AMOFSG/9-IP/8 20/9/11

AERODROME METEOROLOGICAL OBSERVATION AND FORECAST STUDY GROUP (AMOFSG)

NINTH MEETING

Montréal, 26 to 30 September 2011

Agenda Item 5: Observing and forecasting at the aerodrome and in the terminal area 5.1: Observations

AUTO METAR SYSTEM AT CIVIL AIRPORTS IN THE NETHERLANDS: DESCRIPTION AND EXPERIENCES

(Presented by Jan Sondij)

SUMMARY

This paper provides an overview of the AUTO METAR system used in the Netherlands. The system includes the entire technical infrastructure used for the automated generation of all meteorological aeronautical observation reports including baack-up systems and procedures. It also includes the supervision of all issued reports by a remote meteorologist who can provide additional information to ATC. Experiences of the performance and acceptance by ATC of the AUTO METAR system are reported. The process of how this was achieved is presented as well as lessons learned.

1. INTRODUCTION

1.1 The eighth meeting of the Aerodrome Meteorological Observation and Forecast Study Group (AMOFSG/8) led to the formation of an ad-hoc group tasked with reviewing the options for the future reporting of present weather in fully automated weather reports. More background on automated weather observations is provided in this information paper.

1.2 This information paper describes the so-called AUTO METAR system at civil airports in the Netherlands. The process of introduction and approval as well as experiences are presented. Some highlights are reported below. Details are given in the appended document entitled AUTO METAR System at Civil Airports in the Netherlands: Description and Experiences by Wauben and Sondij.

(49 pages) AMOFSG.9.IP.008.5.en.docx AMOFSG/9-IP/8 - 2 -

1.3 Since 15 March 2011, the AUTO METAR system has been operational 24/7 at Rotterdam-The Hague Airport (EHRD). This was the final step of the introduction of the AUTO METAR system at civil airports, military airbases and off-shore structures in the Netherlands, although further improvements to the AUTO METAR system itself are planned for the near future. Professional meteorological observers are currently employed by KNMI only for carrying out aeronautical observations at Amsterdam Airport Schiphol (EHAM). The added value of the observer with respect to the capacity of the Mainport Schiphol makes their presence undisputable for the moment.

1.4 The rationale for introducing the AUTO METAR system is cost savings as local MET offices and local observing staff at airports are no longer required. The current state-of-the-art of observation techniques and technology is such that it is possible to provide an automated observation of good quality if specific measures are taken into account. At the same time, the AUTO METAR system facilitates the possibility to acquire meteorological information from airbases that are closed and unmanned during weekends and from off-shore structures on the North Sea so that a denser, both temporal and spatial, network of aeronautical meteorological observations becomes available to users.

2. DESCRIPTION OF THE AUTO METAR SYSTEM

2.1 The term “AUTO METAR system” is used to denote the entire system used for the automated production of all meteorological aeronautical reports, of which the AUTO METAR is one. The system does not only designate the sensors, the associated technical infrastructure for data acquisition, data processing and data dissemination in suitable formats, but also includes: back-up sensors, systems and procedures; remote monitoring by meteorologists and service staff; communication with users covering daily briefings, intermediate updates and handling of sensor or system maintenance or malfunction; and provisions for local points of contact for the verification of the local meteorological situation. The AUTO METAR system also includes the documentation, the procedures and the service level agreement with ATC.

2.2 It is important to note that the content of the observation reports itself can differ between airports or between states, e. g. an AUTO METAR that contains pressure, wind direction and wind speed, air temperature and dew point only versus an AUTO METAR that contains visibility, clouds, weather and TREND as well. The AUTO METAR system used at civil airports in the Netherlands always contains the full set of parameters although not all weather phenomena and descriptors are included due to limitations of automated observations.

2.3 There are different “flavours” of the AUTO METAR system used in the Netherlands. At off-shore structures, the system has no redundancy and generates only the AUTO METAR every half- hour which is disseminated without human interaction. At civil airports, the AUTO METAR system includes back-up sensors, system redundancy and back-up procedures; AUTO METAR reports (as well as auto local routine and auto special reports which are generated); and all reports are disseminated after verification and complementation remotely by a meteorologist.

2.4 Although the individual components of the AUTO METAR system have proven reliable, redundancy has been taken into account in the set-up used at civil airports and airbases. This includes measures ranging from back-up sensors using independent components of the observation infrastructure to redundant server systems. The set-up used at Rotterdam-The Hague Airport only has single points of failure for visibility and runway visible range, which needs to be assessed in the touchdown zone of the runway in use, and clouds due to the availability of a single sensor for these parameters. In case visibility information at touchdown is not available, it may be possible to approach the runway from the other side which has its own visibility sensor for instrument precision approach and landing operations. - 3 - AMOFSG/9-IP/8

Alternatively, the remote aviation meteorologist, using available information including images of video cameras and or consulting of ATC staff, can advise whether visual flight rules conditions are applicable or not.

2.5 The AUTO METAR system produces meteorological observation reports that meet ICAO requirements and includes coding issues such as: UP (unknown precipitation); TS (lightning); NCD (no clouds detected); and CB/TCU (convective clouds). During the evaluation of the AUTO METAR system various complaints of users were related to the lack of representativeness of automated visibility and cloud observations. The situation experienced by local users conflicted with the observations reported by the AUTO METAR system or the latter showed a large delay. It should be noted that these user complaints were partly related to the unfamiliarity of ATC staff with the details of the measurement systems and internationally agreed observation principles. The measurements principles and data processing algorithms have been documented and the characteristics of the automated results have been provided to users. In some cases, the situation could be improved; for example, by using a so-called marked discontinuity criteria in the averaging of visibility and also the criteria for issuing an auto local special report, such as the delay after reporting an improvement of the situation, have been redefined by mutual agreement.

2.6 The handling and reporting of missing, incomplete or incorrect sensor information has been facilitated and agreed with the users. Missing or incomplete information is indicated in the meteorological reports either by an invalid entry in the corresponding group or by adding a suitable remark. Incorrect or missing sensor information can be overruled orally by the aviation meteorologist, which is logged in the shift reports and voice recorded by ATC and KNMI.

3. EXPERIENCE WITH THE AUTO METAR SYSTEM

3.1 The time between the introduction of the AUTO METAR system and the acceptance by ATC has been used to acquire experience with the AUTO METAR system, its performance and characteristics. In this period changes have been made to the AUTO METAR system, including the documentation, communication and procedures used. Several assessments have been performed, both by KNMI and ATC, which provided useful information on pending issues. Sometimes it turned out that the technical items identified by KNMI were not essential issues for users; instead, users had a need for additional information or service from KNMI. The assessments also gave recommendations that have either been solved or are under investigation with the parties involved.

3.2 Several technical improvements are currently under investigation. However, the usage and usefulness of all the elements contained in the (auto) local routine and (auto) special reports are also being investigated. Issues are, for example: which elements are actually used; whether parameters other than wind and visibility should be runway dependent; which elements should differ for the separate arrival and departure reports that are issued at Amsterdam Airport Schiphol; and whether it is necessary to include a TREND in the auto local routine and auto local special reports.

3.3 The introduction and acceptance of the AUTO METAR system in the Netherlands was a difficult process. Several factors influenced this process, such as the emotions of staff involved; the perception that the system is operated without technical supervision and without monitoring by a meteorologist at a remote location with the possibility to provide additional information; unfamiliarity with the added value of a local observer and characteristics of automated weather observation products; and lack of experience with similar systems and the impact on operations and safety. The auto local routine and auto local special reports turned out to be the most crucial parts of the acceptance process. Open discussions between the parties involved clarified the key issues of the (AUTO) METAR system AMOFSG/9-IP/8 - 4 -

and were beneficial to the acceptance of the AUTO METAR system and its quality. One key issue was that ATC and KNMI agreed on a pro-active role of the aviation meteorologist in case of significant deviations from the expected or perceived meteorological situations or for specific events.

4. ACTION BY THE AMOFSG

4.1 The AMOFSG is invited to note the contents of this information paper.

— — — — — — — —

AMOGSG/9/IP-8 Appendix

APPENDIX

AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

Table of contents

Table of contents ...... i

1. The AUTO METAR system ...... 1 1.1 Introduction ...... 1 1.2 AUTO METAR – a container concept ...... 1 1.3 Description of automated aeronautical meteorological reports at civil airports ...... 2 1.3.1 Definitions ...... 2 1.3.2 Characteristics ...... 3 1.3.3 Coding practices ...... 4 1.3.4 Availability and distribution channels ...... 5 1.4 Description of automated aeronautical meteorological reports at military airbases ...... 6 1.5 Description of automated aeronautical meteorological reports at North Sea off-shore structures ...... 6

2. Observation infrastructure of the AUTO METAR system ...... 7 2.1 Components of the observation infrastructure ...... 7 2.1.1 Meteorological sensors ...... 7 2.1.2 Sensor layout at EHRD ...... 7 2.1.3 Sensor layout at other locations ...... 8 2.1.4 Video camera system at EHRD ...... 9 2.1.5 Video camera system at other locations ...... 10 2.1.6 SIAM sensor interface and multiplexer...... 10 2.1.7 Server systems ...... 11 2.2 System redundancy and backup measures ...... 11 2.2.1 Backup sensors ...... 11 2.2.2 Infrastructure redundancy ...... 11 2.2.3 Backup procedures for visibility and clouds ...... 12 2.2.4 Server redundancy ...... 12 2.3 Optimization of the observation infrastructure ...... 13 2.3.1 Backup measures ...... 13 2.3.2 Sensor issues ...... 13 2.3.2.1 Insects reduced visibility ...... 13 2.3.2.2 Representativeness of visibility and cloud observations ...... 14 2.3.2.3 CB/TCU information ...... 15 2.3.2.4 Video cameras ...... 15 2.3.3 Reporting rules ...... 15 2.3.4 Documentation issues ...... 15

3. Supervision of the AUTO METAR system...... 16 3.1 Monitoring of the AUTO METAR system status ...... 16 3.1.1 Sensor status...... 16 3.1.2 Monitoring by service staff ...... 16 3.1.3 Monitoring by operator ...... 16 3.2 Remote verification and complementation ...... 17 3.2.1 Video camera images ...... 18 3.2.2 Local sensor information ...... 18 3.2.3 Regional sensor information ...... 19 3.2.4 Other sources of information ...... 20 3.2.5 Communication ...... 20 3.2.6 Complementation ...... 21

i AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

4. Introduction process of and experiences with the AUTO METAR system ...... 22 4.1 Task Force Update Criteria (AUTO) SPECIAL ...... 22 4.2 Assessments of the AUTO METAR system ...... 23 4.2.1 Technical assessment by KNMI ...... 23 4.2.2 Safety assessment by KNMI ...... 24 4.2.3 Safety assessment by ATC (LVNL) ...... 25 4.2.4 Assessment of system performance ...... 25 4.2.4.1 Phase 1 availability of sensor information ...... 25 4.2.4.2 Phase 2 availability of (AUTO) METAR and elements ...... 26 4.2.4.3 Phase 3 availability of (AUTO) ACTUAL and (AUTO) SPECIAL and elements ...... 26 4.2.4.4 Phase 4 analysis of situations with non-representative values ...... 27 4.3 Stakeholder Consultation and User Satisfaction ...... 28 4.4 Handling of user complaints...... 30 4.5 Future improvements ...... 31 4.5.1 Video Cameras ...... 31 4.5.2 Representativeness of cloud observations ...... 31 4.5.3 CB/TCU cloud types ...... 32 4.5.4 Representativeness of visibility observations ...... 32 4.5.5 Visibility observations corrected for insects ...... 32 4.5.6 Representativeness of weather observations ...... 32 4.5.7 Improved precipitation type discrimination ...... 32 4.5.8 Content of reports ...... 33 4.5.9 Documentation ...... 33

5. Conclusions and lessons learned ...... 34

6. References ...... 37

Appendix A: Fact sheet update criteria (AUTO) SPECIAL ...... 38

ii AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

1. The AUTO METAR system

1.1 Introduction Since March 15, 2011 the so-called AUTO METAR system is operational at Rotterdam The Hague Airport (EHRD). As a result the weather observations are performed fully automated, but supervised by a meteorologist in the central weather room at the main premises of the Royal Netherlands Meteorological Institute (KNMI) at De Bilt. The automation of meteorological observations started at KNMI in November 2002 when all synoptic meteorological observations were fully automated. Using the same instruments and related equipment, but with algorithms tailored to suit aeronautical requirements, automated aeronautical meteorological observations were introduced at Groningen Airport Eelde (EHGG) and Maastricht Aachen Airport (EHBK) in May 2004. First this so-called AUTO METAR system was used only during closing hours of the airports, but since August 2007, after an evaluation by Air Traffic Control The Netherlands (ATC/LVNL), the AUTO METAR system became operational 24/7. From 2005 onwards the AUTO METAR system was introduced on production platforms at the North Sea. Currently 13 off shore platforms are fully equipped. In 2008 the AUTO METAR system was introduced on nine military airbases. The AUTO METAR system was introduced at Rotterdam The Hague Airport beyond opening hours in December 2010 and has been operational 24/7 since March 2011. The introduction of the AUTO METAR system at Rotterdam The Hague Airport was for the present the final step. Professional meteorological observers are currently employed by KNMI only for carrying out observations at Amsterdam Airport Schiphol (EHAM). The added value of the observer with respect to the capacity of Mainport Schiphol makes their presence undisputable for the moment.

The rationale to introduce the AUTO METAR system is twofold. Developments in observation techniques and technology lead to the belief that it is possible to provide an automated observation of good quality. At the same time the introduction of the AUTO METAR system leads to significant cost savings as local MET offices and local observing staff at an airport are no longer required. As a by-product the AUTO METAR system facilitates the possibility to get meteorological information from airbases that were closed and unmanned during weekends and from productions platforms on the North Sea so that a denser, both temporal and spatial, network of aeronautical meteorological observations became available to the users.

In this document the AUTO METAR system operational at Rotterdam The Hague Airport and the process of how it was achieved and approved are presented.

1.2 AUTO METAR – a container concept It is important to recognize the distinction between the product “AUTO METAR” and the “AUTO METAR system” and to bear this in mind when reading this document.

The product “AUTO METAR” is a routine observation at an aerodrome and is issued as meteorological report for dissemination beyond the aerodrome. The prefix AUTO in the report is to indicate that the report is generated without intervention of a human observer at the aerodrome concerned.

The term “AUTO METAR system” is used to denote the entire system used for the automated production of the meteorological aeronautical reports, and includes all automated observation reports, of which AUTO METAR is one. These reports are: • AUTO METAR; • AUTO SPECI; • AUTO Local Routine Report, in this report referred to as AUTO ACTUAL; and, • AUTO Local Special Report, in this report referred to as AUTO SPECIAL.

1 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

The term “AUTO METAR system” does not only designate the sensors, the associated technical infrastructure for data-acquisition, -processing, and the generation and dissemination of the meteorological aeronautical reports in suitable formats. The AUTO METAR system also includes the information of remote sensing systems for the detection of lightning and CB/TCU, the usage of backup sensors and systems, and the remote monitoring by meteorologists and service staff using suitable tools including e.g. real-time sensor data displays and video camera images. The meteorologist also adds the TREND to the reports. The system also includes the communication with the users covering regular daily briefings; intermediate updates for sudden changes in meteorological conditions; handling of sensor or system maintenance or malfunction; and provisions for local points of contact at air traffic control and the airport itself that can be used for the verification of the local meteorological situation. Finally, the AUTO METAR system also includes documentation, procedures and the service level agreement with ATC.

It is also important to note that the content of the observation reports itself can differ between airports or between states. E.g. an AUTO METAR that contains pressure, wind direction and wind speed, air temperature and dew point only versus an AUTO METAR that contains visibility, clouds, weather and TREND as well.

1.3 Description of automated aeronautical meteorological reports at civil airports A more detailed description of the automated meteorological aeronautical reports at civil airports at The Netherlands is provided underneath.

1.3.1 Definitions The “AUTO METAR” is an aviation routine meteorological report for dissemination beyond the aerodrome. The meteorological information contained in the AUTO METAR is generally representative for the aerodrome and its immediate vicinity. The AUTO METAR is generated at H+20 and H+50 using corresponding sensor information and is disseminated after validation and addition of the TREND by a meteorologist at a remote location. The time label of the AUTO METAR is H+25 or H+55.

The “AUTO SPECI” is a similar meteorological report as the AUTO METAR. The only difference is that the AUTO SPECI is issued in between the fixed time intervals of the AUTO METAR when certain criteria are met, e.g. a change of the visibility exceeding specified thresholds. As the METAR and AUTO METAR are produced half-hourly in The Netherlands the SPECI and AUTO SPECI reports are not produced at civil airports in accordance with ICAO Annex 3 (paragraph 4.4.2 b) standards.

The “AUTO ACTUAL” is an aviation local routine meteorological report for landing air traffic disseminated at the aerodrome. The meteorological information contained in the AUTO ACTUAL is generally representative for the touchdown (take-off) zone or the situation along the runway. The AUTO ACTUAL is generated at H+20 and H+50 using corresponding sensor information and is disseminated after validation and complementation by a remote meteorologist. The meteorologist adds the TREND to the AUTO ACTUAL and can set indicators for specific weather phenomena (cf. section 1.2.1.2). The time label of the AUTO ACTUAL is H+25 or H+55.

The “AUTO SPECIAL” is a similar meteorological report as the AUTO ACTUAL. The only difference is that the AUTO SPECIAL is issued in between the fixed time intervals of the AUTO ACTUAL when certain criteria are met, e.g. a change of the visibility exceeding specified thresholds.

Note that there is a difference in representativeness between AUTO METAR and AUTO SPECI at one side and AUTO ACTUAL and AUTO SPECIAL at the other side. The same applies for several elements in the reports itself. E.g. the 10 minute averaged wind speed and direction is reported in the AUTO METAR whereas in the AUTO ACTUAL the 2 minute averaged wind is reported. Furthermore the wind of a designated wind sensor is used in the AUTO METAR whereas the wind corresponding to the runway in use is reported in the AUTO ACTUAL. Finally it should be noted that the criteria used for issuing a SPECIAL, determined by national agreement, can differ from the international SPECI criteria.

2 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

1.3.2 Characteristics The coding and contents of the automated meteorological reports are basically equal to the manual reports although there are some differences. Apart from the inclusion of the term AUTO itself and specific codes related to sensor limitations like NCD (No Clouds Detected) or UP (Unknown Precipitation) some weather phenomena or descriptors are not reported in the automated reports since there are no suitable sensors to detect them - e.g. patches of fog (BC) or smoke (FU). Also note that the AUTO METAR, AUTO ACTUAL and AUTO SPECIAL are not fully automated since all reports are monitored and complemented by a meteorologist at a remote location, nor are the METAR, ACTUAL and SPECIAL completely manually generated. The so-called visual parameters related to visibility, cloud and weather information are generally entered manually, but most fields in the METAR, ACTUAL and SPECIAL are filled in automatically using processed sensor information. The meteorologist (or observer in case of a METAR, ACTUAL or SPECIAL) can discard the sensor value. In such an event the backup sensor is used automatically for most parameters, but for some, like RVR at touchdown, no alternative can be given.

Identical information in manual and automated meteorological observation reports The following parameters are measured automatically and reported identically in reports compiled by a human observer and in automated reports: • Wind • RVR • Air temperature • Dew point • QNH and QFE In addition the runway in use is indicated by ATC and reported automatically in the (AUTO) ACTUAL and (AUTO) SPECIAL.

Meteorologists at the central forecast office at De Bilt monitor the observation and shall, when appropriate, add the following items to the automated reports: • RSM: runway state message (AUTO METAR only) • LLTI: low level temperature inversion ((AUTO) ACTUAL and (AUTO) SPECIAL only) • Windshear report • Windshear forecast ((AUTO) ACTUAL and (AUTO) SPECIAL only) • TREND: a two hour forecast at the end of the observation report Note that at Amsterdam Airport Schiphol (EHAM) the local observer adds the above items to the reports, but the Windshear forecast and TREND are added after consultation with the aviation meteorologist at De Bilt.

Differences between manual and automated meteorological observation reports

Visibility • The visibility reported in AUTO METAR, AUTO ACTUAL and AUTO SPECIAL may occasionally vary compared to a manual observation in situations where visibility is rather inhomogeneous. This is due to the point measurement principle in automated reports. For example, when fog is reported in automated reports and no fog is present above the runway, visibility at the runway may be higher. The differences can be most pronounced between METAR and AUTO METAR since an observer reports the prevailing visibility, i.e. the greatest visibility which is reached within at least half of the horizon circle or within half of the surface of the aerodrome, whereas the AUTO METAR reports the 10-minute averaged, or 2 to 10 minute averaged in case of a marked discontinuity, visibility assessed by a designated visibility sensor.

3 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

RVR • RVR and visibility for usage in local MET reports are determined automatically for ACTUAL and SPECIAL as well as for AUTO ACTUAL and AUTO SPECIAL. However, the RVR is not included directly in the local MET reports. A RVR indicator is included instead, which informs ATC whether one or more of visibility sensors at the aerodrome reports an RVR or visibility below 1500 m. In that case ATC starts requesting the RVR of all visibility sensors. The RVR indicator is set automatically in the AUTO ACTUAL and AUTO SPECIAL, whereas it is set manually in ACTUAL and SPECIAL. Clouds • Cloud base height reports in automated reports are based on observations by laser ceilometers. Cloud cover (cloud amount) is derived by using a specific algorithm to account for the fact that a ceilometer provides point measurements. The cloud algorithm converts a 10-minute time series of individual cloud bases reported by the ceilometer into cloud layers each with a corresponding cloud base height and cloud amount. • The information of a ceilometer and a nearby visibility sensor are used to distinguish between a cloud base and sky obscured. In the latter case the vertical visibility is reported instead of the cloud base and amount. • Cloud type in the form of cumulonimbus (CB) or towering cumulus (TCU) will be included in automated reports based on an algorithm which uses lightning, weather radar reflectivity and satellite information. • Additional information about the convective characteristics of the clouds through (near) real time weather radar observations is available via the Meteorological Watch Office (MWO) at De Bilt, Meteorological Self Briefing units, ATC centres and via the aviation weather website of KNMI. Present Weather Automated meteorological reports contain a subset of present weather codes, not the complete set as used in METAR, ACTUAL and SPECIAL reports. • Present weather in the vicinity of the aerodrome cannot be observed and therefore the code VC is not used in automated reports. • If precipitation type cannot be determined by the sensor, the code UP (unknown precipitation) is reported. • The following weather phenomena or descriptors are not reported: FC, SS, DS, PO, SA, DU, FU, VA, MI, BC, PR, DR and BL (see ICAO Annex 3, Appendix 3 paragraph 4.4.2 for explanation of codes). • Observations of present weather are carried out by a present weather sensor located on a fixed location. Only weather occurring at that specific location will be reported. • Thunderstorms (TS) are reported when lightning is detected within a distance of 15 km from the aerodrome reference point (ARP) based on information gathered by the national lightning detection network.

1.3.3 Coding practices

Codes not provided in automated meteorological observation reports • FC, SS, DS, PO, SA, DU, FU, VA, MI, BC, PR, DR and BL (see ICAO Annex 3, Appendix 3 paragraph 4.4.2 for explanation of codes). • VC: vicinity. • CAVOK: clouds and visibility OK. • SKC: the term "sky clear" is no longer used in manual and automated MET observations.

Codes used in automated meteorological observation reports only • AUTO: indicator of an automated report. • NCD: no clouds detected. When the sensor does not detect clouds and no CB/TCU are detected, the code NCD is reported due to the point measurement principle. • UP: unknown precipitation. When the type of precipitation cannot be determined by the present weather sensor.

4 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

Codes used in manual and automated meteorological observation reports • NSC: no significant clouds. When no clouds of operational significance (cloud base 5000 ft or more and no CB/TCU) are observed, the abbreviation "NSC" is used.

Additional remarks in the REMARK section AUTO METAR can contain up to three remarks in the REMARK section. The REMARK section is disseminated in the Netherlands only, in line with international data exchange rules. • RMK TS INFO NOT AVBL: in situations where lightning data is not available. • RMK CB INFO NOT AVBL: in situations where weather radar information is not available. • RMK WX INFO NOT AVBL: in situations where the local present weather sensor data is not available.

These REMARKs are for the moment disabled in AUTO ACTUAL and AUTO SPECIAL as the broadcast system of ATC is not yet capable of translating these codes into a voice message. As is the case with ACTUAL and SPECIAL, the meteorologist can add REMARKs on visibility, TS, CB, HAIL or Tornado, SEV SQUALL, Clouds and Turbulence in AUTO ACTUAL and AUTO SPECIAL.

Coding of missing information in automated meteorological observation reports It is essential that the absence of weather or cloud information in an automated report is clearly identified as either the result of the absence of the phenomena itself, or due to failure of the sensor. In the first case the associated group is sometimes omitted in the (AUTO) METAR whereas in the latter case the group is indicated by two or more slashes. When ceilometer information is not available due to technical reasons, the cloud group will be reported by using six slashes "//////". In that case it may still be possible to report “//////CB” or “//////TCU”. The unavailability of weather information will be indicated by "//" and the unavailability of visibility information by "////". Similarly, forward slashes are used to indicate the unavailability of all other parameters.

The AUTO ACTUAL and AUTO SPECIAL are distributed to ATC as a comma separated file, which is then automatically translated into a broadcast via the Automatic Terminal Information Service (ATIS). The available strings have always a distinct content. When an item is not available due to reporting practices the item is indicated by a space “ “. When an item is not available due to technical or other reasons the item is given as “N.A.”.

1.3.4 Availability and distribution channels The AUTO METAR of Rotterdam The Hague Airport is disseminated internationally in the AUTO METAR format via the Aeronautical Fixed Telecommunication Network (AFTN) and available via the regional Operational Meteorological (OPMET) centres. The AUTO METAR is nationally available via Teletext and the aviation website of KNMI. The AUTO METAR is also available via the Amsterdam MET Broadcast (VOLMET) and available on the Closed Information Circuit System (CCIS) of ATC. It should be noted that the METAR and AUTO METAR are made available and disseminated via the main office of KNMI at De Bilt.

The AUTO ACTUAL and AUTO SPECIAL are locally available only via the Automatic Terminal Information Service (ATIS), the system of ATC that broadcasts the meteorological information to the pilots. The reports are also available on the Closed Information Circuit System (CCIS) of ATC. It should be noted that apart from the meteorological information contained in the AUTO ACTUAL and AUTO SPECIAL ATC also receives near real-time sensor information, specifically wind and RVR. The AUTO ACTUAL and AUTO SPECIAL and sensor information is generated on the airport server system and is made available to ATC via a local network connection.

5 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

1.4 Description of automated aeronautical meteorological reports at military airbases Military airbases provide AUTO METAR, AUTO SPECI, AUTO ACTUAL and AUTO SPECIAL. Currently there are seven airbases at The Netherlands: Arnhem/Deelen (EHDL), Eindhoven (EHEH), Gilze-Rijen (EHGR), Leeuwarden (EHLW), Volkel (EHVK), Woensdrecht (EHWO) and De Kooy (EHKD).

At military airbases the AUTO ACTUAL and AUTO SPECIAL are called QAM and contain identical information. The only difference is that a so-called colour state and a colour state forecast, based on NATO specifications, are included in the QAM instead of the TREND. An AUTO QAM is generated on the airport server system and is made available, via a local network connection, to military users.

KNMI is responsible for operating and the maintenance of the AUTO METAR system at the seven military airbases. This includes the sensors at the airport and the hardware and software of the data acquisition, processing and dissemination modules.

It should be noted that the military system also includes presentation systems that display the meteorological reports and sensor information to military air traffic controllers and other local users such as Search and Rescue (SAR) and the fire brigade. Similar presentation systems are also installed at centralized approach of the Air Operations Control Station.

Differences exist between the civil and military AUTO METAR system. The meteorological observation infrastructure differs slightly from the civil aerodromes. For example the data communication lines and power supply facilities, which are provided by ATC and the military, differ and have different redundancy levels. Also the usage of backup sensors is different and no video cameras are used at airbases. Furthermore, the military system uses a single redundant server for data acquisition, processing and dissemination, whereas at civil airports data dissemination is handled by a separate redundant server.

Finally, a meteorological technician with observer skills is available at each airbase during opening hours and can offer assistance to military air traffic controllers and pilots on request.

1.5 Description of automated aeronautical meteorological reports at North Sea off- shore structures On the Netherlands continental shelf are approximately 110 offshore structures with helicopter decks. Approximately 30 of these are manned structures of which 13 provide a half hourly AUTO METAR. The report is produced fully automated. The products AUTO SPECI, AUTO ACTUAL and AUTO SPECIAL are not produced. There are no backup sensors and camera’s installed at the structures. The meteorological sensors and sensor interface used at platforms are identical to the ones used at civil airports and military airbases and are operated and maintained by KNMI. The data-acquisition systems at the platforms are owned by third parties. These systems take care of the local presentation of the meteorological data and make the raw data of the KNMI sensor interface available to the external FTP server of KNMI every minute. The central server system at De Bilt acquires the data of all platforms and generates the AUTO METAR. Note that the AUTO METAR of a platform is not monitored, validated and complemented remotely by a meteorologist before dissemination.

6 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

2. Observation infrastructure of the AUTO METAR system

2.1 Components of the observation infrastructure

2.1.1 Meteorological sensors The meteorological sensors used by KNMI are listed in Table 1. Table 1 reports the sensor, the associated Sensor Intelligent Adaptation Module (SIAM) sensor interface and the meteorological units. All sensors meet the requirements of the World Meteorological Organization (WMO, 2008) and the International Civil Aviation Organization (ICAO, 2010) and are maintained and calibrated by KNMI. Identical sensors are used at civil airports, military airbases, platforms at the North Sea, and at the automated weather observation stations.

Table 1. A list of meteorological sensors used by KNMI. The table also gives the associated SIAM sensor interface and the meteorological units reported by the interface.

SIAM Sensor SIAM Unit description Unit Vaisala Impulsphysik LD40 ceilometer C4 C1 First cloud base C2 Second cloud base C3 Third cloud base CX Vertical range of measurement ZV Vertical visibility Paroscientific Digiquartz 1015A barometer P1 PS Air pressure Kipp & Zonen CM11 pyranometer Q1 QG Global radiation KNMI precipitation gauge R2 NI Precipitation intensity ND Precipitation duration Pt-500 platinum resistor element U1 TG Grass temperature 0.10m Pt-500 platinum resistor element TA Ambient temperature 1.50m Derived by SIAM TD Dew point temperature 1.50m Vaisala HMP-233 capacitive hygrometer RH Relative humidity 1.50m KNMI cup anemometer W0 WS Wind speed KNMI wind vane WR Wind direction Vaisala LM11/21 luminance meter Z4 ZA Background luminance Vaisala FD12P present weather sensor ZM Visibility (MOR) NI Precipitation intensity ND Precipitation duration PW Precipitation type ATC Runway information system B0 BB Runway usage

The last entry of Table 1 is not a meteorological sensor, but a system of ATC that indicates the runway in use. This information is provided to KNMI as a serial string with the format of the SIAM sensor interface so that it can be processed similarly as the sensor information.

2.1.2 Sensor layout at EHRD The positions of the meteorological sensors used at Rotterdam The Hague Airport are shown in Figure 1. Rotterdam has a CAT I runway for instrument precision approach and landing operations. The runway can be used from both sides (06 and 24). Hence according to ICAO (2010) recommendations it is equipped with a wind sensor and a visibility sensor near the touchdown zone at both ends of the runway. Both of these visibility sensors are equipped with a background luminance sensor. The background luminance is used in the derivation of visibility for aeronautical purposes and the Runway Visual Range (RVR) from the Meteorological Optical Range (MOR) reported by the visibility sensor. The pressure

7 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011 sensor is located at the wind mast of runway 24 and the temperature and humidity sensors are situated at the measurement field near 24 touchdown. The measurement field also contains the ceilometer, the rain gauge and a global radiation sensor. The latter two sensors are used for synoptical and climatological purposes only. Backup sensors for pressure, temperature and humidity are located at the wind mast and touchdown zone of 06.

Note that the visibility sensor used at Rotterdam The Hague Airport is in fact a so-called present weather sensor which also reports the precipitation type. The precipitation type in combination with other meteorological information is used to derive the weather information. Lightning, which is also included in the weather information, is not measured locally, but is provided from the central processing unit of the lightning detection system at De Bilt. The lightning information in combination with information from weather radars and METEOSAT satellite is used to provide information on the presence of convective cloud types CB and TCU, which is added to the cloud information. Both the lightning and the CB/TCU cloud type information are provided through the network connection between Rotterdam The Hague Airport and the main offices of KNMI at De Bilt.

KVS2 ↓↑ RUQC Z Threshold

↓↑ WP Threshold Z WPU 24 06 Aiming Aiming point point

KVS1

W = wind Z = visibility / weather Techinical C = cloud Room KNMI U = temperature / humidity P = pressure Q = radiation R = precipitation O = video camera

Figure 1: The position of the meteorological sensors and video cameras at Rotterdam The Hague Airport. The sensors and cameras associated to touchdown 06 and 24 and their respective relay station (KVS) and data communication line to the technical room of KNMI at the airport are shown in red and blue, respectively. The measurement field not only contains temperature and humidity sensors and a ceilometer for cloud observations, but also a rain gauge and a global radiation sensor for synoptic purposes.

2.1.3 Sensor layout at other locations The layout of the meteorological sensors at EHRD is typical for a civil airport with a CAT I runway. A similar layout is used at EHGG. The sensor layout for the civil airport EHBK is nearly identical, but it contains an additional visibility sensor at the mid position as is required for a CAT III runway. A similar instrumentation is used for all CAT III runways of EHAM with 3 visibility sensors per runway and wind measurements representative for the touchdown and take-off position of each runway. Details of the sensor layout of Amsterdam Airport Schiphol can be found in Wauben and Sondij (2009).

The military air bases generally have a sensor layout similar to EHRD with backup sensor for pressure, temperature and humidity, two visibility sensors per runway, but are equipped with 3 wind sensor sets. The measurement field which is located near the middle of the runway contains the primary pressure, temperature and humidity sensors and a wind sensor set on a 10-meter mast. The backup sensors for pressure, temperature and humidity are located at touchdown of the runway together with a transmissometer visibility sensor and wind on a 6-meter mast. The alternate or end position contains the FD12P present weather sensor and wind on a 6-meter mast.

8 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

The platforms on the North Sea and the military airbase EHKD contain only a single sensor set. At platforms a dual wind sensor set is often used. In these cases wind is measured at opposite sides on the vent stack. The upstream sensor set is automatically selected for operational use. Also note that the precipitation gauge and pyrometer are not used at platforms.

2.1.4 Video camera system at EHRD A dual video camera system is mounted on the wind mast of 24 touchdown (cf. Figure 2). The system consists of cameras mounted at 2 and about 9 m, and facilitates monitoring of the representativeness of the visibility measurements at the touchdown zone during daytime. In addition a video camera is located about 500 m before the threshold of runway 06 at a height of 2.5 m and pointed towards 24 touchdown. This camera can be used by the meteorologist to check the general meteorological conditions, particularly of cloudiness and visibility at the airport. Note that no quantitative information on the visibility can be derived from the video camera images. Only a rough check of the measured visibility can be estimated from the images. However, the cameras provide information on the nature of obscuration (shallow fog, patches). The video signal is made available via the data communication network to the central weather room and service staff of KNMI at De Bilt.

camera →

camera ↓

Figure 2: The FD12P visibility sensor and the wind mast with KNMI cup anemometer and wind vane near 24 touchdown at Rotterdam The Hague Airport. Video cameras are mounted at 2 m and near the top (9 m) of the frangible wind mast.

9 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

2.1.5 Video camera system at other locations A dual video camera system as described above for EHRD is also installed near the main touchdown of EHGG; at both ends of the runway of EHBK; and at touchdown of runway 18R of EHAM, which is located about 7 km from the local observer.

2.1.6 SIAM sensor interface and multiplexer All sensors are operated in combination with a so-called SIAM sensor interface, a Sensor Intelligent Adaptation Module. A SIAM communicates with the sensor and converts the sensor output into meteorological quantities in a fixed serial format. A SIAM runs asynchronously and polls the sensor and gets the meteorological as well as the status information. Generally the SIAM polls the sensor every 12 seconds, but if necessary the sensor interface samples the sensor with a higher frequency, e.g. the cup anemometer and wind vane are sampled with 4 Hz and the sensor interface calculates the 3 seconds running average in order to calculate the wind gust and take account of marked discontinuities of the wind. The SIAM performs a format and a range check on the meteorological quantities and generates an output string every 12 seconds. The sensors and SIAM sensor interfaces are installed in the field and are connected via fixed copper lines to a nearby relay station (KVS) of ATC which also supplies the no- break power supply. A relay station typically serves half of the runway, e.g. the sensors associated with 06 touchdown are connected to KVS1 and the sensors associated with 24 touchdown and the measurement field are connected to KVS2 (cf. Figure 1). The SIAM sensor interfaces are situated in the field either directly at the sensor, e.g. in the electronic box of the wind mast or of the visibility sensor or in the central data box at the measurement field. The latter also contains a MUF (MUltiplexing Facility) so that all SIAM data can be sent to KVS2 via a single data line. At the relay station the serial SIAM information is multiplexed on a single serial line and forwarded to the technical room via copper lines. In the technical room all incoming MUF strings are duplicated by splitters, multiplexed on to a single line and given to the server pair for further processing. An overview of the technical observation infrastructure at Rotterdam The Hague Airport is shown in Figure 3. Note that information of the Runway Information System (RIS) of ATC, which indicates which runway is in use, is also fed as a SIAM string into the MUF cascade.

Video De Bilt Cisco modem Schiphol Wind 06 DW0 Video Cisco modem Pressure 06 Technical Room DP1 Wind 24 MUF Cisco switch Temp/Hum 06 DW0 DU1 MUF KNMI LAN Visibility 06 KVS 1 Pressure 24 MOXA DZ4 DP1 KVS 2 Test system Measurement field Visibility 24 MUF Splitter Precipitation Data Box ADCM 0 DZ4 Splitter MIS 0 XR1 Radiation Splitter XQ1 MUF Splitter MUF ADCM 1 Temperature XU1 Humidity Splitter MIS 1 DC4 Splitter VAIS Clouds ATC Runway Information System

Figure 3: An overview of the sensors and SIAMs and splitters in the MUF cascade at Rotterdam The Hague Airport. Black line and boxes denote connections and components which a single point of failure. Blue lines and boxes show the secondary server system with associated sensor data. The video cameras components are denoted in green and the sensor data that is forwarded to the test server system at De Bilt is shown in red. The backscatter information of the ceilometer that is forwarded to De Bilt for monitoring of volcanic ash is also given in red.

10 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

2.1.7 Server systems The 2 outputs of the MUF cascade, each containing all sensor and RIS information, are fed into a redundant server pair. In normal operation one of these ADCM (Aviation Data-acquisition and Communication Module) servers is hot and ingests all sensor data. The SIAM data that is transmitted asynchronously by the SIAM is assigned to a 12 second interval at the hot ADCM. Generally the last SIAM string of each sensor that arrived at the server in a 12 second interval is labeled with the time at the end of the interval. Some processing is involved to handle reception of either none or two SIAM strings in a 12 second interval. The ADCM monitors the status of the sensors and it also takes care of derivations, e.g. RVR and cross wind calculations and the handling of automated backup of sensors, and the generation of meteorological reports. A copy of all raw and derived data is forwarded to the cold server that stores it. The cold server continuously monitors whether the hot server is available. If communication is lost the cold becomes hot and starts processing the data. It is also possible to force a manual failover so that maintenance of the cold server can occur without interruption of the data flow. During a start-up the server checks whether a hot server is present in which case it will go into the cold mode. When 2 servers are hot, e.g. after a failure of the network communication between the 2 servers, the secondary server will automatically switch to cold. Both servers are connected to the KNMI LAN at Rotterdam The Hague Airport. Note that the server systems have 2 network cards and are connected to 2 separate network switches. The sensor and derived data is generally updated every 12 seconds. About 275 out of the total amount of 770 variables available at Rotterdam The Hague Airport update every 12 seconds. The ADCM server performs the crucial tasks of data-acquisition and processing. In order to avoid any loss of performance due to data requests by users a copy of all data is put on the MIS (Meteorological Information Server) server pair which handles the data requests for local users at the aerodrome.

2.2 System redundancy and backup measures

2.2.1 Backup sensors As indicated in Figure 3 the sensors in the field are potential single points of failure. As part of the AUTO METAR system backup sensors for pressure, temperature and humidity were installed at the wind mast and touchdown zone of 06 at EHRD. The backup of pressure, temperature, humidity and wind is automatically taken into account in the processing at the airport server system. Wind can be backed up by the sensor at the opposite end of the runway since their distance is relatively small (about 1 km) and the sheltering factors at both sites are similar. The pair of visibility sensors, however, cannot be used as each others backup as the visibility obtained with the other sensor at a distance of about 1100 m can differ significantly. Since the runway visual range and visibility should be representative for the touchdown zone a backup by the other sensor is not possible in all conditions. Rotterdam The Hague Airport is equipped with a single ceilometer.

2.2.2 Infrastructure redundancy A full redundancy of the system at Rotterdam Airport is available after the splitters in the technical room (cf. Figure 3). In case of a failure or malfunction of a system or communication line after the splitter, the full set of information is still available or after an automated failover to the secondary system. The splitters themselves and the components before the splitters like a sensor are also redundant, but in a different way. In case e.g. a sensor fails a backup sensor will be used automatically. There is, however, no backup for visibility representative for the touchdown position and clouds.

A sensor and its backup sensor are located at different physical locations at the airport and they use other parts of the observation infrastructure such as power supply, multiplexers, splitters, data communication lines and relay stations in order to get the sensor information to the servers systems in the technical room. Hence, should a sensor or an associated component of the observation infrastructure fail then the backup sensor is still available. Even if a connection to a relay station or a MUF at a relay station or a splitter in the technical room fails and all the sensor data of that end of the

11 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011 runway is not available, the backup sensors of pressure, wind, temperature, humidity and weather are still available. There is, however, no automated backup for visibility and clouds.

Note that at Amsterdam Airport Schiphol the redundancy of the infrastructure is at a higher level since sensors and other equipment in the field are connected to a no-break power supply provided by ATC; the sensor information is split at the relay stations in the field and is fed twice to a redundant data communication infrastructure of ATC; the redundant server systems are located at 2 different locations.

2.2.3 Backup procedures for visibility and clouds In case of a failure of a visibility sensor or ceilometer or associated infrastructure the visibility information of the corresponding touchdown position or the cloud information is not available. In such a situation the aviation meteorologist in the central weather room of KNMI at De Bilt can orally provide the missing visibility or cloud information. For that purpose the meteorologist uses the images of the video cameras, the information provided by the meteorological network from nearby stations, information from remote sensing systems, the expected atmospheric conditions and verification of the meteorological situation with local staff of the airport or ATC by means of asking direct questions, e.g. whether visibility markers are visible or not. Note that in such a situation the meteorologist does not make an observation, but indicates whether so-called Visual Flight Rules (VFR) conditions are applicable or not. During VFR conditions the runway can still be used although the air traffic capacity is less than for instrument precision approach and landing operations that apply for Instrument Flight Rules (IFR). The information that is provided orally in case of a malfunction of an observation system or associated infrastructure is logged in the shift reports and voice recorded by ATC and KNMI.

When the information of the visibility sensor at touchdown is not available the runway can only be used if a visibility of 3000 meters or more can be determined by using the cameras or via consulting the ATC staff by telephone. For this purpose Air Traffic Controllers have been provided with a 360 degree overview map of the airport containing reference objects with defined visibility distances. If the visibility cannot be determined or is below 3000 meters the airport cannot be used for instrument landing approaches. However, in such a situation it may be possible to approach the runway from the other side and use the visibility sensor at touchdown.

Naturally any failure is handled by KNMI service staff according to the agreed response time as given in the service level agreement between KNMI and ATC The Netherlands.

2.2.4 Server redundancy The server pair for processing the data (ADCM) and providing the data to the local users (MIS) is duplicated. In case one server fails the other takes over automatically. At Amsterdam Airport Schiphol there is even a third server system with a different operation system, a different application with only the basic functionality and implemented by another manufacturer using different software tools that can provide the most essential data in case the redundant ADCM/MIS fails.

The network components at Rotterdam The Hague Airport are also redundant and each server has 2 network cards and connects to both network switches. Since the servers are located at the airport itself the processing and dissemination of meteorological data to local users runs autonomously and continues uninterrupted if the network connection to De Bilt is lost. In such a situation the airport users still get the local routine and local special reports (AUTO ACTUAL and AUTO SPECIAL) and have access to the processed wind and runway visual range data. During a disruption of the network connection to De Bilt, Rotterdam The Hague Airport receives no lightning and CB/TCU information. This missing information is indicated in the meteorological reports. Loss of the network connection to De Bilt also means that the monitoring of the sensor information by the meteorologist is hampered. For that purpose a backup network connection between Rotterdam The Hague Airport and Amsterdam Airport Schiphol is available. The KNMI observer at Schiphol can monitor the sensor information and add the TREND forecast, as well as Runway State Message, Windshear and Low Level Temperature Information, in close collaboration with the aviation meteorologist at De Bilt when needed.

12 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

2.3 Optimization of the observation infrastructure The period between the first introduction of the AUTO METAR system at Groningen Airport Eelde and Maastricht Aachen Airport in May 2004 and the implementation of the AUTO METAR system at Rotterdam The Hague Airport in March 2011 was used to improve the system based on the experiences and feedback from users. Some improvements to the system are highlighted in the following subsections.

2.3.1 Backup measures The implementation of backup sensors at Rotterdam The Hague Airport has already been mentioned in section 2.2.1. Note that backup sensors for pressure, temperature and humidity have also been installed at Groningen Airport Eelde and Maastricht Aachen Airport. Visibility and runway visual range at the touchdown position have no backup. The reason for this is that due to the large spatial differences that can occur for visibility the sensor at the other end of the runway cannot be used as a backup. A suitable backup sensor for visibility would require an additional visibility sensor near the touchdown position. However, even if such a backup sensor would be available near the touchdown position it would still use the same infrastructure unless the complete chain would be duplicated, which would be very expensive. Visibility at touchdown position is an accepted single point of failure.

Note that in the current situation the same infrastructure as for the ATC systems is partly used. Hence in case of a malfunction at e.g. the relay stations or the connections to the technical room the corresponding ATC systems would also be not available. The introduction of backup measures only makes sense by considering the total context. Note that in case the visibility at a touchdown position is not available, the sensor at the other end of the runway, which uses a different part of the infrastructure, will generally still be available so that the runway can be approached from the other side. When the visibility conditions are critical the wind conditions are usually such that there are no restrictions in that respect either. In case of good visibility conditions the runway can be used under so-called visual flight rules (VFR). The video cameras have been introduced at the airport as a tool for assisting the meteorologist to verify whether VFR conditions apply. The meteorologist can also consult ATC staff by telephone. For this purpose Air Traffic Controllers have been provided with a 360 degree overview map of the airport containing reference objects with defined visibility distances.

2.3.2 Sensor issues

2.3.2.1 Insects reduced visibility During the evaluation of the AUTO METAR system at Rotterdam The Hague Airport situations occurred when the MOR reported by the visibility sensor sometimes reported reduced values around sunset. Investigations showed that insects in the measurement volume can cause a significant decrease of the MOR reported by a forward scatter visibility sensor. Collaboration with manufacturer was established in order to filter the spikes due to insects in the raw sensor signal before calculation of the MOR. A field test evaluation of the MOR reported by an FD12P equipped with firmware that filtered for insects showed good results (cf. Wauben, 2011). KNMI currently prepares the introduction of an updated firmware with insect filtering of the MOR for operational use at civil airports.

13 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

50000 50000

10000

10000 Background luminance (cd/m luminance Background

1000

1000 MOR(m)

MOR FD12P raw (m) 100 2 MOR FD12P corrected (m) ) MOR Reference (m) Background Luminance FD12P (cd/m2)

100 10

40 3 0 3 6 9 12 15 18 21 24 Time (UT)

Figure 4: The 1-minute averaged MOR and background luminance observed by a FD12P forward scatter visibility sensor at De Bilt on August 5, 2010. The background luminance (blue curve with scale on the right) is 4 cd/m2 at nighttime and show a sharp increase near sunrise (3:30 UT) and a decrease near sunset (20 UT). The MOR (black curve with scale on the left) show values exceeding 10 km. Around sunrise (3 to 5 UT) and at night (22-24 UT) low MOR values occur during fog. Similar MOR values are reported by a transmissometer at De Bilt during these periods (not shown). Reduced MOR values due to insects occur around sunset (19:30 to 20:30 UT) with values below 1 km. The MOR of the transmissometer shows no reduced MOR values during this period. The insect filtering of the FD12P mitigates the MOR reduction significantly (red curve) although the corrected MOR still shows MOR reductions compared to a constructed reference (green curve) which is in fact the rescaled MOR of the transmissometer. The corrected MOR is, however, above aeronautical visibility limits.

2.3.2.2 Representativeness of visibility and cloud observations During the evaluation of the AUTO METAR system at the regional airports various complaints of users were related to lacking representativeness of the visibility and cloud observations by visibility sensors and ceilometers. The situation experienced by local users conflicted with the observations reported by the AUTO METAR system or the latter showed a large delay. Hence users complained about faulty sensor observations. Analysis of these situations showed that the sensors and corresponding algorithms worked correctly, but that measurements at a specific location can deviate significantly from that of an observer looking around. For example clouds associated to a front can only be reported by a ceilometer once they are directly overhead and then it takes 10-minutes before the cloud algorithm changes the cloud cover to overcast. Similar deviations can occur for visibility. The reported visibility is generally a 10- minute averaged value. A so-called marked discontinuity criteria is used which reduced the averaging period to 2-minutes when the visibility changes significantly. It should be noted that the user complaints that were related to lacking representativeness of the visibility and cloud observations were partly related to the unfamiliarity of ATC staff with the details of the measurement systems and internationally agreed observation principles.

14 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

The usage of multiple sensors, e.g. ceilometers, has been considered to improve the spatial representativeness of the observations. The cloud algorithm uses a 10-minute time series of individual cloud base height measurements reported by the ceilometer. The fraction of the sky evaluated by the cloud algorithm can be imagined as a narrow strip along the sky, the width of this strip is determined by the opening angle of the ceilometer and the length by the movement of the clouds due to the wind aloft. Several of such strips are considered when multiple ceilometers are combined in the cloud algorithms. Even by using several ceilometers only a small portion of the entire sky will be sampled. Analysis of the use of three instead of one ceilometer at Schiphol versus the observer in 2002 showed therefore only little improvement.

Another observation technique is required to solve the spatial representativeness issue. For that purpose a scanning pyrometer, the so-called NubiScope, has been evaluated (Wauben et al, 2010). The NubiScope performs a scan of the entire sky every 10-minutes during which the thermal infrared sensor determines the sky temperature in 1080 orientations. From the sky temperature the presence of clouds can be determined. The total cloud cover obtained with the NubiScope showed good results, but accurate height information, which is crucial for aeronautical purposes, is unfortunately lacking.

2.3.2.3 CB/TCU information In 2004 CB/TCU information was not included in the AUTO METAR system. At the end of 2006 CB/TCU information derived from lightning and precipitation radar reflectivity data was added to the system. Recently, the CB/TCU product has been improved by using METEOSAT satellite data in addition to radar and lightning data.

2.3.2.4 Video cameras Video camera systems have been installed at civil airports as a tool for the meteorologist to check the meteorological conditions at the airport remotely. The video images can be used as a source of information on cloudiness, visibility and present weather. Although the images give an indication of cloudiness and visibility, the quality of images is generally considered too poor for an accurate estimation of cloudiness and visibility. Options to improve the video camera systems are currently considered.

2.3.3 Reporting rules During the evaluation period some bugs in the software and configuration were identified and solved. Furthermore it was noted that sometimes an undesired delay occurred as a result of a combination of the integration time for the derived variable and the criteria for issuing a local special report, the so-called (AUTO) SPECIAL. The SPECIAL criteria used in The Netherlands were reviewed by a Task Force consisting of experts of KNMI and ATC. The recommendations of the Task Force which consisted of adjustments to thresholds and criteria for issuing a (AUTO) SPECIAL have been accepted and implemented by both organizations. For example the delay in reporting the improvement of the visibility has been specified as 5 minutes, by local agreement. In addition, a marked discontinuity was introduced in the calculation of the 10-minute averaged visibility in order to reduce the response time for sudden changes in visibility conditions.

2.3.4 Documentation issues Several issues reported during the evaluation of the AUTO METAR system at the regional airports could be traced to misunderstanding or misinterpretation of the results or reporting rules. For that purpose a fact sheet on SPECIAL criteria, and other documents explaining specific aspects of the AUTO METAR system, have been created. Special attention was given to the so-called visual parameters - visibility, clouds and weather. For these parameters the introduction of the AUTO METAR system not only led to different reporting rules, but also resulted in changes to the characteristics of the reported parameters. Hence the interpretation of these reported parameters needed to be adjusted in close cooperation with the users. In addition, so-called introduction sessions with air traffic controllers were held on location and aviation meteorologists were instructed in ATC operations.

15 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

3. Supervision of the AUTO METAR system

3.1 Monitoring of the AUTO METAR system status

3.1.1 Sensor status The advanced meteorological sensors perform a real-time check on the sensor status. For that purpose the signal of various sensor modules are monitored internally and warning or errors messages are generated when certain threshold values are exceeded. Internal check can also include measurement of the electronic noise, the stability of an emitter and the sensitivity of a receiver against internal references as well as the contamination of the lenses of optical sensors. The sensor status is passed on to the SIAM sensor interface and is available in near real-time together with the measured values itself. The sensor interface performs a real-time validation on the format of the received sensor messages. It also does internal checks on the quality of the received meteorological data. All parameters are checked against the climatological range. Depending on the expected variability of a meteorological unit under consideration additional checks on a too large variability or sudden jumps or a lack of variability in the observed unit are performed. When more meteorological units are available at a SIAM sensor interface, e.g. wind speed and direction or ambient and dew point temperature, then a cross validation of the units is performed. At the server system additional cross checks are performed. All these checks use warning and error thresholds and provide their results in near real-time to the airport server system. The warnings and errors are uniquely translated into a data quality which is available for each variable. Hence users as well as maintenance staff can see immediately when a warning or error status is issued.

3.1.2 Monitoring by service staff KNMI service staff monitors the sensor status of the entire network on a daily basis using the SIAM status information. Based on this information corrective or preventive maintenance is planned. Preventive maintenance is based on operational experience with the sensors. The calibration interval of the sensors ranges between 8 month for the humidity sensor and 36 month for the pressure and temperature sensor. A sensor is returned to KNMI just before the calibration is expired and replaced by another sensor. At the calibration laboratory it is first checked whether the returned sensor deviates from the reference within the allowed limits. The calibration interval is changed if the allowed limits are exceeded either too often or hardly at all. Next maintenance is performed on the sensor; it is calibrated and put into stock. Preventive maintenance in the field is also based on experience. At civil airports and airbases maintenance is performed every 2 months. The limiting factor here is cleaning of the lenses of the visibility sensors. Corrective maintenance is performed when either the sensor status or users indicate its necessity. If e.g. contamination of the lenses is reported then an additional maintenance visit is made to clean the lenses, or trained local staff is asked to perform the maintenance. A real-time check of the sensor status and output is made during and after maintenance has been performed.

3.1.3 Monitoring by operator Apart from the daily monitoring and planning of maintenance by service staff, an operator (the so-called “procesbewaker”) monitors the correct functioning of all crucial KNMI systems continuously. In case of malfunctions service staff can be alerted or requested. Action is taken according to the service level agreement. The importance and hence the priority of the maintenance of the sensor depends on the (expected) meteorological situation. If applicable corrective maintenance will be applied on a 24*7 basis. The operator has various tools to his disposal that facilitate the monitoring of the correct operation of crucial server systems and the availability of sensors and sensor information. Figure 5 shows some examples of screen available on a GDIS (Graphical DISplay) client system showing the status of sensors and MetNet systems at Rotterdam The Hague Airport.

16 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

Figure 5: Screen shots of MetNet systems giving an overview of the status of Rotterdam The Hague Airport. Going clockwise from top left they show: (i) central server at De Bilt (CIBIL) giving an overview of the entire MetNet including the server and sensor status at Rotterdam; (ii) ADCM server at Rotterdam showing the (derived) sensor data on a map (iii) and in the AUTO METAR and AUTO ACTUAL report generation screen (both use the convention black data is valid; magenta denotes warning or backup; red ? is faulty or missing; and italic is manually confirmed/adjusted); (iv) the final screen shows the incoming sensor data (again color code to indicate good, warning status and error of corrupt data). The latter also shows an overview of the status of the server systems at Rotterdam The Hague Airport.

3.2 Remote verification and complementation A continuous verification of the validity of the meteorological information is performed by the aviation meteorologist who has access to the 12 second meteorological data. The validation can be performed by using the information from other sensors at the airport, by consulting the video camera images at the airport, by using the information of nearby meteorological stations, by considering the general meteorological conditions or by contacting local staff of the airport or air traffic control. The aviation meteorologist has access to near real-time data from other airports, off-shore platforms and automated weather stations that are part of MetNet, as well as satellite and weather model information. Some sources of meteorological information are illustrated below.

17 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

3.2.1 Video camera images Figure 6 gives an example of the images obtained continuously from the video cameras at Rotterdam The Hague Airport. These images are available to service staff and the aviation meteorologist at the central weather room at De Bilt.

Figure 6: Illustration of the images obtained with the video cameras at 24 touchdown at Rotterdam The Hague Airport (top panels) and the camera before the threshold of runway 06.The yellow rectangle indicates the field of view of the video camera at 9 m equipped with a tele lens in the image obtained with the video camera at 2 m equipped with a wide-angle lens.

3.2.2 Local sensor information Figure 7 shows graphs of some meteorological variables centrally available in MetNet with an update every minute. Note that local information at Rotterdam The Hague Airport can also be viewed on a GDIS with a 12 second update by connecting to the ADCM server system. Other screens for displaying the local sensor information are shown in Figure 5.

18 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

Figure 7: Time series of the sensor data that is centrally available in MetNet with an update every minute. The screen has four panels showing information for the civil airports Amsterdam Airport Schiphol (top left), Groningen Airport Eelde (top right), Maastricht Aachen Airport (bottom right), and Rotterdam The Hague Airport (bottom left), respectively.

3.2.3 Regional sensor information Figure 8 presents a geographical overview of the visibility observations centrally available in MetNet with an update every 10 minutes. The graph shows the visibility at the civil airports, military airbases, platforms on the North Sea and automated weather observing stations that are all part of Meteorological Network. The information of other meteorological parameter can also be presented either geographical or in trend curves.

19 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

Figure 8: Illustration of the MetNet visibility information in the Netherlands that is centrally available and can be visualized geographically with an update every 10 minutes.

3.2.4 Other sources of information Apart from the information mentioned above the meteorologist has also access to satellite and weather model information. All information sources combined enables the meteorologist to create a mental image of the meteorological situation which is continuously checked against available information and updated.

3.2.5 Communication Contact between the meteorologist and local staff of the airport or air traffic control can be established in order to give information or feedback on the current and upcoming meteorological conditions. Note that the aviation meteorologist can overrule the sensor derived visibility, clouds and present weather values reported in the aeronautical reports orally or force sensors to fault so that the sensor data is disabled or, if applicable, the backup is used. The meteorologist also adds the TREND, a landing forecast with a validity of 2 hours, to the meteorological reports that are issued at least every half hour. Furthermore the meteorologist adds, if required, the runway state to the AUTO METAR report and can issue other reports (wind shear report, wind shear forecast and low level temperature inversion) as part of the AUTO ACTUAL and AUTO SPECIAL manually. There is regular contact between the aviation meteorologist and air traffic controllers. Each morning a briefing is held during which the expected meteorological situation of the next 24 hours is reported. Furthermore, contact is established when there is a significant deviation in the expected meteorological situation, in case there is a need for an update on behalf of a

20 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

specific event, in case of reasonable doubt concerning the meteorological information provided, or when there is a malfunction in the observation infrastructure.

3.2.6 Complementation Figure 9 illustrates the AUTO METAR system Rotterdam Airport, the redundancy of the system and backup procedures. The products added by the aviation meteorologist are indicated as well as the communication between the aviation meteorologist of KNMI at De Bilt and ATC at Rotterdam Airport. Note that CB/TCU is only provided by the meteorologist orally in case of a network failure between De Bilt and Rotterdam The Hague Airport. In normal situations the CB/TCU information is added automatically to the observation report based on an algorithm that runs at De Bilt and uses satellite and weather radar information.

Figure 9: Illustration of the AUTO METAR system at Rotterdam The Hague Airport in The Netherlands showing the relation between the aviation meteorologist at De Bilt and the air traffic controller at Rotterdam Airport. The effect of a failure of the data communication network is illustrated as well.

21 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

4. Introduction process of and experiences with the AUTO METAR system Rotterdam The Hague Airport is the alternate for Amsterdam Airport Schiphol. As such there is a relationship between the introduction of the AUTO METAR and the Mainport capacity. Hence the AUTO METAR system was introduced at Rotterdam The Hague Airport in line with standard ATC The Netherlands quality and performance analysis procedures. The time period between the first introduction of the AUTO METAR system at Groningen Airport Eelde and Maastricht Aachen Airport in May 2004 and the final implementation of the system at Rotterdam The Hague Airport in March 2011 was used to gain experience and to improve the system. Before the introduction safety assessments of the AUTO METAR system at Rotterdam The Hague Airport were performed and the user requirements with respect to the local routine and local special reports ((AUTO) ACTUAL and (AUTO) SPECIAL) were updated and implemented in the system.

4.1 Task Force Update Criteria (AUTO) SPECIAL The introduction of the AUTO METAR at Maastricht Aachen Airport and Groningen Airport Eelde resulted in several issues or shortcomings of the products AUTO ACTUAL and AUTO SPECIAL. An KNMI analysis showed that part of these shortcomings were not specifically related to the AUTO METAR system but arose from international coding practices and additional user requirements. ICAO Annex 3 explicitly states that the requirements should be established in consultation with the users: ‘’4.4.1 A list of criteria for special observations shall be established by the meteorological authority, in consultation with the appropriate ATS authority, operators and others concerned.’’

In 2009 a Task Force Update Criteria (AUTO) SPECIAL was created, in Dutch the “Werkgroep SPECIAL”. The Task Force (TF) consisted of a wide range of experts from KNMI and ATC The Netherlands (LVNL) in order to ensure that in discussions all aspects were covered. The aim of the TF was to investigate, optimize and harmonize the service provision for civil airports in The Netherlands by creating an optimal set of update criteria for the (AUTO) SPECIAL. The final report of the TF, which contained an integral advice and change proposals, was presented to and subsequently adopted by senior management of KNMI and LVNL in March 2010.

The starting point of the TF was that the current legislation and international regulations were leading. In particular the “Regeling Luchtvaartmeteorologische Inlichtingen 2006” (Dutch Aviation Law) and the ICAO Annex 3 “Meteorological Service for International Air Navigation”. The ICAO Annexes make a distinction between standards and recommendations. It is policy in The Netherlands to comply with both ICAO standards and recommendations.

In principle it is allowed to define national agreements exceeding the standards of the ICAO Annexes. In certain cases this is even promoted by ICAO and the Annex describes this as ‘up to local agreement’. This is specifically the case for the update criteria of the (AUTO) SPECIAL.

ICAO Annex 3; Appendix 3; 2.3.1 (standard): The list of criteria for the issuance of local special reports shall include the following: a) those values which most closely correspond with the operating minima of the operators using the aerodrome; b) those values which satisfy other local requirements of the air traffic services units and of the operators;

It is clear that the update criteria have to be established in consultation with the local users. The objective of the TF to ensure optimal update criteria for the (AUTO) SPECIAL is in line with the ICAO Annex 3 standards.

22 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

Each element of the (AUTO) ACTUAL was discussed in detail in the TF. The aim was to create one set of update criteria applicable for all civil airports in The Netherlands regardless whether the observations were automated or not. This was achieved except for Special VFR conditions, which are not applied at Amsterdam Airport Schiphol. The TF came up with several change proposals, based on consensus, for the update criteria of the (AUTO) SPECIAL, some of which are listed underneath:

• the inclusion of the variation from the mean wind speed (gust) when the maximum wind speed exceeds the mean speed by 5 knots or more; • the inclusion of light intensity precipitation; • an improvement of visibility after a 5 minute prolongation instead of 10 minutes; • an improvement of present weather after a 5 minute prolongation instead of 10 minutes; • an exception is TS, with or without showers, for which a 10 minute prolongation of improvement is required; • special VFR criteria for civil airports; • the usage of a TREND code based on the military coding practices; • the logging and archiving of (AUTO) ACTUALs and (AUTO) SPECIALs in a database in order to generate statistics on the content and update frequency of the (AUTO) SPECIALs; • an update of the back-up procedures for meteorological service provision for the civil airports; and • the generation of a fact sheet update criteria (AUTO) SPECIAL.

All alteration proposals were accepted by KNMI and LVNL management and have been implemented since. An evaluation of the renewed update criteria will take place in autumn 2011.

An overview of all update criteria of the (AUTO) SPECIAL for civil airports at The Netherlands is presented in Appendix A: Fact sheet update criteria (AUTO) SPECIAL.

4.2 Assessments of the AUTO METAR system

4.2.1 Technical assessment by KNMI In 2009 a technical assessment of the AUTO METAR system was performed by KNMI. The assessment took place after the introduction of the AUTO METAR system at Maastricht Aachen Airport and Groningen Airport Eelde. The assessment focused on the technical aspects of the AUTO METAR system including the process of how the system was specified, accepted and handed over to maintenance; whether the system met the specified quality and continuity requirements; and to what extend KNMI and suppliers were able to handle problems with the system. The assessment was performed by TriOpSys, an ICT company specialized in mission critical IT-systems. The assessment was based on interviews of KNMI staff involved in the development and operation of the AUTO METAR system and on available documentation. The main conclusions of the assessment were that the concept of the AUTO METAR system was considered suitable, but safety aspects of the system had been considered insufficiently during the development of the system. After the introduction of the AUTO METAR system during opening hours at Maastricht Aachen Airport and Groningen Airport Eelde KNMI worked on improvements of the system. However, it was unclear whether the problems identified by KNMI were really the essential issues for the users and whether they got the proper attention, resources and budget. This was caused by the way this process was organized by KNMI. The project team working on the handling of problems and improvements had not been given the necessary authority and support from KNMI management, and issues could not be prioritized. Furthermore the assessment indicated that it was unclear whether the technical problems identified by KNMI were really the essential issues for the users. The incidents reported by users often indicated a need for additional information or service. Hence it was recommended to identify and prioritize the problems of the AUTO METAR system from the point of view of safety and of the users. Also it was not clear from the available documentation whether the characteristics of the AUTO METAR system had been communicated adequately to the users.

23 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

4.2.2 Safety assessment by KNMI Under European Law KNMI as Air Navigation Service Provider (ANSP) for meteorology is not obliged to perform a safety assessment. Since the introduction of the AUTO METAR system at EHRD involved safety issues KNMI decided to perform a safety assessment. The safety assessment was lead by an external party, the Air Transport Safety Institute (ATSI) of the National Aerospace Laboratory (NLR) and comprised all users of aeronautical meteorological data at Rotterdam The Hague Airport and all possible error causes except malicious ones. The assessment was executed by several joint consultations with representatives of ATC, Rotterdam The Hague Airport, airlines as KLM, VLM and Transavia, the Aircraft Owners and Pilots Association The Netherlands (AOPA-NL), Dutch Air Line Pilots Association (DALPA), the Royal Netherlands Aeronautical Association (KNVvL or RNAA), the general aviation club Rotterdam and KNMI. During these sessions all possible hazards have been identified and complemented with hazards reported in the literature. The hazards have been classified as: 1. Meteorological report not available 2. Item in meteorological report not available 3. Item in meteorological report incorrect 4. Item in meteorological report unclear 5. Frequency of meteorological reports changes 6. Meteorological service causes additional communication, workload or distraction.

The hazards have been evaluated for all meteorological reports and all meteorological parameters. The users were asked to indicate the consequences and the seriousness of the hazards whereas KNMI provided information on the causes of the hazards, the details and performance of the AUTO METAR system and the current system with a local observer. The current situation at the airport served as the reference and was considered to meet the safety requirements. Hence the hazards of the AUTO METAR system were estimated qualitatively and relatively to the current system. A summary of the significance of the AUTO METAR system on the hazards and their effect is given below.

Hazard: Effect: 1. Report not available not significant, negative 2. Item not available not significant, negative 3. Visibility incorrect not significant, unknown Cloud cover incorrect not significant, unknown Cloud type incorrect uncertain, negative Weather incorrect not significant, negative Other items incorrect no effect 4. Item unclear no effect 5. Frequency changes not significant, unknown 6. Additional workload or distraction not significant, unknown

For all hazards the assessment did not just assume that individual sensors or elements of information were not available or incorrect, but what the impact would be taking into account the available information (observations, forecasts, warnings and advisories) provided as integral part of the meteorological service provision. It turned out that generally the AUTO METAR system, including backup systems and procedures as well as additional information sources, was still able to provide adequate information so that the impact was generally not significant. Only an incorrect cloud type reported by the AUTO METAR system was considered to have a potentially significant effect. However, the safety assessment was based on the old CB/TCU product that only used the lightning and weather radar reflectivity information and had a low detection rate. The recent inclusion of the METEOSAT information to the algorithm improved the CB/TCU product of the AUTO METAR system and by doing so reduced the impact to not significant. It is worth mentioning that the safety assessment included 8 recommendations. Improvement of CB/TCU was one of them. The other 7 recommendations have either been solved, are being developed or investigated, or are under discussion with parties involved.

24 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

4.2.3 Safety assessment by ATC (LVNL) As the introduction of the AUTO METAR system at EHRD is an alteration in the Air Traffic Service Provider (ATSP) operation for LVNL, its safety management system requires a safety assessment. Moreover, this is also required by European Law (EC 2096/2005, Annex II, section 3.2). The standard LVNL developed and applies for changes to the functional system is called a VEM-analysis. This investigates and assesses the impact on safety (in Dutch: Veiligheid), Efficiency and Environment (in Dutch: Milieu). The VEM-analysis for the AUTO METAR system EHRD has been applied in addition to the KNMI safety assessment for several reasons. The first is that LVNL is responsible for the ATC safety (unlike KNMI). This implies that LVNL only can assess whether ATC safety risks are acceptable or not. Moreover, the KNMI safety assessment did not meet all LVNL quality requirements for a safety assessment; the analysis of some hazards was considered as insufficient and/or incorrect. Finally, the impact on efficiency needed to be included. The final overall conclusion of LVNL for the AUTO METAR system was that it was acceptable once all mitigating measures as agreed between KNMI and LVNL had been applied. The analysis was performed internally by ATC and based upon the experiences with the AUTO METAR systems at Maastricht Aachen Airport and Groningen Airport Eelde. LVNL also included information and performance numbers provided by KNMI. The hazards and their impact on safety and efficiency have been determined by ATC by expert judgments. The assessment included hazards with too optimistic information (impact on safety) as well as too pessimistic information (impact on efficiency). It is interesting to note that the ATC hazards did not include CB/TCU information, whereas it included precipitation type near freezing temperatures. Furthermore the delay introduced by the AUTO METAR system on the information was considered a hazard. The outcome of the expert judgments indicated that hazards had either no effect on safety but small to medium effect on efficiency, or no to small effect on safety. Hazards related to availability had no impact if the AUTO METAR system met the requirements. Overall LVNL concluded that the AUTO METAR system was acceptable, once some mitigating measures had been implemented. These measures related to the implementation of recommendations of the Task Force SPECIAL, specifically items related to timeliness of reporting significant deteriorations and improvements. But also that ATC and KNMI agreed on a pro-active role of the aviation meteorologist in case of significant deviations from the expected meteorological situations or for specific events. This was facilitated by introducing a color video camera at the airport, by establishing communication between the aviation meteorologist and the ATC controller with the possibility to ask specific questions related to visibility to the ATCO, and by organizing regular daily telephone briefings between the meteorologist and the ATC controller.

4.2.4 Assessment of system performance The introduction of the AUTO METAR system at EHGG, EHBK and EHRD required information regarding the performance of the system in order to evaluate the safety aspects. This information was also provided to ATC for their VEM analysis. At first ATC required information on (i) the availability of the observations, if possible on element level, (ii) statistics on how often the aviation meteorologist or observer overruled or complemented the observations and (iii) how these changes are handled and recorded. In consultation with ATC it was decided to provide the required performance information in four phases.

4.2.4.1 Phase 1 availability of sensor information The first phase of the VEM analysis addressed the availability of sensor information at the civil airports. The information is readily available since it is part of the ISO certified quality management system of the infrastructure department of KNMI. The availability of the sensor data at airports is derived from the 1- minute data that is available in the data base of the central MetNet system at De Bilt. The availability of the civil airports should be at least 99.8% according to the service level agreement of airport systems. In the evaluation period January - September 2010 the availability was met at EHBK (99.93%) and EHGG (99.98%) but failed at EHRD (99.60%) since for 2 months the availability was very poor (April 98.71% and September 98.42%). In April the ceilometer was temporarily not available as a result of maintenance during which the backscatter profile information was made available to users for monitoring of the volcanic ash cloud and the ceilometer was moved from a location near the middle marker to the

25 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011 measurement field. In fact this maintenance took place when the airport was closed due to volcanic ash. In September a malfunction of the data communication with the relocated ceilometer emerged after replacement of the sensor.

It should be noted that both corrective and preventive maintenance reduced the availability although preventive maintenance is generally performed during marginal conditions and after informing the users. Also the calculated availability includes all sensors at the airport, including the ones that are only used for synoptical or climatological purposes. Furthermore the calculation does not account for the presence of backup sensors. Hence the availability of sensor information for the airport end-user will be larger than the values reported above. It should also be noted that the availability is not only determined by the sensor itself, but also by the infrastructure (power supply, sensor interface, data communication lines, multiplexers, splitters, server systems and network components). Finally it should be noted that the availability is calculated from the data in the central database and not from the data in the airport system itself. Hence communication problems between the airport and De Bilt might affect the numbers, although generally all sensor data is recovered during failovers of server systems and interruptions of network communications.

It was concluded that the AUTO METAR system meets the overall availability requirement for sensor information.

4.2.4.2 Phase 2 availability of (AUTO) METAR and elements The second phase of the VEM analysis addressed the availability of the AUTO METAR and its elements in relation to the performance of the METAR. The evaluation is based on the (AUTO) METARs that have been archived by the Message Switch System (MSS) at De Bilt. The MSS forwards the reports to the users. Note that the actual number of (AUTO) METARs received by ATC could be lower due to network problems. Since KNMI produces an (AUTO) METAR every half hour a total number of 13104 reports should be available for each airport in the evaluation period January - September 2010. 9 reports where entirely missing for all three airports during a malfunction at De Bilt, AUTO METARs were not available in time at the MSS in only a couple of occasions (6 and 9 events) whereas 204 METARs were not available in time. Furthermore it was observed that the number of corrected METARs (316) largely exceeded the number of corrected AUTO METARs (141 and 161).

18 events occurred with elements of the METAR missing (indicated by slashes “//”). 2 events are related to temperature, the other events are related to the visual parameters (cloud and weather). The latter events are probably caused by the fact that the “absence” of cloud and weather has not been confirmed by the observer, but was left at the missing default value. The AUTO METAR showed much more events with missing elements in the report, but most of them (187 for both airports) are related to CB/TCU. On these occasions the CB/TCU algorithm output was probably not available in time. This issue has been solved by optimizing the acceptable delay in which the CB/TCU information is made available. There were 6 and 5 events of missing weather and/or visibility in the AUTO METAR and EHGG had 2 additional events (one of missing humidity and one of missing wind and pressure).

It was concluded that the availability of the AUTO METAR and its elements is very high and is not inferior to the METAR after the timeliness of CB/TCU information had been improved.

4.2.4.3 Phase 3 availability of (AUTO) ACTUAL and (AUTO) SPECIAL and elements The third phase of the VEM analysis addressed the availability of the AUTO ACTUAL and AUTO SPECIAL and its elements in relation to the performance of the ACTUAL and SPECIAL. The evaluation is based on the (AUTO) ACTUALs and (AUTO) SPECIALs that have been archived by the central Aerodrome Database System (ADS) at De Bilt. The ADS stores all MetNet data and reports from all civil airports in The Netherlands for a period of 100 days. The data is extracted manually from the ADS on a daily basis and stored indefinitely on the Mass storage (Opslag) System (MOS). Archiving of airport data on the MOS started on January 1, 2009 for EHBK and EHGG and on January 1, 2010 for EHRD. The evaluation was again performed on data up to September 2010. Due to problems with the ADS and the

26 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

manual extraction, the archive is not complete (92% - 98% is available). The available data showed that zero AUTO ACTUALs were missing whereas 23 ACTUALs were missing. The 23 events are probably related to situations where a SPECIAL occurred before the ACTUAL was disseminated by the observer. The archive also showed that the number of AUTO SPECIALs relative to the number of AUTO ACTUALs (76% and 85%) largely exceeds the relative occurrence of SPECIALs (18%).

The availability of the elements in the (AUTO) ACTUALs and (AUTO) SPECIALs was determined. Out of a total of 14909 ACTUAL and SPECIAL reports 30 showed missing elements. 20 events missed weather information; 2 wind, 2 visibility, 2 cloud, 1 temperature, 1 dew point and 2 times the transition level, which is actually derived at EHAM and forwarded to the other civil airports. Out of a total of 52899 and 51946 AUTO ACTUAL and AUTO SPECIAL reports 63 and 61 showed missing elements. Here visibility is mainly missing (26 and 35) and furthermore wind (2 and 2), cloud (7 and 2), temperature (6 and 6), dew point (6 and 8), QNH (2 and 2) and the transition level (14 and 6). Temperature, dew point and pressure had no backup during the evaluation period and the backup for visibility at EHBK was only implemented for the AUTO METAR, but not for the AUTO ACTUAL and AUTO SPECIAL. Also note that the 2 events of missing wind at each airport occurred in combination with missing visibility and are probably the result of missing information about the runway in use which is the responsibility of ATC.

The (AUTO) ACTUALs and (AUTO) SPECIALs archived in phase 3 have also been used to evaluate the frequency of occurrence of specific weather phenomena. The AUTO ACTUAL reported significant clouds 50 and 60% of the time of which 5 and 6 % contain CB/TCU whereas the ACTUAL reported significant clouds 70% of the time 11% of which contain CB/TCU. A more striking difference was observed in the frequency of occurrence when either the visibility of the RVR is below 1500m. In the AUTO ACTUAL of EHBK this occurred 9.6% of the time, at EHGG 17% but in the ACTUAL of EHRD it occurred on 5.4% of the time. However, these numbers are largely affected by differences in local climatology.

It was concluded that the availability of the AUTO ACTUAL and its elements is very high and is not inferior to the ACTUAL after the backup sensors had been included. The number of AUTO SPECIALs largely exceeds the number of SPECIALs.

4.2.4.4 Phase 4 analysis of situations with non-representative values The fourth phase of the VEM analysis addressed the number of situations where the values of the (AUTO) ACTUAL and the (AUTO) SPECIAL were perceived as being non-representative, in particular those cases were the values were too optimistic. The evaluation is based on the shift reports of KNMI and ATC, and user complaints from ATC and pilots for EHGG, EHBK and EHRD for the years 2009 and 2010. This analysis proved to be the most difficult as it contains a subjective element of analyzing the reported incidents on weather observations. Each incident has been analyzed and categorized into visibility (VIS), clouds (CLD), weather (WX), CB/TCU, wind (FF) and network or system. In some cases it was not possible to determine whether the observation was representative or not. For example visibility perceived from the ATC tower can differ compared to what a pilot perceives, or what is measured at a height of 2.5 meters by the forward scatter meter. Furthermore it cannot be excluded that there could be more situations with non-representative values then those reported by air traffic controllers, pilots and meteorologists. The current understanding is that the numbers provided in Figure 10 constitute a realistic view of the number of situations where there is doubt on the accuracy of the weather observations. Note that the numbers of EHRD relate to manual observations, whereas EHGG and EHBK deal with automated observations.

27 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

2009 2010 Total Station Element Number of reports Number of observations Number of reports Number of observations Number of Number of observations in shift log more positive then in shift log more positive then reports in shift more positive then observation of ATCO observation of ATCO log observation of ATCO EHGG VIS 12 3 21 2 33 5 CLD 1 0 5 1 6 1 WX 3 0 2 0 5 0 TCU/CB 1 1 1 0 2 1 FF 0 0 0 0 0 0 NETWORK/SYSTEM 3 1 4 TOTAL 20 4 30 3 50 7

EHBK VIS 4 1 2 0 6 1 CLD 0 0 6 2 6 2 WX 2 0 1 0 3 0 TCU/CB 1 1 1 0 2 1 FF 1 0 0 0 1 0 NETWORK/SYSTEM 3 4 7 TOTAL 11 2 14 2 25 4

EHRD VIS 2 0 2 0 4 0 CLD 1 1 1 0 1 1 WX 1 0 5 0 6 0 TCU/CB 0 0 0 0 0 0 FF 0 0 0 0 0 0 NETWORK/SYSTEM 1 0 1 TOTAL 5 1 8 0 12 1

Figure 10: Overview of situations of perceived non-representative values of observations for the airports EHGG, EHBK and EHRD over the period 2009 and 2010 based on shift reports and user complaints.

The available data showed that in a two year period there are 50 (EHGG), 25 (EHBK) and 12 (EHRD) occasions where observations were perceived as not being representative. The majority of the occasions are related to visibility, followed by clouds and weather. The visibility cases are partly due to fog patches around the sensor whereas the runway itself is clear of fog, see also section 4.3.

For the safety analysis of LVNL especially the situations where the observations are more positive (impact on safety) than the actual situation as perceived by the air traffic controller are of interest. The analysis showed that this is the case in 7 (EHGG), 4 (EHBK) and 1 (EHRD) situations over the two year period.

An ongoing effort is put into the logging and analysis of situations where the observations reported by the AUTO METAR system deviates from the situation experienced by local users. Section 4.3 provides more details.

4.3 Stakeholder Consultation and User Satisfaction The AUTO METAR system at Rotterdam The Hague Airport was introduced with great care towards both internal and external stakeholders. It was apparent that the interests of the stakeholders were not always alike. For KNMI observers the introduction of automated observations implicated the disappearance of their function. Air traffic controllers faced a change in their working methods and feared a deterioration of the provided meteorological services as air traffic controllers prefer observations by men in which instant consultation is possible. The airlines, who finance the meteorological service provision, were keen to see the introduction reflected in a reduced cost scheme of KNMI.

The regulator, the Directorate General of Civil Aviation of the Ministry of Infrastructure and Environment, assigned KNMI to introduce the AUTO METAR system at Rotterdam The Hague Airport. However, it was made clear from the beginning that the introduction had to be a meticulous process involving the input and commitment of all stakeholders, and in collaboration with the National Supervisory Authority (NSA).

All external stakeholders had been informed of the introduction and the characteristics of the AUTO METAR system at Groningen Airport Eelde and Maastricht Aachen Airport in 2007. With regard to the introduction at Rotterdam The Hague Airport users were informed and involved in many different ways. At the yearly technical and financial user consultation meetings the AUTO METAR system was addressed specifically and ample time was created on the agenda to discuss and evaluate the system. Users were informed on the AUTO METAR system via an Aeronautical Information Circular and

28 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

information on the website of KNMI. The direct stakeholders of Rotterdam Airport were invited to take part in the safety assessment of KNMI which was lead by an external party, the Air Transport Safety Institute of the National Aerospace Laboratory (ATSI-NLR). The assessment was executed by several joint consultations with representatives of ATC, Rotterdam The Hague Airport, airlines as KLM, VLM and Transavia, the Aircraft Owners and Pilots Association The Netherlands (AOPA-NL), Dutch ALPA, the Royal Netherlands Aeronautical Association (KNVvL or RNAA), the general aviation club Rotterdam and KNMI.

Furthermore, intensive consultation has taken place between KNMI and LVNL (ATC) and Rotterdam Airport on the changes in service provision. As input for the discussions served the outcome of the TF SPECIAL, the safety assessments of KNMI and LVNL, and the received user complaints. KNMI has organized dedicated communication sessions on location for ATC explaining the characteristics of the AUTO METAR system. Sessions have also been organized at the KNMI premises at De Bilt for air traffic controllers and airport staff in order to meet the meteorologists who monitor the AUTO METAR system, and to see the observation equipment and the systems in use. The outcome of the consultation lead to the introduction of a daily operational briefing via telephone conference between the meteorologist and the air traffic controller at Rotterdam Airport. Furthermore, backup procedures have been adjusted and the reporting of incidents was formalized and structured in a slightly different way. Most important, however, is a pro-active approach of both meteorologist and air traffic controller whenever there are reasons to believe that the observation may be not representative. Important to recognize is that the automated meteorological observation reports are part of the meteorological service provision comprising of observations, forecasts, warnings and advices, and should be seen from within this context. The automated meteorological observations are not fully able to determine all weather phenomena and descriptors, however, the meteorologists located at De Bilt are available for briefing or consultation.

A lengthy discussion arose with LVNL on the reporting of reduced visibility due to fog patches around the sensor whereas the runway itself is clear of fog. These events occur regularly as the sensor is, according to regulations, situated on a grass surface. Based on observation principles KNMI was of the opinion that the reduced visibility should be reported, especially as the fog patches may roll onto the runway and result in a sudden reduction of the visibility on the runway. It was the understanding of the air traffic controller that the visibility should be reported as seen above the runway itself. It turned out to be that some observers had brought this into a practice, though not in line with standing KNMI regulations. In the following discussion KNMI made it clear that it is possible to change this observing policy, but only after carrying out an impact study in collaboration with users and authorized by the national supervisory authority. For the moment it has been agreed to report the reduced visibility, so not changing the current system, and to evaluate this procedure in 2012.

In general it became apparent that the users were not aware of the meteorological guidelines and regulations on how to perform weather observations at airports. Likewise, it was not always clear for the meteorological community how the meteorological information was used in an air traffic management environment. Learning more about each other roles and responsibilities during the consultation process provided essential insights which lead to improvements of the AUTO METAR system and as such to a better acceptation of the automated observations.

It was interesting to notice that views among stakeholders on the impact of meteorological conditions varied quite significantly. For instance, within KNMI the detection of CB/TCU was considered as critical and crucial to the introduction of the AUTO METAR system. The outcome of the NLR safety assessment showed that the users gave less impact to the consequences of a possible deteriorated detection of CB/TCU. First of all, it was recognized that the manual detection of CB/TCU by an observer is not perfect, for instance during night time, or in situations with embedded CBs. Most important turned out to be that CB/TCU conditions are forecasted well in advance, or are communicated via Airman Meteorological (AIRMET) and Significant Meteorological (SIGMET) reports in case of sudden and

29 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011 unforeseen CB/TCU conditions. In general VFR traffic will not be operating under CB/TCU conditions, and IFR traffic has means to deal with CB/TCU conditions in flight.

Similarly, the detection of Blowing Snow and Drifting Snow was given a much higher weight by users then was foreseen within KNMI. As a result KNMI started a dedicated study into the possibilities to come up with an improved detection of these phenomena. The outcome of the study came available in April 2011. An analysis over the period 2001-2010 showed that there has only been one situation with drifting and blowing snow at Amsterdam Airport Schiphol. Furthermore, only in the USA an algorithm is used for the automated detection of blowing snow based on a combination of observations of temperature, wind speed, present weather, visibility and clouds. The investments to tune this algorithm for The Netherlands situation is costly and does not seem justified given the small number of blowing snow events. As there are no off the shelf solutions for the automated detection of blowing snow KNMI has decided not to report blowing and drifting snow in the AUTO METAR system and has informed the users likewise. However, investments in improved camera systems at the airport continue, and in the regular contact moments between the meteorologist and the air traffic controller these weather conditions will be given proper attention.

From both ATC and airline perspective the detection and reporting of freezing fog and freezing rain was considered to be extremely important. In general discrimination of freezing fog and freezing rain around the touchdown zone by an observer at a relative distance of the runway is challenging. The AUTO METAR system uses an algorithm to report freezing rain and freezing fog. A study, investigating new measurement techniques, is initiated to improve the detection of freezing fog and freezing rain.

With regard to the monitoring of the AUTO METAR system by the meteorologist at De Bilt the choice was made to overrule the information of visibility, clouds and present weather only orally instead of altering the observation reports themselves. This was done for evaluation purposes in order to make a clear distinction between the output of the automated system and any added value provided by the meteorologist. This decision may be reconsidered in the future.

In general there have not been many complaints from users as airlines, pilots and airports on the quality of the automated observation reports. As was shown in section 4.2.4.4. up to 30 incidents per airport per year were reported mainly by air traffic controllers. The fact that the majority of the complaints were received from ATC is in itself not surprising as the air traffic controller is one of the key users of the automated reports. The number of complaints decreased significantly in the last year. This was partly due to the improvements of the AUTO METAR system and the fact that users became more familiar with the characteristics of the automated observations.

4.4 Handling of user complaints The AUTO METAR system, like all operational systems of KNMI, is part of the ISO certified quality management system of KNMI. This means that the performance of the system is monitored and reported and set against the requirements agreed upon with the users. Maintenance, preventive and corrective, on the system is handled according to fixed procedures and within the agreed time interval. Any malfunctions or complaints about the system are logged and handled or addressed by the appropriate staff at KNMI.

During the implementation and evaluation of the AUTO METAR system at EHBK and EHGG various communication channels involving different KNMI departments handled user complaints and incidents concerning the AUTO METAR system. KNMI staff involved were e.g. service staff of the instrumental department, functional maintenance staff of the MetNet systems, account managers, operators, observers, meteorologists and other people who were or had been involved in the AUTO METAR system. All these people had their legitimate and specific role in the AUTO METAR system and external contacts with their counterparts at the user’s organizations. However, due to this diversity there was no general overview of the user’s complaints and no uniformity in handling and sometimes even inconsistent responses to the users. When incidents were handed over between departments within

30 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

KNMI the exact nature of the complaint was often open to interpretation or unclear and tracking them back to the user to get more detail was time consuming and often proved impossible. Monitoring of the status of incidents and possible follow up actions was also sometimes lacking.

In order to improve the handling of incidents they have been channeled in a structured way. All incidents are logged in a central system at KNMI as well as at the users side (ATC). All incidents are accompanied with detailed information on place, date/time, specific system/sensor or report/element involved, description of the problem and how it was observed or what was expected. The incidents were at first analyzed and classified by a project team and currently this is handled by the account manager aviation and handed over to the appropriate department/staff within KNMI. The replies to the users were also channeled through the central contact points. The status of all open incidents was discussed with ATC during regular meetings. At first these meeting occurred every 2 months, but currently every 6 months. Items have also been prioritized together with the users and deadlines have been set.

Since this setup for the handling of incidents related to the AUTO METAR system was introduced in August 2007, nearly 200 incidents have been handled. The incidents revealed several software bugs in the AUTO METAR system, which were then solved with their respective prioritization. Often items required clarifications or documentation of the AUTO METAR system or its characteristics. The number of user complaints decreased drastically in the last year. This was partly due to changes to the AUTO METAR system and the fact that users became more acquainted with the characteristics of the automated observations. However it was decided to keep the handling of incidents related to the AUTO METAR system as it is and make it part of the service level agreements between KNMI and LVNL.

4.5 Future improvements Several further improvements to the AUTO METAR system have been mentioned in previous sections. This section gives an overview of ongoing and outstanding improvements in random order.

4.5.1 Video Cameras The quality of the video camera images is rather poor. As a first step towards a new system the technical requirements of the video cameras are currently being drafted in collaboration with the aviation meteorologists. Also a pilot has been performed with video camera systems with zoom, pan and tilt functionality that is considered for use at the automated weather observing stations. The intended users of this system are technical staff, inspection and data validation, but the system has also been made available in the central weather room for evaluation. Note that video camera can also play a role in the items related to representativeness of clouds, visibility and present weather observations.

4.5.2 Representativeness of cloud observations In certain situations the users experience the lacking spatial representativeness of the cloud observations with a ceilometer as a delay. This delay is caused by the fact that clouds need to be directly overhead the ceilometer before they can be detected and the cloud algorithm introduces an additional delay by evaluating the data over a 10-minute time period. The usage of multiple ceilometers at the airport does not improve the overall results significantly as a previous study at Amsterdam Airport Schiphol showed. In the near future the added value of ceilometers in the intermediate vicinity (the four fog stations around Schiphol are located about 15 km from Schiphol and have recently been equipped with ceilometers) will be evaluated.

Possible future developments in this respect need to be settled with the users. Automated systems that are currently commercially available and that can provide spatial information of clouds have limitations so that they cannot be considered for unattended use. The NubiScope, a scanning pyrometer that determines the presence of clouds from the sky temperature observed at each orientation, gives accurate information on cloud cover but the height information is rather poor, whereas the Total Sky Imager (TSI, a digital camera pointed downward at a hemispheric mirror that determines the presence of clouds from the color of sky at each pixel) gives accurate information on cloud cover only during day time

31 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

with no height information. Both systems could possibly be considered as a first guess cloud product or serve as an alert after which the aviation meteorologist, using e.g. video camera images and other information sources, can complement the cloud information when necessary.

4.5.3 CB/TCU cloud types The CB/TCU product has recently been improved by using lightning and weather radar reflectivity data in combination with METEOSAT information. The weather radar reflectivity data is currently used at a 2.5 by 2.5km resolution. The currently available 1 by 1km resolution data cannot be used for CB/TCU since the clutter leads to many faulty events. The removal or suppression of the clutter in the 1 by 1km resolution radar reflectivity data is under investigation.

Another development that is considered - but is also largely affected by clutter in the radar reflectivity data, the limited range of the weather radar and the lightning detection system, and the lack of a reference data set - is the introduction of the CB/TCU product for the North Sea platforms.

4.5.4 Representativeness of visibility observations Visibility sensors perform a point measurement. Hence the horizontal visibility reported in the METAR can differ significantly from the value reported in the AUTO METAR. The information of all visibility sensors at an airport can be used in order to derive a median visibility, which is more representative of the situation at an airport. This median can be reported in the AUTO METAR. The automated local routine and local special reports, which require the visibility that is representative of the touchdown zone, cannot use the median visibility obtained from several sensors unless they are located near the touchdown zone.

4.5.5 Visibility observations corrected for insects The forward scatter visibility sensors are affected by the presence of insects in the measurement volume. Filtering out the spikes caused by insects in the raw sensor signal before calculation of the MOR diminishes the visibility reductions significantly. The insect filtering will be implemented in the firmware of the visibility sensor at civil airports.

4.5.6 Representativeness of weather observations The visibility sensors are in fact present weather sensors. Hence the precipitation type, and therefore the weather, is also assessed in only a small measurement volume. The information of all present weather sensors at an airport can be used in order to derive the weather that is more representative for the situation at an airport. The current weather algorithm can be used to determine which weather phenomena will be reported. Since there is currently no practice of runway dependent weather, the weather of the combined present weather sensors at an airport can be used in the AUTO METAR as well as in the automated local routine and local special reports.

4.5.7 Improved precipitation type discrimination The present weather sensor has limitations regarding the discrimination of solid precipitation, particularly in situations with mixed precipitation and light precipitation intensities. Also the detection of freezing precipitation is poor. In order to mitigate the first problem a so-called disdrometer has been selected and is currently being evaluated at Amsterdam Airport Schiphol. The aim is to enhance the discrimination of the precipitation type of the present weather sensor by using the information of the disdrometer. The issue of freezing precipitation is addressed at Maastricht Aachen Airport were additional temperature sensors have been placed near the present weather sensors along the runway in order to investigate the sensitivity of the discrimination to local temperature differences.

32 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

4.5.8 Content of reports The content of the local routine and local special reports, i.e. the (AUTO) ACTUAL and (AUTO) SPECIAL, have largely been derived from previous systems that were owned by ATC. Recently the (AUTO) SPECIAL criteria have been updated in close collaboration with the users. Currently the usage and usefulness of all the elements contained in the (AUTO) ACTUAL and (AUTO) SPECIAL are being investigated. Issues are e.g. which elements are actually used; whether parameters other than wind and visibility should also be runway dependent; which elements should differ for the separate arrival and departure local special and local routine reports that are issued at Amsterdam Airport Schiphol; and whether it is necessary to include a TREND in the local special and local routine reports.

4.5.9 Documentation Several documentations have already been taken care of in the past regarding the AUTO METAR system. They are mostly pertained to the differences that could be expected between automated en manual meteorological reports. More recently the focus is more on the overall system and procedures. Currently a document is being drafted in which the entire visibility chain of the AUTO METAR system is presented. The paper covers all aspects from the sensor, sensor calibration and maintenance, to the processing and backup rules and procedures of the variables that are made available to the user. During the evaluation several incidents were related to misconceptions concerning visibility, so it was agreed that the entire visibility chain should be documented. Another document that is being prepared is related to the spatial representativeness of the cloud observations. For the near future a detailed description of the configuration of the processing in the MetNet airport server system is planned including an explanation of the general rules that have been applied.

33 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

5. Conclusions and lessons learned This paper provides an overview of the AUTO METAR system used in The Netherlands. At off-shore structures at the North Sea the AUTO METAR system is operated without redundancy and produces the AUTO METAR automatically every half hour without human intervention. At civil airports the user requirements are more demanding. Hence the AUTO METAR system at civil airports not only includes the entire technical infrastructure used for the automated production of all meteorological aeronautical observations reports including the local routine and local special meteorological reports as well as the AUTO METAR. The system also includes backup sensors, systems and procedures. Furthermore all reports are monitored and complemented by a remote meteorologist who has remote access to local sensor information and video camera images. The communication between the meteorologist and ATC covers daily briefings; pro-active updates when necessary and notifications of sensor or system maintenance or malfunction. Experiences of the performance of the AUTO METAR system and acceptance by ATC are reported, including the process of how this was achieved.

The introduction of the AUTO METAR system was the outcome of a lengthy process of automation of aeronautical weather observations which dated back to the early nineties of the previous century. Notwithstanding that long preparation period the introduction process proved to be extremely difficult and met severe resistance of internal and external stakeholders. In hindsight a significant part of the resistance from stakeholders could have been prevented if the process had been organized in a different way, and definitions had been made clear from the beginning. Underneath a short summary of the most important general lessons learned is given.

• The use of the word “AUTO” in “AUTO METAR” may and has been interpreted by users as if a system will be introduced that produces an observation without technical supervision and without monitoring by a meteorologist at a remote location. This is not the case for the AUTO METAR system employed at civil airports at The Netherlands. However, some users had this impression and it required a significant communication effort to correct this perception. • The use of the word “AUTO METAR”, a product, to describe the “AUTO METAR system” has lead to severe confusion and concomitant discussions. This has led to meetings were the AUTO METAR was discussed and participants were discussing different products unknowingly. And for example led to reports on inaccurate AUTO METARs by users where in reality the product AUTO ACTUAL or AUTO SPECIAL was meant. It is important to discriminate between the container concept AUTO METAR system and the observation product AUTO METAR. • The content and method of production of the “AUTO METAR”, and subsequent products as AUTO SPECI, AUTO ACTUAL and AUTO SPECIAL, may vary in several ways, even within one country. This is not clear to a user when only the coded product is available. The automated observation can contain less detail compared to a manual observation, what can only partly be seen from the code. However, whether for example remote supervision or remote monitoring is applied or if and what algorithms are used to generate CB/TCU information cannot be deduced from the report. • The initial focus of the automation within KNMI was on the METAR and SPECI. However, the real challenge in automating aviation meteorological observations lies not in METAR but in the ACTUAL and SPECIAL since this information is used operationally by ATC and pilots at the airport. This has lead to unforeseen issues when introducing AUTO ACTUAL and AUTO SPECIAL and required additional changes to the system. • Initially the AUTO METAR system setup was such that first the (AUTO) METAR and (AUTO) SPECI was generated and next the (AUTO) ACTUAL and (AUTO) SPECIAL. Since the (AUTO) SPECI are no longer required when producing half hourly (AUTO) METAR it became necessary to switch the order.

34 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

• A general assumption seemed to be that an observation performed by an observer is by definition perfect. And as a consequence the manual observation served as the reference when assessing the automated observation. However, this assumption is not always valid, as is shown in several studies. For example the eye sight of an observer may be impaired without the observer noticing this in daily life, and the location of the observer and the illumination levels at night time are not always optimal. Manual observations are not only always perfect, but in addition it is no guarantee that the meteorological reports are always complete and available in time. • For local users the change from a manned situation to the AUTO METAR system needs to be handled with care as air traffic controllers are used to instant consultation whenever there is a discrepancy in the perceived weather conditions and the reported weather. ATC feared for a deterioration in product quality as well, but it was universally accepted that the introduction of the AUTO METAR system on a location that would otherwise provide no or reduced information had added value. • It is important to realize that the introduction of automated observations is an emotional process. The process and discussions related to the introduction and acceptance of the AUTO METAR system sometimes included emotional aspects that could only be overcome with care. The initial reaction of KNMI was to provide more technical information on the system, which is not surprising given the technical and scientific nature of KNMI. However, change management and open and transparent discussions with LVNL (ATC) based on a partnership relation was needed and has been established in order to overcome the multiple views on the introduction of automated observations, and in order to gain management commitment. • In some cases the actual issues of the users concerning the (AUTO) METAR system differed from the perception of internal stakeholders of KNMI. Open discussions between the parties involved clarified the crucial issues and were beneficial to the acceptance of the AUTO METAR system and its quality. • The TF SPECIAL resulted in several improvements of the update criteria of the weather observation reports and provided a better insight of the use of the meteorological products by ATC. In essence the majority of these improvements were not directly related to automated observations, but were perceived to be related when the AUTO METAR system was introduced. In hindsight this process of updating and fine tuning the update criteria in dialogue with the aviation users should have been performed earlier as part of the regular account management process. It has been agreed upon with ATC that the update criteria will be assessed at least once every year. • The communication about the AUTO METAR system to aviation stakeholders is of major importance and a key factor in the acceptance of automated weather observations by users. In the process of ‘marketing’ the AUTO METAR system it is important to realize that common weather observation principles and the method of how weather observations are performed at civil airports are not widely known by users as for example air traffic controllers (not being meteorologists). It is essential that both the air traffic management and the meteorological community understand each others roles and responsibilities in order to create tailored solutions for the provision of meteorological services. • In the process of improving the automated meteorological observations the experience and knowledge of the professional observers can be extremely helpful. However, it has been proven difficult to gain commitment from all observers given the fact that they were being made redundant. Furthermore, the fact that some observers provided additional services, not in line with KNMI regulations, resulted in the situation that additional requirements of ATC were not identified at an early stage and only became apparent when the AUTO METAR system was introduced. Intensified communications with external stakeholders on the content and quality of the service provision is a way to overcome this.

35 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

• A key element of the acceptance of the AUTO METAR system by ATC was a pro-active approach of both meteorologist and air traffic controller whenever there are reasons to believe that the observation may not be representative. A daily telephone briefing between the meteorologist and the air traffic controller has been initiated, and backup procedures have been updated and adjusted. It is important to recognize that the automated meteorological observation reports are part of the meteorological service provision comprising of observations, forecasts, warnings and advices, and should be seen from within this context. The automated meteorological observations are not fully able to determine all weather phenomena and descriptors, however the meteorologists located at De Bilt are available for briefing or consultation. • Due to the point measurement observation techniques and used algorithms information on clouds and visibility there may be some delay in the update frequency of these elements in the automated weather reports. Research is ongoing in order to improve the quality of the automated weather observations. • Assessments by external parties have proven to be a useful tool for keeping in touch with the outside world by identifying essential issues that were not related to meteorology and technology. They have also facilitated that proper attention and resources were given to communication and documentation as well as the need to provide relevant performance numbers crucial for the evaluation of the AUTO METAR system. • The characteristics of the automated weather observations differ from manual observations and as such may require a change in the operating procedures of aviation users. In this respect it is advised to ICAO to take this into account in future updates of the ICAO annexes and guidance material. • The AUTO METAR system concept is new and could not be verified against experiences with similar systems elsewhere. In order to share experiences KNMI has had several sessions with international parties, both meteorological institutes and ATC organizations, and shared experiences within bodies like the MET Alliance, WMO-CIMO, ICAO-AMOFSG en ICAO-EANPG-METG.

36 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

6. References Aeronautical Information Circular Series A The Netherlands (AIC-A 13/10 24 MAR 11 revised): Automated aeronautical meteorological observations at civil airports Haij M. de: Inventarisatie van de mogelijkheden voor de automatische detectie van driftsneeuw ten behoeve van AUTO METAR, versie 1.0, 15 april 2011 ICAO: Meteorological Service for International Air Navigation, Annex 3, 17th edition, Montréal, Canada, July 2010 Klein Obbink, B.: Safety Assessment AUTO METAR, NLR Air Transport Safety Institute: NLR-CR-2010- 278 KNMI/LVNL: Advies van de Werkgroep Updatecriteria ACTUAL/SPECIAL (WG SPECIAL), versie 1.0. 31 maart 2010 KNMI: Service Level Agreement Weer – Infrastructuur, 15 april 2010 KNMI: Data Analyse “AUTO METAR”, versie 1.1, 8 november 2010 Koetse W. en J. Sondij: KNMI data analyse (AUTO) ACTUAL en (AUTO) SPECIAL ten behoeve van LVNL VEM analyse AUTO METAR, versie 1.0, 8 december 2010 LVNL: VEM Analyse Introductie AUTO METAR Rotterdam, D/R&D 09/D S&P 10 018 versie 1.0, 14-02- 2011 TriOpSys: Audit Verslag AUTO METAR, F815, versie 1.9, 06-07-2009 Wauben, W. and J. Sondij: The meteorological observation infrastructure at Schiphol, Information Paper METG, 19 September 2009 Wauben, W.: Visibility Chain Rotterdam The Hague Airport, internal document, KNMI, June 2011 Wauben, W.: Evaluation of the Vaisala FD12P 1.91S firmware with insect filtering, KNMI Technical Report No. 316, February 16, 2011 Wauben, W. et al.: Laboratory and Field Evaluation of the NubiScope, Paper 1-8, WMO, Technical Conference, Helsinki, 2010 Wauben, W. et al.: On the Generation of an Optimized Fractional Cloudiness Time Series using a Multi- Sensor Approach, Paper 2-4, WMO, Technical Conference, Helsinki, 2010 WMO: Guide to Meteorological Instruments and Methods of Observation, 7th edition, WMO No. 8, Geneva, Switzerland, 2008 WMO: Manual on Codes, International Codes Vol. I.1 Part A – Alphanumeric Codes, 2010 edition, WMO No. 306, Geneva, Switzerland, 2010

37 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

Appendix A: Fact sheet update criteria (AUTO) SPECIAL

Item Variable Definition The criteria for issuing a SPECIAL or AUTO SPECIAL report The runway(s) in use for which the Runway in use Runway in use A change of the runway(s) in use, including opening/closing of runway(s). meteorological observation report is valid.

The 2 minute average wind direction in A change in the mean wind direction of 30 degrees or more, the mean wind speed before and/or Wind Direction degrees with respect to true North. after the change being 10 knots or more. Speed The 2 minute average wind speed in knots. A change in the mean wind speed, being an increase or a decrease, of 10 knots or more. The most backed and veered wind direction A difference in the most backed and most veered wind direction of 60 degrees or more, the mean Directional over the last 10 minutes in degrees with wind speed being 3 knots or more AND the directional wind variation is not reported in the variation respect to true North. previous report. The maximum wind speed (gust) and/or the minimum wind speed (lull) differs 5 knots or more from the average wind speed AND the maximum or minimum wind speed is not reported in the previous report. The gust (maximum wind speed) and the lull

Speed variation (minimum wind speed) over the last 10 minutes in knots. A difference in the maximum wind speed (gust) and/or the minimum wind speed (lull) of 5 knots or more in comparison to the previous reported maximum wind speed (gust) and/or

minimum wind speed (lull), the mean wind speed before or after the change being 7 knots or more. improvement deterioration The horizontal visibility at the touch down After a 5 minute prolongation of a visibility Immediately (bearing in mind processing time) Visibility Visibility (VIS) zone (TDZ) of the runway in use over the last value when reaching or exceeding a when the visibility drops below a visibility 10 minutes in meters. visibility threshold. threshold. The horizontal visibility thresholds are 0800, 1500, 3000, 5000 and 8000 meters.

38 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

Item Variable Definition The criteria for issuing a SPECIAL or AUTO SPECIAL report Onset or cessation of the following weather phenomena: • freezing fog (FZFG); • low drifting (< 2 meters) dust (DU), sand (SA) or snow (SN): DR.. (no intensity !) (not provided in AUTO SPECIAL); • blowing (2 meters or higher) dust (DU), sand (SA) or snow (SN): BL.. (no intensity !) (not provided in AUTO SPECIAL); The observed present weather phenomena at • thunderstorms with or without precipitation: TS or (+, ,-) TS.. (..) (..); Present Present weather an aerodrome that may have an effect on • squall (SQ); weather aviation. A maximum of three present • funnel cloud or water spout on (+) or above ( ) ground or water surface: (+, ,) FC (not provided weather groups are reported. in AUTO SPECIAL). Onset, cessation or change in intensity of the following weather phenomena: • light, moderate or heavy freezing precipitation: (+, ,-) FZ .. (..); • light, moderate or heavy drizzle (DZ), rain (RA), snow (SN), unknown precipitation (UP) with

or without showers: (+, , -) (SH).. (..) (..); • light, moderate or heavy ice pellets (PL), small hail or soft hail (GS) or hail (GR) with or without showers: (+, , -) (SH).. (..) (..). Changes between precipitation types RA and DZ, without change of intensity, do not lead to the issuance of a SPECIAL or AUTO SPECIAL (e.g. -RA becoming -DZ or -RADZ becoming -DZRA) improvement deterioration After a 5 minute prolongation of a weather Immediately (bearing in mind processing time) improvement. Exception is TS, with or when the weather deteriorates and reaches a without showers, for which a 10 minute weather threshold. prolongation of improvement is required.

The clouds (cloud amount, height of cloud base and cloud type) of operational significance and representative of the approach area. A maximum of four cloud Clouds Clouds groups are reported.

When no clouds of operational significance (cloud base 5000 ft or more and no CB/TCU) are observed, the abbreviation "NSC" is used.

39 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011 Item Variable Definition The criteria for issuing a SPECIAL or AUTO SPECIAL report improvement deterioration The amount of cloud coverage reported in After a 10 minute prolongation of "NCD" (AUTO METAR) or "NSC" improvement which is defined as when the Immediately (bearing in mind processing time) Cloud coverage (METAR) (0 okta), "FEW" (1 to 2 oktas), cloud coverage of the cloud layer(s) below when NCD, NSC, FEW or SCT changes to BKN "SCT" (3 to 4 oktas), "BKN" (5 to 7 oktas) or 1500 feet changes from BKN or OVC to or OVC with a cloud base below 1500 feet. "OVC" (8 oktas). SCT, FEW, NSC or NCD. improvement deterioration After a 10 minute prolongation of improvement which is defined as when the Immediately (bearing in mind processing time) The height of the cloud base in hundreds of height of the lowest cloud layer, with a when the height of the lowest cloud layer with a Cloud height feet. height below 1500 feet AND with a coverage of BKN or OVC, drops below one or coverage of BKN or OVC, reaches or more cloud base thresholds. exceeds one or more cloud base thresholds. The height of the cloud base thresholds are 100, 200, 300, 500, 1000 and 1500 feet. Cumulonimbus clouds (CB) or Towering The observation or dissipation of CB and/or TCU clouds at any height irrespective of cloud CB/TCU Cumulus clouds (TCU). coverage.

ATC may, under certain conditions, authorize special VFR flights within a control zone, when the flight visibility is not less than A conditional criterion - for EHBK, EHGG and EHRD only - applies when: specified values. SPECIAL • visibility is between ≥3 and <5 kilometers irrespective of cloudbase, or; A clearance for a special VFR flight may be VFR criteria • visibility is ≥5 kilometers AND the cloudbase, BKN or OVC, is <1500 feet, THEN; granted to pilots of aeroplanes when: (valid for Clouds and • clouds (FEW or more) ≥ 600 feet determine whether Special VFR conditions apply or not.

EHBK, EHGG visibility a. the flight visibility is not less than 3 km; and EHRD Every change in meteorological conditions leading to a change in VFR status (normal VFR, b. the clouds - FEW and SCT included - are only) SPECIAL VFR or below limits) leads immediately to the issuance of a SPECIAL or AUTO not below 600 ft; SPECIAL. c. the VFR flight can be executed clear of clouds and in continuous sight of ground or water surface. Visibility Cloud base (BKN or OVC) Clouds VFR status

≥ 5 km ≥1500 ft all normal VFR ≥ 5 km <1500 ft ≥ 600 ft SPECIAL VFR

40 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011

Item Variable Definition The criteria for issuing a SPECIAL or AUTO SPECIAL report ≥ 5 km <1500 ft < 600 ft below limits ≥3km and <5km ≥1500 ft ≥ 600 ft SPECIAL VFR ≥3km and <5km ≥1500 ft < 600 ft below limits ≥3km and <5km <1500 ft ≥ 600 ft SPECIAL VFR ≥3km and <5km <1500 ft < 600 ft below limits <3km all all below limits improvement deterioration The vertical visibility is defined as the After a 10 minute prolongation of vertical visual range into an obscuring Immediately (bearing in mind processing time) Vertical Vertical visibility improvement when the vertical visibility medium, expressed in hundreds of feet. In when the vertical visibility drops below one or visibility (VV) reaches or exceeds one or more vertical case of vertical visibility the second, third more vertical visibility thresholds. visibility thresholds. and fourth cloud group remains void. The vertical visibility thresholds are 100, 200, 300, 500 and 1000 (1000 in case of precipitation)

feet.

The air temperature in degrees Celsius (M Air temperature when negative). A change of 2 degrees or more from the temperature and/or dew-point reported in the previous Temperature Dew-point The dew-point temperature in degrees report. temperature Celsius (M when negative).

The pressure corrected to mean sea level in Pressure QNH A change of 2 hPa or more from the QNH reported in the previous report. hectoPascals.

A landing forecast which consists of a concise statement of the expected significant If a TREND is amended due to the fact that the current TREND is no longer representative for TREND TREND changes in the meteorological conditions at the expected weather changes. The TREND amendment criteria equal the issuance criteria for the aerodrome with a period of validity of SPECIAL and AUTO SPECIAL. two hours.

41 AUTO METAR system at civil airports in The Netherlands: Description and experiences August 5th, 2011 Item Variable Definition The criteria for issuing a SPECIAL or AUTO SPECIAL report A reported sudden change of wind direction Wind shear Wind shear report A wind shear report is issued or cancelled. and/or wind speed at an airport. Wind shear A forecast for a sudden change of wind A wind shear forecast is issued or cancelled. forecast direction and/or wind speed at an airport.

A layer in the lower atmosphere in which Low Level Low Level temperature increases at least 10 degrees Temperature Temperature Celsius with altitude in the lowest 1000 ft A Low Level Temperature Inversion warning is issued or cancelled. Inversion Inversion (LLTI) (also known as Marked Temperature Inversion, MTI). improvement deterioration The range in meters over which the pilot of The presentation of RVR is ceased when all The presentation of RVR starts when one or an aircraft present over the centre line of a Runway Runway Visual operational visibility sensors at the more of the operational visibility sensors at the runway can see the runway surface markings Visual Range Range (RVR) aerodrome report visibility AND RVR aerodrome report(s) a 10 minute averaged or the lights delineating the runway or values of 1500 meters or more. visibility and/or RVR below 1500 meters. identifying its centre line.

Aviation An aviation incident or accident which Aviation incident incident or occurred at, or in the vicinity of, the If air traffic control reports an incident or accident at, or in the vicinity of, the aerodrome. or accident accident aerodrome.

Interruption of the data delivery of one or The loss or return of data delivery of one or more variables in the meteorological observation Missing data Missing data more variables of the meteorological report. observation report.

The interval at which the SPECIAL criteria Every minute the SPECIAL criteria are assessed on the above mentioned rules and if one or Interval Interval are assessed. more criteria are met a SPECIAL or AUTO SPECIAL is issued. Version 1.0 date 28012011

42