A 3-D ensemble variational (3DEnVar) data assimilation method has been implemented and tested for oceanographic data assimilation of sea surface temperature (SST), sea surface salinity (SSS), sea ice concentration (SIC), and salinity and temperature profiles. To damp spurious long-range correlations in the ensemble statistics, horizontal and vertical localisation was implemented using empirical orthogonal functions. The results show that the 3DEnVar method is indeed possible to use in oceanographic data assimilation. So far, only a seasonally dependent ensemble has been used, based on historical model simulations. Near-surface experiments showed that the ensemble statistics gave inhomogeneous and anisotropic horizontal structure functions, and assimilation of real SST and SIC fields gave smooth, realistic increment fields. The implementation was multivariate, and results showed that the cross-correlations between variables work in an intuitive way, for example, decreasing SST where SIC was increased and vice versa. The profile data assimilation also gave good results. The results from a 25-year reanalysis showed that the vertical salinity and temperature structure were significantly improved, compared to both dependent and independent data.

The ETKF rescaling scheme has been implemented into the HIRLAM forecasting system in order to estimate the uncertainty of the model state. The main purpose is to utilize this uncertainty information for modelling of flow-dependent background error covariances within the framework of a hybrid variational ensemble data assimilation scheme. The effects of rank-deficiency in the ETKF formulation is explained and the need for variance inflation as a way to compensate for these effects is justified. A filter spin-up algorithm is proposed as a refinement of the variance inflation. The proposed spin-up algorithm will also act to prevent ensemble collapse since the ensemble will receive 'fresh blood' in the form of additional perturbation components, generated on the basis of a static background error covariance matrix. The resulting ETKF-based ensemble perturbations are compared with ensemble perturbations based on targeted singular vectors and are shown to have more realistic spectral characteristics.

The subsurface flow of high-saline water masses from the Bornholm Basin through the Stolpe Channel plays an important role for the renewal of the Baltic Central Basin deep waters. In order to determine whether rotating 11/2-layer hydraulic theory is an appropriate tool for describing this process, maximal-transport estimates based on climatological data from the Bornholm and Gdansk Basins have been established. These were found to deviate considerably from observational realities, and hence similar hydraulic considerations were also applied to more-or-less synoptic field data from a Finnish field campaign carried through in the mid-1980s. Also in this case significant differences were found between calculated transport capacity and observations. Since it furthermore was demonstrated that the characteristics of the observed cross-channel hydrographic structure could be explained using a frictional-balance model of the deep-water flow, it has been concluded that a hydraulic framework, although providing an upper bound of the transport, is of limited use when dealing with the Stolpe-Channel overflow. Although it cannot be excluded that the inflow is inviscid, but submaximal, it is more likely that the transport is governed by the combined effects of friction and wind forcing.

We propose to add an extra source of information to the data-assimilation of the regional HIgh Resolution Limited Area Model (HIRLAM) model, constraining larger scales to the host model providing the lateral boundary conditions. An extra term, J(k), measuring the distance to the large-scale vorticity of the host model, is added to the cost-function of the variational data-assimilation. Vorticity is chosen because it is a good representative of the large-scale flow and because vorticity is a basic control variable of the HIRLAM variational data-assimilation. Furthermore, by choosing only vorticity, the remaining model variables, divergence, temperature, surface pressure and specific humidity will be allowed to adapt to the modified vorticity field in accordance with the internal balance constraints of the regional model. The error characteristics of the J(k) term are described by the horizontal spectral densities and the vertical eigenmodes (eigenvectors and eigenvalues) of the host model vorticity forecast error fields, expressed in the regional model geometry. The vorticity field, provided by the European Centre for Medium-range Weather Forecasts (ECMWF) operational model, was assimilated into the HIRLAM model during an experiment period of 33 d in winter with positive impact on forecast verification statistics for upper air variables and mean sea level pressure.

The impact of dense saltwater inflows on the phosphorus dynamics in the Baltic Sea is studied from tracer experiments with a three-dimensional physical model. Model simulations showed that the coasts of the North West Gotland Basin and the Gulf of Finland, the Estonian coast in the East Gotland Basin are regions where tracers from below the halocline are primarily lifted up above the halocline. After 1 yr tracers are accumulated at the surface along the Swedish east coast and at the western and southern sides of Gotland. Elevated concentrations are also found east and southeast of Gotland, in the northern Bornholm Basin and in the central parts of the East Gotland Basin. The annual supplies of phosphorus from the deeper waters to the productive surface layers are estimated to be of the same order of magnitude as the waterborne inputs of phosphorus to the entire Baltic Sea. The model results suggest that regionally the impact of these nutrients may be quite large, and the largest regional increases in surface concentrations are found after large inflows. However, the overall direct impact of major Baltic inflows on the annual uplift of nutrients from below the halocline to the surface waters is small because vertical transports are comparably large also during periods without major inflows. Our model results suggest that phosphorus released from the sediments between 60 and 100 m depth in the East Gotland Basin contributes to the eutrophication, especially in the coastal regions of the eastern Baltic Proper.

This article compares interactively coupled atmosphere-ocean hindcast simulations with stand-alone runs of the atmosphere and ocean models using the recently developed regional ocean-atmosphere model NEMO-Nordic for the North Sea and Baltic Sea. In the interactively coupled run, the ocean and the atmosphere components were allowed to exchange mass, momentum and heat every 3 h. Our results show that interactive coupling significantly improves simulated winter sea surface temperatures (SSTs) in the Baltic Sea. The ocean and atmosphere stand-alone runs, respectively, resulted in too low sea surface and air temperatures over the Baltic Sea. These two runs suffer from too cold prescribed ERA40 SSTs, which lower air temperatures and weaken winds in the atmosphere only run. In the ocean-only run, the weaker winds additionally lower the vertical mixing thereby lowering the upward transport of warmer subpycnocline waters. By contrast, in the interactively coupled run, the ocean-atmosphere heat exchange evolved freely and demonstrated good skills in reproducing observed surface temperatures. Despite the strong impact on oceanic and atmospheric variables in the coupling area, no far reaching influence on atmospheric variables over land can be identified. In perturbation experiments, the different dynamics of the two coupling techniques is investigated in more detail by implementing strong positive winter temperature anomalies in the ocean model. Here, interactive coupling results in a substantially higher preservation of heat anomalies because the atmosphere also warmed which damped the ocean to atmosphere heat transfer. In the passively coupled set-up, this atmospheric feedback is missing, which resulted in an unrealistically high oceanic heat loss. The main added value of interactive air-sea coupling is twofold: (1) the elimination of any boundary condition at the air-sea interface and (2) the more realistic dynamical response to perturbations in the ocean-atmosphere heat balance, which will be essential in climate warming scenarios.

The limited area model forecasting problem is a lateral boundary condition (LBC) problem in addition to the initial condition problem. The data assimilation has traditionally been considered as a process for estimation of the initial condition only, while for the limited area data assimilation this estimation may be extended to include also the LBCs, at least during the data assimilation time window when observations are available. A procedure for such a control of the LBCs has been included in the four-dimensional variational data assimilation (4D-Var) scheme for the HIgh Resolution Limited Area Model (HIRLAM) forecasting system. A description of this procedure is provided together with results from idealised as well as real data experiments. The results indicate that control of LBCs may be important with small forecast domains and in particular for weather disturbances moving quickly into and through the forecast domain.

A 3-dimensional variational data assimilation (3D-Var) scheme for the HIgh Resolution Limited Area Model (HIRLAM) forecasting system is described. The HIRLAM 3D-Var is based on the minimization of a cost function that consists of one term J(b). which measures the distance between the resulting analysis and a background field, in general a short-range forecast. and another term J(o). which measures the distance between the analysis and the observations. This paper is concerned with the general formulation of the HIRLAM 3D-Var and with Jb. while the companion paper by Lindskog and co-workers is concerned with the handling of observations, including the J(o) term, and with validation of the 3D-Var through extended parallel assimilation and forecast experiments. The 3D-Var minimization requires a pre-conditioning that is achieved by a transformation of the minimization control variable. This change of variable is designed as an operator approximating an inverse square root of the forecast error covariance matrix in the model space. The main transformations are the Subtraction of the geostrophic wind increment, the bi-Fourier transform, and the projection on vertical eigenvectors. The spectral bi-Fourier approach allows one to derive non-separable structure functions in a limited area model. in the form of vertically dependent horizontal spectra and scale-dependent vertical correlations. Statistics have been accumulated from differences between +24 h and +48 h HIRLAM forecasts valid at the same time. Results from single observation impact studies as well as results from assimilation cycles using operational observations are presented. It is shown that the HIRLAM 3D-Var produces assimilation increments in accordance with the applied analysis structure functions, that the fit of the analysis to the observations is in agreement with the assumed error statistics. and that assimilation increments are well balanced. It is also shown that the particular problems associated with the limited area formulation have been solved. These results, together with the results of the companion paper, indicate that the 3D-Var scheme performs significantly better than the statistical interpolation scheme.

The tangent-linear and the adjoint of the spectral High Resolution Limited Area Model (HIRLAM) have been derived as a first step in the development of a 4-dimensional variational data assimilation system for HIRLAM. The adjoint of the spectral HIRLAM was applied successfully to test the sensitivity of short-range forecast errors to initial conditions. These sensitivity experiments were carried out for a particular case study in addition to a full 5-day period. The results of the sensitivity experiments indicate an ability of the adjoint model to improve the assimilation of baroclinically developing systems and this may open possibilities for application of the adjoint model in a ''Poor mans 4-dimensional variational data assimilation'' in advance of the implementation of the full 4-dimensional variational data assimilation.

A 4-dimensional variational data assimilation (4D-Var) scheme for the HIgh Resolution Limited Area Model (HIRLAM) forecasting system is described in this article. The innovative approaches to the multi-incremental formulation, the weak digital filter constraint and the semi-Lagrangian time integration are highlighted with some details. The implicit dynamical structure functions are discussed using single observation experiments, and the sensitivity to various parameters of the 4D-Var formulation is illustrated. To assess the meteorological impact of HIRLAM 4D-Var, data assimilation experiments for five periods of 1 month each were performed, using HIRLAM 3D-Var as a reference. It is shown that the HIRLAM 4D-Var consistently out-performs the HIRLAM 3D-Var, in particular for cases with strong mesoscale storm developments. The computational performance of the HIRLAM 4D-Var is also discussed.

The adjoint of a limited area model has been used to study the sensitivity of 12 h forecast errors to initial and lateral boundary conditions. Upper troposphere potential vorticity and mean sea level pressure verification scores for 1 month of operational forecasts from the Swedish Meteorological and Hydrological Institute were used to select 2 cases with particularly poor forecast performance. The sensitivity experiments show that errors in initial data is the most likely explanation for one of the forecast failures, while errors in initial as well as lateral boundary data can explain the 2nd forecast failure. Results from the sensitivity experiments with respect to the lateral boundary conditions indicate that poor quality lateral boundary conditions may be improved by utilizing subsequent downstream observations within the model integration area. This result is of great significance with regard to the possibilities for applying 4-dimensional variational data assimilation (4DVAR) for limited area forecast models. Results from the sensitivity experiments also reveal, however, that the lateral boundary treatment in operational limited area models needs to be improved with regard to the mathematical formulation. It is furthermore shown that modifications to be applied to the lateral boundary conditions need to be determined with appropriate time resolution and that some filtering of these lateral boundary modifications has to be introduced to avoid enhanced high-frequency gravity wave noise in the vicinity of the lateral boundaries.

The katabatic wind system over the Greenland ice sheet is studied using simulations of the hydrostatic Norwegian Limited Area Model (NORLAM) and measurements of an instrumented aircraft. The structure and the dynamics of the katabatic wind over the ice sheet are investigated for a case study of the aircraft-based experiment KABEG (Katabatic wind and boundary layer front experiment around Greenland) in the area of southern Greenland in April/May 1997. Monthly mean Structures and individual contributions of the momentum budget integrated over the boundary layer are examined for one winter month. The NORLAM is able to simulate realistically the Structures of the katabatic wind system in the lowest 400 in. The comparison with KABEG aircraft measurements for a katabatic wind case with strong synoptic forcing shows good agreement for the momentum budget terms. The pure katabatic force represents the main mechanism for the boundary layer wind field. but a considerable influence of the large-scale synoptic forcing is found as well. Acceleration components from the NORLAM forecasts are also presented for the whole month of January 1990. The monthly mean fields show significant regional differences because of different inversion strengths and synoptic forcings. In particular. Southeast Greenland is influenced by transient synoptic cyclones and the associated cloud patterns. All other areas of the slopes of the Greenland ice sheet are characterized by a downslope katabatic acceleration. The pressure gradient force over the northwestern part of the Greenland ice sheet points in the direction of the local katabatic force, which explains the relatively strong monthly mean near surface winds over the ice. Over the southwestern and northeastern parts of Greenland, however, no significant synoptic support of the katabatic winds is present, and the synoptic pressure gradient is even opposed to the katabatic force in some regions.

Precipitation and evaporation over the Baltic Sea are calculated for a one-year period from September 1998 to August 1999 by four different tools, the two atmospheric regional models HIRLAM and REMO, the oceanographic model PROBE-Baltic in combination with the SMHI (1 x 1)degrees database and Interpolated Fields, based essentially on ship measurements. The investigated period is slightly warmer and wetter than the climatological mean. Correlation coefficients of the differently calculated latent heat fluxes vary between 0.81 (HIRLAM and REMO) and 0.56 (SMHI/PROBE-Baltic and Interpolated Fields), while the correlation coefficients between model fluxes and measured fluxes range from 0.61 and 0.78. Deviations of simulated and interpolated monthly precipitation over the Baltic Sea are less than 5 mm in the southern Baltic and up to 20 mm near the Finnish coast for the one-year period. The methods simulate the annual cycle of precipitation and evaporation of the Baltic Proper in a similar manner with a broad maximum of net precipitation in spring and early summer and a minimum in late summer. The annual averages of net precipitation of the Baltic Proper range from 57 mm (REMO) to 262 turn (HIRLAM) and for the Baltic Sea from 96 turn (SMHI/PROBE-Baltic) to 209 rum (HIRLAM). This range is considered to give the uncertainty of present-day determination of the net precipitation over the Baltic Sea.

A system for mesoscale analyses of selected variables has been developed. The analysed parameters are of general interest in operational weather forecasting, but normally not available from NWP systems, or available, but with a significantly lower quality than achieved by the mesoscale analysis system. A supplementary objective is to produce initial information to be used for now-casting techniques. Examples of parameters are precipitation, temperature, humidity, visibility, wind and clouds. The basis of the analysis system is the optimal interpolation technique (OI). The use of observations from automatic stations, radars and satellites have been investigated. The investigation indicates that a dense network of ordinary precipitation gauge measurements can produce more accurate analyses than more elaborate systems like radar that suffers from anomalous echoes and other errors.

An observation operator for Doppler radar radial wind measurements is developed further in this article, based on the earlier work and considerations of the measurement characteristic. The elementary observation operator treats radar observations as point measurements at pre-processed observation heights. Here, modelling of the radar pulse volume broadening in vertical and the radar pulse path bending due to refraction is included to improve the realism of the observation modelling. The operator is implemented into the High Resolution Limited Area Model (HIRLAM) limited area numerical weather prediction (NWP) system. A data set of circa 133 000 radial wind measurements is passively monitored against the HIRLAM six-hourly background values in a 1-month experiment. No data assimilation experiments are performed at this stage. A new finding is that the improved modelling reduces the mean observation minus background (OmB) vector wind difference at ranges below 55 km, and the standard deviation of the radial wind OmB difference at ranges over 25 km. In conclusion, a more accurate and still computationally feasible observation operator is developed. The companion paper (Part II) considers optimal super-observation processing of Doppler radar radial winds for HIRLAM, with general applicability in NWP.

This study investigated the diurnal cycle of precipitation in Sweden using hourly ground observations for 1996-2008. General characteristics of phase and amplitude for the diurnal cycle of precipitation, both in amount and frequency, were identified. In the warm season (April-September), the 'typical' afternoon (14-16 LST) peaks are dominant over inland Sweden, whereas late night to early morning (04-06 LST) peaks with relatively weak amplitude are discernable in the east coast along the Baltic Sea. The diurnal variation is almost negligible in the cold season (October-March), due to the weak solar radiation at high latitudes. The variations of convective activity forced by solar heating and modulated by geographical characteristics were suggested as primarily factors to invoke the cycles and spatial variation identified. The observed cycle was compared with the cycle simulated by a regional climate model. The model fairly well captures the spatial pattern of the phase of the diurnal cycle. However, the warm season afternoon peak is simulated too early and too uniformly across the stations, associated with too frequent occurrences of convective rainfall events with relatively light intensity. These discrepancies point to the need to improve the convection parametrization and geographic representation of the model.

A method to evaluate forecasts of total fractional cloud cover using satellite measurements is demonstrated. Cloud analyses in the form of monthly cloud climatologies are extracted from NOAA. AVHRR data which are compared to corresponding cloud forecast information from the HIRLAM and ECMWF numerical weather prediction models. The satellite-based cloud information is extracted for a summer month in 1994 and a winter month in 1995 by use of the SMHI cloud classification model SCANDIA. Cloud analyses are conducted for an area covering a substantial part of northern Europe. Deficiencies in forecasted cloud amounts are found for both models, especially the underestimation of cloudiness for short forecast lengths with the HIRLAM model. Forecast improvements using the HIRLAM model are indicated when introducing a cloud initialisation technique using cloud fields from initial 6-hour forecasts (first-guess fields). Future systematic validations using this technique are, however, needed to make firm conclusions on the general model behaviour. SCANDIA-derived cloud information is proposed as a valuable complement to other datasets used for cloud forecast validation (e.g., the SSM/I- and ISCCP data sets).

Two different approaches for improving the representation of background error standard deviations have been developed and introduced into the HIRLAM high-resolution limited area model 3-D variational data assimilation scheme. One of the methods utilizes a horizontally varying climatological background error standard deviation field, estimated from a time-series of innovations. The second approach attempts to take temporal and spatial variations of the background error standard deviations into account by applying an Eady instability measure to the background field. The two approaches are described in detail and their functionality is demonstrated. Parallel data assimilation and forecasts experiments indicate a slightly positive impact on average verification scores, and in addition a positive impact is demonstrated for an individual synoptically active case.

A 3-dimensional variational data assimilation (3D-Var) scheme for the HIgh Resolution Limited Area Model (HIRLAM) forecasting system is described. The HIRLAM 3D-Var is based on the minimisation of a cost function that consists of one term, J(b), which measures the distance between the resulting analysis and a background field, in general a short-range forecast, and another term. J(o), which measures the distance between the analysis and the observations. This paper is concerned with J(o) and the handling of observations, while the companion Paper by Gustafsson et al. (2001) is concerned with the general 3D-Var formulation and with the J(b) term. Individual system components. such as the screening of observations and the observation operators, and other issues, such as the parallelisation strategy for the computer code, are described. The functionality of the observation quality control is investigated and the 3D-Var system is validated through data assimilation and forecast experiments. Results from assimilation and forecast experiments indicate that the 3D-Var assimilation system performs significantly better than two currently used HIRLAM systems. which are based on statistical interpolation. The use of all significant level data from multilevel observation reports is shown to be one factor contributing to the superiority of the 3D-Var system. Other contributing factors are most probably the formulation of the analysis as a single global problem, the use of non-separable structure functions and the variational quality control, which accounts for non-Gaussian observation errors.

There are well-known difficulties to run numerical weather prediction (NWP) and climate models at resolutions traditionally referred to as 'grey-zone' (similar to 3-8 km) where deep convection is neither completely resolved by the model dynamics nor completely subgrid. In this study, we describe the performance of an operational NWP model, HARMONIE, in a climate setting (HCLIM), run at two different resolutions (6 and 15 km) for a 10-yr period (1998-2007). This model has a convection scheme particularly designed to operate in the 'grey-zone' regime, which increases the realism and accuracy of the time and spatial evolution of convective processes compared to more traditional parametrisations. HCLIM is evaluated against standard observational data sets over Europe as well as high-resolution, regional, observations. Not only is the regional climate very well represented but also higher order climate statistics and smaller scale spatial characteristics of precipitation are in good agreement with observations. The added value when making climate simulations at similar to 5 km resolution compared to more typical regional climate model resolutions is mainly seen for the very rare, high-intensity precipitation events. HCLIM at 6 km resolution reproduces the frequency and intensity of these events better than at 15 km resolution and is in closer agreement with the high-resolution observations.

The impact of assimilating temperature, salinity, oxygen, phosphate and nitrate observations on marine ecosystem modelling is assessed. For this purpose, two 10-yr (1970-1979) reanalyses of the Baltic Sea are carried out using the ensemble optimal interpolation (EnOI) method and a coupled physical-biogeochemical model of the Baltic Sea. To evaluate the reanalyses, climatological data and available biogeochemical and physical in situ observations at monitoring stations are compared with results from simulations with and without data assimilation. In the first reanalysis, only observed temperature and salinity profiles are assimilated, whereas biogeochemical observations are unused. Although simulated temperature and salinity improve considerably as expected, the quality of simulated biogeochemical variables does not improve and deep water nitrate concentrations even worsen. This unexpected behaviour is explained by a lowering of the halocline in the Baltic proper due to the assimilation causing increased oxygen concentrations in the deep water and consequently altered nutrient fluxes. In the second reanalysis, both physical and biogeochemical observations are assimilated and good quality in all variables is found. Hence, we conclude that if a data assimilation method like the EnOI is applied, all available observations should be used to perform reanalyses of high quality for the Baltic Sea biogeochemical state estimates.

A model for the parameterization of lake temperatures and lake ice thicknesses in atmospheric models is presented. The model is verified independently, and it is also tested within the framework of the High Resolution Limited Area Model(HIRLAM), applied operationally for short range weather forecasting at the Swedish Meteorological and Hydrological Institute (SMHI). The lake model is a slab model based upon energy conservation and treats the lakes as well mixed boxes with depths represented by the mean depths. The model is forced by near surface fluxes calculated from total cloudiness, air temperature, air humidity and low-level winds. A data base, describing 92000 Swedish lakes. provides the model with lake mean depths, areal sizes and locations. When the model is used for parameterization of lake effects in the atmospheric model, all the smaller lakes and the fractions of larger lakes within each horizontal grid square of the atmospheric model are parameterized by four model lakes, representing the lake size distribution. The verification of the lake model is done by comparing it with a more advanced, vertically resolved model, including parameterization of turbulent mixing processes, as well as by comparison with observations. A sensitivity test shows great interannual variations of the ice-covered season, which implies that lake models should be used instead of climate data. The results from an experiment with two-way coupling of the lake model to the atmospheric model are verified by comparing forecasted weather parameters with routine meteorological observations. These results show that the impact of lake effects can reach several degrees C in air temperatures close to the surface.

Salinity and temperature variations in the Baltic proper and the Kattegat have been analyzed with a numerical ocean model and a large amount of observational data. In the model, the Baltic Sea is divided into 13 sub-basins with high vertical resolution, horizontally coupled by barotropic and baroclinic flows and vertically coupled to a sea-ice model which includes dynamics as well as thermodynamics. The model was integrated for a 15-year period (1980-1995) by using observed meteorological forcing data, river-runoff data and sea-level data from the Kattegat. The calculated 15-year median profiles of salinity and temperature in the different sub-basins are in good agreement with observations. However, the calculated mid-depth salinities in the Arkona Basin and Bornholm Basin were somewhat overestimated, and the calculated deep-water temperatures in the Arkona Basin and the Bornholm Basin are somewhat lower than the observed values. Frontal mixing and movements in the Kattegat and the entrance area of the Arkona Basin were important to consider in the model. Water masses were simulated well, and prescribing constant deep-water properties in the Kattegat proved to be a reasonable lateral boundary condition. Further, comparisons were made between observed and calculated seasonal and interannual variations of the hydrographic properties in the Eastern Gotland Basin, as well as the interannual variations of the annual maximum ice extent. We conclude that the model can simulate these variations realistically. The major Baltic inflow of 1993 was also simulated by the model, but the inflowing water was 1-2 degrees degrees too cold. Finally, the response times to changes in forcing of the Baltic proper and the Kattegat were investigated by performing the so-called lock-exchange experiment. Typical stratification spin-up times were of the order of 10 years for the Kattegat, and 100 years for the Baltic proper.

The objectives of the present paper are to formulate and explore a coupled sea ice-ocean model and to examine the sensitivity of ice in the Baltic Sea to climate change. The model treats the Baltic Sea as 13 sub-basins with vertical resolution, horizontally coupled by estuarine circulation and vertically coupled to a sea ice model which includes both dynamic and thermodynamic processes. The reducing effect on the barotropic exchange due to sea ice in the entrance area is also added. The model was first verified with data from 3 test periods representing one mild, one normal and one severe ice winter. The maximum seasonal ice extent was then examined on the basis of simulated and observed data for the period 1980-1993. After that, some climate scenarios (both warm and cold) were examined. The seasonal, regional and interannual variations of sea ice were well described by the model, and the thermal response in the Baltic Sea can be realistically simulated applying forcing data from rather few stations. The Baltic Sea system is highly sensitive to climate change, particularly during the winter season, Warming may drastically decrease the number of winters classified as severe, forcing the climate towards more oceanic conditions. On the other hand, cooling will increase the number of severe winters, forcing the climate towards more sub-arctic conditions.

The ice-ocean response to variable winds is analysed based upon two types of models. An analytical ice-ocean model with linear stress laws and forced by periodic winds is first derived. Secondly a numerical, vertically resolved ice-ocean model is introduced. In the numerical model, the ice-water stress law is calculated from a turbulence model and the wind stress is calculated on the basis of a square law formation. By comparing the ice-ocean stress law formulations, it is illustrated that the numerical model predicts an ice-ocean stress law that has a power slightly less than 2 compared to 1 for the analytical model. The numerical prediction is in good accordance with field observations and the slight deviation from 2 is due to wall effects close to the ice-water interface. It is then demonstrated that the ice-ocean response to variable winds could be well simulated by both models, but the analytical model did not capture the wind dependency properly (because of the linear approach). The ice and current factors are amplified at wind frequencies close to inertial (omega = -f) and damped at high positive and negative frequencies. The maximum ice and current factors at a wind frequency equal to the inertial oscillation are shown to be dependent only on the friction coefficients. With the constants applied in the present study, the maximum ice drift and current speed are equal to 7.8% and 5.5% of the wind speed, respectively. These steady state values are however quite unrealistic as they would require a uniformly changing wind direction for many inertial periods.

Recent observations and modelling studies suggest that the Arctic climate is undergoing important transition. One manifestation of this change is seen in the rapid sea-ice cover decrease as experienced in 2007 and 2012. Although most numerical climate models cannot adequately reproduce the recent changes, some models produce similar Rapid Ice Loss Events (RILEs) during the mid-21st-century. This study presents an analysis of four specific RILEs clustered around 2040 in three transient climate projections performed with the coupled Rossby Centre regional Atmosphere-Ocean model (RCAO). The analysis shows that long-term thinning causes increased vulnerability of the Arctic Ocean sea-ice cover. In the Atlantic sector, pre-conditioning (thinning of sea ice) combined with anomalous atmospheric and oceanic heat transport causes large ice loss, while in the Pacific sector of the Arctic Ocean sea-ice albedo feedback appears important, particularly along the retreating sea-ice margin. Although maximum sea-ice loss occurs in the autumn, response in surface air temperature occurs in early winter, caused by strong increase in ocean-atmosphere surface energy fluxes, mainly the turbulent fluxes. Synchronicity of the events around 2040 in the projections is caused by a strong large-scale atmospheric circulation anomaly at the Atlantic lateral boundary of the regional model. The limited impact on land is caused by vertical propagation of the surface heat anomaly rather than horizontal, caused by the absence of low-level temperature inversion over the ocean.

Freshwater (FW) induced transformations in the upper Arctic Ocean were studied using a coupled regional sea ice-ocean model driven by winds and thermodynamic forcing from a reanalysis of data during the period 1948-2011, focusing on the mean state during 1968-2011. Using passive tracers to mark a number of FW sources and sinks, their mean composition, pathways and export were examined. The distribution of the simulated FW height reproduced the known features of the Arctic Ocean and volume-integrated FW content matched climatological estimates reasonably well. Input from Eurasian rivers and extraction by sea-ice formation dominate the composition of the Arctic FW content whilst Pacific water increases in importance in the Canadian Basin. Though pathways generally agreed with previous studies the locus of the Eurasian runoff shelf-basin transport centred at the Alpha-Mendeleyev ridge, shifting the Pacific-Atlantic front eastwards. A strong coupling between tracers representing Eurasian runoff and sea-ice formation showed how water modified on the shelf spreads across the Arctic and mainly exits through the Fram Strait. Transformation to salinity dependent coordinates showed how Atlantic water is modified by both low-salinity shelf and Pacific waters in an estuary-like overturning producing water masses of intermediate salinity that are exported to the Nordic Seas. A total halocline renewal rate of 1.0 Sv, including both shelf-basin exchange and cross-isohaline flux, was estimated from the transports: both components were of equal magnitude. The model's halocline shelf-basin exchange is dominated by runoff and sea-ice processes at the western shelves (the Barents and Kara seas) and Pacific water at the eastern shelves (the Laptev, East Siberian and Chukchi seas).

A dispersion model for the stratified benthic boundary layer is formulated. It is based on "small-scale" vertical dispersion and a "large-scale" horizontal flow field. A modified Langevin equation governs the stochastic vertical migration of an ensemble of marked fluid elements. These elements are spread out by the horizontal flow, determined by a one-dimensional model, which includes a two-equation (k - epsilon) turbulence scheme. The later yields statistical information necessary for the stochastic process. Statistical properties of the dispersion process are then calculated from the evolution of the ensemble of elements. A rather idealized case with a linearly stratified fluid subject to a suddenly imposed barotropic pressure gradient is considered. A quasi-geostrophic interior flow is formed with a benthic boundary layer at the bottom. Marked fluid elements are released at the bottom and then followed for several pendulum days. It is found that the dispersion process is well characterized by K = Cu(*)l/(where u(*) and l are the friction velocity at the bottom and the layer thickness, respectively), and where C approximate to 15. A similar relation but based on external parameters only, becomes: K = C-b vertical bar partial derivative P/partial derivative y vertical bar(2)/rho(2) f(5/2) N-1/2, where C-b approximate to 0.11 in the range N/f = 28- 88

The decade 1991-2000 was warm and wet in Sweden, with 10-station mean temperature 0.8 degreesC above and 20-station mean precipitation 6% above the mean for 1961-1990. Here we study the question if such changes should be seen as a symptom of anthropogenic climate change or if they might be of purely natural origin. Using the control simulations of 19 atmosphere-ocean general circulation models and taking into account difference's between the simulated and observed interannual variability, we estimate that the recent increase in temperature and that in precipitation had both about a 6-7% chance to occur solely as a result of natural variability. Using the corresponding simulations with increasing CO2, we further estimate that the anthropogenic forcing raised the probability of the observed changes to occur to 23% for the increase in temperature and to 14% for the increase in precipitation. About half of the warming and about 30% of the increase in precipitation appear to be explained by anthropogenic forcing. The seasonal aspects of observed and simulated climate change are also discussed, with special emphasis on winter, when the observed warming has been much larger than expected from the model simulations. Finally, a probabilistic forecast for the Swedish climate in the first decade of the 21st century suggests a 95% (87%) possibility of warmer (wetter) annual mean conditions than in 1961-1990 on the average. One of the caveats in our analysis is that the model simulations exclude variations in solar and volcanic activity, the effects of which might not be fully covered by our resealing of interannual variability.

Two regional climate model experiments for northern and central Europe are studied focussing on greenhouse gas-induced changes in heavy precipitation. The average yearly maximum one-day precipitation P-max shows a general increase in the A hole model domain in both experiments, although the mean precipitation P-mcan decreases in the southern part of the area, especially in one of the experiments. The average yearly maximum six-hour precipitation increases even more than the one-day P-max suggesting a decrease in the timescale of heavy precipitation. The contrast between the P-max, and P-max changes in the southern part of the domain and the lack of such a contrast further north are affected by changes in wet-day frequency that stem, at least in part. from changes in atmospheric circulation. However, the yearly extremes of precipitation exhibit a larger percentage increase than the average wet-day precipitation. The signal-to-noise aspects of the model results are also studied in some detail. The 44 km grid-box-scaie changes in P-max are very heavily affected by inter-annual variability, with an estimated standard error ;of about 20% for the 10-year mean changes. However. the noise in P-max decreases sharply toward larger horizontal scales, and large-area mean changes in P-max can be estimated with similar accuracy to those in P-mcan Although a horizontal averaging of model results smooths out the small-scale details in the true climate change signal as well, this disadvantage is, in the case of P-max changes, much smaller than the advantage of reduced noise.

Two 2 x 10-year climate change experiments made with the Rossby Centre regional Atmospheric climate model(RCA) are reported. These two experiments are driven by boundary data from two global climate change simulations, one made with HadCM2 and the other with ECKAM4/OPYC3, in which the global mean warming is virtually the same, 2.6 degreesC. The changes in mean temperature and precipitation show similarities (including broadly the same increase in temperature and in northern Europe a general increase in annual precipitation) as well as differences between the two RCA experiments. These changes are strongly governed by the driving GCM simulations. Even on the RCA grid box scale, the differences in change between RCA and the driving GCM are generally smaller than the differences between the two GCMs. Typically about a half of the local differences between the two RCA simulations are attributed to noise generated by internal variability, which also seems to explain a substantial part of the RCA-GCM differences particularly for precipitation change. RCA includes interactive model components for the Baltic Sea and inland lakes of northern Europe. The simulated changes in these water bodies are discussed with emphasis on the wintertime ice conditions. Comparison with an earlier RCA experiment indicates that a physically consistent treatment of these water bodies is also of importance for the simulated atmospheric climate change.

Doppler radar radial wind observations are modelled in numerical weather prediction (NWP) within observation errors which consist of instrumental, modelling and representativeness errors. The systematic and random modelling errors can be reduced through a careful design of the observation operator (Part I). The impact of the random instrumental and representativeness errors can be decreased by optimizing the processing of the so-called super-observations (spatial averages of raw measurements; Part II). The super-observation processing is experimentally optimized in this article by determining the optimal resolution for the super-observations for different NWP model resolutions. A 1-month experiment with the HIRLAM data assimilation and forecasting system is used for radial wind data monitoring and for generating observation minus background (OmB) differences. The OmB statistics indicate that the super-observation processing reduces the standard deviation of the radial wind speed OmB difference, while the mean vector wind OmB difference tends to increase. The optimal parameter settings correspond at a measurement range of 50 km (100 km) to an averaging area of 1.7 km(2) (7.3 km(2)). In conclusion, an accurate and computationally feasible observation operator for the Doppler radar radial wind observations is developed (Part I) and a super-observation processing system is optimized (Part II).