Verification of wind forecasts for the airports
Honningsvåg, Hammerfest, Hasvik, Tromsø,
Evenes, Narvik, Mosjøen, Sandnessjøen,
Brønnøysund, Værnes, Ørsta-Volda,
Sandane, Førde and Fagernes for the period
1.5.2009 to 30.4.2010.
Knut Helge Midtbø, Mariken Homleid and Viel Ødegaard
(P.O. Box 43, N-0313 OSLO, NORWAY)
ABSTRACT
Atmospheric turbulence still accounts for a significant percentage of weather-related aircraft incidents in
Norway. A turbulence modelling project is run by the Norwegian Meteorological Institute in cooperation
with SINTEF. A nested model system designed for forecasting atmospheric turbulence has been developed.
From mid 2007 the system has been operated for seven airports in Norway. Report 2, 2008 (Midtbø et. al.,
2008) presents the system and gives verification results for the last months of 2007.
Previous notes presented similar verification results for 2008 up to 30th of April 2009. The present note is a
follow up covering the one year period from 1st of May 2009 till 30th of April 2010.
In the project it has been decided to do a special study for Værnes airport. The airport is chosen as the one
of the seven for which there are available AMDAR reports from aircrafts on a regular basis.
The models validated include HIRLAM 8-km (8-km resolution), UM 4-km and UM 1-km (Unified Model,
4-km and 1-km horizontal resolution) and the SIMRA model (approximately 100m resolution). Verification
results are presented.
No. 8
Oslo, 31.05.2010
1. Model description
Below is included a short description of the model system.
The global ECMWF model provides lateral boundary conditions for the nested turbulence model system.
Those boundary values are used for a first nest of the HIRLAM8 hydrostatic 8-km model covering
Scandinavia and adjacent sea areas. From 27.3 2008 increased computer resources made it possible to
upgrade to 8 km resolution (the model was renamed to HIRLAM8). The upgrade also included introduction
of data assimilation in the new HIRLAM8. In HIRLAM10 no such assimilation was operational due to
limited computer power. Nested inside this HIRLAM10 (from 27.3.2008 HIRLAM8) model is a version of
the non-hydrostatic Unified Model (UM 4-km) developed at Met Office set up with 4-km resolution. (See
areas in figure 2). The airport specific part of the turbulence system consists of two models covering an
area surrounding the airport.
Figure 1 Computation points in the UM. Wind components u and v are computed one half of a grid length
displaced compared to ŋ that stands for T, q and all (left) Distribution of UM vertical layers
illustrated with wind arrows in all grid points. T, q and all are computed on the interphones
between these layers (right).
A: 1-km resolution of the Unified Model (UM 1-km)
B: approximately 100-m turbulence model Simra
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
2
A is nested inside the 4-km run. Initial values and boundary conditions for B is taken from model A.
Surrounding the airports the airports which during the period has increased to 14 are set up several UM 1km models. The set up of UM1 has changes during the period covered. The changes are described briefly in
section 2 below. The set up used at the end of the one year period is presented in figures 2,3,4 and 5.
All the models involved are regular numerical weather prediction models with exception for SIMRA,
which is a turbulence model. The model areas for the UM 1-km and Simra for the 14 airports are given in
figures 6-19. The figures also include model topography and surface projection of the runway. For
examples of results as presented to the ATM users we refer to IPPC.
In the present model set up input to the SIMRA turbulence forecasts are produced by running UM 1-km
model with an interpolated +3 hour forecast from UM 4-km as initial values. Boundary values are specified
from the same UM 4-km every hour, and the model is integrated from +6 hours up to + 18 hours. The
procedure is repeated based on 4-km forecasts starting at 00UTC and 12 UTC. 1-km model results are
produced for both 00UTC and 12 UTC from +6 up to + 18 hours.
Data are interpolated from the spherical rotated UM 1-km grid to UTM coordinates used by SIMRA.
Vertical interpolation is needed to adapt to the higher resolution topography in the SIMRA model.
Weather parameters transferred from UM to SIMRA are given in the following list:
Surface pressure 2d
Horizontal wind 3d
Potential temperature 3d
Specific humidity 3d
Pressure at model levels 3d
SIMRA is a model based upon Reynolds equations with a standard (K, ε) - turbulence closure and
boundary conditions. It has the capability of predicting flows with separation, attachment, hydraulic
transitions and internal wave breaking.
The SIMRA model has a dynamic estimation of turbulent kinetic energy and dissipation and predicts
turbulent kinetic energy (TKE). The square root of TKE has the dimension of velocity and is used as an
indicator of turbulence intensity.
Initial data for the SIMRA model are interpolated from the 1-km runs valid every hour (+6, +7, +8, +18).
For each of those points of time the SIMRA model is run with constant boundaries until a stationary
solution is achieved. Rather than an ordinary forecast dataset, SIMRA supplies data that can be used to
estimate quantitatively the strength of the turbulence in each time interval (1 hour) (Eidsvik and Utnes,
1997).
The reason for running with a lead-time of 6 hours is that the total time from taking the observations up to
finished computations is 4 hours. As the system is operated today, turbulence forecasts valid every hour
between +6 and +18 based on 00UTC and 12UTC are available at approximately 04UTC and 16UTC.
2. Updates during the period
During the 1st May 2009 till 30th April 2010 SIMRA computations for several new airports has been
introduced. The model set up of the UM 1-km has been changed as upgrades to version 7.3 and as revision
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
3
of the areas and inclusion of areas for additional (new) airports. UM1FOS, UM1BMS, UM1NN and
UM1SN are areas for UM1 and are shown in figures 2-5.
17/9:
Start of SIMRA forecast for Førde airport Bringeland
Upgrade to UM 7.3 for UM1FOS
21/9:
Start of SIMRA forecast for Brønnøysund airport, Brønnøy
Upgrade to UM 7.3 for UM1BMS
21/10:
Start of SIMRA forecast for Mosjøen airport, Kjærstad and Harstad/Narvik airport, Evenes
28/10:
Upgrade to UM 7.3 for UM1NN
18/12
SIMRA forecast for Honningsvåg, Hammerfest, Narvik nested in UM1NN
7/1:
Start of SIMRA forecast for Hasvik airport and Tromsø airport, Langnes
8 /4
Start of SIMRA forecast for Fagernes airport, Leirin
Start of UM1SN (version 7.3)
SIMRA for Værnes airport nested in UM1SN.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
4
Figure 2. Model domains for UM 1-km (named UM1NN) for parts of Northern Norway. The Simra areas
for the airports Honningsvåg, Hammerfest, Hasvik, Tromsø, Evenes and Narvik are shown in blue.
Topography given by a colour from green at 0 meters to brown above a few hundred meters.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
5
Figure 3. Model domains for UM 1-km (named UM1BNS) for parts of Mid-Norway. The Simra areas for
the airports Sandnessjøen, Mosjøen and Brønnøysund are shown in blue. Topography given by a colour
from green at 0 meters to brown above a few hundred meters.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
6
Figure 4. Model domains for UM 1-km (named UM1FOS) for parts of Western Norway. The Simra areas
for the airports Ørsta/Volda, Sandane and Førde are shown in blue. Topography given by a colour from
green at 0 meters to brown above a few hundred meters.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
7
Figure 5. Model domains for UM 1-km (named UM1SN) for parts of Southern Norway. The Simra areas
for the airport Værnes and Fagernes are shown in blue. Topography given by a colour from green at 0
meters to brown above a few hundred meters.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
8
Figure 6. Model area for SIMRA at Honningsvåg airport, Valan. Model topography colour coded from
green to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
9
Figure 7 Model area for SIMRA at Hammerfest airport. Model topography colour coded from green to
brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
10
Figure 8 Model area for SIMRA at Hasvik airport. Model topography colour coded from green to brown.
Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
11
Figure 9 Model area for SIMRA at Tromsø airport, Langnes. Model topography colour coded from green
to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
12
Figure 10 Model area for SIMRA at Harstad/Narvik airport, Evenes.Model topography colour coded from
green to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
13
Figure 11 Model area for SIMRA at Narvik airport, Framnes. Model topography colour coded from green
to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
14
Figure 12. Model area for SIMRA at Mosjøen airport, Kjærstad. Model topography colour coded from
green to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
15
Figure 13 Model area for SIMRA at Sandnessjøen airport, Stokka. Model topography colour coded from
green to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
16
Figure 14 Model area for SIMRA at Brønnøysund airport, Brønnøy. Model topography colour coded from
green to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
17
Figure 15 Model area for SIMRA at Trondheim airport, Værnes. Model topography colour coded from
green to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
18
Figure 16 Model area for SIMRA at Ørsta-Volda airport, Hovden. Model topography colour coded from
green to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
19
Figure 17 Model area for SIMRA at Sandane airport, Anda. Model topography colour coded from green to
brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
20
Figure 18 Model area for SIMRA at Førde airport, Bringeland. Model topography colour coded from green
to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
21
Figure 19 Model area for SIMRA at Fagernes airport, Leirin. Model topography colour coded from green
to brown. Isolines for each 80 meters are also given.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
22
3. Observations for verification
We have used three standard meteorological observation types in this study. We include a description.
The model system is designed for forecasting turbulent conditions. We have addressed the need for
turbulence measurement for verification/validation. So far the data for verification is limited to
observations, which are unfortunately rather sparse. In May 2008 a trial period started. During this period
Widerøe pilots delivered special turbulence reports on how they felt the quality of the forecasts were
compared to the real experience of the conditions. After the trial period the feedback from the pilots was
evaluated and the overall impression was that it was good correlation between the forecast and the pilots
experience of the actual wind conditions.
3.1 SYNOP
SYNOP is an acronym for surface based reports. A SYNOP report contains wind measurements at 10 meter
using standard traditional surface measurements for the airport at standard height 10 meter above land
surface.
3.2 TEMP
TEMP is an acronym for vertical soundings as measured from a balloon released and ascending due to lift
of the balloon. A measurement device is attached to the balloon. Data are transmitted to the ground from
this device in real time. The wind is calculated from the observed drift of the balloon relative to the ground.
The observational error of wind from TEMP reports is of the order of a few m/s.
3.3 AMDAR
There are available data from higher levels in the surrounding areas of some of the Norwegian airports.
Some aircrafts are equipped with installations necessary to transmit meteorological observations known as
AMDAR (Aircraft Meteorological DAta Relay).
The following is referred from AMDAR REFERENCE MANUAL prepared by Derek Painting, World
Meteorological Organization (WMO).
“Wind speed and direction are computed by resolution of the vectors:
V = Vg – Va
Where V is the wind vector, Vg (ground velocity) is the velocity of the aircraft with respect to the earth and
Va is the velocity of the air with respect to the aircraft. Va is calculated from true airspeed and heading.
Heading and ground velocity are derived from the inertial reference unit (IRU).
True airspeed is a function of Mach number and static air temperature. Errors in Mach number are the
most significant. For example with a Mach number error of 0.5% at cruise level, airspeed error is
some1.2m/s. Thus with zero error from the navigation system, wind vector errors up to 1.2m/s are to be
expected and are also dependent on the angle between the wind at flight level and the aircraft heading.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
23
Errors in true airspeed combine with errors from the IRU. The basic calculations assume perfect
alignment of the aircraft with the airstream and zero roll, pitch, yaw and perfect inertial platform
alignment. At high pitch/roll angles wind vector errors, which are proportional to true airspeed can be
significant. For example at 150kt airspeed with 5 degrees pitch and 10 degrees roll a wind vector error of
some 1m/s can be expected regardless of the true wind vector. At low wind speeds vector errors can lead to
large errors in wind direction. Thus a more useful indication combining wind speed and direction error as
vector error would suggest a typical uncertainty of 2-3m/s.”
In the AMDAR format there are possibilities for reporting turbulence measures. Turbulence is reported in
one or more of three ways:
(i)As variation in vertical acceleration experienced by the aircraft.
(ii)As 'derived equivalent vertical gust'.
(iii)As an index related to eddy dissipation rate (EDR).
For more details see the reference manual referred to above.
In Europe quite many AMDAR reports includes the index listed EDR (see list above). We have
investigated the AMDAR reports presently available from aircrafts operating in Norway for some arbitrary
days in January 2008. We have for those aircrafts found only reports with this index given as missing. For
the project there would be of great value if actions could be taken in order to include such an index in the
AMDAR reports for Norwegian areas.
4. Measures used for verification
Basis for the verification work at the Norwegian Meteorological Institute is operational storage of both
observations and model output in secured databases. In this way all data for validation and verification is
available for further investigation if necessary.
All model data used in the following validation is interpolated horizontally to the position of the
observation with Bessel interpolation with good interpolation accuracy. In the vertical data are interpolated
as linear in the logarithm of p. To compare with observations at times off the hourly available model data a
linear interpolation in time is carried out.
In order to summarize the results for the wind in some way, we have computed the square root ( suv ) of the
mean of the squared standard deviations of the component errors of the two wind components u and v
( su and sv ). The formula is:
suv 
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
1
2
2
  su  sv
2
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00

Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
24
4.1 Verification of horizontal wind forecasts against SYNOP reports
for the airports Honningsvåg, Hammerfest, Hasvik, Tromsø, Evenes,
Narvik, Mosjøen, Sandnessjøen, Brønnøysund, Værnes, ØrstaVolda, Sandane, Førde and Fagernes.
We have included figures showing monthly means of suv and Mean Error based on SYNOP reports for all
airports (with exception of Fagernes airport, Leirin for which data covers a few weeks of April only). Those
two parameters are used as a quality indicator for the wind forecasts.
An overall result is that the errors are smaller in the summer season as is the wind force as a mean. This is a
typical result, which is also clearly seen, in the lower panel of figure 1 for several models.
The results are presented for one and one airport starting in Northern Norway with Honningsvåg ending up
in Western Norway.
For Honningsvåg the errors are quite large for all models. Simra and also UM 1-km underestimates the
wind while UM 4-km does not.
At Hammerfest Simra and UM 1-km has smaller errors than UM 4-km most of the year and it is clear that
the forecasts here is improved with the highest resolution. There is however an overestimation in the UM 4km at this place while the Simra and UM 1-km underestimates also here.
At Hasvik airport the period covers only four months. Here there is however a reduced error in Simra
compared to the UM 1-km. Results for UM 4-km is missing due to a technical problem storage problem.
At Tromsø airport there is also only four months of data. The error is almost equal in magnitude for the
different models at this place; Underestimation of wind in Simra is rather significant also here.
At Harstad/Narvik airport, Evenes error measures is remarkable similar for all the models.
The results for Narvik airport, Framnes represent a location where the UM 4-km has the lowest error. This
result is somewhat outstanding and is caused by very small-scaled features in terms of terrain and shoreline
at this place. Because of the details, the models perform rather different. A closer inspection should
probably be carried out for this location.
For Sandnessjøen airport, Stokka Simra and UM 1-km is better than UM 4-km in summer and winter
respectively. This airport represents one of the airports for which there is a clear advantage from using
resolution better than 4 km.
For Mosjøen airport, Kjærstad the errors are as there is often gentle wind conditions here. For this place
UM 4-km has the lowest error level.
For Brønnøysund airport, Brønnøy 8 months of data are available. The results are that there is a rather
complicated error structure here and none of the models demonstrates clear advantages over the others.
For Trondheim airport, Værnes the errors are small connected to generally weak surface winds. At this
place all models has remarkable similar results.
For Ørsta-Volda the highest resolution models has the largest errors at surface. This is probably caused by
detailed and complicated terrain rather close to the runway. This might in principle be a sign of not so good
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
25
turbulence forecasts at this airport. Since we do not have turbulence measurements it might as well be
turbulence measurements of about a quality similar to the airports for which the pilot reports (see earlier
reports) demonstrated that the model gave important warnings of turbulence.
For Sandane airport, Anda there are rather similar errors for all models. There is however a clear lower
error in UM 4-km in September. (These results are also found for September although not that clear for
Ørsta-Volda) This is probably due to atmospheric circulation in favour of the coarser 4 km grid for this
location in the complex terrain.
Førde airport, Bringeland is another of the airports where the errors are small and where all models show
up rather similar results.
The results discussed here are, however, consistent with the findings reported in the previous notes. There
are seasonal variations and also other differences from year to year as one might expect, but nevertheless
monthly means are rather stable from year to year. This is comforting as it indicates that the quality of the
forecasts is stable in time.
As discussed in Midtbø (2006), the results must be used with much care. The model is designed for
forecasting turbulence in the upper air. A verification of forecasts of surface winds should be quite relevant
as it can be argued that good surface wind forecast could be used as an indication of the models wind
forecasts in general. We do not, however, think that this can be fully justified without turbulence
measurements. The forecasts of wind at higher levels are the basis for the turbulence forecasts and in order
to say something about the turbulence forecast, turbulence and wind at higher levels has to be verified
against observations from those levels. The turbulence computed by the model is a result of complicated
structures and gradients in wind and temperature and thus there are those structures rather than wind near
surface, which is the key to good turbulence estimates.
We have also included in table 1 results showing the availability of forecasts on our system as numbers and
as percentage of the possible forecasts (two computational runs each day). We see that there have been
available complete data sets for more than 90% of the possible runs. Most of the missing runs have been
caused by some computational weaknesses in the UM model. As we could not be satisfied with availability
we have therefore pushed forward updating UM in order to improve it. Introduction of version 7.3 of UM
has been done in steps during the year (the update sequence is given in section 2. In January 2010 all UM
1-km except UM1SN had been upgraded. For the Simra runs the set up was at that time so that all models
except the one for Værnes (this model was upgraded in April) was run nested in version UM 7.3 from 1st of
February 2010. As a result of the upgrades the availability has increased to a total 13 missing forecast out
of 14x178 possible forecast for the last three months of the year (February, March and April) giving a
percentage above 99,0 which we regard as a satisfactory number in this respect.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
26
Figure 20. Verification results for Honningsvåg airport, Valan. Upper panel shows plot of
suv
(see section 4) for SIMRA, UM 1-km and UM 4-km
while lower panel shows mean error. Colour code given in the plot.
Figure 21. Verification results for Hammerfest airport. Upper panel shows plot of
suv (see section 4) for SIMRA, UM 1-km and
UM 4-km while lower panel shows mean error. Colour code given in the plot.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
27
Figure 22. Verification results for Hasvik. Upper panel shows plot of
suv (see section 4) for SIMRA, UM 1-km and UM 4-km
while lower panel shows mean error. Colour code given in the plot.
Figure 23. Verification results for Tromsø airport, Langnes. Upper panel shows plot of
suv (see section 4) for SIMRA, UM 1-km
and UM 4-km while lower panel shows mean error. Colour code given in the plot.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
28
Figure 24. Verification results for Harstad/Narvik airport, Evenes. Upper panel shows plot of
suv (see section 4) for SIMRA,
UM 1-km and UM 4-km while lower panel shows mean error. Colour code given in the plot.
Figure 25. Verification results for Narvik airport, Framnes. Upper panel shows plot of
suv (see section 4) for SIMRA, UM 1-km
and UM 4-km while lower panel shows mean error. Colour code given in the plot.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
29
Figure 26. Verification results for Sandnessjøen airport, Stokka. Upper panel shows plot of
suv (see section 4) for SIMRA, UM
1-km and UM 4-km while lower panel shows mean error. Colour code given in the plot.
Figure 27. Verification results for Mosjøen airport, Kjærstad. Upper panel shows plot of
suv (see section 4) for SIMRA, UM 1-
km and UM 4-km while lower panel shows mean error. Colour code given in the plot.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
30
Figure 28. Verification results for Brønnøysund airport, Brønnøy. Upper panel shows plot of
suv (see section 4) for SIMRA,
UM 1-km and UM 4-km while lower panel shows mean error. Colour code given in the plot.
Figure 29. Verification results for Trondheim airport, Værnes. Upper panel shows plot of
suv (see section 4) for SIMRA, UM 1-
km and UM 4-km while lower panel shows mean error. Colour code given in the plot.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
31
Figure 30. Verification results for Ørsta-Volda airport, Hovden. Upper panel shows plot of
suv (see section 4) for SIMRA, UM
1-km and UM 4-km while lower panel shows mean error. Colour code given in the plot.
Figure 31. Verification results for Sandane airport, Anda. Upper panel shows plot of
suv (see section 4) for SIMRA, UM 1-km
and UM 4-km while lower panel shows mean error. Colour code given in the plot.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
32
Figure 32. Verification results for Førde airport, Bringeland. Upper panel shows plot of
suv (see section 4) for SIMRA, UM 1-
km and UM 4-km while lower panel shows mean error. Colour code given in the plot.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
33
Airport
All data
missing
Sum data
missing
21
Some
data
missing
41
% Operated with
all or some data
62
Number
% Operated with
of forecast all data
possible
730
91,5
Honningsvåg
airport, Valan
Hammerfest
airport.
Hasvik airport.
Tromsø airport,
Langnes
Harstad/Narvik
airport, Evenes
Narvik airport,
Framnes.
Mosjøen
airport,
Kjærstad
Sandnessjøen
airport, Stokka.
Brønnøysund
airport,
Brønnøy.
Trondheim
airport, Værnes
Ørsta-Volda
airport, Hovden
Sandane airport,
Anda
Førde airport,
Bringeland
Fagernes
airport, Leirin
21
39
60
730
91,8
97,1
1
1
1
1
2
2
226
226
99,1
99,1
99,6
99,6
4
17
21
382
99,0
95,1
8
19
27
730
96,3
98,9
10
10
20
382
94,8
97,4
15
19
34
730
95,3
97,9
11
9
20
424
97,6
95,0
28
20
48
730
93,4
96,2
26
22
48
730
93,4
96,4
26
22
48
730
93,4
96,4
23
19
42
424
90,1
94,6
0
0
0
28
100,0
100,0
97,1
Table 1
Table showing available forecast from the SIMRA runs. The two last columns show the percentage of
forecasts available either complete or the sum of complete results and the results when some of the data are
missing.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
34
4.2 Verification of horizontal wind forecasts against upper air TEMP
reports for Ørlandet airport in Trøndelag
The model system is nested. The quality of wind forecasts in the coarser mesh models are verified against
TEMP reports on a routine basis. In this report we have computed verification against TEMP reports for
Ørlandet airport. This verification is done for UM 4-km and HIRLAM10 / HIRLAM8 because neither UM
1-km nor SIMRA covers the location.
The observations can not be viewed upon as what is called the true value for the model data. This has two
causes: measurement errors and the model not representing a mean value for the model grid box. This
means that the model represents the grid box and not always the measured quantity regardless of the
observation accuracy.
When we in the following compare the two values (observations and model results) we will in short call the
difference for model errors. It should be kept in mind that this error has originates from observational errors
and errors connected to difference in what the two values represent as explained above.
We have done a separate investigation of the model errors measured by the TEMP reports for Ørlandet. The
reason for this is that we want to monitor the model error in the Trøndelag area relative to an observing
system like the TEMP reports. This is because a well-known and well-documented error level characterizes
those reports. The station closest to Værnes is Ørlandet. As seen in figure 8 this is close to the lateral
boundary for the UM 1km model for Værnes and outside the SIMRA area. Out of the results we have
chosen to display data for 850 hPa, which is a level around 1500 meters above sea level. We have
compared the wind from UM 4-km and HIRLAM10 / HIRLAM8 to the wind from TEMP reports. UM 1km results are not displayed, as they are quite similar to UM 4-km results.
A selection of verification results is shown in table 2a-c. In this table we have show the verification
parameters for +12 hour prognosis. We see that the error using this measure at the start is around 2 m/s. As
explained above this result are a combination of observation accuracy and the model representing grid
boxes rather than the TEMP measurement. This is in agreement with findings in previous.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
35
ME
MAE
RMSE
SDE
Wind
FF.UM4
-0,22
2,00
2,36
2,35
speed
in
FF.Hirlam8 FF.ECMWF
-0,34
-0,57
1,86
1,65
2,31
2,05
2,28
1,97
850 hPa
00+12
component
U.Hirlam8
1,98
3,90
4,85
4,43
in
U.ECMWF
0,96
1,78
2,17
1,95
850 hPa
00+12
ME
MAE
RMSE
SDE
U
U.UM4
1,33
2,67
3,36
3,09
component
V.Hirlam8
-1,64
3,85
4,96
4,68
in
V.ECMWF
0,24
1,62
2,08
2,06
850 hPa
00+12
ME
MAE
RMSE
SDE
V
V.UM4
-1,12
3,14
4,12
3,96
Table 2a
Verification results for wind speed (ff) and components u and v of wind for 1st of May till 31st of August
2009.
Verification results of interpolated model data compared to horizontal wind from TEMP reports observed at
Ørlandet airport. The observation level is at 850 hPa. The model data are 12 hour forecasts valid 12 UTC.
The models are UM4 and HIRLAM8 and ECMWF (left to right). The parameters computed are standard
statistical parameter for the value of model value minus observation. We see in the three upper tables
values for wind force (FF), u-component and v-component of the wind. The parameters are mean error
(ME); mean absolute error (MAE), root mean square error (RMSE) and standard deviation of error (SDE).
All results are in m/s. Interpolation methods for model data are described in the text. The results are
computed for models based on the 00 UTC only. The TEMP reports used for this verification are close to
100.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
36
ME
MAE
RMSE
SDE
Wind
FF.UM4
-0,55
2,09
2,70
2,64
speed
in
FF.Hirlam8 FF.ECMWF
-0,04
-0,61
2,11
1,82
2,67
2,23
2,67
2,15
850,00
hPa
00+12
component
U.Hirlam8
3,40
4,68
6,06
5,02
in
U.ECMWF
0,57
1,83
2,20
2,12
850,00
hPa
00+12
ME
MAE
RMSE
SDE
U
U.UM4
2,18
3,43
4,55
3,99
component
V.Hirlam8
-3,31
6,02
7,41
6,63
in
V.ECMWF
-0,03
1,85
2,22
2,22
850,00
hPa
00+12
ME
MAE
RMSE
SDE
V
V.UM4
-2,05
4,33
5,52
5,13
Table 2b
Verification results for wind speed (ff) and components u and v of wind for 1st of September till 31st of
December 2009.
Verification results of interpolated model data compared to horizontal wind from TEMP reports observed at
Orland airport. The observation level is at 850 hPa. The model data are 12 hour forecasts valid 12 UTC.
The models are UM4 and HIRLAM8 and ECMWF (left to right). The parameters computed are standard
statistical parameter for the value of model value minus observation. We see in the three upper tables
values for wind force (FF), u-component and v-component of the wind. The parameters are mean error
(ME); mean absolute error (MAE), root mean square error (RMSE) and standard deviation of error (SDE).
All results are in m/s. Interpolation methods for model data are described in the text. The results are
computed for models based on the 00 UTC only. The TEMP reports used for this verification are close to
100.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
37
ME
MAE
RMSE
SDE
Wind
FF.UM4
-0,53
1,74
2,15
2,08
speed
in
FF.Hirlam8 FF.ECMWF
-0,19
-0,81
2,23
1,70
2,82
2,13
2,81
1,97
850,00
hPa
00+12
component
U.Hirlam8
0,71
3,79
4,71
4,65
in
U.ECMWF
-0,01
1,86
2,29
2,29
850,00
hPa
00+12
ME
MAE
RMSE
SDE
U
U.UM4
0,52
2,88
3,66
3,63
component
V.Hirlam8
-2,34
4,70
6,04
5,57
in
V.ECMWF
-0,05
1,83
2,30
2,30
850,00
hPa
00+12
ME
MAE
RMSE
SDE
V
V.UM4
-1,71
3,60
4,82
4,51
Table 2c
Verification results for wind speed (ff) and components u and v of wind for 1st of January till 30th of April
2010.
Verification results of interpolated model data compared to horizontal wind from TEMP reports observed at
Ørlandet airport. The observation level is at 850 hPa. The model data are 12 hour forecasts valid 12 UTC.
The models are UM4 and HIRLAM8 and ECMWF (left to right). The parameters computed are standard
statistical parameter for the value of model value minus observation. We see in the three upper tables
values for wind force (FF), u-component and v-component of the wind. The parameters are mean error
(ME); mean absolute error (MAE), root mean square error (RMSE) and standard deviation of error (SDE).
All results are in m/s. Interpolation methods for model data are described in the text. The results are
computed for models based on the 00 UTC only. The TEMP reports used for this verification are close to
100.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
38
4.3 Validation of horizontal wind forecasts against upper air AMDAR
reports for the SIMRA Værnes area.
Among the seven airports Værnes airport is the only one where there are available AMDAR reports. This is
because not all aircrafts are equipped with the necessary systems to produce and transmit AMDAR reports.
The data forming basis for those reports are produced continuously during the flight including approach
and take off. The observed wind is subject to a sampling procedure and AMDAR reports are produced at
specified time intervals. An example of measured data is shown in figures 8 and 9. The observed data are
irregularly distributed in space and time as a consequence of the flight regulations and the variation in the
air traffic. For the data used for this verification we have selected data in the volume covered by the
SIMRA model for Værnes. The model volume extends up to approximately 1840 meters above sea level. In
this way the same observations are used for validating HIRLAM8, UM 1-km and SIMRA.
Results are presented tables 3a-f. The tables covers the first four months (a and b), the next for (c and) and
the last four months (e and f). a, c and e covers verification based on computations from 00 UTC while the
other three tables covers verification from 12 UTC. For details about the results in the tables, see the table
text.
The results are as already mentioned confined to the volume box covered by SIMRA. From the previous
chapter we have the result that a 12-hour prognosis of horizontal wind has an error of 2- 3 m/s at 1500
meters height when compared to TEMP reports for Ørlandet. The results in tables 3 a-f are for forecast
lengths of a few hours and the errors displayed are of the same order of magnitude. But, conclusions about
the value of the inner SIMRA nest can first be drawn when we have observations of turbulence available
for verification.
Værnes is situated closer to the large mountain areas in Southern Norway than Ørlandet. We have seen that
the errors are on overall a little larger than the errors presented in chapter 3.2.
Following up this investigation can draw more distinct conclusion on a larger data set. In a larger data set
there also be possible to sample data for more distinct height levels for the AMDAR investigation.
The results shows the view that the UM 1-km and SIMRA are able to reduce horizontal wind error
compared to the coarser mesh model HIRLAM8.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
39
Figure 33.
Model domains for UM 1-km for Værnes airport used until April 2010. Topography with 70 m spacing
between isolines. The SIMRA domain is shown with red wind arrows every second grid point. In black a
wind arrow (in 850 hPa) from the radiosonde at Ørlandet airport. Also in black, wind arrows from AMDAR
for two different aircrafts.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
40
Figure 34
The SIMRA domain around Værnes. The topography is shown with 70 m between isolines. The straight
black line indicates the projection of the sectors for take off and landing. The red wind arrows show the
wind in the in-flight levels for every second grid point. The black wind arrows are from AMDAR for two
different aircrafts.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
41
Statistics
for
00+9,+10,+11,+12,+13,+14,+15
Number
of
cases:
998
FF.H8
FF.UM1
FF.SIMRA
ME
0,62
0,82
0,71
MAE
2,04
2,42
2,33
RMSE
2,64
3,13
3,07
SDE
2,56
3,02
2,98
U.H8
U.UM1
U.SIMRA
ME
0,69
1,46
1,24
MAE
3,35
3,19
2,90
RMSE
4,17
3,98
3,71
SDE
4,11
3,70
3,50
V.H8
V.UM1
V.SIMRA
ME
1,10
-0,67
-0,82
MAE
3,14
1,61
1,70
RMSE
4,06
2,27
2,36
SDE
3,91
2,17
2,21
Table 3a
Verification results for wind speed (ff) and components u and v of wind for 1st of May till 31st of August
2009. The computations are for 00UTC +9,+10,+11,+12,+13,+14,+15. The table contains results for wind
force ff, and for the components u and v of the wind. The error measures are mean error (ME); mean
absolute error (MAE), root mean square error (RMSE) andd standard deviation of error (SDE).
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
42
Statistics
for
12+3,+4,+5,+6,+7,+8,+9
Number
of
cases:
324
ME
MAE
RMSE
SDE
FF.H8
0,45
1,94
2,47
2,44
FF.UM1
0,04
2,12
2,67
2,68
FF.SIMRA
-0,21
2,00
2,65
2,64
ME
MAE
RMSE
SDE
U.H8
0,53
3,25
4,22
4,20
U.UM1
0,80
2,23
2,73
2,62
U.SIMRA
0,09
2,24
2,87
2,87
ME
MAE
RMSE
SDE
V.H8
0,17
3,14
3,99
3,99
V.UM1
-0,96
1,95
2,63
2,45
V.SIMRA
-1,10
2,01
3,05
2,85
Table 3b
Verification results for wind speed (ff) and components u and v of wind for 1st of May till 31st of August
2009. The computations are for 12UTC +3,+4,+5,+6,+7,+8,+9. The table contains results for wind force ff,
and for the components u and v of the wind. The error measures are mean error (ME); mean absolute error
(MAE), root mean square error (RMSE) and standard deviation of error (SDE).
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
43
Statistics
for
00+9,+10,+11,+12,+13,+14,+15
Number
of
cases:
906
ME
MAE
RMSE
SDE
FF.H8
1,65
2,77
3,63
3,23
FF.UM1
0,90
2,59
3,33
3,21
FF.SIMRA
0,94
2,61
3,38
3,25
ME
MAE
RMSE
SDE
U.H8
-2,14
3,85
4,86
4,37
U.UM1
-0,02
2,94
3,61
3,61
U.SIMRA
0,26
2,79
3,47
3,46
ME
MAE
RMSE
SDE
V.H8
2,09
4,19
5,57
5,16
V.UM1
0,11
1,88
2,58
2,58
V.SIMRA
0,09
1,99
2,68
2,68
Table 3c
Verification results for wind speed (ff) and components u and v of wind for 1st of September till 31st of
December 2009. The computations is for 00UTC +9,+10,+11,+12,+13,+14,+15. The table contains results for
wind force ff, and for the components u and v of the wind. The error measures are mean error (ME); mean
absolute error (MAE), root mean square error (RMSE) and standard deviation of error (SDE).
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
44
Statistics
for
12+3,+4,+5,+6,+7,+8,+9
Number
of
cases:
363
ME
MAE
RMSE
SDE
FF.H8
1,58
2,80
3,57
3,21
FF.UM1
0,47
2,03
2,58
2,54
FF.SIMRA
0,39
2,34
3,24
3,22
ME
MAE
RMSE
SDE
U.H8
-1,65
3,74
4,72
4,43
U.UM1
-0,01
2,53
3,06
3,06
U.SIMRA
-0,14
2,47
3,16
3,16
ME
MAE
RMSE
SDE
V.H8
1,88
3,92
5,17
4,82
V.UM1
-0,41
1,41
1,94
1,90
V.SIMRA
-0,39
1,89
2,91
2,89
Table 3d
Verification results for wind speed (ff) and components u and v of wind for 1st of September till 31st of
December 2009. The computations is for 12UTC +3,+4,+5,+6,+7,+8,+9. The table contains results for wind
force ff, and for the components u and v of the wind. The error measures are mean error (ME); mean
absolute error (MAE), root mean square error (RMSE) and standard deviation of error (SDE).
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
45
Statistics
for
00+9,+10,+11,+12,+13,+14,+15
Number
of
cases:
1295
ME
MAE
RMSE
SDE
FF.H8
1,68
2,60
3,34
2,89
FF.UM1
1,08
2,25
2,91
2,71
FF.SIMRA
1,12
2,18
2,80
2,57
ME
MAE
RMSE
SDE
U.H8
-0,91
3,67
4,63
4,54
U.UM1
0,48
2,50
3,08
3,04
U.SIMRA
0,54
2,24
2,75
2,70
ME
MAE
RMSE
SDE
V.H8
1,59
3,14
4,44
4,14
V.UM1
-0,24
1,62
2,23
2,22
V.SIMRA
-0,20
1,78
2,40
2,39
Table 3e
Verification results for wind speed (ff) and components u and v of wind for 1st of January till 30th of April
2010. The computations is for 00UTC +9,+10,+11,+12,+13,+14,+15. The table contains results for wind force
ff, and for the components u and v of the wind. The error measures are mean error (ME); mean absolute
error (MAE), root mean square error (RMSE) and standard deviation of error (SDE).
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
46
Statistics
for
12+3,+4,+5,+6,+7,+8,+9
Number
of
cases:
704
ME
MAE
RMSE
SDE
FF.H8
2,09
2,80
3,51
2,83
FF.UM1
0,97
1,88
2,42
2,21
FF.SIMRA
1,32
2,05
2,87
2,55
ME
MAE
RMSE
SDE
U.H8
-1,44
3,28
4,28
4,03
U.UM1
-0,09
1,98
2,47
2,47
U.SIMRA
0,18
2,01
2,53
2,52
ME
MAE
RMSE
SDE
V.H8
1,28
3,11
4,22
4,02
V.UM1
-0,18
1,43
1,91
1,90
V.SIMRA
0,26
1,56
2,43
2,42
Table 3f
Verification results for wind speed (ff) and components u and v of wind for 1st of January till 30th of April
2010. The computations is for 12UTC +3,+4,+5,+6,+7,+8,+9. The table contains results for wind force ff, and
for the components u and v of the wind. The error measures are mean error (ME); mean absolute error
(MAE), root mean square error (RMSE) and standard deviation of error (SDE).
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
47
5. DISCUSSION
As discussed above, the modelling of the winds above surface in SIMRA are for most places improved in
this nested system when compared to coarser mesh meteorological models.
The findings in the present report show that the accuracy of the forecasts is of the same type and shows
similar local variations as demonstrated in the reports covering previous periods. The UM model has been
upgraded during the period but since the SIMRA model version has been kept in the period, this is what
should be expected.
The regularity has been improved after introduction of upgrades of the UM model.
Since SIMRA computes the turbulence from wind distribution, a good verification of the wind especially
higher up in the atmosphere is a necessary condition for producing good turbulence forecast. Surface wind
conditions will in this context be a weaker measure of ability to produce good turbulence estimates.
In the project we want still to give high priority to getting turbulence measurements from the flight
recorders. In this way we will expect important verification, which we can use for improving the system
further. There is also a need for looking closer into the accuracy of the AMDAR data.
Since the start of the project we have carried out validation the SIMRA model by comparing to standard
SYNOP reports. For the new airports the results are comparable to the results gained at the airports
Hammerfest and Sandnessjøen for which the model has been operated since 2005/2006. We have by using
available SYNOP and AMDAR reports and comparing to Hammerfest and Sandnessjøen (where we in the
test phase received pilot reports), demonstrated that the system operates with similar quality at the all
airports. Bearing in mind the evaluation we did of the pilot reports collected during the early years of the
project, we can conclude that we have established turbulence forecast of satisfactory accuracy at the new
airports also. . For Fagernes airport there is however yet a too short period covered to estimate the accuracy
of the forecasts.
Above we used the term similar and not equal about the quality at the different airports. There is still a
degree of uncertainty connected to the quality of forecasts for the new airports. There has as discussed
earlier not been any test phase with pilot reports at those sites. Such reports would have been very helpful.
We will stress that CFD-models (as SIMRA) and meteorological models are generalized models. As such
models they do not need turbulence observations for the sites where they are applied. Local variations in
quality caused by weaknesses in modelling of local forcing must be expected in addition to local variations
caused by frequency of phenomena not described by the models. We have, however chosen SIMRA
models grids and resolution with emphasis on resolving the features causing the turbulence at each airport.
Therefore we will conclude that the turbulence forecasts should be expected to be of similar quality at all
airports.
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
48
References
Haugen, J. E., Aspelien, T., Homleid, M., Køltzow, M. and Vignes, O.(2008)
Operasjonell Hirlam med 12km, 8km og 4 km gitter. (in Norwegian)
Met.no note No. 1, The Norwegian Meteorological Institute, 2008
Midtbø, K.H., Bremnes, J.B., Homleid, M., Ødegaard, V (2008) Verification of wind forecasts for the
airports Hammerfest, Honningsvåg, Sandnessjøen, Ørsta, Værnes, Sandane and Narvik.
Met.no note No. 2, The Norwegian Meteorological Institute, 2008
Midtbø, K.H., Bremnes, J.B., Homleid, M., Ødegaard, V (2008) Verification of wind forecasts for the
airports Hammerfest, Honningsvåg, Sandnessjøen, Ørsta, Værnes, Sandane and Narvik for the period
1.1.2008 to 30.4.2008.
Met.no note No. 10, The Norwegian Meteorological Institute, 2008
Midtbø, K.H., Bremnes, J.B., Homleid, M., Ødegaard, V (2008) Verification of wind forecasts for the
airports Hammerfest, Honningsvåg, Sandnessjøen, Ørsta, Værnes, Sandane and Narvik for the period
1.5.2008 to 31.8.2008.
Met.no note No. 18, The Norwegian Meteorological Institute, 2008
Midtbø, K.H., Bremnes, J.B., Homleid, M., Ødegaard, V (2008) Verification of wind forecasts for the
airports Hammerfest, Honningsvåg, Sandnessjøen, Ørsta, Værnes, Sandane and Narvik for the period
1.9.2008 to 31.12.2008.
Met.no note No. 20, The Norwegian Meteorological Institute, 2009
Midtbø, K.H., Bremnes, J.B., Homleid, M., Ødegaard, V (2008) Verification of wind forecasts for the
airports Hammerfest, Honningsvåg, Sandnessjøen, Ørsta, Værnes, Sandane and Narvik for the period
1.1.2009 to 30.04.2009.
Met.no note No. 21, The Norwegian Meteorological Institute, 2009
Midtbø K. H. (2006) Validation of modelled surface winds for Hammerfest, Sandnessjøen and Molde
airports, Research Report No. 11, The Norwegian Meteorological Institute, 2006,
Kaplan et al. (2005) Characterizing the severe turbulence environments associated with commercial
aviation accidents. Part 2: Hydrostatic mesoscale numerical simulations of supergradient wind flow and
streamwise ageostrophic frontogenesis. Meteorology and Atmospheric Physics 88, 153-173
Eidsvik K.J and T.Utnes (1997) Flow separation and hydraulic transitions over hills modelled by the
Reynolds equations. Journal of Wind Engineering and Industrian Aerodynamics 67&68, 403-413
Hertenstein R. F. and J.P. Kuettner (2005) Rotor types associated with steep lee topography: influence of
the wind profile. Tellus 57A, 117-135.
Hurley, P. J (1997) An evaluation of several turbulence schemes for the prediction of mean and turbulent
fields in complex terrain. Boundary-Layer Meteorology 83, 43-73.
Sharman, R., Tebaldi, C., Wiener, G., Wolff, J.
An Integrated Approach to Mid- and Upper-Level Turbulence Forecasting
Weather and Forecasting 2006 21: 268-287
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
49
Dannevig, P. Hoem, V. (1978)
Turbulensforholdene ved norske flyplasser (in Norwegian), Technical Report No. 30, The Norwegian
Meteorological Institute
Postal address
P.O.Box 43, Blindern
NO-0313 OSLO
Norway
Office
Niels Henrik Abelsvei 40
Telephone
+47 22 96 30 00
Telefax
+47 22 96 30 50
e-mail: met@met.no
Internet: met.no
Bank account
7694 05 00628
Swift code
DNBANOKK
50