Storm surge A2-scenario for Hamburg St. Pauli, 2030; version 17/10/200712/07/200503/07/2005. Storm surge scenarios for Hamburg

Transcription

1 Storm surge A-scenario for Hamburg St. Pauli, 00; version 7/0/007/07/0050/07/005 Storm surge scenarios for Hamburg Abstract A scenario for future storm surge heights for the tide gauge of Hamburg St. Pauli is constructed on the basis of results from a regional model for Cuxhaven. Under the A emission scenario, an increase of the mean annual maximum water level of about 0.74 m appears possible and plausible for the time horizon of 00. In Cuxhaven, an increase of 0.4 may be expected in this scenario. In 085 the expected increase for St. Pauli is m. These values are uncertain, not only because of the employed emission scenarios but also because of a series of downscaling steps, which describe the chain of processes from increased emissions and local climate change impact. Introduction Since both, the disastrous flood of 96 and the well-managed 976 surge, flood protection in the city of Hamburg and the area downstream between Hamburg and the Elbe mouth has been constantly adapted to changes in the height of high floods. Such changes may be due to river construction measures (Freie und Hansestadt Hamburg 005, Arbeitsgemeinschaft für die Reinhaltung der Elbe 984, Siefert et al. 988) or to changes in the global climate (Freie und Hansestadt Hamburg 005). In the present study, the influence of the possible future climate change on water levels at high tide in Hamburg St. Pauli is investigated. This possible future climate change is described by scenarios of future climate change. These scenarios present possible, consistent, plausible but not necessarily probable futures (e.g., Schwartz 99). They have been prepared by first envisaging emissions of climatically relevant substances into the atmosphere, and by then simulating the effect of these emissions with numerical models. Towards this end, results from an A -scenario (Houghton et al. 00) of storm surge levels at the North Sea coast between 070 and 00 are projected for Hamburg St. Pauli for the time horizon 00. This scenario is one of a series of scenarios which have been considered in the EU project PRUDENCE (Christensen et al., 00). They are all derived from a base climate change simulation with the global General Circulation Model HadAMH of the Page

2 Storm surge A-scenario for Hamburg St. Pauli, 00; version 7/0/007/07/0050/07/005 Hadley Center processing the IPCC A SRES scenario. This scenario envisages an increase of atmospheric greenhouse gas concentrations at the end of the st century which corresponds roughly to a tripling of pre-industrial levels. A is a relatively pessimistic scenario. A series of North Sea storm surge scenarios (Woth et al., 005) is constructed in two steps. First, the HadAMH global results on a 00x400 m grid are dynamically downscaled to the Northern European atmosphere. Then the barotropic hydrodynamic model TRIMGEO of the North Sea is exposed to the downscaled wind and air pressure data on a grid of about 50 m. TRIMGEO simulates water levels and currents on a grid of about 0 m (e.g., Aspelien und Weisse, 005) for decades of years. The dynamical downscaling is achieved with four different regional models but a major result of Woth et al. (005) is that the eventual storm surge scenarios depend only wealy on the regional climate model used (Figure ). Therefore in this analysis we use only the downscaled results obtained with the regional climate model CLM. The results of the downscaling chain described above leads to an estimate of the expected changes from to the time horizon given the emission scenario A. It is not possible to use the simulation for the directly as a possible future for this time. This is because of the systematic errors in the simulations when simulating the time horizon, the simulated high water levels are underestimated a bit, which originates mainly from the global climate change simulation. Therefore it is common in climate research to consider only the change, assuming that the relatively small systematic errors cancel out. In the following we present a simple statistical method to derive estimates for the site St. Pauli in Hamburg for the foreseeable future of 00 from these North Sea storm surge A scenarios. Page

3 Storm surge A-scenario for Hamburg St. Pauli, 00; version 7/0/007/07/0050/07/005 Figure : Projected changes of the inter-annual mean of the 99.5 th percentile of storm surge heights along the coastal 0 m depth line (see red dots in inset) for the four models. The changes are compared to the 95% confidence interval (depicted as grey shaded band) of inter-annual natural variability which is inferred from the hindcast (From Woth et al., 005). Methodology We need to introduce two empirically based approximations: A lin relating water levels at coastal sites near the Elbe mouth and two sites on the 0 m bathymetry line (which is simulated by the TRIMGEO model) with the water level at St. Pauli in the port of Hamburg. This approach has been suggested by Langenberg et al. (999) An estimation of the situation at the midterm 00 from the two available time horizons and The 99.5 th percentile is the threshold selected such that 99.5% of the storm surge values are less than this threshold. Page

4 Storm surge A-scenario for Hamburg St. Pauli, 00; version 7/0/007/07/0050/07/005. Lining North Sea surge levels and St. Pauli surge levels To derive projections of water levels in Hamburg St. Pauli from water levels at coastal grid boxes, a statistical function describing the relationship between water levels in the two locations is needed. For this purpose, data on historical high tide water levels in Hamburg St. Pauli between 980 and 990 is used. This particular interval has been chosen because river deepening measures which might influence water levels have not been carried out during this time (compare figure ). This data set is compared with the high tide water levels of the hindcast run during the same time period. This hindcast run was made with the TRIMGEO model, forced with highresolution analysed wind and air pressure. Analysed means a best guess of the synoptic situation derived from observations. Figure : Mean low water (MTnw, green line) and mean high water (MThw, red line) at Hamburg St. Pauli The mean low water is not stable during the interval of interest ( ) but the mean low water is stable. High tide water levels for different grid boxes located at the coast close to the Elbe mouth and for grid boxes located on the 0m bathymetry-line close to the Elbe mouth are considered. Page 4

5 Storm surge A-scenario for Hamburg St. Pauli, 00; version 7/0/007/07/0050/07/005 A preliminary comparison of the two data sets on the basis of scatter diagrams suggested that a curve consisting of a linear component f and a quadratic component f would provide a good fit (Figure ): f (x) = f f ( x) = ax + b, x < κ ( x) = cx + dx + e, x κ ' ' with f κ) = f ( ) and f ( κ) = f ( ). ( κ κ Later, we want to describe the change in storm surge heights in terms of the multiyewar mean of annual maxima. Therefore, we add the constraint f ( µ ) = µ C SP. Here SP µ = 4.56 m represents the multiyear annual maximum at St. Pauli and µ C the multiyear annual maximum at the grid box close to the Elbe mouth. Starting with a sufficiently big interval x, ] we determine the coefficients a, b, c, d, e, λ λ as those which minimize for, λ, [ x ε ( s, x ) = ( f( xi ) yi ) + ( f i= + λ γ ( f ( x n i= ) f ( x ( x ) y ) i )) + λ γ ( f i ' ( x ) f ( x ' )) + λ γ ( f ( µ ) µ C SP ) = min! for each x [ x, x ] at the site s. The numbers γ i are weights given to the constraints. In our case, we have γ = 750, γ = and γ =, i.e., maximum weight is given to the continuity of the fit and minimum weight to the equivalence of the multiyear annual maximum heights at St. Pauli and at the gridboxes at the mouth of the Elbe, and to the continuity of the derivative of the fit. This process is successively repeated for smaller intervals of which the x determined in the previous step constitutes the middle. The algorithm terminates if no more improvements are achieved. The smallest value for _ is reached for the coastal grid box s centered at 5.9 N 8.9 E with a=0.8, b=0.64, c= 0.9, d=-0.8, e=.67 and x =. 79. Figure shows the linear/quadratic fit for this set of parameters and the scatter cloud of pairs of high tide values at St.Pauli and at the gridpoint s. The constraints of continuity of the function and its derivative is satisfactorily fulfilled, also the condition that the mean maximum of.67 m at s is mapped on the mean maximum of 4.56 m at St. Pauli is met. Page 5

6 Storm surge A-scenario for Hamburg St. Pauli, 00; version 7/0/007/07/0050/07/005 Finally we determine the root mean square error of the fit, i.e., n ( f ( xi ) yi ) + ( f ( xi ) yi ) i= n + i= ζ =, which amounts to 0. m for the selected optimal set of parameters.. Temporal interpolation As outlined in the introduction, the simulations provide at this time only a projection of the expected change from the control period until , given scenario A and the global HadAMH simulation. To establish a projection of the results onto our time horizon 00, we assume a development of storm surge heights parallel to the increase in temperature (Houghton et al., 00). The expected increase _ from 990 to 00 is 0.7 K which is about 0% of the increase from the interval to the interval (.5 K). Thus, we assume that the mean maximum surge height at the location at the mouth of the Elbe is increased by about 0% of the increase derived from the TRIMGEO scenario for the time horizon. For the mean sea level rise D we use the projection provided by the IPCC (Houghton et al., 00) for 00, which amounts to 0. m. Page 6

7 Storm surge A-scenario for Hamburg St. Pauli, 00; version 7/0/007/07/0050/07/005 6,00 f ( M ( H ) +?( M ( S ) - M ( C ) ) ) 5,00 Mean ann. max St Pauli = f(m(h)) Hamburg St Pauli 4,00 M ( H ) +?( M ( S ) - M ( C ) ),00 M(H),00,00 0,00 0,00,00,00,00 4,00 5,00 6, n / 8.9e Figure : Linear-quadratic fit for water levels at Hamburg St. Pauli and at the coastal grid point 54.0 N / 8.9ºE. The yellow line indicates the multiyear means of annual maxima derived from the hindcast for the coastal grid point and derived form the observations at St. Pauli. The light blue line indicates the mean annual maximum in 00 given changes in storm surge levels, and the projection for the St. Pauli tide gauge. Results We consider the multiyear mean of the annual maximum M, specifically for the hindcast simulation H960-90, the control simulation C and the A-Scenario S In the following we drop the indices. The projected mean annual maximum high tide water level P at St. Pauli is estimated as P = f ( M ( H ) + ϕ[ M ( S ) M (C )]) + D The difference M(S)-M(C) of the mean annual maximum high tide at the coastal point s = (5.9 N 8.9 E) in the Scenario S and Control-Run C amounts to 0.9 m. The present mean Page 7

8 Storm surge A-scenario for Hamburg St. Pauli, 00; version 7/0/007/07/0050/07/005 annual maximum M(H) is.5967 m. The expected contribution by global mean sea level pressure is D = 0. m. Adding this all together, we have m for the projected mean annual maximum high tide at St. Pauli in 00, which represents an increase of 0.74 m. Without sea level rise the projected mean maximum high tide would be 4.67, which represents an increase of 0.. For the time horizon 085 the expected increase in mean sea level is 0. m, and the effect of stronger surges amounts to an increase of 0.58 m. In total, the mean annual maximum is expected to be 5.,47 m in Hamburg St. Pauli, which is 0.9 m higher than presently. 4 Discussion and caveats We have presented a simple approach to estimate changes in extreme water levels at the tide gauge of Hamburg St. Pauli. This method relates scenarios for North Sea near-coastal locations to the highly location specific conditions far inside the Elbe estuary. This lin taes the form of a transfer function, which maps coastal high water levels simulated in a hindcast with a hydrodynamical model, on observations taen at the tide gauge. This transfer function is valid only for the specific hydrodynamical model TRIMGEO which has been employed in the hindcast and in the scenario simulations. The resulting values are uncertain, not only because of the employed emission scenarios but also because of a series of downscaling steps, which describe the chain of processes lining increased emissions and local climate change impact. In a further step, we will examine the projected increases in storm surge heights not only under the A emission scenario but also under the less severe B conditions. Also different combinations of global and regional climate models will be employed. 5 References Arbeitsgemeinschaft für die Reinhaltung der Elbe, 984. Gewässeröologische Studie der Elbe. Arbeitsgemeinschaft für die Reinhaltung der Elbe, Hamburg. Page 8

9 Storm surge A-scenario for Hamburg St. Pauli, 00; version 7/0/007/07/0050/07/005 Aspelien, T. and R. Weisse (005), Assimilation of Sea level Observations for Multi-Decadal Regional Ocean Model Simulations for the North Sea, GKSS report 005/ Christensen, J.H., T. Carter, F. Giorgi, 00: PRUDENCE employs new methods to assess european climate change, EOS, Vol. 8, p. 47. Freie und Hansestadt Hamburg, Behörde für Stadtentwiclung und Umwelt, 005. Hochwasserschutz in Hamburg: Stand des Bauprogramms. Amt für Bau und Betrieb, Hamburg. Houghton, J.T., Y. Ding, D.J. Griggs, M. Noguer, P. J. van der Linden and D. Xiaosu (Eds.), 00. Climate Change 00: The Scientific Basis. Contribution of Woring Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC). Cambridge University Press, UK. Langenberg, H., A. Pfizenmayer, H. von Storch and J. Sündermann, 999: Storm related sea level variations along the North Sea coast: natural variability and anthropogenic change.- Cont. Shelf Res. 9: 8-84 Siefert, W. Havnoe, K., 988. Einfluss von Baumassnahmen in und an der Tideelbe auf die Höhen hoher Sturmfluten, Die Küste 47. Schwartz, P., 99: The art of the long view. John Wiley & Sons, 7 pp Woth, K., Weisse, R., von Storch, H., 005. Dynamical modelling of North Sea storm surge extremes under climate change conditions an ensemble study. Ocean Dyn. (in press) Page 9