The applied correction for transport processes had a varying impact. Especially for high speed, non-circular transport regimes these advection processes have to be taken into account, although the principle differences in mortality rates can also be seen without a corresponding correction (see Table 1). It is well known that the flow dynamics in the Baltic Sea are highly complex. They are mainly determined by the ephemeral character of wind stress, the baroclinic mass field and the complicated bottom topography . However, the Baltic Sea Model used for the simulations has generally proven to be applicable for simulating egg and larval drift processes [8, 14]. To explicitly test the applicability of the model for predicting the horizontal distribution patterns of ichthyoplankton during the investigated time periods, separate model runs were performed during an earlier study . The simulated results of this study were in good accordance with field observations, confirming the applicability of the Baltic Sea Model. This good accordance could be reached, although, as in the present study, simplistic vertical distribution patterns had to be used. The vertical position of the eggs and larvae does have an influence on their horizontal transport. Unfortunately, no direct measurements for the time periods under investigation are available. While stage-specific vertical distribution patterns of cod eggs are well investigated [6, 7, 15], corresponding information for cod larvae is rather scarce. The available information on stage-specific vertical distribution of cod larvae in the Baltic [5, 16] can only be used to extract rough distribution patterns. If older cod larvae perform diurnal vertical migrations, as observed in other cod stocks  can so far not be answered. As a first approximation of the vertical distribution vertically resolving samplings 4 weeks apart from the periods under investigation were used. It is known that the vertical distribution patterns can change over the spawning period, but strong changes within 4 weeks are not to be expected. Furthermore, a sensitivity analysis conducted for the transport processes in the Bornholm Basin  revealed no significant differences in the drift patterns of cod larvae within a vertical range of 12 m. As a conclusion, we are quite confident, that the modelled development of the horizontal distribution patterns are, despite the various shortcomings, close to reality.
The importance of corrections for transport processes will even increase for planned studies on sprat larval mortality, as these are distributed shallower in the water column [15, 19] and therefore are subject to higher current speeds.
There is no indication for a strong small-scale patchiness of cod eggs or larvae, which would not be resolved by the station grid and thereby considerably bias the abundance estimates. As well the results from the 15 years ichthyoplankton time series as results from high-resolution samplers like the Longhurst-Hardy Plankton Recorder (Raab, pers. comm., Inst. Marine Sciences, Kiel) and the Ichthyoplankton Recorder (Mees, pers. comm., Inst. Marine Sciences, Kiel) rather led to the assumption, that the distribution and abundance of ichthyoplankton is well covered by the station grid in use. The uncertainties in the abundance estimates arising from the non-synoptic sampling procedure as well as the spatial sampling resolution can so far not be quantified, but a theoretical study dealing with this problem has been conducted by Voss and Hinrichsen (in prep.) showing mean errors in abundance estimates of 10–20%.
Even after accounting for these uncertainties, the results presented here for cod larvae suggest a high variability in the importance of different 'critical periods' for recruitment. In May 1988 the period between hatch and the onset of feeding was characterised by high mortalities (~20 % day-1), whereas during the same period in August 1991 very low mortality rates were observed. Different processes have been proposed to influence survival on this stage. Viable hatch requires a minimum oxygen level of 2 ml/l . To account for year to year variations in the hydrographic environment occupied by the eggs, the so called Reproductive Volume (RV) has been defined [21, 22]. The RV defines the volume of water suitable of successful development of cod eggs by threshold levels in temperature (>1.5°C), salinity (>11 psu) and oxygen content (> 2 ml/l) as experimental studies have shown that below these limits development of cod eggs ceases before hatch [23, 24, 20]. In August 1991 the RV in the Bornholm Basin was 127 km3, substantially higher than in May 1988 with only 93 km3. Unfortunately, no field sampling for the vertical distribution of eggs was performed during the two periods, but these results as well as modelling results  suggest less favourable hydrographic conditions for the late egg and early larval stages in May 1988. Even if hatching occurs, low oxygen concentrations negatively impact larval activity . Under such circumstances hatched cod larvae might sink to deeper, less oxygenated water layers, as specific gravity increases after hatch and the larvae are not able to counterbalance the sinking rate. Another potentially important source of mortality is predation by the clupeids sprat and herring. Performed stomach content analysis of sprat and herring indicate considerable predation pressure on all egg stages in May 1988 , while for August 1991 the consumption of cod eggs by the clupeids was low . This may have the potential to explain the strong differences in mortality rates during the period from hatch to first feeding encountered in this study.
The second investigated 'critical period' ranges from the onset of feeding (larval stage 5) to the state of an established feeder (larval stage 8). During this phase the larvae have to migrate vertically into upper water layers with sufficient light conditions and higher prey concentrations for successful feeding . Predation on these larvae is of less importance, as the vertical overlap between prey and predator is rather limited, with the clupeids feeding within and below the permanent halocline [26, 27]. Correspondingly, in May 1988 almost no larvae were identified in the diet of herring and sprat. In a time series principally covering the spawning seasons from 1988 to 1996, only in August 1991 relatively high numbers of cod larvae were found in the stomachs of herring , indicating an increased predation pressure. Another possible source of variation in survival are differences in prey availability for the larvae. Unfortunately, there are no data on abundance and distribution of suitable prey organisms, like copepod nauplii and small copepodite stages [28–30], for these time periods. The only indication available is that the total meso-zooplankton biomass in the Bornholm Basin was higher in August 1991 than in May 1988 . However, it has been suggested, that prey abundance in the Bornholm Basin is always sufficient to ensure feeding success of cod larvae . Therefore the potential influence of changing prey availability remains to be clarified.
The ability of the larvae to migrate vertically and to establish as feeding larvae might be influenced by their energy reserves, i.e. the size and quality of their yolk sac. Egg size is a function of fish size and the batch number spawned . In the beginning and at the end of the spawning season the egg size is smallest. While it was main spawning time in May 1988, August 1991 corresponds to the late spawning season . There are no direct egg diameter measurements available, but due to the relative spawning time egg diameter and therefore yolk reserves should have been higher in May 1988, favouring survival of the larvae.
The different mortality rates found for the two investigated periods can not easily be attributed as inter-annual or seasonal differences. Due to the complex abiotic and biotic relationships in the Baltic Sea, it is rather a combination of both effects. However, the two time periods, May 1988 and August 1991 revealed the peak larval abundance in each year, so that changes in mortality rates probably had a comparable impact on the resulting recruitment.
Baltic cod recruitment decreased strongly in the beginning of the 1980ies. Since 1985 it fluctuated around a rather low level compared to the late 1970's and early 1980's. Both investigation dates exhibited outstanding high abundance values for cod larvae in the last 15 years . While the spawning stock biomass as well as the potential egg production were higher in 1988 than in 1991 , recruitment in 1988 reached only an intermediate level (compared to the 'low recruitment phase' since 1985). The 1991 year-class was above average, but by far also not reaching the values from the beginning of the 1980ies. It becomes obvious that also after reaching the state of an established feeder processes must act as bottlenecks of survival, limiting year-class strength. Besides others, transport processes have been shown to influence survival probability [35, 8].