The Influence of Topography on East African October to December Climate : Sensitivity Experiments with RegCM 4

The influence of topography on east African climate is investigated using the International Centre forTheoretical Physics Regional ClimateModel, with focus onOctober to December season. Results show that themean rainfall (temperature) significantly reduces (increases) over the region when topography elevation is reduced. Based on the model, when topography over the selected region (KTU) is reduced to 25%, the mean rainfall (temperature) over east Africa is reduced (increased) by about 19% (1.4C). The maximum rainfall (temperature) reduction (increase) is however observed around the region over which topography is reduced. The reduction in topography elevation resulted in an anomalous moisture divergence at low level and descending motion over the region. KTU topography enhances the surface heat flux over KTU region and tends to enhance convection over both KTU and the east African region. The topography also helps in the generation of the high frequency mesoscale and subsynoptic disturbances over the region.These disturbances produce precipitation over the region and may also enhance precipitation systems over remote areas due to propagation of the disturbances. The magnitude of the zonal wind speed at 850 hpa increases with the decrease in topography elevation.


Introduction
Topography plays an essential role in determining a planet's atmospheric circulation.It has long-range dynamical effects; a mountain can perturb a uniform flow to produce a train of waves downstream.It acts not only as an obstacle to the flow of air, but also as a source of heat.Variations in topography can establish horizontal temperature gradients which can generate a flow.Both dynamical and thermal manifestations of topography represent a problem of great complexity (Blumsack [1]).
The influence of topography (mountains) on different aspects of climate has been studied in different regions.Hahn and Manabe [2] noted that the influence of Tibetan Plateau (TP) on the south Asian monsoon circulation is attributed to mechanical and thermodynamic effects of TP.The presence of mountains is noted to be instrumental in maintaining the south Asian low pressure systems.Dickinson and Knight [3] studied the frontal interaction with mesoscale topography and showed that mountains retard and block the approaching front at the surface while the upper-level potential vorticity anomaly associated with the front moves across the domain unaffected.
Recent studies have made use of different models to examine the influence of topography on climate (e.g., Lee et al. [4] (Tibetan Plateau), [5,6] (Rocky Mountains)).For example, Flesch and Reuter [6] noted that when topography elevation is reduced, rainfall amount reduces.The maximum reduction in rainfall amount is registered over the mountains and foothills and is associated with the reduction in orographic lifting and the associated vertical water vapor flux.
It has been noted that the seasonal rainfall patterns over east Africa are very complex due to the existence of complex topography and large inland water bodies (Indeje et al. [7]).Research on the influence of topography on the east African

Model Description
The Regional Climate Model (RegCM4) is the latest version of the Abdus Salam International Center for Theoretical Physics (ICTP) Regional Climate Model [35,36].It is an evolution of the previous version (RegCM3), described by Pal et al. [37].The dynamical structure of RegCM4.0 is the same as that of the hydrostatic version of the mesoscale model, version 5 (MM5) of the National Center for Atmospheric Research (NCAR) and Pennsylvania State University (Grell et al. [38]).RegCM4.0 and the available model options are briefly summarized as follows.
It is a hydrostatic, compressible, sigma-p coordinate model run on Arakawa B-grid in which wind and thermodynamical variables are horizontally staggered (Giorgi et al. [39]).The radiative transfer scheme used is that of the global model CCM3 [40,41].The planetary boundary layer (PBL) used is the modified Holtslag (Holtslag et al. [42]) and a new PBL scheme of the University of Washington [43,44] is implemented in the model.
Three cumulus convection schemes exist.The simplified version of the Kuo-type scheme of Anthes [45] is described by Anthes et al. [46] and has been present since the earliest version (RegCM1).The second and the mostly used scheme is that of Grell [47] in the implementation of Giorgi et al. [48].Two different closure assumptions can be adopted (Arakawa-Schubert or Fritsch-Chappell).The third is the MIT scheme, introduced in RegCM3 [37,49,50].
The resolved scale precipitation scheme is essentially based on the SUBEX (subgrid explicit moisture scheme) parameterization of Pal et al. [51] and includes a prognostic equation for cloud water (the scheme is not significantly changed in RegCM4.0 compared to RegCM3, other than in some parameter settings).
The land surface processes are described via the Biosphere-Atmosphere Transfer Scheme (BATS) of Dickinson et al. [52], Subgrid BATS Giorgi et al. [53], and the Community Land Model version CLM3.5 [54,55].Compared to BATS, CLM is a more advanced package described in detail by [56,57].For the ocean fluxes, it makes use of the dragcoefficient parameterization included in the BATS package (Dickinson et al. [52]), and to improve the calculation of diurnal fluxes over the ocean, the prognostic sea surface temperature (SST) scheme described by Zeng and Beljaars [58] was implemented in the model.In RegCM3, Pal et al. [37] implemented the scheme of Zeng et al. [59] which is based on a Monin-Obhukov turbulence representation.This scheme was added in order to improve the excessive evaporation over warm tropical oceans found in the BATS option.
A simplified aerosol scheme specifically designed for application to long-term climate simulations has been incrementally developed within the RegCM system.Solmon et al. [60] first implemented a first-generation aerosol model including SO 2 , sulfates, organic carbon, and black carbon.Zakey et al. [61] then added a 4-bin desert dust module, and Zakey et al. [62] implemented a 2-bin sea salt scheme.Additionally in RegCM4, the dust emission scheme accounts for subgrid emissions by different types of soil, and the soil texture distribution has been updated according to Laurent et al. [63].The dust emission size distribution can now also be treated according to Kok [64].When all aerosols are simulated, 12 additional prognostic equations are solved in RegCM4, including transport by resolvable scale winds,   Grell (Grell, 1993 [47]) Planetary boundary layer Holtslag et al. (1990 [42]), Bretherton et al. (2004 [44]) Microphysics SUBEX (Pal et al. 2000 [51]) Land Surface Processes CLM3.5 (Steiner et al. 2009 [54]; Tawfik and Steiner, 2011 [55]) turbulence and deep convection, sources, and wet and dry removal processes (Giorgi et al. [36]).The RegCM system includes an interactive 1-dimensional thermal lake model which has been applied in different regional settings, such as [65,66].The tropical band configuration is also implemented in RegCM4.In this configuration, the model uses a Mercator projection centered over the equator for a band covering the entire tropical region, from 45 ∘ S to 45 ∘ N. The use of the Mercator projection allows the model grid to exactly cover the tropical band with the end points in the longitudinal direction exactly overlapping (Coppola et al. [67]).

Data and Model Setup
In this study, Grell convection scheme with Fritsch-Chappell closure assumption (Grell-FC) is used, as in Sylla et al. [24].Adeniyi [68] similarly observed that Grell-FC offers a better simulation over west Africa.The land surface processes are described via the Community Land Model, version CLM3.5.
It is chosen based on the model performance in this particular domain, as in Yu and Wang [69].The model was run at a resolution of 50 km (Table 1) over the period 1999-2008.The experiments were done with the initial and lateral boundary conditions obtained from ERA-interim gridded reanalysis data at 1.5 degree resolution (Uppala et al. [70]).The data is a third generation of the European Centre for Medium-Range Weather Forecast (ECMWF) reanalysis product.The sea surface temperatures are obtained from the National Oceanic and Atmospheric Administration (NOAA).It is the Optimum Interpolated Sea Surface Temperature (OISST), produced weekly on a 1 degree resolution (Reynolds et al. [71]).The SST and the boundary conditions are updated 6 hourly in the model.
Three sets of experiments were done.The control experiment with actual topography (TP100) and sensitivity experiments with topography reduced to 75% (TP75) and 25% (TP25).The area over which topography is reduced lies between longitudes 34 ∘ E-38 ∘ E and latitudes 6 ∘ S-2 ∘ N. It will hereafter be referred to as KTU since it covers high mountains in parts of Kenya, Tanzania, and Uganda (KTU) including Mt. Elgon, Mt.Kenya, Mt.Kilimanjaro, Mt.Meru, Usambara Range, and Ngorongoro Crater, among others.Researchers such as those in [4][5][6] used a similar approach in their studies.Precipitation datasets used to evaluate the model performance over east Africa are the Global Precipitation Climatology Project (GPCP) version 2.2 combined precipitation dataset, gridded at 2.5 degree resolution (Adler et al. [72]) provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their web site at http://www.esrl.noaa.gov/psd/and that of Climate Research Unit, CRU TS3.10 dataset [73,74].CRU TS3.10 (CRU TS 3.10.01for precipitation) is a monthly gridded climatology of station data for the period 1901-2009 at a resolution of 0.5 degree.This is because the African rain-gauge data observed during the postindependence era of the 1970s through to recent years has many spatial and temporal discontinuities over large sections of east Africa (Schreck III and Semazzi [75]).The wind, relative humidity, and temperature used to determine and evaluate moisture transport are those of ERA-interim, gridded at 0.75 degree resolution (Dee et al. [76]).The National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis zonal and meridional wind (Kalnay et al. [77]) were used to compute eddy kinetic energy.The first year of the model output was dropped as part of the model spin-up.Analyses were then done over 9 years (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008).

Results and Discussion
In this section, the model results are provided and discussed.Figure 1 shows the model domain and elevation in meters, the study area (EA), and the region KTU over which topography is reduced in the sensitivity experiments.The mean annual cycle of precipitation (in mm/day) for the period 2000-2008, averaged over longitudes 28 ∘ E-42 ∘ E and latitudes 12 ∘ S-5 ∘ N. TP100 is due to the simulation with actual topography (control experiment), TP75 is due to the simulation with topography over KTU reduced to 75%, and TP25 represents simulation when the topography is reduced to 25%.
among the various climatological sources of climate variability over the region (Owiti and Zhu [14]).For example, in the period 1996 to 2005, 9 out of the 10 years are among the years with the highest annual temperature on record prior to the Intergovernmental Panel on Climate Change report (IPCC [78]).Knowledge of variability of rainfall and temperature in terms of annual cycle are therefore important components in the understanding of climate variability.
The climatology of the annual cycle is characterized by unimodal, bimodal, and trimodal rainfall regimes in different zones over east Africa (Section 1).However, the climatology of the observed area average annual cycle of rainfall over the region [28 ∘ E-42 ∘ E, 12 ∘ S-5 ∘ N] has higher amounts of rainfall received in October through May, with lower values received between June and September.This climatology is well captured by the model over the period 2000-2008 (Figure 2), where the simulation with actual topography (TP100 or control experiment) exhibits higher amount of rainfall received over east Africa, followed by simulation with the topography over the region KTU reduced to 75% (TP75).The least rainfall amounts are received when the topography is reduced further to 25% (TP25), implying that rainfall over the study area decreases with decreasing KTU topography.The presence of KTU topography thus enhances precipitation over the study area.
Analysis of the annual cycle of temperature (Figure 3) reveals that temperature generally increases with reducing topography over KTU.

Seasonal Rainfall and Temperature.
As discussed in Section 1, the ITCZ is recognized for defining the bimodal rainfall regime experienced in most parts of the region during March to May (MAM) and October to December (OND).Rainfall is modulated from one season to another by the interaction between the ITCZ and perturbations in the global climate circulation, as well as with changes in the local circulation systems initiated by land surface heterogeneity induced by variable vegetation characteristics, large inland lakes, and topography (Owiti and Zhu [14]).
In this section, the influence of KTU topography on OND rainfall and temperature over east Africa is examined.Results (Figure 4) show that the model captures the observed pattern of OND rainfall (Figures 4(a) and 4(b)) over the region, with maximum rainfall over the western sector as opposed to the northeastern sector which tends to receive less rainfall.The sensitivity tests, Figures 4(c) and 4(d), reveal that there is a general reduction in rainfall in most parts of east Africa when topography over KTU is reduced to 25% (Figure 4(d)).There exists a significant reduction in rainfall around the region over which topography is reduced (Figure 4(g)), with a small section over northern Kenya (Turkana region) exhibiting an increase in rainfall.This corroborates the findings by Flesch and Reuter [6] that the maximum reduction in rainfall amounts is registered over the mountains and foothills.The reduction in rainfall is associated with the reduction in orographic lifting and the associated vertical water vapor flux (Flesch and Reuter [6]).
Further analysis of the remote influence of KTU topography in locations which are not part of KTU (Figure 11) reveals that the reduction in KTU elevation leads to a reduction in rainfall in locations UG (30 ∘ E-33 ∘ E, 1 ∘ N-4 ∘ N) and TZ (33 ∘ E-39 ∘ E, 10 ∘ S-7 ∘ S), whereas over KE (37 ∘ E-39 ∘ E, 3 ∘ N-5 ∘ N) there is an increase in rainfall, which agrees with the observed increase in Figure 4(g).KTU topography thus has influence on remote locations over east Africa.In order to understand the influence of topography on the interannual variability of rainfall, we examine the standard deviation of the mean OND rainfall for the simulation with actual topography and that with topography reduced to 25%.Results (Figure 12) indicate that the standard deviation decreases over all the analysis regions, except KE region, which shows an increase when topography is reduced.This reveals that the interannual variability of rainfall over the region is similarly influenced by topography.
Analysis of OND temperature reveals that the model realistically reproduces the observed temperature (Figures 5(a show that there is a general increase in temperature in most parts of east Africa when topography over KTU is reduced to 25%.The maximum increase in temperature is observed in the region over which topography is reduced (Figure 5(d)).Further analysis reveals that the area average temperature over EA (KTU), Table 3, increases by about 1.4 ∘ C (8 ∘ C) when topography over KTU is reduced to 25%.The increase in temperature may also be associated with the observed reduction in rainfall over the region (Afiesimama et al. [79]).

Wind and Moisture
Transport.Atmospheric circulation is important to precipitation because its ultimate effect is to increase the water vapor in the local air and lower the air temperature by transporting water vapor and invading cold air, leading to air saturation (Lu [80]).In order to offer possible explanation in regard to the observed reduction in rainfall over the region, we investigate the moisture transport at 850 hpa and the associated moisture convergence/divergence during OND season.The climatology of the observed moisture transport is characterized by a high moisture convergence to the east of Lake Victoria (around 35 ∘ E, 1 ∘ S), covering part of KTU (Figure 6(a)).This convergence is well captured by the control experiment (Figure 6(b)).However, in the sensitivity experiment where the topography over KTU is reduced to 25%, the high moisture convergence zone shifts to the west of Lake Victoria, around longitude 30 ∘ E. Figure 6(d) exhibits anomalous moisture divergence over most parts, with maximum divergence over the region between longitudes 33 ∘ E-37 ∘ E and latitudes 6 ∘ S-2 ∘ N, covering most parts of KTU where a significant rainfall reduction is observed.
The magnitude of the mean OND zonal wind speed (mainly easterlies) at 850 hpa over the EA region is observed to increase with decreasing topography elevation (Figure 7).This may be responsible for the observed shift in the maximum moisture convergence zone at this level from the east of Lake Victoria (around 35 ∘ E, 1 ∘ S) to far west.
Further analysis of the vertical velocity (omega) reveals that the climatology of OND omega is characterized by rising motion over the region (Figure 8(a)), particularly between longitudes 28 ∘ E-38 ∘ E (which includes KTU topography region).This climatology is realistically captured by the model when the actual topography is used (Figure 8(b)).
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Zonal wind (m/s) Month TP100 TP75 TP25 Figure 7: The mean annual cycle of the zonal wind magnitude (in ms −1 ) for the period 2000-2008, averaged over longitudes 28 ∘ E-42 ∘ E and latitudes 12 ∘ S-5 ∘ N. TP100 is due to the simulation with actual topography, TP75 is due to the simulation with topography reduced to 75%, and TP25 represents simulation when the topography is reduced to 25%.
When KTU topography is reduced to 25% (Figure 8(c)), the rising motion is suppressed.This implies that the observed rising motion over the region is associated with orographic lifting due to KTU topography.
The vertical velocity difference [(c) minus (b)] in Figure 8(d) shows that there is a net subsidence over the region.The anomalous sinking motion indicates that convection is suppressed as a result of reducing topography to 25%.According to Flesch and Reuter [6], the reduction in rainfall is associated with the reduction in orographic lifting and the associated vertical water vapor flux.It thus explains why rainfall over KTU reduced (by about 60%), which in turn reduced the area average rainfall amount recorded in the entire region of east Africa by about 19% (Table 2).
According to Shi et al. [5], the activities of high frequency mesoscale and subsynoptic disturbances can be measured by the eddy kinetic energy (EKE) in the simulations.EKE is defined as where  and V are zonal and meridional wind components at 500 hPa and  and V are their respective time means.In order to examine the contribution of topography in the generation of the high frequency mesoscale and subsynoptic disturbances over the region, we compute EKE based on November 2006 (wet year) with the time mean values determined over the period 2000-2008.Results (Figure 10) indicate that the simulation with actual topography (TP100, Figure 10(b)) has a zonally oriented high EKE band between 28 ∘ E-43 ∘ E and 10 ∘ S-3 ∘ S (covering the southern sector of KTU), signifying active high frequency mesoscale and subsynoptic disturbances during the simulation period.On the other hand, when topography elevation is reduced to 25% (TP25, Figure 10(c)), a similar high EKE band appears, but the absolute values are much lower, indicating a suppression of the mesoscale and subsynoptic activities.The NCEP reanalysis generally underestimates the EKE (Figure 10(a)) due to the relative coarse resolution, as in Shi et al. [5].TP100 experiment shows that the high frequency mesoscale and subsynoptic disturbances spread over most parts of east Africa, which produces precipitation in the region and may enhance precipitation systems over other regions as they propagate downstream to the inner sectors of east Africa.

Surface Heat Fluxes.
In order to understand the thermal influence of topography over the region, we examine the response of surface heat flux in association with the observed rainfall anomaly due to the reduction in topography.As shown in Figure 9(a), the surface (sensible and latent) heat flux is generally larger in the control experiment (TP100) than in the sensitivity experiment (TP25), particularly over KTU region where the maximum heat flux is observed.

Advances in Meteorology
Figure 9(b) shows that the sensible heat flux increases over the region between latitudes 5 ∘ S-2 ∘ N when topography is reduced to 25% (SHF is more in TP25 than in TP100), implying that the sensible heat flux increases with decreasing topography elevation over this region.However, over the region between longitudes 35 ∘ E-42 ∘ E and latitudes 2 ∘ N-5 ∘ N, the sensible heat flux is greater in TP100 than in TP25.This coincides with the observed significant increase in rainfall in this region (Figure 4(g)), resulting from the reduction in topography.
Analysis of the surface latent heat flux (LHF, Figure 9(c)) over the region reveals that LHF in TP100 is higher than LHF in TP25, implying that LHF deceases when topography is reduced.This is consistent with the observed reduction in rainfall over KTU region due to the reduction in topography.On the other hand, over the region between longitudes 35 ∘ E-42 ∘ E and latitudes 2 ∘ N-5 ∘ N, the latent heat flux is greater in TP25 than in TP100.This shows that LHF increases over this region as KTU topography is reduced.The increase in LHF over this region is similarly consistent with the observed increase in rainfall in the region (Figure 4(g)).
Generally, the presence of topography over KTU in TP100 experiment enhances the surface heat flux over KTU.This may enhance convection both over KTU and the region and  could explain why rainfall over the region is higher in TP100 than in TP25, as in Shi et al. [5].

Summary and Conclusion
In this study, we use the International Centre for Theoretical Physics (ICTP) Regional Climate Model (RegCM4.0) to examine the influence of topography on the east African climate, with focus on October to December (OND) season.
The control simulation was done with actual topography and sensitivity experiments were carried out with topography elevation reduced to 75% and 25%.
Results show that the OND rainfall (temperature) significantly reduces (increases) over the region when topography elevation is reduced.Based on the model, when the topography over KTU is reduced to 25%, the mean OND rainfall (temperature) over EA region reduces (increases) by about 19% (1.4 ∘ C).The maximum rainfall (temperature) reduction (increase) is, however, observed around KTU region, where rainfall (temperature) over KTU reduces (increases) by about 60% (8 ∘ C).This may be explained by the anomalous moisture divergence exhibited over the region at low level and the associated anomalous sinking motion.Divergence at low level results in vertical shrinking which suppresses convection due to subsidence [81,82].11), and EA (28 ∘ E-42 ∘ E, 12 ∘ S-5 ∘ N).TP100 is due to the simulation with actual topography and TP25 represents simulation when the topography is reduced to 25%.
The interannual variability of OND rainfall over the east African region is influenced by KTU topography.
The presence of topography over KTU enhances the surface heat flux over the region.
The reduction in topography leads to an increase in the magnitude of the zonal wind speed (easterlies) at 850 hpa.
The eddy kinetic energy (EKE) shows that topography over KTU helps in the generation of the high frequency mesoscale and subsynoptic disturbances over the region.These disturbances produce precipitation over the region and also enhance precipitation systems over remote areas as a result of the propagation of the disturbances downstream.
The findings from this study and the biases identified call for more detailed investigative studies, particularly in the region.These results may motivate researchers and modeling centers to further improve the performance of the model over east Africa.

Figure 1 :
Figure 1: The model domain and elevation (in m).The rectangle (EA) is the area of study (east Africa), located over longitudes 28 ∘ E-42 ∘ E and latitudes 12 ∘ S-5 ∘ N. The inner rectangle (red) is the area (KTU) over which topography is reduced [34 ∘ E-38 ∘ E, 6 ∘ S-2 ∘ N] in the sensitivity experiments.

4. 1 . 1 2 3Figure 2 :
Figure2: The mean annual cycle of precipitation (in mm/day) for the period 2000-2008, averaged over longitudes 28 ∘ E-42 ∘ E and latitudes 12 ∘ S-5 ∘ N. TP100 is due to the simulation with actual topography (control experiment), TP75 is due to the simulation with topography over KTU reduced to 75%, and TP25 represents simulation when the topography is reduced to 25%.

Figure 3 :
Figure3: The mean annual cycle of the 2 m temperature (in ∘ C) for the period 2000-2008, averaged over longitudes 28 ∘ E-42 ∘ E and latitudes 12 ∘ S-5 ∘ N. TP100 is due to the simulation with actual topography (control experiment), TP75 is due to the simulation with topography over KTU reduced to 75%, and TP25 represents simulation when the topography is reduced to 25%.
) and 5(b)), with the highest temperature values observed in the northeastern sector of EA.Sensitivity tests (Figures 5(b) and 5(c))

Figure 11 :Figure 12 :
Figure11: The mean OND rainfall (mm) over selected regions which are not part of KTU, refered to as remote regions, averaged over KE (37 ∘ E-39 ∘ E, 3 ∘ N-5 ∘ N) in Kenya, TZ (33 ∘ E-39 ∘ E, 10 ∘ S-7 ∘ S) in Tanzania, and UG (30 ∘ E-33 ∘ E, 1 ∘ N-4 ∘ N) in Uganda.TP100 is due to the simulation with actual topography, TP25 represents simulation when the topography is reduced to 25%, CRU refers to the Climate Research Unit data, and GPCP is the Global Precipitation Climatology Project data.

Table
: Model configuration used in the study.

Table 2 :
Statistics of the model simulated mean OND rainfall (RF) in mm over east Africa (RF EA), averaged over longitudes 28 ∘ E-42 ∘ E and latitudes 12 ∘ S-5 ∘ N, and that over KTU (RF KTU), averaged over longitudes 34 ∘ E-38 ∘ E and latitudes 6 ∘ S-2 ∘ N for the control experiment (TP100), sensitivity experiments with topography reduced to 75% (TP75) and 25% (TP25), and the observed CRU and GPCP datasets.B EA and B KTU are the rainfall biases with respect to TP100, over EA and KTU, respectively.The respective percentage biases due to EA and KTU are denoted by PB EA and PB KTU.