As mentioned previously, the quantity of available AMDAR data has been steadily growing since its inception. However, as seen in Figure 1, the number of observations varies with economic conditions and as airlines and fleets change. Figure 4 shows the variation in the number of observations, by type of observation, over a month.
Table 3 indicates the number of observations available during April 2012 by data type and pressure level. Table 4 gives the numbers of observations that were assimilated during April 2012 by observation type and pressure level. It is important to note that not all observations received are assimilated due to quality issues or data redundancy.
Figure 4: Number of AMDAR Observations, by Observation Source, during the April 19 to May14, 2012 time period. Source: CMC
FORMAT
|
LEVEL (hPa)
|
NUMBER
|
AMDAR
|
100-300
|
440512
|
AMDAR
|
301-700
|
495528
|
AMDAR
|
701-SFC
|
543248
|
BUFR
|
100-300
|
2449334
|
BUFR
|
301-700
|
2330572
|
BUFR
|
701-SFC
|
2305930
|
ADS
|
100-300
|
327312
|
AIREP
|
100-300
|
94930
|
Total
|
all
|
8,987,368
|
Table 3: Available observations for April 2012. Source: CMC
Level (hPa)
|
Number of Observations
|
Number Wind
|
% Rejected
|
Number Temperature
|
% Rejected
|
Number Humidity
|
%Rejected
|
100-300
|
1485271
|
1466810
|
0.3
|
1459681
|
0.1
|
28959
|
3.0
|
301-700
|
909101
|
898451
|
0.3
|
895373
|
0
|
74897
|
5.3
|
701-surface
|
445141
|
442082
|
0.4
|
438141
|
0.2
|
41434
|
4.0
|
Table 4: QC-Var on assimilated observations - April 2012. Source: CMC
In general, the quality of AMDAR observations is considered to be good. There is a relatively low rejection rate based on the QC criteria developed by the AMDAR community. Of interest is the higher rejection rate for humidity data. This reflects the continued development of the WVSS – II humidity sensor by Spectra Sensors Inc., with the involvement of the European and American AMDAR community.
Aircraft and Data Quality
Considerable work has been done on assessing the quality of the meteorological data provided by aircraft within the AMDAR system. Painting (2003) provides a good overview of aircraft instrumentation and the errors associated with each system.
Various researchers have attempted to quantify the error or bias in AMDAR measurements, the greatest difficulty being in quantifying a “truth” measurement. The work by Schwartz and Benjamin focused on colocation studies, with aircraft within a minimum time and horizontal and vertical distance of rawinsondes (Schwartz 1995) and other AMDAR aircraft (Benjamin et al, 1999). The basic assumption in colocation studies is that the air mass is stable within these constraints, and that mesoscale effects are minimal.
Drue (2008) used data from various Lufthansa aircraft types, descending into Frankfurt Rhein/Main airport. Descent profile data from aircraft were compared to the hourly mean of the data. Ballish and Kumar ( 2008 ) compared AMDAR data to the corresponding 6 hour forecast, interpolated to the observation locations.
Each of these studies showed that there are biases within the AMDAR data. Although the magnitude and sign of the bias vary by aircraft type and model, pressure and phase of flight, there is thought to be an overall slight warm bias to the AMDAR temperature data (Ballish and Kumar, 2008).
It must be remembered that the instrumentation systems developed for aircraft use have been developed to be adequate for the safe and efficient operation and navigation of the aircraft, not for the acquisition of meteorological data. As Drue pointed out, a 0.2K temperature error corresponds to a 0.7 hPa error in static pressure, a value within the allowable tolerance for static pressure sensors on commercial aircraft.
Clearly, not all aircraft or instrumentation are suitable for AMDAR use. It is well known within the AMDAR community that even certain commercial aircraft types may not be capable of producing in-flight meteorological data of sufficient quality to satisfy the AMDAR standards. For example, the DHC-8 aircraft have been shown to have issues with temperature bias in flight due to the physical type and location of the sensor on the aircraft, as related to the airflow around the sensor.
Older aircraft also present unique challenges. While some aircraft have been updated with the most recent, fully integrated avionics, other aircraft of the same vintage have the original equipment or a mix of old and new instrumentation and avionics which may not be fully compatible with air data computers or ACARS equipment.
Geographic Coverage of AMDAR
Figure 5 shows the geographic coverage of AMDAR data on May 14, 2012, 00Z over a 6 hour period. This figure shows a typical pattern of AMDAR observations across the globe. This is also shown in Figure 6, which indicateds the average number of observatons assimilated per 24 hour period, within a 10 degree by 10 degree grid during April 2012. This coverage is shown for all levels, from the surface to 100 hPa.
Clearly, there are areas in which aircraft flights are very rare, such as areas of the South Pacific, South Altantic and Indian Oceans. Surprisingly, much of Northern Canada and the Arctic region are not well covered. Additionally, and not surprisingly, as there is virtually no air traffic in the Antarctic region, there is no AMDAR coverage.
Note that Northern Canada presents a special challenge, one that Environment Canada has been struggling with for many years. While there is considerable overflight traffic , the local air traffic is minimal, a result of the extremely low population density in the region. Consequently, the area is typically served by smaller airlines, operating smaller and/or older aircraft types. These aircraft are generally not suitable for AMDAR. Additionally, there is no VHF ACARS coverage in the region.
However, globally, there are clearly large areas of the continents that do not get AMDAR coverage while considerable air traffic is known to exist.
This report will identify 6 geographical regions with sparse AMDAR data. Inclusion of these regions in the global AMDAR program would enhance the forecast accuracy locally, regionally and globally. The Data Sparse Regions are:
South America
Central America/Caribbean
Africa
Middle East and Central Asia
Asia Pacific
Former Soviet Republic Countries
Figure 5: Global distribution of AMDAR data – all levels. Source:CMC
Figure 6: Global distribution of assimilated AMDAR observations. The average number of observations per 24 hour period during April 2012 within a 10 degree by 10 degree grid. Source: CMC
Share with your friends: |