The interaction between climate and the African great lakes: A study with a regional climate model.
In this PhD project, use will be made of the regional climate model CCLM, in which only very recently, the freshwater lake parameterisation scheme FLake has been implemented. The research will be build around three main axes: (i) improvement of the treatment of turbulent exchanges and lake dynamics in FLake, (ii) climate simulations –and evaluation– for Central and Eastern Africa using CCLM driven by Global Climate Model (GCM) output, and (iii) assessing the uncertainty range of these simulations by developing and implementing a new method, the so-called physically-based statistical downscaling. The second axis again falls into two parts, as runs will be performed to reconstruct both present (2001-2012) and future (2071-2100) climatic conditions.
1. Improving CCLM’s FLake module
The interaction of the atmosphere with an underlying lake surface strongly depends upon the lake’s surface temperature and its time-rate-of-change. Moreover, lakes strongly modify the structure and transportproperties of the atmospheric surface layer and therefore the surface fluxes of heat, moisture and momentum. The two layer freshwater lake model FLake attempts to address both issues by solving a set of differentialequations while using a parameterised vertical temperature structure. However, although the model performs well for small and shallow lakes, ithas difficulties reproducing both the near-surface air temperature and the thermal structure of large lakes. Three major model deficiencies canbe identified in the current version: the representation of turbulent fluxes, the parameterisation of temperature profiles and the spin-up time, each of which will be investigated within the research proposed.
First, the parameterisation of surface water roughness lengths with respect to wind and scalar quantities such as potential temperature and specific humidity will be investigated. For example, the effect of a limited wind fetch over inland water surfaces compared to sea surfaces on the momentum transfer equation needs to be assessed as it is still lacking in the present model version and biases the parameterisation of turbulent flux exchanges. Second, the temperature profile parameterisation will be improved and extended as to include the abyssal layer. This issue is especially important in the context of the African rift lakes, given that primary production highly depends on it, and hence a good representation of the lake’s thermal structure is of primary importance for predicting future ecosystem productivity. One possible way of improvement would be to allow for horizontal heat and water transfers to influence the water’sthermal structure. For the evaluation of FLake’s ability to represent lake Kivu’s thermal structure, use will be made of lake temperature profiles collected during multiple past and near-future field campaigns in the Kivu region organised by the EAGLES consortium. Finally, the technicalissue of the lake temperature spin-up following a cold start of FLake will be addressed. Since lakes have a long memory, erroneous initial conditions lead to wrong lake surface temperatures until the memory is faded. A way out of this problem could be to determine a climatological mean state of the lake and to use this as initial conditions for the Flake module. These mean conditions can be derived from an offline Flake integration.
2.1. Simulating present conditions
A present-day simulation will be performed with CCLM for the period 2001-2012, using the lateral boundaries from the European Centre for Medium-range Weather Forecasting (ECMWF) re-analyses. The domain will enclose the larger Central andEastern Africa, using 150 x 150 x 32 grid points with a resolution of ~0.0625° (7 kilometres). This spatial resolution is sufficient to take into account the effect of Lake Kivu on the local climate, as is shown ina modelling study over Lake Chad in West Africa.
Before futureclimate predictions can be performed, it is important to evaluate the model performance for the present-day climate. The ability of CCLM to reproduce the present-day central and eastern African climate will be evaluated using in situ measurements available for the period 2002-2010 from local meteorological stations around the lake and gathered in the EAGLESproject’s Kivu database. Satellite data from the Tropical Rainfall Measurement Mission (TRMM) are available since 2000 on a 0.25° resolution and will be used to spatially evaluate 3 hourly or monthly accumulated precipitation. Cloud cover will be evaluated using satellite observations from the Moderate-resolution Imaging Spectroradiometer on board the Terrasatellite (MODIS-Terra) or the Meteosat Second Generation (MSG) program, available every 15 minutes at a 1km spatial resolution, and atmospheric soundings available from the University of Wyoming.
2.2. Understanding and attributing climate change around Lake Kivu
The climate change signal around Lake Kivu will be simulated using the CCLM model for the period 2071-2100. The Hamburg GCM (ECHAM5) will deliver initial and lateral boundary conditions. These fields are available from international intercomparison exercises contributing to the Intergovernmental Panel on Climate Change (IPCC) 5th Assessment report, more specifically the Coordinated Regional Climate Downscaling Experiment (CORDEX). Their aim is to provide higher-resolution climate information than is available directly from contemporary global climate models (~ a few hundred kilometres). For Africa, the standard spatial resolution of the experiments of CORDEX is 0.44° (50 kilometres). Again, the simulation will be doneon a resolution of 0.0625° for the same domain as the present-day climate simulations.
Based on these simulations, changes in atmospheric circulation and their effect on local climate around the African rift lakes will be investigated. For example, the potential effect ofchanges in large-scale climate oscillations (e.g. El Niño-Southern Oscillation) on the lake physics will be examined, by analogy with what has been reported in Lake Tanganyika.
3. Assessing the uncertainty range by means ofphysically-based downscaling
It is important to emphasize that one RCM simulation for the future provides no information on the range of uncertainty. In order to deal with this critical issue, one can perform multiple dynamical downscaling experiments, i.e. conduct several integrations with the RCM, each time initiated and fed at the boundaries by output from a different GCM. However, although much is to say infavour of this approach, the multi-model ensemble method is computationally very expensive. Given the presence of abundant in-situ observations, one could also opt for statistical downscaling, wherein one searches for linear relationships between e.g. observed surface temperature and precipitation and a range of atmospheric predictor variables. However thisapproach is tenuous as good observational data are very sparse in the present Central African context.
In the PhD project, it is proposed to develop a physically-based statistical downscaling method that uses linear relationships between output both from global and high-resolution regional climate model integrations. Hence, transfer functions will be derived that calculate the temperature and precipitation distributionas a function of, on the one hand, geophysical features (orography, vegetation type, soil characteristics, background albedo, etc.) and, on theother hand, large-scale meteorological conditions like circulation patterns, temperature, humidity, convective available potential energy and other variables derived from ECMWF re-analyses data. Also, the implementation of more advanced statistical techniques is envisaged, such as principal component analysis and maximum covariance analysis, as they allow to find patterns within and between large datasets, and the Mann-Kendall non-parametric trend test, as it is able to detect non-linear relationships. The ability of the physically-based statistical downscaling model to reproduce present climate will be evaluated in similar ways as explained in §2.1. This off-line model will subsequently be driven by the large-scale atmospheric conditions from different GCM climate scenarios available from CORDEX to obtain different possible realisations of future near-surface climate. Hence, variation in the sensitivity of African climate under different scenario’s of future changes in large-scale atmospheric circulation will be taken into account. Moreover, this off-line model will even allow to assess the uncertainty related to the magnitude of the expected global warming –i.e. the choice of the radiative forcing pathway– as it will also be run for different ‘Representative Concentration Pathways’ (RCPs).