Use of High-Resolution Dynamical Adaptation in Operational Suite
and Research Impact Studies
Stjepan Ivatek-Šahdan and Martina Tudor
Meteorological and Hydrological Service, Grič 3, HR-10000 Zagreb, Croatia
ABSTRACT
The article presents a very high-resolution dynamical adaptation of the wind field with the
hydrostatic version of the ALADIN model used in the operational suite, including a sensitivity study
of a recent case of bura and a MAP IOP 15 case study. Bura is a strong wind with severe gusts; it
occurs suddenly and affects traffic on the eastern Adriatic. It has a large spatial variability and
depends on an upstream terrain configuration. The described approach makes it possible to predict
the occurrence, strength and spatial variability of the wind field in the mountainous area of the
eastern Adriatic coast. Although this area is often affected by severe wind events (bura), measured
wind data do not provide adequate coverage and are usually not representative enough. Good
performance of the operational wind prediction encouraged its use for estimating the expected
extreme wind speeds on the route of a new highway. Two sensitivity tests were carried out to
explore the impact of the neglected processes. It was found that the impact is negligible in the case
of severe bura event.
Keywords: bura, dynamical adaptation
1. Introduction
The main focus of this article is to present the ability of the recently established operational forecast
model to predict the occurrence, strength and spatial variability of bura (the Croatian expression for
bora). There have been numerous theories developed about the nature of bura. The so called “fall
wind” theory was not supported by ALPEX field experiment data (Smith, 1987). The hydraulic theory
(Smith, 1985) and the wave breaking theory presented by Peltier and Clark (1979) were applied to
some ALPEX cases in Durran (1986). Klemp and Durran (1987) investigated the hydraulic theory
applied to bura in the framework of a numerical model. There are numerous case studies (in
Croatia) related to the bura wind (Brzović 1998/99, 1999, Brzović and Jurčec, 1997, Ivančan-Picek
and Glasnović 2000, Tutiš et al. 2000). For example, Jurčec and Brzović, (1995) discuss 2 cases
and mention 3 mechanisms of bura; fall wind, mountain wave breaking and the internal hydraulic
theory. They found that extreme wind speeds were associated with the surface pressure
perturbation and superadiabatic layers in the low troposphere, and concluded that a non-hydrostatic
fine mesh model with a realistic topography and sophisticated parameterizations of moist and
turbulent processes is necessary to predict the onset, duration and strength of bura. Numerical
simulations of the Adriatic cyclone often associated to bura were performed (Brzović and Jurčec,
1997), including sensitivity studies, where the Dinaric Alps or moist processes were removed from
the model (Brzović, 1999). Removal of the Dinaric Alps significantly weakened the cyclone and bura
diminished. Enger and Grisogono (1998) have shown that the enhanced pressure gradient due to
low pressure above the sea behind the orographic obstacle, caused by the temperature difference
between sea and land brings bura further from the coast.
Although the theoretical background and mechanisms that control the development of bura were
explored in numerous studies, its strength, occurrence and spatial variability have not been
operationally well predicted. Theory suggests that a sophisticated model with a very high resolution
is necessary for this purpose. This would also require a powerful computer.
Žagar and Rakovec (1999) have published the results where the ALADIN model was used for a
high-resolution dynamical adaptation of the wind field to orography. Instead of the 48-hour evolution
of the processes in the dynamical adaptation using the full model, they reduced the number of
vertical levels in the upper troposphere, excluded the moist processes and ran the model for 30 min
(30 time steps) using the same data as initial and coupling (large scale forecast of the driving
model) data. This was enough for the wind field to adapt to high-resolution orography. The method
is expected to perform better when the pressure gradient and orography play the most important
role and when the flow is strong.
1
Bura occurs when there is a strong pressure gradient across the coastal mountains. It is connected
with advection over the coastal mountains towards the sea, with anticyclone over relatively cold
ground and/or cyclone development over the relatively warmer sea and is therefore more severe
and frequent during the cold part of the year. Bura is the consequence of synoptic forcing
interaction with a mountain in the immediate vicinity of the sea. The flow is blocked or significantly
modified by the mountains. The pressure gradient wants to push the air further but the mountain
blocks the flow allowing the air to find its way through the gaps in the mountain range. Bura is
characterized by its sudden occurrence, spatial variability and dependence on upstream terrain
configuration. Since bura is a strong wind which depends on the synoptic forcing that is modified (or
shaped) by local terrain, the method introduced by Žagar and Rakovec (1999) should improve its
forecast. It is computationally cheap, so it was introduced in the operational suite for 5 2-km
resolution domains covering the mountainous parts of Croatia.
A strong wind with severe gusts that often affects transport requires good forecasting. As the
forecast has performed well in practice (see later), the method has been used to estimate the
expected extreme wind speed along the route of a highway, since measurement data were scarce.
Data from several new automatic stations along the highway path are now available and will be
used to show that the simplifications introduced in the operational suite were justified.
The operational suite is described in Section 2. The process of obtaining a high resolution wind
forecast is explained in Section 3, followed by the results of the operational suite, impact and
sensitivity study presented in Section 4 with conclusions in Section 5.
2. The operational suite
ALADIN (Aire Limitee Adaptation Dynamique developement InterNational) is a limited-area model
(LAM) built on the basis of the global model IFS/ARPEGE (ARPEGE - Action de Recherche Petite
Echelle Grande Echelle, IFS - Integrated Forecast System). ALADIN keeps the same vertical
discretisation, grid-point dynamics and physics as the global ARPEGE model. ALADIN uses the
spectral technique for the horizontal representation of fields. A double Fourier representation is
used, limited by an elliptic truncation (Machenhauer and Haugen, 1987). A hybrid pressure-type 
coordinate (Simmons and Burridge, 1981) on 37 model levels is used in the vertical with the finite
difference method. Primitive prognostic equations are solved for the wind components, temperature,
specific humidity and surface pressure, using the two-time-level semi-implicit semi-lagrangian
integration scheme.
The physical parameterization package includes vertical diffusion parameterization (Louis et al.,
1982) with shallow convection (Geleyn, 1987); convective and stratiform processes are treated
separately by a Kessler-type large-scale precipitation scheme and a modified Kuo-type deep
convection scheme. Radiation is parameterized according to Geleyn and Hollingsworth (1979) and
Ritter and Geleyn (1992). The transport of moisture and heat vertically in the soil are parameterized
in two layers (Giard and Bazile, 2000). To ensure a continuous transition from large-scale to smallscale data, an intermediate zone is defined - the Coupling zone - where a large-scale solution
computed with the global ARPEGE, or a bigger LAM model is mixed with the solution resulting from
the ALADIN integration (Radnoti, 1995) following the relaxation technique (Davies, 1976). The
central zone represents the region of meteorological interest, where the forecast is fully adapted to
small-scale conditions.
In the Croatian Meteorological Service, ALADIN runs operationally for 00 and 12 UTC. First, the 48hour forecast is obtained using ALADIN on the LACE domain, which covers almost the whole
Europe with 12.2-km resolution. The initial and boundary conditions are obtained from the analysis
and forecasts of the ARPEGE global model run in Meteo-France, with DFI (Digital Filter
Initialization, Lynch and Huang, 1994) on the analysis. Afterwards, the output fields are dynamically
adapted to the Croatian domain with 8-km resolution using the 48-hour integration of the ALADIN
model with a full physics package. The output wind field of the latter is finally dynamically adapted
to a 2-km resolution using the procedure described in the following Section. The model is run on 
levels and, prior to the visualization, the fields are interpolated to pressure levels. Wind field is also
interpolated to the measurement level (10 m agl) and this is the surface wind forecast.
2
3. The operational high-resolution dynamical adaptation
Dynamical adaptation is run sequentially for each output file. The procedure takes the output fields
from the operational ALADIN with 8-km resolution, interpolates them to the 2-km resolution,
perturbs the fields and destroys the quasi-stationary state. Also, the terrain representation in the
model is different, as shown in Figure 1. New valleys open and peaks rise higher so that the flow
needs to adapt to these new terrain features. The model fields adapt to high-resolution relief by
running the numerical model from this initial state with the new resolution for the time necessary to
achieve the quasi-steadiness (30-min, with 60 sec time step). Part of the model physics, describing
the moist and radiation processes is excluded and the number of levels in the upper troposphere
and stratosphere is reduced. This approach can not predict local thermal circulation or convectioninduced circulation. The method might also fail in the case of front propagation because a stationary
state is not reached, or when large-scale forecast of the driving model fails. The method performs
much better when the wind is strong enough to overcome the circulation induced by the local
thermal or convection-induced circulation. However, when the main wind forcing is the pressure
gradient over mountains, as it is the case with bura, the neglected contributions are far less
important than those retained in the model (this will be shown in the final part of Section 4).
The model is run on 15 levels in the vertical (the lowest model level for both the 8-km resolution
model and then 2-km resolution model is approximately 17 m above terrain), using only the vertical
diffusion and GWD (gravity wave drag) parts of the physical parameterizations package. The
vertical resolution in the lower troposphere is the same for all domains. The operational suite takes
about 20 minutes on a LINUX PC for a 48-hour forecast with a 3h interval using the described highresolution dynamical adaptation for one domain (80x80 points).
3.1. Representation of the orography
Figure 1. Orography representation in the ALADIN/Croatia 8-km resolution domain (left), zoomed to
the area covered by the dynamical adaptation (center) and in the dynamical adaptation 2-km
resolution domain (right). The scale is the same for all figures. The black lines represent the
land/sea mask and political borders from the graphic software.
Terrain height is represented as smoothed in the Croatian domain, shown in Figure 1. The same
scale is used in all 3 figures. For example, the Zavižan station is placed at a height of 1594 m, the
altitude of the nearest grid-point in the LACE domain with 12-km resolution is 715 m, in the Croatian
domain with 8-km resolution it is 888 m and in the Senj domain for dynamical adaptation with 2-km
resolution 1512 m. The increase in resolution produces a large impact on the spatial variability of
the surface wind field and the maximum strength it reaches.
4. Results of the operational dynamical adaptation
The results of the operational forecast using the ALADIN model in extreme weather situations
concerning a strong flow across the Dinaric Alps will be presented. The performance of the
operational high-resolution dynamical adaptation will be illustrated with 3 examples; first the MAP
IOP 15 case, followed by a more recent case of bura that is documented by additional
measurement data and a sensitivity study of the impact of the different simplifications introduced in
the operational version.
3
4.1. MAP IOP 15
The weather situation is described according to the model output from the 00 UTC run on
November 6th on the Croatian domain and the high-resolution dynamical adaptation of the wind field
from the same model run. All fields are 42-hour forecasts valid at 18 UTC on November 7th 1999.
The 850-hPa wind field (Figure 2, top, left) has a strong ageostrophic component. The flow is
stronger as it approaches the coastline but then it suddenly slows down at this height, while the flow
close to the surface becomes much stronger, as can be seen from the surface wind field shown in
the same figure, bottom-left. Behind the mountain ridge, the air is dry and relatively warm compared
to the air in the flow that approaches the mountains from inland (same figure, top right). The
movement of the air immediately downstream of the mountain is mostly downward and adiabatic.
The vertical cross-section of the wind speed and potential temperature fields, in Figure 2 (bottom
right), shows a stable layer of air close to the ground, a less stable air above and the maximum
vertical wind speed just below the ridge.
Figure 2. AT850 (absolute topography of the 850 hPa isobaric surface) and the wind field on the
same surface (top left), RT500/1000 (relative topography of the 500 hPa surface over the 1000 hPa
surface) and vertically averaged relative moisture between those layers (top right), surface wind
field and mean sea level pressure (bottom left) and vertical cross-section of the vertical wind speed
(omega - Pa/s, shaded) and potential temperature (isolines) along lat=44.24 (bottom right).
A comparison of the vertical cross-sections of the horizontal wind speed and direction for the 8-km
and 2-km resolutions (Figure 3) shows that the wind maximum is much closer to the ground and to
the mountain slope in the high-resolution model. The arrows show the direction of the horizontal
wind. Wind velocity changes significantly with height above the sea. This is first observed as a
change in direction, but as the flow close to the surface strengthens, it weakens at the upper levels.
4
The weakening of the flow aloft (about the 850 hPa isobaric surface height) and the strengthening
of the flow adjacent to the surface suggests that at least part of the energy of the flow is transported
downwards and confined to a shallow layer, resulting in a low-level jet, by stably stratified air
adjacent to the ground.
Figure 3. Vertical cross-sections along lat=44.24 with 8-km resolution (top row) and 2-km resolution
(bottom row) for November 7th 1999, 12 UTC (left) and 18 UTC (center) and for November 8 th 1999,
00 UTC (right). The arrows on all pictures represent the horizontal wind direction, and the shaded
areas correspond to the wind speed in (m/s).
The location of the Maslenica Bridge, south of the Velebit Mountain, is very important for road
traffic. An automatic station has been installed there. With measured wind speeds of the magnitude
shown in Figure 4, the bridge is closed for traffic. Four consecutive forecasts at 8- and 2-km
resolutions are shown in the same figure. The 8-km resolution forecast produces too weak winds (in
the case of bura) for this bridge just downstream of the mountain pass on the Velebit Mountain (part
of the Dinaric Alps). The 2-km resolution forecast gives wind speeds that correspond much better to
the measurements. The new method performs well when the forecast is compared to these and
other measurement data (not shown here) for bura events. Since a new highway was under
construction, more automatic stations were introduced providing more data for verification. But for
this, a more recent case should be studied.
5
Figure 4: Wind measurement data, 8-km resolution forecasts (full markings and 00 UTC run in the
legend) and 2-km resolution dynamical adaptation (open markings and dada in the legend) of the
surface wind for Maslenica bridge for the period from 6th to 11th November 1999, 00 UTC. The 8km resolution forecast data are from the ALADIN 48-h forecast 00 UTC runs for the 6th, 7th, 8th
and 9th of November 1999 and DADA data are from the Dynamical adaptation of the forecasts on
the Maslenica domain.
4.2. February 2003 case
The first study of the expected wind speeds along the route of the new road was based on a few
cases of bura, from the winters of 2000/01 and 2001/02, (not shown), when the wind measured on
the Maslenica Bridge was strong to severe and well predicted. New automatic stations were
established in the autumn of 2002 and the beginning of 2003. Measurements from those stations
present another justification for the operational configuration of the dynamical adaptation to 2-km
resolution using the ALADIN model.
The weather situation is characterized by a wide cyclone in the southeastern edge of the Pannonian
valley and a smaller cyclone above the Adriatic with a ridge stretching over the coastal part of
Croatia forming a strong pressure gradient with isobars parallel to the coast. Both low-pressure
systems have a precipitation associated to them.
The zoom of the 8-km resolution surface wind field forecast and the surface wind field as predicted
by the high-resolution dynamical adaptation for 3 domains is shown in Figure 5. The forecasts
obtained by both are compared to the measurements on 2 meteorological stations in Figure 6. The
operational 2-km resolution forecast is compared to two alternatives in Figure 7. and all three setups are compared to the measured data in Figure 8. The comparison of the wind fields shows that
the wind field obtained by the 2-km resolution dynamical adaptation is much stronger in certain
areas than predicted by the 8-km resolution forecast (Figure 5), and corresponds better to the
measured data (Figure 6 and Figure 8). The data at the edges of the dynamical adaptation domain
(the intermediate or coupling zone) remain the same as in the driving model, while the impact of the
adaptation is strongest in the central part of the domain (Figure 5). The increase in the resolution
produces a large impact on the spatial variability of the surface wind field. The comparison of the
wind fields obtained with 8-km and 2-km resolutions shows that the main characteristics of the bura
wind situation are already captured by the 8-km resolution forecast. However, the maximum wind
strength and spatial variability are underestimated.
6
In Figures 6 and 8, the measurements of wind speed and the results of the simulation with 8-km
and 2-km grid models are compared with measured data for the eastern part of the Adriatic Sea.
Figure 6 shows a comparison with measured data for February 2003 on the Baričević Viaduct and
at the exit from the Ledenik Tunnel, giving an overview of the forecast behavior in strong wind
conditions in that area. Figure 8 is focused on one forecast run using different configurations of the
2-km resolution dynamical adaptation, showing a similar model behavior for 2 more stations – two
bridges: Maslenica and Pag. All the instruments are situated 10 m agl in the vicinity of the
construction. The instruments measure wind speed in the range of 0.2 to 70 m/s, with 0.1 m/s
precision, and wind direction ranging from 0° to 360°, with 5° precision, in 1-sec intervals. The 10minute average wind speed and direction and the strongest wind gust in the 10-minute interval and
its direction are stored. Only the wind speed is used in the comparisons below. The model output is
the wind speed taken usually at the nearest point of the model (in the horizontal) from the 10 m agl
wind forecast.
Figure 5. Zoom of the 8-km resolution forecast (top row), and the 2-km resolution dynamical
adaptation (bottom row) of the 10 m agl wind field for 3 domains, 30 hours forecast, valid at 06
UTC, February 5th 2003.
Figure 6. Measured wind speed for the Baričević Viaduct (left) and the Ledenik Tunnel (right)
automatic stations and modeled data from the closest model point for February 2003. Measured 10
min average wind speed (dark blue), 10 min maximum (light blue), all model forecasts for February
2003 (00 and 12 UTC runs) with 8-km resolution (orange) and 2-km resolution operational
7
dynamical adaptations (yellow). The 2-km resolution predicts the occurrence and strength of the 10min average wind speed well.
Recent model data from new automatic stations shows that the 2-km resolution dynamical
adaptation corresponds well with the observed data. The use of dynamical adaptation with the
ALADIN model for the impact study and the use of ALADIN operational forecast products have
proven to be a good choice (see later). In Figure 6, 8-km and 2-km resolution forecast data for each
operational model run (every day, 00 and 12 UTC run) are compared to the measurements, the 10min average and 10-min maximum wind speed. When the model starts, it can be seen that the
surface wind speed in the analysis is much lower than the measured one, because it is originally
from the global analysis. The wind adapts to the 8-km orography during the first 3 hours of
integration, but the measured strength is not reached, because the 8-km resolution representation
of orography is insufficient. Further dynamical adaptation of the wind field to 2-km resolution
orography gives much better results. The obvious failure of predicting wind speed for the period
from 6th to 7th February 2003 was caused by a wrong large-scale forecast, as can be seen in detail
in Figure 8. Only one forecast run is shown there, but the discrepancy in the forecasted wind speed
on 6th February that is not measured persists during the following forecast runs.
4.3. Sensitivity study
When compared to the model version used at 8-km resolution, the 2-km resolution version does not
have thermal nor moist processes included, except that the temperature and moisture fields do exist
in the input data. It also lacks the evolution of the state of the atmosphere. Therefore operational
wind speed forecasts were compared with the two alternatives:
- the dynamical adaptation run using the full physical parameterization package and
- the 48-hours forecast run with 2-km resolution, using the full physical parameterization package.
In the case analyzed, bura was strong, and there was precipitation inland. In Figure 7, the surface
wind fields and the vertical cross-sections of the wind field for the 3 set-ups of the 2-km model are
compared. Although the operational set-up in general gave slightly higher wind speeds than the
other two, a comparison with the measurements shows that the choice of the operational dynamical
adaptation set-up for the estimation of the expected surface wind speed in extreme cases was good
(in the sense that it requires less computer resources giving the same forecast quality). Same
analysis on other cases gives similar results. A previous sensitivity study (Brzović and Jurčec, 1997)
using the same numerical model (with 10.6-km resolution) showed that the wind is weaker when the
moist processes are omitted, while this one has the opposite result. The explanation for this
apparent discrepancy could be: the moist and thermal processes were already accounted for by the
driving model; including them in the high-resolution adaptation model modifies the process of
adaptation. However, the differences in forecasted surface wind are very small as can be seen in
Figure 7 and Figure 8.
8
Figure 7. Surface wind (10 m agl) for the operational dynamical adaptation (top, left), dynamical
adaptation using the whole physics package (top, center) and full forecast integration on the
dynamical adaptation domain (top, right) for 4th February 2003, 00UTC forecast. Vertical wind
cross-section along lat=44.24 for the operational dynamical adaptation (bottom, left), using the
whole physics package (bottom, center) and forecast with 2-km resolution (bottom, right). The
arrows on all pictures represent the direction of the horizontal wind, and the shaded areas
correspond to the wind speed in (m/s). The vertical cross-section is along lat=44.24, and the
highway stretches along this cross-section from lon=15.52 to lon=15.65.
The model dynamics used in the 8-km resolution run and the 2-km resolution run is the same. The
part of physics omitted in the latter one does not contribute significantly to the forecasted wind
speed, as shown in Figure 8. The difference between them is the resolution of orography – the way
the upstream terrain configuration is represented in the model. The main features of the actual
topography exist in the 2-km resolution representation in the sense that the mountain peaks and
passes have suitable heights and positions.
9
Figure 8. Measured wind speed from automatic stations and modeled data for the 00 UTC run on
February 4th 2003 from the closest model point. Measured 10 min average wind speed in m/s (dark
blue), 10 min maximum (light blue), 8-km resolution model forecast (red), 2-km resolution dynamical
adaptations; operational (orange), using whole physics package (yellow) and 2-km resolution 48hour model integration (green).
5. Conclusions
The operational forecast of the occurrence, strength and spatial variability of bura, using the
described method, has proved successful. This suggests that not all of the above mentioned
processes are important for the successful prediction of bura. The MAP IOP 15 case study (Tudor
and Ivatek-Šahdan, 2002) and the sensitivity study presented here show that this approach may be
used to estimate the expected extreme wind speeds in areas where measurements are inadequate.
The understanding of the bura phenomenon has been among the most important issues in Croatian
meteorology as this phenomenon has significant impact on local infrastructure and road
communications. The prediction of severe bura events used to be based on the experience of the
forecaster, who would recognize a synoptic situation that would lead to a severe bura event. But the
strength of such event was a matter of speculation. Numerous cases were analyzed and theories
developed that were tested on data obtained during the ALPEX and MAP experiments. However,
available operational models have not predicted the occurrence, speed, spatial variability and
maximum speed of the bura wind. It was expected that this can be accomplished using a
nonhydrostatic model at a very high resolution with sophisticated physical parameterizations.
The local version of the operational forecast using the ALADIN model has been established
including a new approach of the high-resolution dynamical adaptation for the mountainous parts of
Croatia. Special care was taken that the terrain height in the model is close to the actual height of
the mountain peaks and passes. Although the model did not include all the processes, anticipated
to be important for bura in previous studies, the forecast was found very successful. The model was
not designed for very high resolution, it is a limited-area version of the global model. It is also
hydrostatic, the non-hydrostatic effects are omitted completely. The physical parameterizations of
moist, thermal and radiation processes were omitted. The effects of moisture and heating/cooling
are already included in the input fields and do not require a high-resolution representation for this
phenomenon. They are important when the synoptic situation is developing - in the large-scale
model. Local wind, however, is the result of the interaction of this large-scale pressure field and
small-scale orography. The 8-km resolution model already predicts the moment when bura will start
and end, while the local wind speed is much better predicted by the 2-km resolution dynamical
adaptation.
Acknowledgements:
This study was partly supported by the Croatian Ministry of Science and Technology (project:
0004001). The authors would like to thank the anonymous reviewers for their very valuable
comments and suggestions.
LITERATURE
Brzović, N., 1998/99: A case study of a cold air outbreak on 20th April 1997 over Croatia. Cro. Met.
J., 33/34, 1-18.
Brzović, N., 1999: Factors affecting the Adriatic cyclone and associated windstorms. Contr. Atmos.
Phys., 72, 51-65.
Brzović, N. and V. Jurčec, 1997: Numerical simulation of the Adriatic cyclone development.
Geofizika, 14, 29-46
Courtier, P. and J.-F. Geleyn, 1988: A global spectral model with variable resolution – application to
the shallow-water equations. Quart. J. Roy. Meteor. Soc., 114, 1321-1346.
Courtier, P., C. Freydier, J.-F. Geleyn, F. Rabier and M. Rochas, 1991: The ARPEGE project at
METEO-FRANCE. Proceedings from the ECMWF workshop on numerical methods in
atmospheric models. 9-13 September 1991, 2, 193-231.
10
Davies, H.C., 1976: A lateral boundary formulation for multi-level prediction models. Quart. J. Roy.
Meteor. Soc., 102, 405-418
Durran, D.R., 1986: Another look at downslope windstorms. Part I: The development of analogs to
supercritical flow in an infinitely deep continuously stratified fluid. Journal of Atmospheric
Sciences, 43, 2527-2543.
Enger, L. and B. Grisogono, 1998: The response of bora type flow to the sea surface temperature.
Q.J.R. Meteorol. Soc., 124, 1227-1244.
Geleyn, J.-F., 1987: Use of a modified Richardson number for parameterizing the effect of shallow
convection. In: Matsuno Z. (ed)., Short and medium range numerical weather prediction,
special volume of J. Meteor. Soc. Japan, 141-149.
Geleyn, J.-F. and A. Holligsworth, 1979: An economical analytical method for computation of the
interaction between scattering and line absorption of radiation. Contr. Atmos. Phys., 52, 1-16
Giard, D. and E. Basile, 2000: Implementation of a New Assimilation Scheme for Soil and Surface
Variables in a Global NWP Model. Mon. Wea. Rev., 128, 997-1015
Ivančan-Picek, B. and D. Glasnović, 2000: Capability of the ALADIN model to predict heavy
precipitation and strong wind events during the MAP, Proceedings of the 9th International
ALADIN Workshop, Brussels, 6-8 March 2000.
Jurčec, V. and N. Brzović, 1995: The Adriatic bora: special case studies, Geofizika, 12, 15-32
Klemp, J.B. and D.R. Durran, 1987: Numerical modeling of bora winds. Meteorol. Atmos. Phys., 36,
215-227
Louis, J.-F., M. Tiedke, and J.-F. Geleyn, 1982: A short history of PBL parametrisation at ECMWF.
Proceedings from the ECMWF Workshop on Planetary Boundary Layer Parametrisation, 2527 November 1981, 59-79
Lynch, P. and X.-Y. Huang, 1994: Diabatic initialization using recursive filters.Tellus, 46A, 583-597.
Machenhauer, B. and J.E. Haugen, 1987: Test of a spectral limited area shallow water model with
time-dependent lateral boundary conditions and combined normal mode/semi-Lagrangian
time integration schemes. Proceedings from the ECMWF Workshop on techniques for
horizontal discretisation in numerical weather prediction models. 2-4 November 1987, 361377
Peltier, W.R. and T.L. Clark, 1979: The evolution and stability of finite-amplitude mountain waves.
Part II: Surface drag and severe downslope windstorms. Journal of Atmospheric sciences, 36,
1498-1529.
Radnoti, G., 1995: Comments on ‘A Spectral Limited-Area Formulation with Time-dependent
Boundary conditions to the Shallow-Water Equations’. Mon. Wea. Rev., 123, 3122-3123.
Ritter, B. and J.-F. Geleyn, 1992: A comprehensive radiation scheme for numerical weather
prediction models with potential applications in climate simulations, Mon. Wea. Rev., 120,
303-325.
Schmidt, F., 1977: Variable fine mesh in a spectral global model. Beitr. Phys. Atmos., 50, 211-217
Simmons, A. J. and Burridge D.M., 1981: An energy and angular momentum conserving vertical
finite-difference scheme and hybrid vertical coordinate. Mon. Wea. Rev., 109, 758-766
Smith, R.B., 1985: On severe downslope wind. Journal of Atmospheric Sciences, 42, 2597-2603.
Smith, R.B., 1987: Aerial observations of Yugoslavian bora. Journal of Atmospheric Sciences, 44,
269-297.
Tudor, M. and S. Ivatek-Šahdan, 2002: MAP IOP 15 case study. Cro. Met. Jour., 37, 1-14.
Tutiš, V., B. Ivančan-Picek and D. Glasnović, 2000: MAP SOP severe Adriatic bora event. 26th
International Conference on Alpine Meteorology, 11-15 September 2000., Innsbruck,
Oesterrechische Beitraege zu Meteorologie und Geophysik. Heft Nr.23/Publ.Nr.392. (CD
ROM)
Žagar, M. and J. Rakovec, 1999: Small-scale surface wind prediction using Dynamical adaptation.
Tellus, 51A, 489-504.
11