antenna has to be mounted below the “ring beam” of the radar installation so as to avoid interference and microwave energy saturation. To further increase the receiving capability, a 12-dB high-gain low-noise preamplifier specifically tuned to 1090 MHz is fitted below the antenna. All sites have the same equipment installed, although the height and position of the antenna can vary slightly due to structure, local planning, and physical building constraints. Further details of each site are given in Table 2. The first site [Exeter (EXT)] has been running since July 2013, with some short periods of downtime. Thurnham (THU) has been running since April 2014. The remaining sites came online in March 2015. As the network has been slowly deployed, we have paused to analyze some aspects of the data before deciding to proceed. This is reflected in the choice of statistics shown and the dates chosen.

Table 2.

A description of the receiver sites comprising the Met Office network. The number of observations per day is an approximate estimate given the performance of each receiver if it were operating on its own.

Table 2.

The U.K. network is currently operating with a timeliness of less than 20 min for more than 90% of observations. This is mainly due to data packaging and could be improved if required, providing a suitable source for nowcasting models.

Quality control parameters

The raw Mode-S messages are packaged and sent to a central processing site in 10-min intervals. The central processing system evaluates the data; when all four of the required messages have been received within 10 s of each other (two broadcast DF-17 and two requested DF-2x messages), using time stamps assigned by the receiver, wind and temperature data are generated. SSR have a rotation time of between 5 and 8 s; assuming a single message type can be requested per rotation, and two requested messages are required, it seems reasonable to have a time limit of 10 s for all messages to be received. No smoothing of the aircraft data is performed at any stage in the Met Office processing. Care is taken in identifying the DF-2x subtype, as it is possible for the misidentification of messages. A comparison of the bits of the message to the known prescribed format is not always sufficient, and the decoded data must be compared to a table of known failure conditions. These failure conditions are compiled over time when gross errors are identified and investigated.

As part of the wind processing, some quality control checks are performed, which are summarized in Table 3. These conditions have been chosen such that maneuvering aircraft and incorrectly identified/partially corrupted messages can be removed from the processing. All of the checks act as confirmation of a correctly decoded parameter set with those related to roll angle and heading also acting to remove maneuvering aircraft [these are identified with an asterisk (*) in the table]. When aircraft maneuver the airflow over the sensors changes, which reduces the accuracy of the observations. Where the quality checks are not passed, the data are discarded. Before assimilation into any NWP model, it is anticipated that similar checks to those used on AMDAR will be applied. Future work may provide some refinement of these QC parameters.

Table 3.

The QC parameters used by the Met Office Mode-S EHS–derived wind and temperature derivation system. The asterisk (*) indicates the conditions used to remove maneuvering aircraft. There are additional sanity checks of the binary format of DF-2x messages to confirm their subtype which are not listed.

Table 3.

4. Data quantities and coverage

In this section data from May 2015 has been used. This period has been chosen, as all five of the operational receivers were operating continuously. The number of observations has a seasonal dependence with more flights during the summer school holidays (July–August in the United Kingdom).

The locations of observations are limited by three factors: the flight paths of aircraft, the line of sight from aircraft to the receiver, and the line of sight from aircraft to ATM SSR. It is therefore not a trivial task to predict the likely impact of a single receiver, or even a network. Figure 1a is a map of the United Kingdom with 0.025° longitude × latitude squares showing where Mode-S EHS data have been received. The color of the square indicates the lowest observation GNSS altitude within that box between 21 and 31 May 2015, inclusive. This processing is relatively computationally intensive, analyzing periods greater than 10 days is not currently feasible with the available resources. The bright red boxes indicate an altitude between 0 and 100 m, suggesting that there are observations very close to or at the surface. The major U.K. airports are clearly visible with low altitudes surrounding the London airports of Heathrow, City, Gatwick, Luton, Southend, and Stanstead. There are also low altitudes at Exeter Airport, Manchester Airport, and Belfast International Airport. Further from the receivers, there are areas of airspace where aircraft are descending/ascending into/out of other airports can be seen. This is most noticeable around the midlands airports of Birmingham and East Midlands, Bristol airport, and Bournemouth/Southampton; England; Dublin, Ireland; and Scotland. These areas will be targeted for future receiver deployments. The apparent cutoff of data on the northern edges of the data that are apparent in Fig. 1a is due to the limitations of the location processing method. This becomes more apparent the farther north the receiver is located. The cutoff of data should never appear within 250 km of a receiver.

Fig. 1.
Fig. 1.

Two maps of the United Kingdom. (a) Map has 0.025° longitude × latitude squares, where the colors indicate the lowest observation GNSS altitude within that box between 21 and 31 May 2015, inclusive. The bright red boxes indicate an altitude between 0 and 100 m, suggesting that there are observations very close to the surface. (b) Map has nonopaque circles for each observation between 0600 and 0700 28 May 2015. The colors represent which receiver was used; during this time 310 000 observations were derived.

Citation: Journal of Atmospheric and Oceanic Technology 33, 4; 10.1175/JTECH-D-15-0184.1

Figure 1b shows the locations of all observations during one hour (0600–0700 UTC 28 May 2015) with the colors indicating which receiver gathered the data. Figure 1b shows some areas with gaps that do not appear in Fig. 1a. This implies that the distribution of observations for short time periods is not consistent. While it may be tempting to target all regional airports with receivers, it is important to understand the technical abilities of the aircraft that use them. This is discussed further below. Southend airport is a similar size to the Exeter airport in terms of passenger numbers but provides significantly more profile data. Each receiver also does not perform as may be expected. EXT yields very little data apart from being close to Exeter airport. This is due to its position and surroundings. EXT’s antenna is mounted on the side of a metal mast (approximately 5 cm from the mast), surrounded on two sides by trees with a radar dome at approximately the same level, further restricting the view of the sky. THU provides data for a smaller geographical area to the north than one may expect. This is due to duplicate removal between different receivers; THU produces larger files, which take slightly longer to arrive at the processing server and are therefore processed last. This can be mitigated by transferring files at a higher frequency.

Figure 2a shows the number of observations per day during May 2015; the number of observations fluctuates between 4.0 and 5.7 million. The slowly increasing trend in peak number of observations is probably due to the increase in flights as airlines introduce summer schedules. It is interesting to note that midweek appears to have fewer observations than the average, although there is very little consistency in the number of observations. On 22 May 2015, when the availability of Mode-S EHS–derived winds was highest during this period, the E-AMDAR program received 53 057 observations in total from all participating aircraft (KNMI 2014). There were more than 5.7 million Mode-S EHS–derived wind observations, more than two orders of magnitude higher for a significantly smaller geographical area. This density of ABO is unprecedented over the United Kingdom.

Fig. 2.
Fig. 2.

(a) Number of observations derived per day from aircraft Mode-S messages received using the five Met Office receivers. (b) Average number of observations per hour derived from aircraft Mode-S messages received using the five Met Office receivers using the same data as in (a).

Citation: Journal of Atmospheric and Oceanic Technology 33, 4; 10.1175/JTECH-D-15-0184.1

The average number of observations per hour during a one-month period (May 2015) is shown in Fig. 2b. Overnight the number of observations drops to around 50 000 per hour; during these very low-data-quantity hours, the amount of AMDAR data normally drops to nearly zero observations per hour. It is important to note that the Mode-S overnight data are not all high-altitude overflight data but typically includes 5–10 profiles per hour during the quietest hours into airports within range of the network. Mode-S clearly has great potential for providing data during these data-sparse periods, complementing existing ABO. The data time profile is dependent on the nature of the local airports and will vary greatly depending on region and country. For example, some airports in the United Kingdom are allowed only to operate between set times; therefore, there are no profiles to these airports outside of their operating hours.

Mode-S EHS data are received from aircraft using Exeter airport. Currently there are no AMDAR data reported from this regional international airport. Although, given the number of commercial airline movements, the quantity of data is not as high as may be expected with fewer than 10 profiles per day compared to approximately 50 scheduled flights per day. This is due to most of the aircraft using the airport not generating wind and temperature observations. For one airline this is simply due to their aircraft not being equipped with Mode-S EHS–capable transponders. For regional airports, such as Exeter, where a single airline makes up the majority of commercial movements, such details can have a significant impact on the quantity of available data. A second airline’s aircraft presented a different problem, as they do carry the required transponders. Of the four recorded messages, both of the broadcast ADS-B DF17 messages were received and recorded for several of their aircraft but one of the two required DF-2x messages was missing from them. They are requested by ATM but there are missing parameters within the messages. They therefore fail the QC/message-type tests but are still conforming to the standard.

5. Receiver network design

Based on the data received and processed from the five operating receiver sites (before duplicate removal), it was decided that a range of 250 km from a well-placed antenna for high altitude (above 8500 m) is appropriate. This is consistent with the ATM literature on Mode-S and ADS-B and slightly below the “horizon” for that altitude. The smaller dashed circles in Fig. 1a show this range. Circles of 250-km radius have been placed over the United Kingdom to show that the entire United Kingdom could be covered by a network of only four receivers if optimally situated. This is not practical due to the availability of sites, topography, and limitations of local installations (antenna may not have a clear view of the entire sky). Weather radar sites are secure, with good communications and power facilities and very good horizons. Using the locations of the existing radar sites, a network of five or six receivers would be required to cover all of the airspace over the United Kingdom, which has a land area of 243 610 km2. Six receivers have a maximum coverage (assuming a nominal range of 250 km) of 1.2 Mm2. Because of the shape of the United Kingdom, a significant overlap of data and large areas of sea would be covered. With the current network, a single receiver in Scotland would complete the U.K. upper-air network using six receivers, with EXT being largely redundant for high-altitude observations. Land area is not the best metric for network design but without knowing the number of Mode-S/ADS-B-equipped aircraft movements in the United Kingdom, it is a convenient simplification.

In the AMDAR community, an emphasis is placed on observations from aircraft profiles. In discussions the authors believe that this is likely due to the importance of profiles for forecasters to ensure the model output is consistent and to meet the requirements of convective-scale NWP (Cress and Wergen 2001; Petersen 2016). It is therefore necessary to consider the locations of major airports when designing a network. A basic qualitative line-of-sight analysis has been conducted between each of the Met Office radar sites in the south of the United Kingdom and the nearby commercial airports. This analysis uses freely available terrain data for a straight line between the radar and the center of the runway along with calculated distance to the horizon based on the likely antenna altitude and runway altitude. This analysis has shown that seven Mode-S receivers at operational weather radar sites could provide most of the profile data from the major airports in the south of the United Kingdom, and a significant amount of cruise-level data over the whole United Kingdom, with intermediate altitude data from farther airports. This work is not precise due to local differences at each site. It is impossible to precisely predict minimum altitudes from any given airport; therefore, it is recommended to conduct local testing. Two of the current five sites are not ideal locations but were the only pragmatic options at the time of installation.

Because of local conditions at the receiver sites, airports, and ATM SSR sites, it is sensible to take a pragmatic approach to network design using the existing data to identify the best locations for future receivers. Figure 1a identifies several areas where it is apparent that aircraft are descending into airfields out of range of the receivers, allowing for the next phase of receiver locations to be linked to these data. The Midlands, Bristol, and Portsmouth areas have all been identified as good locations for additional receivers. After the next phase of deployment, a similar map to that presented in this paper will be generated, allowing the identification of holes in the low-altitude data around commercial airports. It is anticipated that the final U.K. network will consist of fewer than 20 receivers, mostly at the radar sites, which will provide a near-complete observations network. Optimizing any network for profile data will also provide significant en route data, as each receiver has a range of 250 km; aiming for profiles from regional airports at a smaller separation than that will provide complete coverage.

6. Data quality

The data presented here have been selected at various times from the currently operating network of receivers; therefore, the number of receivers and observations used varies. This represents an approximate road map of the decisions on proceeding with the deployment of the network, ensuring that we had the required computational power and storage for the data processing.

Data from EXT for 7 days from 9 August 2013 have been processed using the Met Office observations processing system (OPS). A week is a very short period, but the aim of this was to show similar data quality (and improvements after heading correction) to that reported previously in de Haan (2011) to decide on further deployment of receivers (from the one operating at this time) and what processing may be required. The OPS allows for a comparison between observations and the Met Office operational NWP models. For each observation a bilinear interpolation of the nearest background field data values from the Unified Model UKV high-resolution (1.5-km horizontal resolution with 70 vertical levels) model creates a value at the location (in time and space) of the observation. The difference between the observation and the background value is then found (o−b value). The model fields used were T + 2-, 3-, 4-h forecasts retrieved from the operational archive. The wind vector was split into u and υ components prior to processing. The o−b values for every observation have been found and then placed into 250-m high-altitude bins. The average (mean, black lines) and root-mean-square (RMS, gray lines) o−b values for all of the observations in each bin were then calculated and plotted in Fig. 3 as solid lines. The number of observations is significantly higher around altitudes of 10 000 and 12 000 m, the typical cruise altitude for long-haul flights, due to most observations being from aircraft transiting between North America and either London or continental Europe. Aircraft flying these routes tend to maintain a constant altitude within an air corridor. The o−b values are similar to those as would be expected from the de Haan (2011) results, as neither smoothing nor correction has been applied to these data. The temperatures show a variable mean bias in the range from −3 to 2 K. The RMS o−b values are large at low altitudes, with values up to 12 K; these values are similar to those produced by de Haan (2010) and A. K. Mirza et al. (2016, unpublished manuscript). Error analysis suggests these higher values arise due to lower aircraft speeds at low altitudes (A. K. Mirza et al. 2016, unpublished manuscript). The RMS and mean o−b values for the wind components show better agreement for the u component with RMS o−b values between 0.5 and 5 m s−1 compared to between 0.75 and 10 m s−1 for the υ component. This is explained by the majority of flights traveling east–west or west–east; the wind component orthogonal to the direction of air movement vector will show a larger uncertainty if a heading correction is required. Applying a correction method affects the υ component more than the u component, as shown in Fig. 3 and discussed below.

Fig. 3.
Fig. 3.

The o−b mean values (gray lines) and RMS (black lines) for 7 days of data from 9 to 15 Aug 2013 in 250-m altitude bins. The solid lines show the uncorrected data; the dashed lines show the uncorrected data after a heading correction has been applied. The 7 days of 1–7 Aug 2013 data have been used to create the corrections. (a) The u wind component, (b) the υ wind component, (c) the number of observations per 250-m high altitude bin, and (d) the temperature.

Citation: Journal of Atmospheric and Oceanic Technology 33, 4; 10.1175/JTECH-D-15-0184.1

Background wind values were found from the o−b values from the OPS for the week of 1–7 August 2013, inclusive. These dates were chosen as they were close to the analysis week (7 days from 9 August 2013), but distinct. Assuming that the wind vector, ground movement vector, and TAS are correct, the heading was found for each observation. The difference between the two headings (observed and calculated) was calculated for each observation. The heading differences were then averaged for each aircraft, with the average defined as the heading correction. Figure 4 shows the distribution of heading corrections with a bin width of 0.1°. As can be seen, the corrections are not distributed around zero and there are three peaks. The reasons for this distribution are currently unclear, although they may be related to the age of the magnetic heading lookup table being used (and indirectly aircraft type). de Haan (2013) has previously reported a similar NWP-based correction method, where similar distributions of heading corrections are also identified. This simple approach may not provide a complete solution, as the heading correction is expected to be a function of heading, and potentially location and altitude (Jacobs et al. 2014; Collinson 2011). The number of observations/aircraft averaged to estimate the correction varies greatly from 20 to over 15 000 individual observations with most having just a few hundred observations. This implies that the corrections for most aircraft are based on only a single observed flight. Single flights are not ideal, as currently unmodeled local variabilities in the atmosphere may be significant. It is also likely that the individual o−b values will increase with time between hourly background files (increasing the RMS o−b not the mean o−b). It is difficult to distinguish this effect in the data due to variability in the atmosphere.

Fig. 4.
Fig. 4.

A histogram of the heading corrections generated from Mode-S EHS–derived data generated between 1 and 7 Aug 2013, and applied to the data received for 9–15 Aug 2013.

Citation: Journal of Atmospheric and Oceanic Technology 33, 4; 10.1175/JTECH-D-15-0184.1

The corrections were applied to all data in the 7 days beginning 9 August 2013. After the heading corrections have been applied, winds and o−b values were recalculated. Again, the o−b values have been placed into altitude bins with a height of 250 m, and the mean and RMS of the o−b values calculated for each bin. Figure 3 presents the recalculated data as dashed lines. There is a reduction in both the mean and RMS o−b values found for the recalculated data. Below 12 000 m for both of the u and υ wind components, the RMS o−b values after correction are below 5 m s−1, and are mostly around 2.5 m s−1. The reduction is not symmetric though; the υ component is improved to a greater extent than the u component as to be expected given the predominant flying directions of the aircraft. It should be noted that these errors also include model errors that cannot be distinguished from the observational errors.

To facilitate observation monitoring and to provide a steer on data assimilation, a near-real-time calculation of o−b statistics is generated. This uses a subtly different method compared to the one used above to replicate the o−b errors that would be generated for an operational observation source before data assimilation. Figure 5 shows 28 days of o−b means and RMS values calculated for each model run; the solid black circles are Mode-S EHS–derived data, and the solid gray triangles are AMDAR. The statistics are generated using the UKV model T = 0 background fields generated 8 times per day. For each observation the nearest model background field is selected (within 1.5 h of the observation). A linear interpolation is then made to the location of the observation to generate the background value to subtract from the observation. The o−b values are processed in such a way as to match as closely as possible the statistics that are produced from data sources that are assimilated into the model, where the o−b values are calculated in the same way. This work was undertaken when two receivers were online (EXT and THU). The date is chosen at the beginning of the operational archiving and real-time monitoring, as this is the only period when heading corrections were turned off on the operational archive. From this it was decided to recalculate and review heading corrections every 2 weeks using the previous 6 weeks’ worth of o−b data to generate the new heading corrections. The number of AMDAR observations per model run ranged between 78 and 1570 observations; they are therefore shown separately in Fig. 5d. For each observation a heading difference is calculated. When more than 3000 observations are recorded over at least a 2-week period, the individual differences are averaged to find a heading correction per aircraft. These heading corrections were introduced on 28 August 2014. Before heading corrections are turned on, the o−b Mode-S EHS–derived data show significantly more scatter in the mean values and larger RMS values compared to those for AMDAR. Subsequently, the data show a significant improvement with statistics close to those of AMDAR. Table 4 shows the average means and RMS values for the u and υ wind components for AMDAR and Mode-S. The AMDAR wind data are of slightly better quality than the Mode-S EHS–derived data, although further refinement to the heading corrections may reduce the Mode-S errors. For the temperatures the Mode-S EHS–derived data are of a lower quality; although, because of the high number of aircraft in some airspace and the frequency of reporting, an improved observation may be found by averaging. Please note that the AMDAR observations are assimilated into the model, whereas the Mode-S EHS–derived data are not. The improvement in o−b RMS in the u component is also represented in the υ component of the wind; the figures for this are not included.

Fig. 5.
Fig. 5.

The o−b statistics for Mode-S EHS–derived (solid black circles) and AMDAR (solid gray triangles) u components using the UKV model. (a) The mean o−b for each model run, (b) the RMS o−b for each model run, (c) the number of Mode-S o−b values, and (d) the number of AMDAR o−b values. Heading corrections were introduced on 28 Aug 2014.

Citation: Journal of Atmospheric and Oceanic Technology 33, 4; 10.1175/JTECH-D-15-0184.1

Table 4.

Averaged o−b statistics for Mode-S and AMDAR. Both the averaged o−b means and the averaged RMS values are shown. The averaged o−b mean is the bias with respect to the NWP model.

Table 4.

All five receivers have been online only since March 2015; therefore, there are not enough data to report on the seasonal variability of the o−b statistics. The real-time monitoring is ongoing, and we hope to have the data to perform this analysis in the future.

de Haan (2011, 2013) also presented an alternative method to calculate aircraft heading corrections, where alignment to a runway shortly after landing is used to find the aircraft heading correction. de Haan (2013) showed that these methods provide corrections that were consistent with each other, although the runway method allows only for aircraft that have used specific airports a large number of times, limiting its potential use. It should also be noted that heading corrections will change with time due to aircraft maintenance schedule. It is therefore suggested that any operational system be regularly updated. It is known that some aircraft manufactures require a physical change to the avionics to update the tables. It is therefore likely that the tables will remain static for the lifetime of the aircraft. It is intended that when the Met Office data are operationally assimilated, these processes will be automated with updated corrections being produced at least monthly using a rolling method, taking into account only the most recent data. In monitoring heading corrections over several months, no significant shifts have yet been identified, although de Haan (2013) does report on some significant shifts.

Gross errors

There are some cases of aircraft either consistently or intermittently reporting winds or temperatures that are obviously gross errors. These errors represent less than 0.1% of the aircraft data. In some cases the errors appear to be due to incorrect message identification. Such errors occur not only in the parameters used for generating the meteorological data but also in the aircraft position parameters, where some aircraft appear to be landing with an offset to the runway. There may be some potential in both correcting for the positional offsets and reporting back to the airlines.

7. Conclusions

This paper has presented the initial findings of the first phase of the Met Office Mode-S receiver network in the United Kingdom. This method has been shown to be a feasible alternative to receiving the data directly from national ATMs. The prospective extended network of fewer than 20 Mode-S receivers (15 operational weather radar sites, two research and development radar sites, and other sites near major airports) covering all commercial aircraft movements over the United Kingdom would be low cost to implement and provide of the order 7 million observations (a conservative estimate) of winds and temperatures per day. Where available, weather radar sites should be the first-choice locations for Mode-S receivers due to their excellent horizons and existing infrastructure. Other sites where they are near major airports should also be considered. The current Mode-S network is already delivering orders of magnitude more observations than E-AMDAR for a significantly lower cost. AMDAR will continue to have value, as the data are of higher quality and it has the potential in the future to provide humidity data. While it would be feasible to cover the majority of the airspace using fewer receivers (6), it is perceived that there is significant value in the profile data that can be obtained by a denser network. It has been shown that the wind observations are of similar quality to AMDAR observations and are therefore of adequate quality for NWP and aviation forecasting. The quality of the wind observations can be improved significantly through the application of a simple heading correction regime. Work on the correction regime of headings and airspeeds is currently ongoing within the Met Office, with both the underlying science and the operational correction methods being investigated. It is the view of the authors that this observation system shows significant promise for both operational meteorology and for providing observations for the ever-increasing resolution of numerical weather prediction models.

Acknowledgments

The authors thank Andrew Mirza for the useful discussions throughout this project, Adam Maycock for developing the OPS tools used for calculating o−b values, and members of the Met Office IT teams for the continued support in moving and storing the data. We also thank Steve Addy from Flarehill Computers Limited for his help in the early development of this technique.

REFERENCES

  • Cardinali, C., , Isaksen L. , , and Andersson E. , 2003: Use and impact of automated aircraft data in a global 4DVAR data assimilation system. Mon. Wea. Rev., 131, 18651877, doi:10.1175//2569.1.

    • Search Google Scholar
    • Export Citation
  • Collinson, R. P. G., 2011: Introduction to Avionics Systems. 3rd ed. Springer, 530 pp., doi:10.1007/978-94-007-0708-5.

  • Cress, A., , and Wergen W. , 2001: Impact of profile observations on the German Weather Service’s NWP system. Meteor. Z., 10, 91101, doi:10.1127/0941-2948/2001/0010-0091.

    • Search Google Scholar
    • Export Citation
  • de Haan, S., 2010: Quality assessment of high resolution wind and temperature observations from ModeS. Rev. ed. KNMI Scientific Rep. WR 2009-07, 31 pp.

  • de Haan, S., 2011: High-resolution wind and temperature observations from aircraft tracked by Mode-S air traffic control radar. J. Geophys. Res., 116, D10111, doi:10.1029/2010JD015264.

    • Search Google Scholar
    • Export Citation
  • de Haan, S., 2013: An improved correction method for high quality wind and temperature observations derived from Mode-S EHS. KNMI Tech. Rep. TR-338, 43 pp.

  • de Haan, S., , and Stoffelen A. , 2012: Assimilation of high-resolution Mode-S wind and temperature observations in a regional NWP model for nowcasting applications. Wea. Forecasting, 27, 918937, doi:10.1175/WAF-D-11-00088.1.

    • Search Google Scholar
    • Export Citation
  • de Haan, S., , de Haij M. , , and Sondij J. , 2013: The use of a commercial ADS-B receiver to derive upper air wind and temperature observations from Mode-S EHS information in The Netherlands. KNMI Tech. Rep. TR-336, 44 pp.

  • de Leege, A., , van Paassen M. , , and Mulder M. , 2013: Using automatic dependent surveillance-broadcast for meteorological monitoring. J. Aircr., 50, 249261, doi:10.2514/1.C031901.

    • Search Google Scholar
    • Export Citation
  • EUMETNET, 2015: E-AMDAR. Accessed 7 July 2015. [Available online at http://www.eumetnet.eu/e-amdar.]

  • Eyre, J., , and Reid R. , 2014: Cost-benefit studies for observing systems. Met Office Forecasting Research Tech. Rep. 593, 11 pp.

  • Finlay, C. C., and et al. , 2010: International geomagnetic reference field: The eleventh generation. Geophys. J. Int., 183, 12161230, doi:10.1111/j.1365-246X.2010.04804.x.

    • Search Google Scholar
    • Export Citation
  • Graham, R. J., , Anderson S. R. , , and Bader M. J. , 2000: The relative utility of current observation systems to global-scale NWP forecasts. Quart. J. Roy. Meteor. Soc., 126, 24352460, doi:10.1002/qj.49712656805.

    • Search Google Scholar
    • Export Citation
  • ICAO, 2012: Technical provisions for Mode S services and extended squitter. 2nd ed. International Civil Aviation Organization Doc. 9871, AN/460, 326 pp.

  • Jacobs, N. A., , Mulally D. J. , , and Anderson A. K. , 2014: Correction of flux valve–based heading for improvement of aircraft wind observations. J. Atmos. Oceanic Technol., 31, 17331747, doi:10.1175/JTECH-D-13-00175.1.

    • Search Google Scholar
    • Export Citation
  • KNMI, 2014: QEvC info. Accessed 10 June 2015. [Available online at http://www.knmi.nl/samenw/geoss/eumetnet/E-Amdar/QEvC/.]

  • Köllner, G., 2013: Mode-S Beast. Accessed 8 July 2014. [Available online at http://www.modesbeast.com/.]

  • Lorenc, A. C., , and Marriott R. T. , 2014: Forecast sensitivity to observations in the Met Office global numerical weather prediction system. Quart. J. Roy. Meteor. Soc., 140, 209224, doi:10.1002/qj.2122.

    • Search Google Scholar
    • Export Citation
  • Maus, S., , Macmillan S. , , McLean S. , , Hamilton B. , , Thomson A. , , Nair M. , , and Rollins C. , 2010: The US/UK World Magnetic Model for 2010-2015. British Geological Survey, NOAA Tech. Rep. NESDIS/NGDC, 98 pp.

  • Petersen, R. A., 2016: On the impact and benefits of AMDAR observations in operational forecasting. Part I: A review of the impact of automated aircraft wind and temperature reports. Bull. Amer. Meteor. Soc., doi:10.1175/BAMS-D-14-00055.1, in press.

  • Stone, E. K., , and Kitchen M. , 2015: Introducing an approach for extracting temperature from aircraft GNSS and pressure altitude reports in ADS-B messages. J. Atmos. Oceanic Technol., 32, 736743, doi:10.1175/JTECH-D-14-00192.1.

    • Search Google Scholar
    • Export Citation
  • Strajnar, B., 2012: Validation of Mode-S Meteorological Routine Air Report aircraft observations. J. Geophys. Res., 117, D23110, doi:10.1029/2012JD018315.

    • Search Google Scholar
    • Export Citation
2

Mode-S Beast, firmware version 1.40, Günter Köllner Embedded Development GmbH.

Save