On the application of the Unified Model to produce finer scale climate information for New Zealand

Transcription

1 Weather and Climate 22,19-27 (2002) On the application of the Unified Model to produce finer scale climate information for New Zealand B. Bhaskaran, J. Renwick and A.B. MuIlan National Institute of Water and Atmo,Oberic Research Wellington, New Zealand Abstract Successful application of a one-way nested regional climate model to simulate finer scale climate details of a region depends on several factors. One of the critical factors relates to the geographical location and circulation features of the region. Here we have investigated the suitability of a domain containing the Islands of New Zealand and adjoining seas for a successful application of the one-way nested regional climate model configuration of the unified model (UM) to simulate high-resolution surface variables. It appears that the geographical location of the New Zealand domain is well suited for such studies, if we carefully choose the size of the domain. Other factors that influence the quality of a regional model simulation, such as external forcing conditions, horizontal resolution, and relaxation zone, are also discussed. An example of regional model simulation using the UM is also shown. 1. Introduction General circulation models (GCMs) based on the dynamical and physical laws of the atmosphere and ocean have been largely used to simulate global climate, and global climate change under a variety of external forcing conditions (IPCC, 2001). The horizontal resolution ( (m) of these GCMs is inadequate to represent the mesoscale forcing in the spatial scale of 10 50km associated with: (1) rough terrain; (2) vegetation characteristics; (3) inland water basins; and (4) complex coastline (Figure 1). Therefore the spatial characteristics of surface variables simulated by the GCMs tend to have insufficient detail, and hence cannot be effectively used as input, for regional climate impacts models. The horizontal resolution of current GCMs cannot be increased to describe the local climate forcing effectively because this would require enormous computing resources (cost increases linearly with vertical resolution and quadratically with horizontal resolution). Therefore the spatial gap between GCMs and climate impact models needs to be bridged using "downscaling" techniques. Downscaling refers to the use of largescale information from GCM output, through dynamical or statistical approaches, to produce local climate details which are not adequately modelled by the global GCMs Corresponding author: Dr. B. Bhaskaran, National Institute of Water and Atmospheric Research, PO Box 14901, Wellington, New Zealand. b.bhaskaran@niwa.crinz

2 20 Weather and Climate, Volume 22 't; (a) GCM orography (b) RCM orography Figure 1: Distribution of orographic height in meters: (a) general circulation model; and (b) regional climate model. The contour intervals are 10, 100, 200, 300, 400, 600, 800, 1000, 1500, and 2000 m. In (a) the global model orography is overlaid over the true coastline. Statistical downscaling methods involve the construction of empirical relationships between observed surface weather variables and observed (or modelled) large-scale free atmospheric variables. These relationships are then applied to the large-scale variables simulated under a variety of forcing conditions to forecast local climate variables. The basic limitation in these techniques is that the forcing and their interactions with the large-scale variables are not described in physically based terms in the empirical relationships and therefore there is no assurance that the empirical relationships developed for a particular situation will also work under different forcing conditions (for example, under climate change scenarios). Modelling approaches try to explicitly describe the local climate forcing by increasing the horizontal resolution over the specific regions of interest. One way of achieving increased resolution over a particular region is to employ a variable resolution global spectral model in which the horizontal resolution is higher over the region of interest and coarser elsewhere in the globe (Deque and Piedelievre, 1995). One of the severe limitations of this technique is that while we increase horizontal resolution over the region of interest, a further decrease in resolution elsewhere becomes inevitable in spectral models. This may lead to poor simulation of planetary scale circulations which is not desirable for a realistic simulation of local climate variables. Another modelling approach is to use a high-resolution limited area model driven at its boundaries by analysed initial and lateral boundary conditions or global GCMs (Jones et al., 1995; Bhaskaran et al., 1998). In case of GCM driven experiments, the highresolution limited area model is embedded in a coarse resolution GCM over the area of interest. At lateral boundaries either two-way interacting nesting or one-way nesting can be employed. Here we consider using one-way nesting technique in which the initial and lateral boundary conditions needed to run the regional model are provided by the output

3 Bhaskaran, Renwick, Mullan: Finer scale climate information for New Zealand 2 1 of a GCM simulation. The circulation of the regional model, in turn, does not influence the global circulation of the GCM. In the following sections we describe the suitability of the NZ domain for the UM nested modelling approach to simulate high-resolution surface variables. We will also demonstrate its suitability for the New Zealand region with an example. 2. Regional climate model and New Zealand climate 2.1 Regional Climate Model The Regional climate model considered here is based on the United Kingdom Met. Office (UKMO) unified model (UM) software (Cullen, 1993). The UM is designed to allow the users to configure its atmospheric component as a regional climate model locatable over any part of the globe. The global configuration of the atmospheric component can also be used to drive the regional model at its lateral boundaries. The driving global model and the regional climate model are both hydrostatic, primitive equation grid point models. The distribution of 19 vertical levels is the same in both models. The driving model employs 2.5' x 3.75' horizontal resolution, while the regional model used here has a 0.15' x 0.15' horizontal grid. The models use identical representation of subgrid-scale processes. For further information the reader is referred to Bhaskaran et al., (1996). In summary, the regional model differs from the driving atmospheric GCM only in the horizontal resolution and hence the time-step of the integration. The philosophy underlying nested regional modelling is that while the coarse resolution driving atmospheric GCM simulates the response of the general circulation to global forcing, the high resolution regional model simulates the effects of local climate forcing on the regional circulation and distribution of climatic variables over the area of interest (Giorgi and Mearns, 1991). Simulation of the New Zealand climate is very well suited to this philosophy, since the seasonal mean climate is strongly influenced by a combination of global and regional scale forcing. 2.2 New Zealand Climate The seasonal mean atmospheric circulation of New Zealand may be split up into two main components: the planetary scale circulation associated largely with the El Nino Southern Oscillation (ENSO), High Latitude Mode (HLM), and wave 3 location; and a regional mesoscale circulation induced by local forcing. Through the Hadley circulation, interannual variations in the tropical Pacific sea surface temperature (SST) anomalies exert a major influence on interannual variations in regional synoptic circulation features such as the surface westerlies and the split in the upper-level westerly jet stream in austral summer (Bhaskaran and MuIlan, 2003) and therefore on precipitation (MuIlan, 1995). In the regional climate model, the planetary scale forcing is supplied by a coarser resolution atmospheric GCM which simulates the time-averaged Hadley circulation and its

4 22 W e a t h e r and Climate, Volume 22 response to tropical Pacific SST anomalies reasonably well (Bhaskaran and Mullan, 2003). On the other hand, the regional component of the circulation is driven by sharp gradients in the topography, inland water basins, and complex coastline. The regional model should be able to simulate these regional features better than the driving GCM due to its finer resolution. 3. Nesting technique One-way nesting, in which the driving atmospheric GCM passes on its information to regional model but does not receive from it, is employed in the UM, as in many other model configurations (e.g., Renwick et al., 1998). At the lateral boundaries of the RCM, the prognostic variables (surface pressure, horizontal wind components, and temperature and humidity variables adjusted to account for cloud water content (see Smith, 1990)) are relaxed towards atmospheric GCM values at each model level across a four-point boundary buffer zone (Figure 2). For this purpose an increment I5XRi cm is added to the value X; cm simulated by the RCM at each point in the buffer zone (before the next RCM time-step). MR' cm is defined by: 15X1Rcm oti(xbovi X IRcm ); where a. = 1 i 0,1, 2, 3 ( 1 ) XGi cm is the GCM forcing value and i the number of grid points away from the outermost point. The spatial dependency of the relaxation coefficient, a, contributes to the smooth transition from the interior flow field values to those of the field implied by the external data. The nature of this one-way nesting technique forces the following two assumptions: 1) Global forcing external to the regional model domain influences the regional circulation anomalies; and 2) Regional circulation anomalies, in turn, do not influence the global circulation anomalies. The first assumption is quite consistent with the nature of regional synoptic circulations in the New Zealand region, which are strongly affected by the ENSO events taking place external to the region in the tropical Pacific. However, the second assumption may become a severe constraint on using the one-way nested regional climate models, since regional circulations, depending on their nature and geographical locations, may influence the global circulation anomalies. For example, the large-scale circulation initiates mesoscale convection during the monsoon onset over the Indian sub-continent. The convection grows drawing energy from the large-scale circulation. This is desirable as it is consistent with the first assumption. However, the latent heat of condensation resulting from the convection is thought to intensify the strength of the large-scale circulation (Goswami, 1994). This violates the second assumption. Similar arguments can be made for extratropical limited area domains. That is, the extratropical high-frequency transients, resolvable within the regional model domain, maintain the

5 Bhaskaran, Renwick, Mullan: Finer scale climate information for New Zealand OS 45S GOS 150E W Figure 2: Regional model domain for New Zealand. The boundary relaxation zone is shaded. large-scale quasi-stationary circulation anomalies external to the domain (Hoerling and Ting, 1994; Bhaskaran and Mullan, 2003). The domain containing New Zealand and adjoining seas is, however, well suited for the application of the one-way nested regional model (Figure 2). It is in proximity to the subtropical belt, away from the convectively active Southern Pacific Convergence Zone (SPCZ). At the same time the southern flank of the domain does not extend far enough to include areas of the active extratropical high-frequency transients (Trenberth, 1991). By not placing the western boundary over the Australian land-mass, we avoid the regional model generating a sea-breeze. This is desirable because the RCM cannot communicate the massive sea-breeze arising from the land-sea temperature contrast at its western boundary to the driving model. The location of the eastern boundary is restricted only by the requirement of consistency between the regional model circulation and the global model circulation external to the regional model domain (see below). 4. Domain size considerations The above careful selection of the locations of the lateral boundaries does not completely remove the problems associated with the second assumption. For example, the southern boundary does not completely exclude the areas of the high-frequency transients. Similarly the location of the northern boundary will not completely exclude the convectively active subtropical regions. Nevertheless it minimizes the chances of the regional model circulation growing in such a way that it can affect the stability of the regional model solution. This increases the chances of running the regional model smoothly over a reasonable length of time for a desirable size of the domain, without any numerical problems associated with the physics and dynamics of the model. However, the size of the domain needs to be selected carefully.

6 24 W e a t h e r and Climate, Volume 22 Previous applications of the regional model configurations of the UM (Jones et al, 1995; Bhaskaran et al., 1996) suggest that the influence of sub-grid scale forcing on the large-scale circulation of the regional model increases with the domain size, even after carefully locating the lateral boundaries. A larger domain may enable the regional model to evolve on its own and deviate significantly from the large-scale circulation of the driving model to such an extent that the driving model circulation external to the regional model domain may not be considered physically consistent with the regional model solution. This is not desirable as the regional model depends on the driving model for global information. In other words, the regional model cannot simulate a large-scale circulation on its own which is different from the driving model large-scale circulation over the regional model domain. At the same time the domain should be sufficiently large that the mesoscale circulations simulated within the regional model are not undesirably damped (see Kidson and Thompson, 1998). This suggests the importance of selecting an optimum domain size for a successful application of the one-way nested regional model configuration of the UM. The constraint of the driving model on the regional model simulation depends not only on the regional model domain size, but also on the location of the domain. This is where, we believe, the New Zealand domain has a considerable advantage. The driving model constraint on a variety of regional model domains in the tropics and northern mid-latitudes (Europe) is demonstrated in Figure 3. The correlation between interannual variations in the large-scale circulations of the regional model and the driving model is higher and varies relatively less strongly with domain size for the cases considered in tropics. For Europe the correlations are lower and vary more strongly with domain size. These results presumably reflect differences between mid-latitude and tropical dynamics (see Bhaskaran etal., 1996 for details). For the New Zealand domain, due to the fact that it is in proximity of the subtropical belt, the constraint curve in Figure 3 is expected to be reasonably flat and lie between the curves for the tropical and mid-latitude domains. This will reduce the constraint on selecting the size of the domain. Tests are currently under way to investigate this issue. Our preliminary study using one of the smaller domains for the New Zealand region shows that the model captures well the observed orographically-induced temperature gradients (Figure 4). 5. Other issues 5.1 Resolution The vertical resolution of the regional model is the same as is used in the driving model. This allows us to assess the impact of increased horizontal resolution in the regional model. The fineness of horizontal resolution is only limited by the nature of the regional model formulation (the UM uses hydrostatic primitive equations). However a set of recent experiments (Bhaskaran et al, 1996) with the UM suggested a ratio of the driving model to the regional model grid size of 6:1 for a reasonable regional model simulation.

7 Bhaskaran, Renwick, MuIlan: Finer scale climate information for New Zealand Ce c't o Indian Monsoon Region A A European Region Domain Size 10 71(m2 6 7 Figure 3: Mean correlation between driving model and regional climate model 850 hpa anomaly patterns for June-August versus size of the regional model domain for Indian monsoon region and European region. Correlations are calculated over the area of the smallest domain in each case (after Bhaskaran et al., 1996). In the regional model quasi-uniform resolution is achieved by shifting the coordinate pole so that the domain appears as a rectangular equatorial segment on the rotated grid. This transformation helps to avoid the need for Fourier filtering of shorter wavelengths, since most of the regional model domains fall between 25 degrees latitude on either side of the equator in the rotated coordinate system. 5.2 Relaxation Zone A four-point relaxation zone is generally used to relax the driving model data at the rim of the regional model domain. Orographic heights in the regional model are set equal to those of the driving model in the relaxation zone and also in the four rows/columns immediately inside it. The width of the relaxation zone may have to be changed to maintain the consistency between the regional and driving model solutions in the rim, and to smooth the transition of the driving model signal propagating into the interior of the regional model domain. 5.3 Frequency of External Forcing In a standard regional model configuration the forcing data is updated at each time-step by linear interpolation from the driving model output saved every six hours. This is sufficient to resolve the diurnal cycles in the forcing data. Increasing the frequency of external forcing may force the regional model circulation to closely follow that of the

8 26 W e a t h e r and Climate, Volume 22 Figure 4: Regional model simulation of surface temperature for 1-5 June Contours start from 268K at 2K interval. Note the model's ability to simulate the spatial variation of surface temperature associated with the mesoscale orographic forcing. driving model. This may be necessary when the circulation features of the regional model are complex and deviate significantly from the driving model, even for a smallest possible domain. However this may require enormous online storage space for longer simulations. Since it increases I/O tasks, the model run-time may increase further. 6. Summary We have discussed the suitability of the New Zealand domain for a successful application of the unified model (UM) to simulate high-resolution surface variables. The necessary (though not sufficient) conditions for implementing one-way nested regional climate models are highlighted. It appears that the sub- and extra-tropical location of the NZ domain and its circulation features are well suited to apply a one-way nested regional model configuration of the unified model (UM). Earlier simulations of the UM to produce local climate information for the Indian subcontinent and the European region have been largely successful. Our preliminary results for the NZ region show that the model captures well the observed orographicallyinduced temperature gradients (Figure 4). Therefore we are confident that the UM will perform well to provide accurate finer scale climate information under a variety of past, present, and future climate forcing conditions for the NZ region, such as: simulation of New Zealand climate during the last glacial maximum; extended seasonal forecasts downscaled from the global forecast model output; projected increases in atmospheric CO2 and SO2 concentrations; and as input to climate impacts models for formulation of regional response strategies.

9 Bhaskaran, Renwick, MuIlan: Finer scale climate information for New Zealand 2 7 Acknowledgements This research was funded by the New Zealand Foundation for Research, Science and Technology under Contract C01)(0030. The U.K. Meteorological Office supplied the unified model software to NIWA for research and development. References Bhaskaran, B., Jones, R.G., Murphy, J.M. and Noguer, M. 1996: Simulations of the Indian summer monsoon using a nested regional climate model: domain size experiments. Climate Dynamics, 12, Bhaskaran, B., Murphy, J.M. and Jones, R.G. 1998: Intraseasonal oscillation in the Indian summer monsoon simulated by global and nested regional climate models. Mon. Weather Rev., 126, Bhaskaran, B. and Mullan, A.B. 2003: El Nino related variations in the southern Pacific atmospheric circulation: model versus observations. Climate Dynamics, 20, Cullen, M.J.P. 1993: The unified forecast/climate model. Meteorological Magaine, 122, Deque, M. and Piedelievre, J.Ph. 1995: High resolution climate simulation over Europe. Climate Dynamics 11, Giorgi, F. and Mearns, L.O. 1991: Approaches to the simulation of regional climate change. Rev. Geophys. 29, Goswami, B.N. 1994: Dynamical predictability of seasonal monsoon rainfall: problems and prospects. Proceedings of Indian National Science Academy, 60A(1) sensitivity to location of lateral boundaries. Quarterly J. Royal Meteorol. Soc., 121, Kidson, J.W. and CS. Thompson, 1998: A comparison of statistical and model-based downscaling techniques for estimating local climate variations. J. Climate,11, Mullan, A.B. 1995: On the linearity and stability of Southern Oscillation climate relationships for New Zealand. Intl. J. Climatology, 15, Renwick, J. A., J. J. Katzfey, K. C. Nguyen, and J. L. McGregor, 1998: Regional model simulations of New Zealand climate. Geophys. Res. 103 (D6), Trenberth, K.E. 1991: Storm tracks in the Southern Hemisphere. J. Atmos. Sci., 48, Submitted to Weather and Climate, 10 November 2000; Revised: 12 May 2003 Hoerling, M.P. and Ting, M. 1994: Organization of extratropical transients during El Nino. J. Climate, 7(5), IPCC, 2001: Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 881 pp. Jones, R.G., Murphy, J.M. and Noguer, M. 1995: Simulation of climate change over Europe using a nested regional climate model. Part I: assessment of control climate, including