|
UGAMP Research
|
|
Use of the IFS Model in studying the impact of land-surface changes on the climate
The aim of this work is to investigate further the regional and non-regional effects of tropical deforestation. A modified version of the IFS12r1 GCM is used to carry out a full tropical deforestation experiment. The analysis includes using the column model version of IFS12r1 to improve the understanding of the role surface parametrizations in the predicted local changes. The far field effects of the deforestation are also investigated; a topic on which there have been few studies (e.g. Zhang et al., 1996).
The IFS GCM version 12r1 has been modified to include surface vegetation and soil classification schemes (Dorman and Sellers, 1989 and Wilson et al., 1987). These have been added to so that the geographically dependent land surface variables are consistent with each other. Geographic variation over land for leaf cover, stomatal resistance functioning, rooting depth and soil hydraulic properties have been included. This is in addition to the present inclusion of albedo, surface roughness and vegetation cover. Overall this provides a more detailed surface description for investigating the significance of deforestation at regional and non-regional scales.
Using this version of the model with 19 vertical levels on a T42 reduced grid, a deforestation experiment has been carried out where all areas classified as tropical rainforest are replaced with impoverished grassland (mainly Amazonia, Central Africa and Asia).
Summary of modelled regional and non- regional changes
Qualitatively similar changes are seen over all the deforested areas, although the largest and most coherent modelled perturbations are over Amazonia. The changes in the annual means of various surface fields over Amazonia are shown in Fig. 1 (these are statistically significant at a 95% confidence level over a majority of Amazonia). The modelled surface evaporation is reduced over Amazonia. The applied vegetation modifications reduce the ease with which water is recycled back into the atmosphere and the surfaceÕs ability to tap into deep water stores. Loss of the tree canopy also increases surface runoff, tending to reduce evaporation further. In addition to this, the albedo increase and reduction in roughness reduce the total energy released into the atmosphere. Both these effects are potentially significant in the reduction of diabatic heating in the troposphere, thereby tending to reduce the low level moisture convergence (Charney, 1975 and Charney et al., 1977). This explains why the reduction in precipitation is larger than that of the evaporation. The modelled 2 metre air temperature is also increased because of the reduced evaporation efficiency.
Significant changes are seen in regions remote from the perturbation sources in the tropics. An example of this is given in Fig. 2, which shows the N.H. winter mean of the difference in the deviation of the stream functions from their zonal means. The planetary waves appear to have been modified in both hemispheres in this season. Preliminary investigation suggests that there also is a correlation between these and the precipitation anomalies over the North Atlantic.
Further work
Work is continuing on the investigation of the remote response of the model to tropical deforestation.
The role of the surface parametrization in the prediction of climate change is also underway with IFS15r1. This is part of a EC collaborative project.
References
Charney, J. (1975). Dynamics of desert and drought in the Sahel. Quarterly Journal of the Royal Meteorological Society, pp 193-202
Charney, J. et al. (1977). A comparative study of the effects of albedo change on drought in semi- arid regions. Journal of the Atmospheric Sciences, pp 1366-1385
Dorman, J.L. and Sellers, P.J. (1989). A global climatology of albedo, roughness length and stomatal resistance for atmospheric general circulation models as represented by the Simple Biosphere Model (SiB). Journal of Applied Meteorology, pp 833-855
Wilson, M. et al. (1987). Sensitivity of the Biosphere Atmosphere Transfer Scheme (BATS) to the inclusion of variable soil characteristics. Journal of Climate and Applied Meteorology, pp341-362
Zhang, H. et al. (1996). Impacts of Tropical Deforestation. Part II: The role of large-scale Dynamics. Journal of Climate, pp 2498-2521
Climate Integrations of IFS Cycle 13R4
Since the 1996 UGAMP Summer meeting, climate integrations of ECMWF's forecast model (IFS Cycle 13r4) have been extended to 5 years, to give stable climate statistics for the model. Two such integrations have been completed, using Eulerian and semi-Lagrangian dynamics.
The integrations are of the atmospheric model only, with a horizontal (spectral) resolution of T42 and 31 levels, and using a prescribed SST climatology from AMIP. The model options for the dynamics and physical parametrizations are identical to those used at ECMWF, with the exception of the full radiation calculation, which is performed at every longitude rather than every fourth longitude at ECMWF, and the timestep. The Eulerian integration uses a timestep of 20 minutes, to limit CFL violations in the polar night jet, while the semi-Lagrangian integration allows a longer timestep of 45 minutes. Earlier tests confirmed a much smaller sensitivity of the semi- Lagrangian model to timestep than to the change from Eulerian dynamics. In addition, a global mass correction is included to prevent long-term trends in surface pressure, of magnitude 2hPa/year with Eulerian dynamics and 2hPa/month with semi-Lagrangian dynamics.
(contd. at top of next column)
(contd. from previous column)
The 5-year integrations essentially confirm findings from the previous single-year integrations, notably the large differences in tropopause behaviour between the two dynamical schemes. This is the subject of a separate article in this newsletter (see following article).
The two figures are examples of seasonal diagnostics from the semi-Lagrangian integration. Fig. 3 shows the high pass transient kinetic energy at 250hPa, with the same (crude) high pass filter as used by the Joint Diagnostics Project (JDP). The semi-Lagrangian model exhibits improved asymmetry of the Southern hemisphere storm-track over the Eulerian model (not shown), particularly in southern spring and summer. The orientation of the North Atlantic storm-track is too zonal, as in the T42 UGCM, and consistent with the resolution sensitivity study of Déqué & Piedelievre (1995, Climate Dynamics, 11, 321-339). Improved orientation in the Eulerian integration, together with evidence that the IFS semi-Lagrangian dynamics is more dissipative at small scales than the Eulerian dynamics, also suggests that inclusion of smaller scale features in the storm-track improves its orientation.
The seasonal velocity potential at 200hPa, shown in Fig. 4 for the semi-Lagrangian integration, reveals some systematic errors of the model which ECMWF are currently working to improve. The seasonal evolution of the main climatological features is captured, but the precipitation and divergent flow in the West Pacific are not well simulated, particularly in the solstitial seasons. In DJF there is a relative minimum in precipitation over Indonesia, while in JJA there is a split ITCZ, with a minimum instead of maximum precipitation at the latitude of the Philippines. These problems are not particularly sensitive to dynamical formulation.
The IFS modelled climate and variability will be documented in a forthcoming UGAMP Technical Report. Monthly, seasonal and time-series data from the integrations will be made available for further analysis within UGAMP.
Mike Blackburn
(University of Reading)
Advection of water vapour and the cold polar tropopause bias in Eulerian GCMs
Climate integrations of ECMWF's forecast model (IFS Cycle 13r4) using Eulerian and semi- Lagrangian advection schemes (Ritchie et al. 1995, modified) demonstrate a similar sensitivity of high latitude tropopause temperature errors to dynamical formulation as found by Chen and Bates (1996). The ubiquitous cold bias around 200hPa in Eulerian models is largely corrected by using 3D interpolating semi-Lagrangian dynamics, though it is replaced in the ECMWF model by a cold bias of 3-5K throughout the depth of the troposphere at high latitudes.
In the spun-up seasonal climate of the model, the tropopause temperature difference between the two model versions is of opposite sign to the difference in total heating from parametrized processes, also as found by Chen and Bates (1995). However it is impossible to determine cause and effect from such a long-term seasonal diagnostic balance.
The spin-up evolution of the integrations, from a January initial state, presents a very different picture. In the first 15 days, differing temperature trends are evident at the high latitude tropopause, the Eulerian integration cooling at roughly 0.1K/day more than its semi-Lagrangian equivalent (Fig.5a). Both the pattern and magnitude of the difference is mirrored in the total parametrized heating (Fig. 5b) and this is dominated by radiation (Fig. 5c).
In this version of the model, the prognostic cloud variables are not advected, so an obvious hypothesis is that differences in water vapour advection are leading to differences in clear-sky radiative forcing which drive the temperature differences, with cloud radiative effects modifying the clear-sky forcing effect. In order to test this hypothesis, single column clear-sky radiative heating calculations have been performed, using the narrow band model of Shine (1991) and Forster & Shine (1997). Zonally averaged temperature and moisture profiles at 60S averaged over the first 16 days of the integrations were used, initially with fixed ozone and CO2 and no cloud. The profile of differences in net radiative heating seen in the GCM is reproduced, with a prominent dipole between 200hPa and 400hPa (Fig. 6). The dipole is enhanced if the semi-Lagrangian temperature profile is retained and only the water vapour profile is varied, suggesting that the different temperature trends act to reduce the radiative forcing differences due to water vapour. Varying only water vapour above 400hPa shows that the radiative signal is dominated by the upper tropospheric water vapour differences.
The results indicate that initial errors in Eulerian advection of water vapour in the upper troposphere, tending to raise the hydropause, lead to errors in clear-sky radiative forcing which act to raise the thermal tropopause too. Large scale dynamical processes then appear to produce a poleward shift in the tropospheric jet, particularly in the Summer hemisphere, and the storm track response seen in the spun-up climate. Large differences in the seasonal evolution of the Southern hemisphere polar vortex are an associated feature of the climate integrations.
A paper is in preparation on this work, and details of the water vapour advection processes involved are under investigation.
References
Chen, M. and Bates, J. R. 1996 A comparison of climate simulations from a semi-Lagrangian and an Eulerian GCM. J. Climate, 9, 1126-1149.
Forster, P. M. de F. and Shine, K. P. 1997 Radiative forcing and temperature trends from stratospheric ozone depletion. J. Geophys. Res. (accepted).
Ritchie, H, Temperton, C, Simmons, A. J., Hortal, M., Davies, T., Dent, D. and Hamrud, M. 1995 Implementation of the Semi-Lagrangian method in a high-resolution version of the ECMWF forecast model. Mon.Wea.Rev., 123, 489-514.
Shine, K. P. 1991 On the cause of the relative greenhouse strength of gases such as the halocarbons. J. Atmos. Sci., 48, 1513-1518.