1. Introduction
The relationship between oxygen and hydrogen isotopic compositions and precipitation in paleoclimate archives, such as in ice cores, cave speleothems, and lake and ocean sediments, has been debated for several decades (e.g., Gat 1996; Sachse et al. 2012). Indeed, the knowledge of the isotopic composition of paleo precipitation provides a solid basis to investigate past climate conditions (e.g., Dansgaard 1964; Baldini et al. 2008; Bowen 2008; Liu et al. 2010). For example, in the midlatitudes a larger isotopic depletion of precipitation is likely related to lower temperatures, whereas in the tropics a larger isotopic depletion of precipitation is rather related to larger amounts of precipitation (e.g., Bowen 2008). Furthermore, these so-called temperature and precipitation amount effects are modulated by a moisture source effect caused by changes in atmospheric circulations patterns, which makes the interpretation of paleoclimate isotope data more complex (e.g., Baldini et al. 2008; Liu et al. 2010).
Located in a transition zone between midlatitude westerlies and tropical easterlies, southern Africa is known to have experienced dramatic atmospheric circulation changes during the last 20 000 years (Truc et al. 2013; Chase et al. 2015). The westerly and easterly flow regimes divide southern Africa in two distinct rainfall zones, namely the winter rainfall zone (WRZ) near the southwest coast, which receives most precipitation during winter, and the summer rainfall zone (SRZ) across the central and western parts, which receives most precipitation during summer (e.g., Roffe et al. 2019). According to Chase et al. (2015), the spatial extent of the WRZ and SRZ during the last 7000 years has been primarily modulated by changes in the tropical easterly flow during summer. A southward shift of the Southern Hemisphere westerlies may explain a drying of the WRZ that possibly occurred between 9000 and 5500 years ago according to Weldeab et al. (2013), and between 7800 and 2400 years ago according to Zhao et al. (2016). There are indications that the region near the west coast in the SRZ may have been relatively dry between 7000 and 2000 years ago and wetter afterward (Norström et al. 2014), with the timing of these dry and wet phases depending on the paleo archive analyzed (e.g., Neumann et al. 2014), which would be the effect of a varying strength of the tropical easterly flow caused by summer insolation fluctuations (e.g., Zhao et al. 2016). Indeed, other climate driving factors such as the shape of the Earth orbit and interhemispheric teleconnections may also have played a large role in past changes of the southern African climate, so that the nature and role of regional atmospheric circulation changes that happened in southern Africa are still debated (Truc et al. 2013). A better knowledge of past atmospheric circulation changes in southern Africa can be gained thanks to the moisture source effect on isotopic signatures in paleoclimate archives, as was done for example by Dupont et al. (2013) and Schefuß et al. (2011) using plant wax hydrogen isotope changes in marine sediment cores.
In Eq. (1), D and H stand for deuterium and hydrogen compositions and subscript VSMOW refers to the Vienna Standard Mean Ocean Water.
Several paleoclimate reconstructions based on plant wax hydrogen isotope changes are now available for southern African and neighboring regions (e.g., Schefuß et al. 2005, 2011; Tierney et al. 2008; Dupont et al. 2013). For example, in a marine sediment core close to the Congo River mouth in central Africa, Schefuß et al. (2005) interpreted abrupt δD decreases and increases as markers for moist and dry periods that have happened during the last 20 000 years, with a highest precipitation rate around 9000 years ago and an aridification afterward. In a marine sediment core close to the Orange River mouth in southern Africa, Burdanowitz et al. (2018) also interpreted a δD minimum around 3900 years ago as a marker for a wet period, followed by a gradual aridification. And similarly, in a marine sediment core close to the Limpopo River mouth in southern Africa, Miller et al. (2020) considered the relatively high δD values between 7000 and 3000 years ago as a marker for a dry phase, followed by a wet phase.
The interpretation of δD in paleoclimate archives becomes more intricate when the effect of moisture source change is considered. For example, in a marine sediment core 300 km away from the mouth of the Orange River, Dupont et al. (2013) found a strong δD decrease that was dated between 7 and 4.5 million years ago. Confronted with stable carbon isotopic compositions and pollen traces from the same sediments, Dupont et al. (2013) suggested that this deuterium depletion did not reflect an increase in southern African precipitation amounts, but rather corresponded to larger contribution of Indian Ocean moisture sources with lengthened continental moisture transport pathways.
The above-cited studies highlight the difficulty to disentangle the contribution of moisture source changes to isotopic signatures in paleoclimate archives, which can be partly alleviated with a climate model (e.g., Sachse et al. 2012). Several authors have shown the ability of global circulation models equipped with stable water isotopologues to reproduce paleo isotope records (e.g., Pausata et al. 2011; Battisti et al. 2014; Caley et al. 2014; Liu et al. 2014; Feng et al. 2016; Thompson et al. 2021), thus allowing us to elaborate relationships between isotopic variations in paleoclimate archives and circulation patterns changes in the models. Tabor et al. (2018) even further equipped the isotope-enabled version of the Community Earth System Model version 1 (iCESM; Brady et al. 2019) with a tagging procedure in order to more precisely analyze the contribution of moisture sources to isotopic variations in paleoclimate archives. Applying the tagging-enhanced version of iCESM with different precession-eccentricity orbital parameters, Tabor et al. (2018) found that isotopic variations in speleothem records of the South Asian summer monsoon were related to variations in precipitation amounts sourced from different regions.
Global models are limited by their coarse resolution, which complicates a realistic description of isotopic fractionation processes related to cloud physics (e.g., Nusbaumer et al. 2017). A realistic description of cloud physics is particularly important in tropical regions in order to properly represent the precipitation amount effect (Worden et al. 2007; Risi et al. 2008). Regional models, with their finer resolution of clouds, have demonstrated their ability to better represent observed isotopic compositions in precipitation, such as in Moore et al. (2016) with the Weather Research and Forecasting (WRF) Model (Skamarock and Klemp 2008), Pfahl et al. (2012) and Aemisegger et al. (2015) with the Consortium for Small-scale Modeling (COSMO; Steppeler et al. 2003), and more recently Arnault et al. (2021) with the hydrologically enhanced version of the WRF Model (WRF-Hydro; Gochis et al. 2018).
Based on the above-cited studies, it is questionable whether applying the isotope tracing procedure of Tabor et al. (2018) with an isotope-enabled regional model would help to better explain regional circulation effects on isotopic compositions in precipitation and improve the interpretation of isotope signatures in paleoclimate archives. This question is especially relevant for the interpretation of southern African paleoclimate archives, where it is known that regional circulation changes may have had a large effect on the isotope records (e.g., Dupont et al. 2013; Schefuß et al. 2011), although this effect cannot be disentangled with the isotope records alone (e.g., Burdanowitz et al. 2018; Miller et al. 2020).
The aim of this study is to evaluate the contribution of moisture source change to the δD signatures in the southern African sediment cores presented by Burdanowitz et al. (2018) and Miller et al. (2020), in order to complement these authors’ work and in the end assess the benefit of the isotope tracing procedure in a particular case. The method to achieve this aim is to enhance the isotope-enabled version of the coupled atmospheric hydrological model (WRF-Hydro-iso; Arnault et al. 2021) with an isotope tracing procedure, and apply the so-called isotope tagging-enabled WRF-Hydro-iso-tag to southern Africa for selected time slices in the past. The choice of the time slices, with a first one during the mid-Holocene 6000 years ago and a second one during preindustrial times, is conditioned by the availability of global model data to drive an isotope-enabled regional model (Otto-Bliesner et al. 2017), so that the following analysis of the δD signatures in the sediment cores of Burdanowitz et al. (2018) and Miller et al. (2020) is restricted to the difference between the mid-Holocene and preindustrial climates.
The isotope tagging-enabled model version WRF-Hydro-iso-tag, which has been tailored for this study, is presented in section 2. The isotope records presented in Burdanowitz et al. (2018) and Miller et al. (2020) are discussed in section 3. Model application details for the southern African Holocene case study are provided in section 4, and results are discussed in section 5. A summary and perspectives are finally given in section 6.
2. Deuterium tracing modeling procedure
a. About WRF-Hydro-iso
WRF-Hydro-iso is a version of the regional coupled model WRF-Hydro (Gochis et al. 2018; Skamarock et al. 2019) which has been enhanced with two additional water cycles and isotopic fractionation processes in order to describe the fate of stable water isotopologues 1H218O and 1H2H16O in the coupled land–atmosphere system (Arnault et al. 2021). Like traditional meteorological variables in the model, the initial and lateral boundary conditions of the isotopic water variables in WRF-Hydro-iso need to be prescribed from the driving data. As such, WRF-Hydro-iso allows for a consistent representation of regional climate and isotopic compositions in terrestrial and atmospheric water compartments.
The original version of WRF-Hydro-iso from Arnault et al. (2021) is based on the WRF version 4.0 and the hydrological module of WRF-Hydro version 5.0, and a specific set of physics parameterization options, which are the six-class WSM6 microphysics scheme of Hong and Lim (2006), the ACM2 atmospheric turbulence scheme of Pleim (2007), and the Noah-MP community Noah land surface model (LSM) with multiparameterization options of Niu et al. (2011) and the terrestrial water routing schemes of Gochis et al. (2018). Technically, the terrestrial and atmospheric water variables of the model are duplicated for the description of the stable water isotopologues cycles in WRF-Hydro-iso. The fate of the isotopic water variables is resolved with so-called tagged prognostic equations, as detailed by Arnault et al. (2016) for the atmospheric part, and by Arnault et al. (2019) for the terrestrial part.
In Eq. (2), Miso is the “normalized” isotopic water variable that gives the quantity of 1H2H16O in a given water reservoir or water flux described by the water variable M, and δMD is the deuterium composition related to M.
Furthermore, WRF-Hydro-iso includes fractionation processes that occur with water phase changes involving the gaseous phase (Merlivat and Nief 1967; Majoube 1971a,b), at the surface with seawater evaporation, bare soil evaporation, and canopy water evaporation, and in the atmosphere with cloud droplet condensation or evaporation, ice nuclei generation or deposition, and rain droplet evaporation. Implementation details about the WRF-Hydro-iso fractionation processes are provided in Arnault et al. (2021).
b. New developments in WRF-Hydro-iso
The tracing of stable water isotopologues in WRF-Hydro-iso has recently been extended to a subset of the atmospheric turbulence schemes available with WRF (Skamarock et al. 2019), following the tracing procedure of Arnault et al. (2016) that consists in assuming that the total and isotopic water are completely mixed and that the physical processes acting on isotopic water are determined with isotopic ratio weights. Seven atmospheric turbulence schemes can now be used with WRF-Hydro-iso, namely YSU (Yonsei University; Hong et al. 2006), MYJ (Mellor–Yamada–Janjić; Janjić 1994), QNSE (quasi-normal scale elimination; Sukoriansky et al. 2005), MYNN2.5 (Mellor–Yamada–Nakanishi–Niino level-2.5 version; Nakanishi and Niino 2004), ACM2 (Asymmetric Convective Model version 2; Pleim 2007), BouLac (Bougeault and Lacarrère 1989), and UWMT (University of Washington Moisture Turbulence; Bretherton and Park 2009). The aim of the atmospheric turbulence parameterization is to determine the vertical diffusion occurring in the atmospheric boundary layer, depending on the weather condition. The seven schemes that have been selected differ by their method to evaluate the turbulent mixing coefficients used in the vertical diffusion formulations. In particular, the turbulent mixing coefficients can be evaluated with a turbulent kinetic energy-based local closure (MYJ, MYNN2.5, BouLac, UWMT), a nonlocal closure (YSU, ACM2), or a spectral closure (QNSE).
These WRF-Hydro-iso model extensions have two advantages. Primarily, having a set of model options allows the model setup to be tuned to a specific application. Secondarily, it is now possible to generate a physics-based WRF-Hydro-iso ensemble, which allows us to evaluate the contribution of model physics uncertainty to the results.
c. Model enhancement for paleoclimate applications
Otto-Bliesner et al. (2017) elaborated simulations protocols for the Paleoclimate Modeling Intercomparison Project, with the consideration of changes in orbital forcing parameters, namely eccentricity, obliquity, and perihelion longitude, and changes in atmospheric greenhouse gas concentrations, namely carbon dioxide, methane, and nitrous oxide. Parameter values to consider for a mid-Holocene time period (MH), around 6000 years ago, as well as for a control time period during the preindustrial period (PI), defined as year 1850 of the current era, are provided by Otto-Bliesner et al. (2017) and repeated in Table 1.
Orbital forcing parameters and greenhouse gas concentrations conditions for the preindustrial period and mid-Holocene period (about 6000 years ago), as provided by Otto-Bliesner et al. (2017).
This detailed formulation of efac, which is not considered in standard WRF, is implemented in WRF-Hydro-iso in order to explicitly represent the change in insolation resulting from a change in eccentricity e, obliquity ϵ, and perihelion longitude ω.
Concerning the changes in atmospheric greenhouse gas concentrations, it is noted that the Rapid Radiation Transfer Model (RRTM) of Mlawer et al. (1997) uses these gas concentrations values as parameters. It is therefore straightforward to update them with the mid-Holocene and preindustrial values given in Table 1, and thereby adapt the RRTM scheme to a paleoclimate application. Accordingly, the current version of WRF-Hydro-iso for paleoclimate application requires the selection of RRTM.
Aiming at facilitating the change of orbital forcing and greenhouse gas parameters in WRF-Hydro-iso, the model code lines where these parameters are specified have been featured with switches, which allow us to select the climate period at the execution of WRF-Hydro-iso. Finally, WRF-Hydro-iso is compiled with the no-leap-year calendar option, in order to handle paleoclimate model forcing data employing this calendar format, such as the ones used in this study.
d. Adaptation of WRF-Hydro-iso to WRF-Hydro-iso-tag
The focus is on tracing the water isotopologue containing an atom of deuterium 1H2H16O from two moisture source regions, the Atlantic and Indian Oceans, in order to assess the contribution of midlatitude westerlies and tropical easterlies to southern African precipitation deuterium in past climates (Truc et al. 2013; Chase et al. 2015). For this purpose, the water tagging procedure of Arnault et al. (2016, 2019) is adapted to the tracing of stable water isotopologues in WRF-Hydro-iso-tag, based on WRF-Hydro-iso. Technically, following Arnault et al. (2016, 2019), an additional set of tagged isotopic water cycles is considered in WRF-Hydro-iso-tag with tagged isotopic variables and tagged isotopic prognostic equations. The tagged isotopic variables are initialized to zero except for the source region where specified tagged isotopic water variables are set equal to their corresponding isotopic water variables. The current version of WRF-Hydro-iso-tag supports three different types of moisture source, namely isotopic surface evaporation or isotopic precipitation in a specified region, or isotopic water vapor inflow from the domain’s boundaries.
To minimize computing time, the version of WRF-Hydro-iso-tag developed for this study is limited to the description of one water isotopologue, namely 1H2H16O, and includes two new tagged isotopic water cycles to trace 1H2H16O from two different moisture sources. This means that a single WRF-Hydro-iso-tag simulation allows to represent deuterium compositions in the coupled land–atmosphere system, as well as the deuterium ratios from two moisture sources.
Specifically for this study, the Atlantic Ocean moisture source is set as the sum of the isotopic sea surface evaporation occurring west of 20°E and the isotopic water vapor at the initial condition and at the lateral boundaries of the simulation domain west of 20°E. Setting the lateral boundaries as a moisture source allows us to also trace the flow of the isotopic water vapor entering the simulation domain from the Atlantic Ocean sector. Reciprocally, the Indian Ocean moisture source is set as the sum of the isotopic sea surface evaporation occurring east of 20°E and the isotopic water vapor at the initial condition and at the lateral boundaries of the simulation domain east of 20°E. The location of the tagged oceanic areas is indicated in Fig. 1.
(a) Terrain elevation of the simulation’s domain, given in meters above sea level. The area in the square encompasses the study region enlarged in (b). The shaded oceanic areas define the two oceanic moisture sources selected for the tagging procedure, namely the Atlantic Ocean in light blue and the Indian Ocean in light red, the boundary between these two source regions being arbitrarily fixed at longitude 20°E. (b) The study region; the blue lines show the river streams and main tributaries of the Orange River basin and Limpopo River basin. Basin boundaries are delineated by the black contour line. The Limpopo basin delineated here also includes three smaller river basins toward the south, namely the Incomati, Matola, and Lusuftu basins. The red X indicates the location of the marine sediment core GeoB8331-4 (Orange basin) and the red plus sign indicates the location of the marine sediment core GeoB20610-2 (Limpopo basin).
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
3. Presentation of selected southern African Holocene sediment archives
The analysis of southern African Holocene climate is narrowed to the difference between the MH and PI periods, using orbital parameters and greenhouse gas concentrations as defined for modeling purposes by Otto-Bliesner et al. (2017; Table 1). Two hydrogen isotope records from marine sediment cores are selected to test the method developed in section 2 for disentangling the moisture source effect on isotopic signatures, namely GeoB8331-4 and GeoB20610-2. First, the marine sediment core GeoB8331-4 presented by Burdanowitz et al. (2018) is located close to the Orange River mouth in a Holocene sediment package (e.g., Meadows et al. 2002), as indicated by the red cross in Fig. 1b. The main sediment supplier of this offshore mudbelt is the Orange River (Compton et al. 2010), so that GeoB8331-4 reflects a spatially integrated climate signal in the region of the Orange River basin. Second, the marine sediment core GeoB20610-2 presented by Miller et al. (2020) is located northeast of the Limpopo River mouth, as indicated by the red plus sign in Fig. 1b. A strong oceanic current off the Limpopo River mouth is known to transport fluvial sediments eastward, so that GeoB20610-2 is assumed to reflect a spatially integrated climate signal in the region of the Limpopo River basin and three other smaller river basins toward the south, namely the Incomati, Matola, and Lusuftu basins (Schüürman et al. 2019; Miller et al. 2020).
The time series of deuterium compositions δD measured on plant waxes of GeoB8331-4 and GeoB20610-2 are shown in Fig. 2 as a function of age given in thousands of years ago (kya). The δD differences between the MH, around 6 kya, and the PI, in this case approximately set to 0.4 kya, are highlighted in Fig. 2. It is argued that the agricultural practice changes by European settlers in the last 300 years (e.g., Guelke 1976) may have changed isotope signals through old soil material remobilization and erosion, so that the period around 0.4 kya is assumed to best represent PI in these isotope records.
Records of deuterium composition δD of plant waxes from marine sediment cores (a) GeoB8331-4 and (b) GeoB20610-2. The locations of these two marine sediment cores are indicated in Fig. 1b. The x axis gives the age in thousands of years ago (kya) and the y axis gives the δD scale (‰). The δD scale is inverted in order to facilitate the interpretation of δD as a precipitation proxy. The deuterium composition changes between the mid-Holocene period, around 6 kya, and the preindustrial period, approximately set to 0.4 kya, are highlighted in both panels.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
According to Fig. 2, the difference between MH and PI climates in southern Africa is associated with a δD enrichment of 4‰ in the region of the Orange basin and a δD depletion of 4.3‰ in the region of the Limpopo basin. The enhanced temporal variability of the δD measurements in the last 1000 years highlights the large uncertainty associated with these numerical values of δD change between MH and PI. Changing the definition of the PI period by ±0.05 kya leads to a δD enrichment in the region of the Orange basin varying between 2‰ and 4‰, and a δD depletion in the region of the Limpopo basin varying between −4.9‰ and −8.3‰. The magnitude of isotopic signal change between MH and PI is therefore quite uncertain, although the sign of isotopic signal change appears to be relatively robust. This is a clear indication that MH and PI were associated with different climates.
Interpreting δD as a proxy for precipitation, as suggested by Burdanowitz et al. (2018) and Miller et al. (2020), these isotope records would indicate that the region of the Orange basin was more arid and the region of the Limpopo basin more humid at PI in comparison to MH. However, these isotope records could also have been impacted by the moisture source effect, which was not addressed by these authors. It is the aim of this study to assess the potential effects that atmospheric circulation changes may have had on these isotopic signatures, using regional paleoclimate simulations with WRF-Hydro-iso-tag.
4. Model simulation design
a. Driving datasets for Holocene climate
The WRF-Hydro-iso-tag driving dataset is obtained with the isotope-enabled Community Earth System Model iCESM (Brady et al. 2019). For this application, iCESM is used in an atmosphere–land-only configuration, that is, with the isotope-enabled Community Atmosphere Model version 5 (iCAM5; Nusbaumer et al. 2017) coupled with the isotope-enabled Community Land Model version 4 (iCLM4; Wong et al. 2017), and with a prescribed condition of the sea surface. In particular, sea surface temperature is derived from the dataset of Hurrell at al. (2008) for PI, and from a coupled simulation with the Community Climate Model System (Braconnot et al. 2007; Gent et al. 2011) for MH. The same seawater isotopic condition is used for MH and PI, based on the ocean oxygen isotope dataset of LeGrande and Schmidt (2006) and assuming a deuterium excess equal to zero, as shown in Fig. 3.
Map of seawater deuterium composition (‰) from LeGrande and Schmidt (2006) projected on the simulation domain shown in Fig. 1a. The area in the square encompasses the study region.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
Two 12-yr runs with iCESM have been conducted, one for MH and another for PI, using the orbital and greenhouse gas forcing information of Table 1 from Otto-Bliesner et al. (2017). The iCESM grid employed is a regular latitude–longitude grid with a spatial resolution of 0.9° in latitude and 1.25° in longitude. The 12-yr time length of forcing dataset is chosen in order to generate 10-yr average results with WRF-Hydro-tag-iso after a 2-yr spinup time period, as in Arnault et al. (2021).
The 10-yr length for the study period is a trade-off between computational cost with WRF-Hydro-iso-tag and the climatological relevance of period-average results. The global energy imbalance at the top of the atmosphere (e.g., Trenberth et al. 2014) for the 10-yr study period with iCESM is +0.4 W m−2 for MH and +0.7 W m−2 for PI. Accordingly, the 10-yr length is too short to reach radiative equilibrium, and the differences between MH and PI 10-yr average simulation results may be partly affected by model internal variability. This is an indication that the robustness of the present study could be improved with a longer length of study period, which was however too computationally expensive with WRF-Hydro-iso-tag to be achieved here.
The iCESM outputs are saved at a 6-h time interval, which is needed to correctly specify the lateral boundary condition of a regional simulation. For the sake of storage space saving, the iCESM outputs have been postprocessed to only keep a subregion encompassing the WRF-Hydro-iso-tag domain shown in Fig. 1a.
The final iCESM dataset contains the necessary atmospheric variables to drive WRF-Hydro-iso-tag, which are the horizontal wind components, temperature, geopotential heights, water vapor and isotopic water vapor on 30 pressure levels for the initial and lateral boundary conditions, soil temperature, soil moisture, isotopic soil moisture, snow cover and isotopic snow cover for the initial soil condition, and sea surface temperature and seawater isotopic compositions for the lower boundary condition. In the following, the iCESM variables for MH are labeled as “CESM,MH” and the iCESM variables for PI are labeled as “CESM,PI”.
b. WRF-Hydro-iso-tag simulations ensemble setup
The above-discussed iCESM datasets are downscaled with WRF-Hydro-iso-tag using the domain shown in Fig. 1a. The horizontal resolution is set to 9 km, as a compromise between computational cost and clouds details resolution. The resulting horizontal grid size is relatively big, with 1251 × 750 grid points, in order to reduce the influence of the boundary condition on WRF-Hydro-iso-tag results in southern Africa. The number of vertical levels is set to 50, with a model top at 10 hPa. The time step to resolve the equations of atmospheric motions is set to 40 s for stability reasons. The simulation time period is 12 years including a 2-yr spinup time, as mentioned above.
The two sets of tagged deuterium variables, representing the deuterium moisture fractions originating from the Atlantic and Indian Oceans, are initialized following the procedure of section 2D. The choice of parameterized physics options is limited to the schemes for which the isotopic enhancement in WRF-Hydro-iso-tag has been implemented, as listed in sections 2a and 2b. In particular, the selected options are the longwave radiative flux scheme RRTM, the shortwave radiative flux scheme of Dudhia (1989), the WSM6 microphysics scheme, and the Noah-MP land surface model. No terrestrial water routing option is considered in the land surface modeling, which would require a much finer grid spacing than the 9 km employed here (e.g., Gochis et al. 2018; Fersch et al. 2020). Concerning, the atmospheric turbulence scheme, seven options are considered, namely YSU, MYJ, QNSE, MYNN2.5, ACM2, BouLac, and UWMT, in order to generate a physics-based ensemble and evaluate model physics uncertainty in the WRF-Hydro-iso-tag simulations results.
This makes a total of fourteen 12-yr WRF-Hydro-iso-tag simulations, seven for MH and seven for PI, which gives a MH–PI pair of simulations for each of the seven atmospheric turbulence schemes considered. For each simulation, two-dimensional surface variables are saved at a daily interval in the model outputs. In the following, WRF-Hydro-iso-tag ensemble-mean variables for the MH subensemble are labeled as “WRF,MH” and WRF-Hydro-iso-tag ensemble-mean variables for the PI subensemble are labeled as “WRF,PI”. The following analysis focuses on the study region displayed in Fig. 1b, which represents a southern African region encompassing the Orange and Limpopo basins.
c. Model validation strategy
The skill of iCESM and WRF-Hydro-iso-tag in producing realistic climate conditions for southern Africa is assessed with present-day observational datasets, chosen as the gridded product of monthly averaged temperature at a spatial resolution of 0.5° from the Climate Research Unit (CRU; Harris et al. 2014), and the gridded product of daily precipitation sums at a spatial resolution of 0.1° from the Integrated Multi-Satellite Retrievals for the Global Precipitation Measurement mission (IMERG; Huffman et al. 2019). The IMERG product is selected because of its relatively high spatial resolution, despite its availability only from June 2000. In the following, CRU and IMERG datasets are averaged for a 10-yr reference period set to 2001–10, and the iCESM and WRF-Hydro-iso-tag results are compared to this climatological reference.
In Eqs. (5) and (6), P represents the total precipitation, whereas Pwinter stands for the precipitation falling between April and September and Psummer stands for the precipitation falling between October and March.
Isotopic results from iCESM and WRF-Hydro-iso-tag are finally assessed with monthly observation at three stations from the Global Network of Isotopes in Precipitation (GNIP) of the International Atomic Energy Agency and the World Meteorology Organization (IAEA/WMO 2020), namely the Cape Town Airport station at 33.97°S, 18.60°E, the Pretoria station at 25.73°S, 28.18°E, and the Windhoek station at 22.95°S, 17.15°E. (The location of these stations is visualized on the maps shown in Fig. 8.) GNIP observations allow us to derive station-based climatological values of precipitation deuterium composition δPD, which are compared with model results using Eq. (2).
d. Climate change evaluation strategy
Both the iCESM and WRF-Hydro-iso-tag datasets allow us to investigate the interdependencies between the MH-to-PI changes in temperature T, precipitation P, and precipitation deuterium compositions δPD in southern Africa. This is achieved by comparing 10-yr average maps of the CESM,MH; CESM,PI; WRF,MH; and WRF,PI variables. The WRF-Hydro-iso-tag simulations provide further information with respect to changes in deuterium source ratios.
Consequently, the effect of a change in the amount of winter and summer precipitation from a source area on δPD is analyzed with 10-yr average maps of WRF,MH and WRF,PI deuterium source ratios.
Equation (11) is not exact due to numerical uncertainties in the isotope tracing procedure. Assuming small numerical uncertainties, Eq. (11) allows us to quantify the contribution of moisture source changes to a δPD change. For clarity, it is emphasized that these moisture source changes include the both source-specific changes in precipitation amount and temperature, so that the deuterium source ratios, as defined in Eq. (7), are still needed to disentangle the precipitation amount effect associated with a moisture source.
Finally, in order to link WRF-Hydro-iso-tag results with the marine sediment core measurements discussed in section 3, the T, P, rAtlantic,winter, rIndian,winter, rAtlantic,summer, rIndian,summer, and δPD quantities are spatially averaged for the Orange basin and the Limpopo basin using precipitation weights. The aim of the weighted averaging is to account for the fact that regions receiving more rainfall probably deliver more sediments to the river that are finally stored in the marine sediment cores.
The MH-to-PI changes in weighted average quantities, namely ΔT, ΔP, ΔrAtlantic,winter, ΔrIndian,winter, ΔrAtlantic,summer, ΔrIndian,summer, and ΔδPD, are considered for each MH–PI pair of the WRF-Hydro-iso-tag ensemble, a MH–PI pair being defined as a set of MH and PI simulations with the same turbulence parameterization option. Uncertainties in the WRF-Hydro-iso-tag setups with respect to atmospheric turbulence parameterization lead to different pair values of ΔT, ΔP, ΔrAtlantic,winter, ΔrIndian,winter, ΔrAtlantic,summer, ΔrIndian,summer, and ΔδPD, which allows us to better appreciate the relationships between δPD changes and climate changes in the Orange and Limpopo basins.
5. Modeling results
a. Modeled climate biases
Maps of 10-yr average temperature and precipitation from iCESM and WRF-Hydro-iso-tag are compared with the CRU and IMERG climatological references in Figs. 4 and 5, respectively.
(a) Map of CRU temperature T (K) averaged between 2001 and 2010, projected on the study region shown in Fig. 1b. (b),(c) Differential maps between CRU temperature shown in (a) and a 10-yr averaged temperature from the iCESM simulation for mid-Holocene and preindustrial period, respectively. (d),(e) As in (b) and (c), but with the ensemble-mean temperature from the WRF-Hydro-iso-tag simulations.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
(a) Map of IMERG precipitation P (mm day−1) temporally averaged between 2001 and 2010, projected on the study region shown in Fig. 1b. (b),(c) Percentage bias maps between IMERG precipitation shown in (a) and a 10-yr averaged precipitation from the iCESM simulation for the mid-Holocene and preindustrial period, respectively. (d),(e) As in (b) and (c), but with the ensemble-mean precipitation from the WRF-Hydro-iso-tag simulations.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
The CRU temperature in Fig. 4a shows that present-day climate is associated with warmer temperatures toward the northern tropical region and colder temperatures in the uphill regions of the Orange and Limpopo basins. The temperature difference between CRU and iCESM in Figs. 4b and 4c and between CRU and WRF-Hydro-iso-tag in Figs. 4d and 4e is mostly negative, except near the northwest coast. The iCESM negative temperature difference is largest close to the east coast, with difference values mostly between −1 and −2 K, whereas the WRF-Hydro-iso-tag negative temperature difference is generally much larger, especially toward the west, with difference values mostly between −3 and −5 K. The fact that the temperature difference patterns for each model remain comparable between MH and PI suggests that these differences are related to systematic model temperature biases, and that the smallest temperature bias is obtained with iCESM. The relatively higher WRF-Hydro-iso-tag temperature bias shown in Figs. 4d and 4e could be related to the choice of WRF-Hydro-iso-tag model physics, since WRF-Hydro-iso-tag is driven by iCESM. The fact that MH and PI have considerably less greenhouse gas concentration than the 2001–10 period chosen for the reference dataset may also partly explain the cold bias.
The IMERG precipitation in Fig. 5a shows that present-day climate is associated with enhanced precipitation in the northern and western parts of southern Africa, as well as near the south coast. The region near the northwest coast is particularly dry, associated with the Namibian desert. The precipitation percentage difference between IMERG and iCESM in Figs. 5b and 5c is mostly very positive, except along the west coast, whereas the precipitation percentage difference between IMERG and WRF-Hydro-iso-tag in Figs. 5d and 5e is mostly negative, except near the west, south, and east coasts. The iCESM positive precipitation percentage difference is largest close to the east coast with percentage difference values above 100%, where the climatological precipitation is largest, and toward the west as well where the climatological precipitation is relatively low. Such a widespread wet bias in southern Africa seems to be a recurrent feature with iCESM (e.g., Brierley et al. 2020).
With WRF-Hydro-iso-tag, the area of positive precipitation percentage difference is much smaller compared to that with iCESM, with percentage difference values below 50% near the east coast, and above 50% near the southwest coast. The WRF-Hydro-iso-tag negative percentage difference that covers a broad central region is largest toward the northwest, with percentage difference values mostly between −60% and −80%.
As for temperature in Fig. 4, the fact that the precipitation percentage difference patterns for each model remain comparable between MH and PI in Fig. 5 suggests that these differences are related to systematic model precipitation percentage biases. Accordingly, WRF-Hydro-iso-tag displays the smallest precipitation percentage bias in the region receiving the largest climatological precipitation amount near the east coast, and iCESM displays a particularly large positive percentage bias in the central and western parts of southern Africa, which receive less precipitation climatologically.
WRF-Hydro-iso-tag additionally displays a large positive percentage bias close to the southwest coast, where iCESM rather shows a negative percentage bias. Since rainfall near the southwest coast is predominantly orographic, resulting from the interaction between midlatitude westerlies and mountain ranges (e.g., Kalognomou et al. 2013), it is argued that the resolution of iCESM is too coarse to properly resolve such orographically induced precipitation, leading to its underestimation. The fact that WRF-Hydro-iso-tag overestimates the IMERG precipitation near the southwest coast may indicate that WRF-Hydro-iso-tag resolves more orographic precipitation details than IMERG in this region, as shown in Figs. 5d and 5e.
The representation of the WRZ and SRZ is assessed with rwinter [Eq. (5)] and rsummer [Eq. (6)] in Figs. 6 and 7, respectively. As documented in literature (e.g., Roffe et al. 2019), the present-day precipitation climatology, in this case from IMERG, gives a WRZ near the southwest coast in Fig. 6a, and a SRZ across the central and western parts of southern Africa in Fig. 7a. Figures 6b, 6c, 7b, and 7c show that iCESM systematically misrepresents the WRZ with a smaller spatial extent toward the east in comparison to IMERG. This means that the positive percentage bias of precipitation with iCESM in the western and central parts of southern Africa (see Figs. 5b,c) is related to an overestimated contribution of summer precipitation. With WRF-Hydro-iso-tag, the winter and summer precipitation contribution differences compared to IMERG climatology are much reduced for PI in Figs. 6e and 7e, but not for MH in Figs. 6d and 7d. The better resemblance of the WRF-Hydro-iso-tag PI seasonal precipitation maps with the present-day climatology suggests that WRF-Hydro-iso-tag better captures the southern African seasonal precipitation climatology in comparison to iCESM.
(a) Map of IMERG winter precipitation ratio rwinter (%), as defined in Eq. (5), temporally averaged between 2001 and 2010, projected on the study region shown in Fig. 1b. (b),(c) Differential maps between IMERG winter precipitation ratio in (a) and a 10-yr averaged winter precipitation ratio from the iCESM simulation for the mid-Holocene and preindustrial period, respectively. (d),(e) As in (b) and (c), but with the ensemble-mean winter precipitation ratio from the WRF-Hydro-iso-tag simulations.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
(a) Map of IMERG summer precipitation ratio rsummer (%), as defined in Eq. (6), temporally averaged between 2001 and 2010, projected on the study region shown in Fig. 1b. (b),(c) Differential maps between IMERG summer precipitation ratio in (a) and a 10-yr averaged summer precipitation ratio from the iCESM simulation for the mid-Holocene and preindustrial period, respectively. (d),(e) As in (b) and (c), but with the ensemble-mean summer precipitation ratio from the WRF-Hydro-iso-tag simulations.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
The impacts of these temperature and precipitation biases on modeled isotopes are evaluated with maps of 10-yr average δPD and climatological GNIP station values in Fig. 8. Both iCESM and WRF-Hydro-iso-tag show much smaller δPD at the location of the GNIP stations. With iCESM, the positive percentage bias of precipitation in the central and western parts of southern Africa appears to induce a large δPD decrease that is unrealistic according to the GNIP observation, especially for PI (Figs. 8a,b). With WRF-Hydro-iso-tag, the lower δPD values in Figs. 8c and 8d could be a consequence of the much enhanced negative temperature bias discussed above (see Figs. 4d,e).
(a),(b) Maps of precipitation deuterium δPD (‰) temporally averaged for a 10-yr time period derived from the iCESM simulation for the mid-Holocene and preindustrial period, respectively, projected on the study region shown in Fig. 1b. (c),(d) As in (a) and (b), but with the ensemble-mean precipitation deuterium from the WRF-Hydro-iso-tag simulations. The circles indicate multiyear averaged GNIP observations at Cape Town Airport station (southwestern circle), Pretoria station (eastern circle), and Windhoek station (northwestern circle).
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
The model biases displayed in Figs. 4–8 highlight the discrepancies in the representation of southern African climate with both iCESM and WRF-Hydro-iso-tag, with a more realistic representation of precipitation seasonality in the case of WRF-Hydro-iso-tag, but a more realistic representation of temperature in the case of iCESM.
b. Climate change from driving data
According to the iCESM climatological maps of geopotential and winds at the 800-hPa level displayed in Fig. 9, but also at the 750- and 850-hPa levels (not shown), the MH-to-PI changes in southern Africa are associated with the enhancement of an anticyclonic circulation above the study region in winter (Figs. 9a,c) and the enhancement of a cyclonic circulation above the study region in summer (Figs. 9b,d). In particular, there is a weakening of the winter westerlies in the southwest, as shown in Fig. 9c, and a weakening of summer easterlies in the northwest, as shown in Fig. 9d. This confirms that southern Africa has experienced moisture source shifts between MH and PI (e.g., Zhao et al. 2016).
(a) Map of geopotential height (Z800; m; color shading) and winds (m s−1; arrows) at the 800-hPa pressure level temporally averaged for the Southern Hemisphere winter months April–September of a 10-yr time period, derived from the iCESM simulation for the mid-Holocene and projected on the simulation domain shown in Fig. 1a. The square area encompasses the study region. (b) As in (a), except for the Southern Hemisphere summer months October–March. (c),(d) As in (a), but for the differential maps of geopotential height and winds in winter and summer, respectively, derived from the difference between the iCESM simulation for the mid-Holocene and another iCESM simulation for the preindustrial period.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
The stronger winter westerlies in the southwest during MH suggest that the southward shift of the Southern Hemisphere westerlies during mid-Holocene hypothesized by Weldeab et al. (2013) and Zhao et al. (2016) was not yet efficient 6000 years ago. On the one hand, stronger summer easterlies in the northeast during PI confirm a wetter phase driven by enforced tropical easterly flow in the last 2000 years (Zhao et al. 2016). On the other hand, weaker summer easterlies in the northwest during PI have not been reported yet in the literature. The weaker summer easterlies in the northwest and the weaker summer westerlies in the southwest during PI are viewed as a circulation re-equilibrium process resulting from the southward shift of the Southern Hemisphere westerlies and the weakening of the South Atlantic anticyclone between MH and PI. The associated enhancement of the summer cyclonic circulation above the study region corresponds to a strengthening of the Angola low, which is suspected to modulate summer precipitation in southern Africa by controlling moisture transport from the South Atlantic (e.g., Pascale et al. 2019).
The TCESM and PCESM changes between MH and PI in southern Africa are displayed in Figs. 10a and 10b. According to Figs. 10a and 10b and Fig. 5b, there is a temperature increase in the coastal regions where the precipitation is lowest and precipitation change is small. However, Figs. 10a and 10b also show a temperature decrease in most inland parts of southern Africa, in association with much increased precipitation. These iCESM modeled patterns are interpreted as the result of two counteracting processes. On the one hand, global models of the climate change between MH and PI generally predict a warming and wetting of the southern African region (e.g., Brierley et al. 2020), a large-scale process that is assumed to take place with iCESM as well. On the other hand, enhanced precipitation leads to wetter soils, enhanced evaporation, and lower temperatures, a process that also takes place in iCESM through the Community Land Model (Wong et al. 2017). The later process would overbalance the former process in the areas with sufficiently large precipitation increase, which in this case corresponds to most inland parts of southern Africa as shown by the MH-to-PI increase in latent heat flux displayed in Fig. 11a.
Differential maps of (a) temperature T (K), (b) precipitation P (%), and (c) precipitation deuterium δPD (‰), derived from the difference between the iCESM simulation for the mid-Holocene and another iCESM simulation for the preindustrial period, temporally averaged for a 10-yr time period and projected on the study region shown in Fig. 1b. (d)–(f) As in (a)–(c), but with the WRF-Hydro-iso-tag ensemble means for the mid-Holocene and preindustrial period.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
(a) Differential maps of latent heat flux LH (W m−2) derived from the difference between the iCESM simulation for the mid-Holocene and another iCESM simulation for the preindustrial period, temporally averaged for a 10-yr time period and projected on the study region shown in Fig. 1b. (b) As in (a), but with the WRF-Hydro-iso-tag ensemble means for the mid-Holocene and preindustrial period.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
c. Isotopic change from driving data
Changes in δPDCESM between MH and PI in southern Africa are provided in Fig. 10c. As shown in Figs. 10a–c, the southwest area of southern Africa experiences an increase in δPDCESM together with an increase in TCESM. This localized positive change in deuterium compositions is interpreted as a temperature effect, especially because PCESM does not change much in this southwest region.
In most other parts of southern Africa, however, there is a decrease in δPDCESM associated with a decrease in TCESM and an increase in PCESM, as illustrated in Figs. 10a–c. The spread-out negative change in δPDCESM is interpreted as a precipitation amount effect, which usually dominates the temperature effect in tropical regions (e.g., Bowen 2008). This iCESM result would corroborate the marine sediment core interpretation of Miller et al. (2020) for the Limpopo basin region, with relatively lower δD values in the sediments during the last 3000 years standing for a wet phase. Quantitatively, the much-enhanced negative change in δPDCESM in comparison to the deuterium record suggests that the magnitude of the MH-to-PI precipitation increase in Fig. 10b is too large.
However, there is a mismatch for the Orange basin, where δPDCESM mainly decreases in association with an increase in PCESM, although according to the marine sediments analyzed by Burdanowitz et al. (2018) an increase in δPD instead happened in this region, in association with a possible aridification.
d. Rainfall zone change from driving data
Southern African rainfall zone changes between MH and PI are evaluated with
Differential maps of (a) winter and (b) summer precipitation ratios rwinter and rsummer (%) as defined in Eqs. (5) and (6), derived from the difference between the iCESM simulation for the mid-Holocene and another iCESM simulation for the preindustrial period, temporally averaged for a 10-yr time period and projected on the study region shown in Fig. 1b. (c),(d) As in (a) and (b), but with the WRF-Hydro-iso-tag ensemble means for the mid-Holocene and preindustrial period.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
In the region of the Orange basin, the previously discussed MH-to-PI precipitation increase (see Fig. 10b) is not associated with much change in the contribution of winter and summer precipitation, as shown in Figs. 12a and 12b, which is related to the fact that the contribution of winter precipitation in this region is already very low during MH with iCESM (see Figs. 6b,c). In the region of the Limpopo basin the contribution of the winter precipitation is not that low, at least during MH as shown in Fig. 6b, so that the previously discussed MH-to-PI precipitation increase in this region is associated with a clear increase of summer precipitation contribution shown in Fig. 12b, which is in line with other studies (e.g., Chase et al. 2015; Zhao et al. 2016).
e. Added value of regional downscaling
The WRF-Hydro-iso-tag ensemble results with regard to TWRF, PWRF, and δPDWRF changes between MH and PI are provided in Fig. 10d–f. Contrary to the above discussed iCESM results, the WRF-Hydro-iso-tag downscaling suggests that a warming happened in entire southern African region, as shown in Fig. 10d, which better resembles what is usually found with global models (e.g., Brierley et al. 2020). The difference between iCESM and WRF-Hydro-iso-tag temperature changes is assumed to be related to the fact that WRF-Hydro-iso-tag does not produce such a large MH-to-PI increase in precipitation in most areas, as shown in Fig. 10e, so that the precipitation cooling effect with WRF-Hydro-iso-tag is much reduced. Still, TWRF displays a smaller increase where PWRF is also much enhanced, especially in the eastern and northern parts of southern Africa (see Figs. 10d,e). The resemblance between PWRF change patterns in Fig. 10e and latent heat change patterns in Fig. 11b shows that the precipitation cooling effect resulting from enhanced precipitation is also happening with WRF-Hydro-iso-tag, to a lesser extent.
The MH-to-PI drying of the WRZ resolved with iCESM is much smaller in terms of amplitude and spatial extent in comparison to WRF-Hydro-iso-tag, as shown in Figs. 10b, 10e, 12a, and 12c. Since iCESM only partially resolves the WRZ orographic precipitation, as discussed in section 5a (see Figs. 5b,c), it is argued that iCESM is not suitable to resolve a precipitation change in this region, thus explaining the magnitude difference of WRZ drying between iCESM and WRF-Hydro-iso-tag in Figs. 10b, 10e, 12a, and 12c.
Nevertheless, the MH-to-PI precipitation decrease in the WRZ, as simulated with WRF-Hydro-iso-tag, is consistent with the above-discussed pattern of reduced winter westerlies in the driving iCESM data (see Fig. 9c) and interpretation of paleo archives in the region (Weldeab et al. 2013; Zhao et al. 2016). This suggests that the large-scale atmospheric circulation changes simulated with iCESM are realistic, and that regional downscaling, such as with WRF-Hydro-iso-tag, is necessary to represent regional-scale climate change features related to orographic precipitation in the WRZ.
The reduced MH-to-PI precipitation increase with WRF-Hydro-iso-tag, in comparison to iCESM, leads to a reduced precipitation amount effect in WRF-Hydro-iso-tag, and δPDWRF in Fig. 10f does not decrease as much as δPDCESM in Fig. 10c. Note that δPDWRF in Fig. 10f is rather increased in the southern and eastern parts of southern Africa, which better matches the marine sediment cores for the Orange basin (Burdanowitz et al. 2018) and to some extent for the Limpopo basin (Miller et al. 2020) as well. However, the MH-to-PI δPDWRF increase in the southern and eastern parts of southern Africa is not associated with a precipitation decrease (see Fig. 10e), which contradicts the aridification process proposed by Burdanowitz et al. (2018). The MH-to-PI δPDWRF increase is partly collocated with an increase of the summer precipitation contribution shown in Fig. 12d, which suggests that the two processes are related.
These results highlight the sensitivity of δPD to climate change, with much different δPD pattern changes between iCESM and WRF-Hydro-iso-tag that reflect much different climate pattern changes between the two models. Recalling that iCESM predicts a MH-to-PI δPD increase in the region of the Orange basin that does not match the isotopic signal in the marine sediment core analyzed by Burdanowitz et al. (2018) (see section 5c), and WRF-Hydro-iso-tag improves this data matching, it could be that the large MH-to-PI increase in precipitation in most areas of southern Africa, as simulated with iCESM, is not realistic, and that this can be corrected with regional downscaling, such as with WRF-Hydro-iso-tag. The reason for a potentially too large MH-to-PI precipitation increase with iCESM may be related to the fact that iCESM generally overestimates precipitation in this region (see Figs. 5b,c; Brierley et al. 2020), so that it would also overestimate a precipitation increase in this region.
f. About Southern African moisture source change
WRF-Hydro-iso-tag allows us to assess the MH-to-PI changes in moisture sources, which also contribute to the MH-to-PI changes in δPDWRF.
Figure 13 displays the contributions of the winter and summer Atlantic and Indian moisture sources to precipitation in southern Africa for MH in Figs. 13a–d, as well as the change in these contributions between MH and PI in Figs. 13e–h. Figures 13a and 13d clearly show a WRZ originating from the Atlantic Ocean moisture source, and an SRZ originating from the Indian Ocean moisture source, a well-known precipitation pattern in southern Africa (e.g., Roffe et al. 2019). Still, Figs. 13b and 13c show that, to some extent, the Indian Ocean moisture source also contributes to the winter precipitation and the Atlantic Ocean moisture source also contributes to the summer precipitation, as highlighted in a recent study by Rapolaki et al. (2020).
Maps of ensemble mean-deuterium source ratios (a) rAtlantic,winter, (b) rIndian,winter (c) rAtlantic,summer, and (d) rIndian,summer as defined in Eq. (7), temporally averaged for a 10-yr time period derived from the WRF-Hydro-iso-tag simulations for the mid-Holocene. (e)–(h) As in (a)–(d), but for the ensemble-mean differential maps of rAtlantic,winter, rIndian,winter, rAtlantic,summer, and rIndian,summer derived from the difference between the WRF-Hydro-iso-tag simulations for the mid-Holocene and the WRF-Hydro-iso-tag simulations for the preindustrial period.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
Compared to MH, PI is characterized by a decrease in winter moisture sources in Figs. 13e and 13f and an increase in summer moisture sources in Figs. 13g and 13h. This confirms that the precipitation change pattern in Fig. 10e is related to a poleward shift of the subtropical westerly-flow and tropical easterly-flow regimes, as also illustrated in Fig. 9 and discussed in Weldeab et al. (2013) and Zhao et al. (2016). Of particular interest is the MH-to-PI increase in summer Atlantic moisture source in the region of the Orange River basin displayed in Fig. 13g, which is interpreted as the consequence of the weakening of the summer easterlies in the northwestern part of southern Africa and the strengthening of a cyclonic circulation, namely the Angola low (e.g., Pascale et al. 2019), bringing more moisture from the northwest as displayed in Fig. 9d.
The role of moisture source changes on δPD is further assessed with the deuterium delta ratios defined in Eq. (10) and displayed in Fig. 14. The sum of the MH-to-PI changes in δPDAtlantic,winter, δPDAtlantic,summer, δPDIndian,winter, and δPDIndian,summer shown in in Fig. 14e is relatively close to the MH-to-PI change in δPD shown in Fig. 10f, which demonstrates that the isotope tracing procedure developed in this study is accurate enough to allow the assessment of moisture source change effects on modeled δPD according to Eq. (11). Strikingly, the patterns of the MH-to-PI changes in δPDAtlantic,winter, δPDAtlantic,summer, δPDIndian,winter, and δPDIndian,summer in Figs. 14a–d are almost identical to the patterns of the MH-to-PI changes in rAtlantic,winter, rAtlantic,summer, rIndian,winter, and rIndian,summer in Figs. 13e–h. Recalling that the change in deuterium delta ratio combines the dual effect of precipitation amount and temperature changes associated with a moisture source, these almost identical patterns between the δPDsource terms (Figs. 14a–d) and rsource terms (Figs. 13e–h) suggest that the effect of precipitation amount changes between moisture sources is dominating the δPD changes in this case.
Differential maps of ensemble mean-deuterium delta ratios (a) δDAtlantic,winter, (b) δDIndian,winter (c) δDAtlantic,summer, and (d) δDIndian,summer as defined in Eq. (10), temporally averaged for a 10-yr time period derived from the difference between the WRF-Hydro-iso-tag simulations for the mid-Holocene and the WRF-Hydro-iso-tag simulations for the preindustrial period. (e) Sum of the mean-deuterium delta ratios shown in (a)–(d).
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
The fact that the δPDsource terms (Figs. 14a–d) and the rsource terms (Figs. 14e–h) display almost identical patterns does not mean that the differentiation between these two sets of terms is useless. In particular, the sum of the rsource terms cannot be equalized to δPD, but the sum of the δPDsource terms instead. If the deuterium source ratio changes were much smaller, the deuterium delta ratio changes would not only emphasize the temperature effect but also indicate which moisture source contributes to it.
Noting that the present and previous sections’ discussions are based on WRF-Hydro-iso-tag ensemble mean results (Figs. 10–14), we argue that additional information concerning the relationship between climate changes and precipitation deuterium changes can be obtained by comparing single MH–PI pairs of the WRF-Hydro-iso-tag ensemble, which is elaborated in the following section (section 5g).
g. Interpretation of basin-average isotopic signal change
Following the analysis strategy of section 4d, the precipitation weighted average MI-to-PI changes in temperature, precipitation amount, moisture sources, and precipitation deuterium composition, referred to as ΔT, ΔP, ΔrAtlantic,winter, ΔrIndian,winter, ΔrAtlantic,summer, ΔrIndian,summer, and ΔδPD, are visualized as scatterplots of the MH–PI pairs results, an MH–PI pair being defined as the set of MH and PI simulations with the same turbulence parameterization option, for the Orange basin in Fig. 15 and for the Limpopo basin in Fig. 16. Indeed, the different turbulence parameterization options generate different realizations of MH-to-PI changes, which allows us to derive empirical relationships between ΔT, ΔP, ΔrAtlantic,winter, ΔrIndian,winter, ΔrAtlantic,summer, ΔrIndian,summer, and ΔδPD.
Precipitation-weighted Orange River basin-average of precipitation deuterium change ΔδPD from MH to PI, plotted as a function of (a) temperature change, (b) winter Atlantic moisture source change, (c) winter Indian moisture source change, (d) precipitation amount change, (e) summer Atlantic moisture source change, and (f) summer Indian moisture source change, for each MH–PI pair of WRF-Hydro-iso-tag members. Each blue plus sign refers to an MH–PI pair, an MH–PI pair being defined as a set of MH and PI simulations with the same turbulence parameterization option. In each panel the red dotted line shows the linear fit with the fitting parameters and correlation coefficient indicated in the caption.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
As in Fig. 15, but for the Limpopo basin.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-22-0041.1
Each scatterplot of Figs. 15 and 16 features a linear fit, in order to evaluate the linear relationship between a change in a climate variable and a change in precipitation deuterium composition. For each panel, the slope and intercept of the linear fit are indicated in the caption as “y = slope x + intercept” as well as the correlation coefficient R. The slope shows how δPD reacts to the change in a climate variable, and the intercept shows the concurrent effect of other climate variables, with a low intercept denoting a low concurrent effect and a high intercept denoting a high concurrent effect. The value of the correlation coefficient shows how consistent the linear response of δPD is to the climate variable.
The comparison between ΔT and ΔδPD in Figs. 15a and 16a tells about the temperature effect, while the comparison between ΔP and ΔδPD in Figs. 15d and 16d tells about the precipitation amount effect. The moisture source effects can be evaluated with the comparisons between the deuterium source ratio changes Δr and ΔδPD in Figs. 15b,c,e,f and 16b,c,e,f.
1) The Orange basin case
Strikingly, the linear fits in the scatterplots of Fig. 15 are associated with positive correlation coefficient values above 0.6 for the temperature effect (Fig. 15a), the precipitation amount effect (Fig. 15d), the summer Atlantic source effect (Fig. 15e), and the summer Indian source effect (Fig. 15f), but much lower correlation coefficient values, below 0.3, for the winter Atlantic source effect (Fig. 15b) and the winter Indian source effect (Fig. 15c). It is remarkable that the linear fit with ΔrAtlantic,summer in Fig. 15e has the lowest intercept, followed by ΔP in Fig. 15d, ΔT in Fig. 15a, and ΔrIndian,summer in Fig. 15f. This suggests that ΔδPD variations among MH–PI pairs in the Orange basin are tightly connected to rAtlantic,summer, and to some extent to P, T, and rIndian,summer as well. Interestingly, the increase of rIndian,summer is associated with a decrease of ΔδPD, which can be related to the well-known precipitation amount effect, although such an effect does not directly apply in the case of the relationship between rAtlantic,summer and ΔδPD. The fact that an increasing contribution of summer Atlantic contribution increases ΔδPD would indicate that the moisture originating from the summer Atlantic Ocean is less isotopically depleted in comparison to the other sources. This is supported by a more isotopically enriched Atlantic Ocean water in comparison to the Indian Ocean water according to the seawater isotopic condition dataset from LeGrande and Schmidt (2006), as displayed in Fig. 3.
Since most ΔP values in Fig. 15d are positive, it is argued that the δD increase of the order of 4 ‰ measured on plant waxes in the GeoB8331-4 core between MH and PI, as displayed in Fig. 2a, rather represents a larger contribution of the summer Atlantic moisture source, a temperature increase, and a smaller contribution of the summer Indian moisture source, rather than an aridification as suggested by Burdanowitz et al. (2018). Such a MH-to-PI summer precipitation increase in the Orange River basin would be coherent with a long-term trend of increasing vegetation cover and fluvial activity over the past 10 000 years deduced from terrigenous materials in GeoB8331-4 (Hahn et al. 2016), and with other paleo archives in the region (Brook et al. 2010; Scott et al. 2005).
2) The Limpopo basin case
All linear fits in the scatterplots of Fig. 16 are associated with positive correlation coefficient values above 0.5. This suggests that ΔδPD variations among MH–PI pairs in the Limpopo basin are tightly connected to T, P, and the four moisture sources considered. The lowest linear fitting intersect is obtained for ΔrAtlantic,winter and ΔrAtlantic,summer in Figs. 16b and 16e, followed by ΔrIndian,winter and ΔrIndian,summer in Figs. 16c and 16f, ΔT in Fig. 16a, and ΔP in Fig. 16d.
Since most ΔT values in Fig. 16a are positive, and ΔrAtlantic,winter and ΔrAtlantic,summer in Figs. 16b and 16e are rather close to zero, we argue that the δD depletion of the order of 4 ‰ measured in plant waxes in the GeoB20610-2 core between MH and PI, as displayed in Fig. 2b, represents a precipitation increase, as originally stated by Miller et al. (2020), this precipitation effect being nevertheless partly counterbalanced by a temperature increase. The deuterium tracing analysis further indicates that the precipitation increase is associated with a temporal shift of the Indian moisture source from winter to summer.
6. Summary and perspective
The isotope-enabled coupled atmospheric-hydrological WRF-Hydro-iso model from Arnault et al. (2021) has been enhanced with deuterium tracers, and this newly developed isotope tracing-enabled model has been referred to as WRF-Hydro-iso-tag. The aim was to quantify the contributions of two moisture source regions to precipitation deuterium in southern Africa, the Atlantic and Indian Oceans, which may help us better understand the relationships between moisture source changes in past climates and southern African paleoclimate archives. For this application, two sets of 12-yr iCESM global simulation outputs, one for the mid-Holocene and another for the preindustrial period, were downscaled with a physics-based WRF-Hydro-iso-tag ensemble. Seven options of atmospheric turbulence parameterizations were considered, which made a set of seven mid-Holocene–preindustrial pairs of simulations using the same physics each.
Comparing model results with present-day climatological datasets highlighted a larger cold bias with WRF-Hydro-iso-tag and a larger wet bias with iCESM. The more realistic climatological precipitation with WRF-Hydro-iso-tag was related to a better representation of the southern African precipitation seasonality, with more winter rainfall in the southwest. The potential reasons for the enhanced cold bias with WRF-Hydro-iso-tag were not deeply investigated, which could be the topic of a future study.
Comparing the climate change results between mid-Holocene and preindustrial subensemble means, it was found that WRF-Hydro-iso-tag generated a much reduced precipitation increase and much reduced precipitation deuterium decrease in comparison to iCESM, which better matched deuterium records in two marine sediment cores, GeoB8331-4 near the mouth of the Orange River and GeoB20610-2 near the mouth of the Limpopo River.
WRF-Hydro-iso-tag further allowed us to quantify the changes in moisture sources between the mid-Holocene and preindustrial subensembles. In particular, the subensemble-mean precipitation changes were associated with a temporal shift of Atlantic and Indian Ocean moisture sources from winter to summer. The variability of the results among the mid-Holocene–preindustrial pairs of the WRF-Hydro-iso-tag ensemble allowed to derive linear relationships between changes in temperature, precipitation amount, precipitation deuterium sources, and precipitation deuterium compositions. For the Orange River basin, it was inferred that a deuterium increase in plant waxes from core GeoB8331-4 may not result from an aridification, as proposed by Burdanowitz et al. (2018), but rather from a temperature increase and a larger contribution from a more isotopically enriched moisture source, namely the summer Atlantic moisture source, to precipitation in the Orange basin area. For the Limpopo River basin, it was found that a deuterium decrease in plant waxes from core GeoB20610-2 may be due to enhanced precipitation amounts, thus confirming the interpretation of Miller et al. (2020), and that this would be related to an enhanced contribution of the summer Indian Ocean moisture source.
To conclude, this study shows the potential of regional downscaling with-WRF-Hydro-tag to improve the comparison to isotope records in climate archives, as well as the usefulness of the isotope tracing capability to clarify the physical interpretation of such isotope records. Accordingly, the use of WRF-Hydro-iso-tag can be advised for further paleoclimate reconstructions (e.g., Cheddadi et al. 2021).
Acknowledgments.
This research is funded by the German Science Foundation (DFG; Grant AR 1183/2-1). The iCESM datasets were generated at the National Center for Atmospheric Research (NCAR) high-performance computational cluster. The WRF-Hydro-iso-tag model was developed at the IMK-IFU Linux cluster. The WRF-Hydro-iso-tag simulations ensemble was computed at the Gauss Centre for Supercomputing e.V. (www.gauss-centre.eu) with the Supercomputer SuperMUC-NG at Leibniz Supercomputing Centre (www.lrz.de). The acquisition of the marine sediment cores was funded by the German Federal Ministry of Science and Education (BMBF) within the project ‘Regional Archives for Integrated Investigation’ (RAiN, 03G0840A). Special thanks go to Benjamin Fersch, Christoph Sörgel, Uwe Hientz, and Frank Neidl for the computer support, and three anonymous reviewers for their contribution to improve the manuscript.
Data availability statement.
The WRF-Hydro-iso-tag source code used in this study, together with a test application case, can be downloaded at https://doi.org/10.6084/m9.figshare.18523733.v1. The iCESM model is available at https://github.com/NCAR/iCESM1.2. The present-day observational datasets used in this study are available online, the temperature dataset from the Climate Research Unit at http://www.cru.uea.ac.uk/cru/data/hrg/, the precipitation dataset IMERG from the NASA/Goddard Space Flight Center’s Mesoscale Atmospheric Processes Laboratory and PPS at https://gpm1.gesdisc.eosdis.nasa.gov/data/GPM_L3/GPM_3IMERGHH.06/, and the Global Network of Isotopes in Precipitation dataset of the International Atomic Energy Agency and the World Meteorology Organization at https://nucleus.iaea.org/wiser/index.aspx. The marine sediment cores datasets shown in this study can be downloaded from the Pangaea database at https://doi.pangaea.de/10.1594/PANGAEA.873743 (GeoB8331-4) and https://doi.pangaea.de/10.1594/PANGAEA.937743 (GeoB20610-2).
REFERENCES
Aemisegger, F., J. K. Spiegel, S. Pfahl, H. Sodemann, W. Eugster, and H. Wernli, 2015: Isotope meteorology of cold front passages: A case study combining observations and modeling. Geophys. Res. Lett., 42, 5652–5660, https://doi.org/10.1002/2015GL063988.
Arnault, J., R. Knoche, J. Wei, and H. Kunstmann, 2016: Evaporation tagging and atmospheric water budget analysis with WRF: A regional precipitation recycling study for West Africa. Water Resour. Res., 52, 1544–1567, https://doi.org/10.1002/2015WR017704.
Arnault, J., J. Wei, T. Rummler, B. Fersch, Z. Zhang, G. Jung, S. Wagner, and H. Kunstmann, 2019: A joint soil–vegetation–atmospheric water tagging procedure with WRF-Hydro: Implementation and application to the case of precipitation partitioning in the upper Danube River basin. Water Resour. Res., 55, 6217–6243, https://doi.org/10.1029/2019WR024780.
Arnault, J., G. Jung, B. Haese, B. Fersch, T. Rummler, J. Wei, Z. Zhang, and H. Kunstmann, 2021: A joint soil–vegetation–atmospheric modeling procedure of water isotopologues: Implementation and application to different climate zones with WRF-Hydro-iso. J. Adv. Model. Earth Syst., 13, e2021MS002562, https://doi.org/10.1029/2021MS002562.
Baldini, L. M., F. McDermott, A. M. Foley, and J. U. L. Baldini, 2008: Spatial variability in the European winter precipitation δ18O–NAO relationship: Implications for reconstructing NAO-mode climate variability in the Holocene. Geophys. Res. Lett., 35, L04709, https://doi.org/10.1029/2007GL032027.
Battisti, D. S., Q. Ding, and G. H. Roe, 2014: Coherent pan-Asian climatic and isotopic response to orbital forcing of tropical insolation. J. Geophys. Res. Atmos., 119, 11 997–12 020, https://doi.org/10.1002/2014JD021960.
Berger, A., and M. F. Loutre, 1994: Precession, eccentricity, obliquity, insolation and paleoclimates. Long-Term Climatic Variations: Data and Modelling, J.-C. Duplessy and M.-T. Spyridakis, Eds., Springer, 107–151.
Bougeault, P., and P. Lacarrère, 1989: Parameterization of orography-induced turbulence in a mesobeta-scale model. Mon. Wea. Rev., 117, 1872–1890, https://doi.org/10.1175/1520-0493(1989)117<1872:POOITI>2.0.CO;2.
Bowen, G. J., 2008: Spatial analysis of the intra-annual variation of precipitation isotope ratios and its climatological corollaries. J. Geophys. Res., 113, D05113, https://doi.org/10.1029/2007JD009295.
Braconnot, P. B., and Coauthors, 2007: Results of PMIP2 coupled simulations of the Mid-Holocene and Last Glacial Maximum—Part 1: Experiments and large-scale features. Climate Past, 3, 261–277, https://doi.org/10.5194/cp-3-261-2007.
Brady, E., and Coauthors, 2019: The connected isotopic water cycle in the Community Earth System Model version 1. J. Adv. Model. Earth Syst., 11, 2547–2566, https://doi.org/10.1029/2019MS001663.
Bretherton, C. S., and S. Park, 2009: A new moist turbulence parameterization in the Community Atmosphere Model. J. Climate, 22, 3422–3448, https://doi.org/10.1175/2008JCLI2556.1.
Brierley, C. M., and Coauthors, 2020: Large-scale features and evaluation of the PMIP4-CMIP6 mid-Holocene simulations. Climate Past, 16, 1847–1872, https://doi.org/10.5194/cp-16-1847-2020.
Brook, G. A., L. Scott, L. B. Railsback, and E. A. Goddard, 2010: A 35 ka pollen and isotope record of environmental change along the southern margin of the Kalahari from a stalagmite and animal dung deposits in Wonderwerk Cave, South Africa. J. Arid Environ., 74, 870–884, https://doi.org/10.1016/j.jaridenv.2009.11.006.
Burdanowitz, N., L. Dupont, M. Zabel, and E. Schefuß, 2018: Holocene hydrologic and vegetation developments in the Orange River catchment (South Africa) and their controls. Holocene, 28, 1288–1300, https://doi.org/10.1177/0959683618771484.
Caley, T., D. M. Roche, and H. Renssen, 2014: Orbital Asian summer monsoon dynamics revealed using an isotope-enabled global climate model. Nat. Commun., 5, 5371, https://doi.org/10.1038/ncomms6371.
Chase, B. M., S. Lim, M. Chevalier, A. Boom, A. S. Carr, M. E. Meadows, and P. J. Reimer, 2015: Influence of tropical easterlies in southern Africa’s winter rainfall zone during the Holocene. Quat. Sci. Rev., 107, 138–148, https://doi.org/10.1016/j.quascirev.2014.10.011.
Cheddadi, R., M. Carré, M. Nourelbait, L. François, A. Rhoujjati, R. Manay, D. Ochoa, and E. Schefuß, 2021: Early Holocene greening of the Sahara requires Mediterranean winter rainfall. Proc. Natl. Acad. Sci. USA, 118, e2024898118, https://doi.org/10.1073/pnas.2024898118.
Compton, J. S., C. T. Herbert, M. T. Hoffman, R. R. Schneider, and J.-B. Stuut, 2010: A tenfold increase in the Orange river mean Holocene mud flux: Implications for soil erosion in South Africa. Holocene, 20, 115–122, https://doi.org/10.1177/0959683609348860.
Dansgaard, W., 1964: Stable isotopes in precipitation. Tellus, 16, 436–468, https://doi.org/10.3402/tellusa.v16i4.8993.
da Silveira Lobo Sternberg, L., 1988: D/H ratios of environmental water recorded by D/H ratios of plant lipids. Nature, 333, 59–61, https://doi.org/10.1038/333059a0.
Dudhia, J., 1989: Numerical study of convection observed during the Winter Monsoon Experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 3077–3107, https://doi.org/10.1175/1520-0469(1989)046<3077:NSOCOD>2.0.CO;2.
Dupont, L. M., F. Rommerskirchen, G. Mollenhauer, and E. Schefuß, 2013: Miocene to Pliocene changes in South African hydrology and vegetation in relation to the expansion of C4 plants. Earth Planet. Sci. Lett., 375, 408–417, https://doi.org/10.1016/j.epsl.2013.06.005.
Estep, M. F., and T. C. Hoering, 1980: Biogeochemistry of the stable hydrogen isotopes. Geochim. Cosmochim. Acta, 44, 1197–1206, https://doi.org/10.1016/0016-7037(80)90073-3.
Feng, R., C. J. Poulsen, and M. Werner, 2016: Tropical circulation intensification and tectonic extension recorded by Neogene terrestrial δ18O records of the western United States. Geology, 44, 971–974, https://doi.org/10.1130/G38212.1.
Fersch, B., A. Senatore, B. Adler, J. Arnault, M. Mauder, K. Schneider, I. Völksch, and H. Kunstmann, 2020: High-resolution fully-coupled atmospheric–hydrological modeling: A cross-compartment regional water and energy cycle evaluation. Hydrol. Earth Syst. Sci., 24, 2457–2481, https://doi.org/10.5194/hess-24-2457-2020.
Garcin, Y., and Coauthors, 2012: Hydrogen isotope ratios of lacustrine sedimentary n-alkanes as proxies of tropical African hydrology: Insights from a calibration transect across Cameroon. Geochim. Cosmochim. Acta, 79, 106–126, https://doi.org/10.1016/j.gca.2011.11.039.
Gat, J. R., 1996: Oxygen and hydrogen isotopes in the hydrological cycle. Annu. Rev. Earth Planet. Sci., 24, 225–262, https://doi.org/10.1146/annurev.earth.24.1.225.
Gent, P. R., and Coauthors, 2011: The Community Climate System Model version 4. J. Climate, 24, 4973–4991, https://doi.org/10.1175/2011JCLI4083.1.
Gochis, D. J., and Coauthors, 2018: The WRF-Hydro modeling system technical description, version 5.0. NCAR Tech. Note, 107 pp., https://ral.ucar.edu/sites/default/files/public/WRFHydroV5TechnicalDescription.pdf.
Guelke, L., 1976: Frontier settlement in Early Dutch South Africa. Ann. Assoc. Amer. Geogr., 66, 25–42, http://www.jstor.org/stable/2562017.
Hahn, A., J. S. Compton, C. Meyer-Jacob, K. L. Kirsten, F. Lucasssen, M. Pérez Mayo, E. Schefuß, and M. Zabel, 2016: Holocene paleo-climatic record from the South African Namaqualand mudbelt: A source to sink approach. Quat. Int., 404, 121–135, https://doi.org/10.1016/j.quaint.2015.10.017.
Harris, I., P. D. Jones, T. J. Osborn, and D. H. Lister, 2014: Updated high-resolution grids of monthly climatic observations—The CRU TS3.10 dataset. Int. J. Climatol., 34, 623–642, https://doi.org/10.1002/joc.3711.
Hong, S.-Y., and J.-O. J. Lim, 2006: The WRF Single-Moment 6-Class Microphysics scheme (WSM6). J. Korean Meteor. Soc., 42, 129–151.
Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, https://doi.org/10.1175/MWR3199.1.
Hou, J., W. J. D’Andrea, and Y. Huang, 2008: Can sedimentary leaf waxes record D/H ratios of continental precipitation? Field, model, and experimental assessments. Geochim. Cosmochim. Acta, 72, 3503–3517, https://doi.org/10.1016/j.gca.2008.04.030.
Huffman, G. J., E. F. Stocker, D. T. Bolvin, E. J. Nelkin, and J. Tan, 2019: GPM IMERG final precipitation L3 1 day 0.1 degree × 0.1 degree V06, Goddard Earth Sciences Data and Information Services Center (GES DISC), accessed 8 February 2021, https://doi.org/10.5067/GPM/IMERGDF/DAY/06.
Hurrell, J. W., J. J. Hack, D. Shea, J. M. Caron, and J. Rosinski, 2008: A new sea surface temperature and sea ice boundary dataset for the Community Atmosphere Model. J. Climate, 21, 5145–5153, https://doi.org/10.1175/2008JCLI2292.1.
IAEA/WMO, 2020: Global Network of Isotopes in Precipitation. International Atomic Energy Agency, accessed 23 September 2020, https://www.iaea.org/services/networks/gnip.
Janjić, Z. I., 1994: The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945, https://doi.org/10.1175/1520-0493(1994)122<0927:TSMECM>2.0.CO;2.
Kalognomou, E.-A., and Coauthors, 2013: A diagnostic evaluation of precipitation in CORDEX models over southern Africa. J. Climate, 26, 9477–9506, https://doi.org/10.1175/JCLI-D-12-00703.1.
Kanner, L. C., N. H. Buenning, L. D. Stott, A. Timmermann, and D. Noone, 2014: The role of soil processes in δ18O terrestrial climate proxies. Global Biogeochem. Cycles, 28, 239–252, https://doi.org/10.1002/2013GB004742.
Konecky, B., S. G. Dee, and D. C. Noone, 2019: WaxPSM: A forward model of leaf wax hydrogen isotope ratios to bridge proxy and model estimates of past climate. J. Geophys. Res. Biogeosci., 124, 2107–2125, https://doi.org/10.1029/2018JG004708.
LeGrande, A. N., and G. A. Schmidt, 2006: Global gridded data set of the oxygen isotopic composition in seawater. Geophys. Res. Lett., 33, L12604, https://doi.org/10.1029/2006GL026011.
Liu, Z., G. J. Bowen, and J. M. Welker, 2010: Atmospheric circulation is reflected in precipitation isotope gradients over the conterminous United States. J. Geophys. Res., 115, D22120, https://doi.org/10.1029/2010JD014175.
Liu, Z., and Coauthors, 2014: Chinese cave records and the East Asia summer monsoon. Quat. Sci. Rev., 83, 115–128, https://doi.org/10.1016/j.quascirev.2013.10.021.
Majoube, M., 1971a: Fractionnement en 18O entre la glace et la vapeur d’eau. J. Chim. Phys., 68, 625–636, https://doi.org/10.1051/jcp/1971680625.
Majoube, M., 1971b: Fractionnement en oxygene 18 et en deutérium entre l’eau et sa vapeur. J. Chim. Phys., 68, 1423–1436, https://doi.org/10.1051/jcp/1971681423.
Meadows, M. E., J. Rogers, J. A. Lee-Thorp, M. D. Bateman, and R. V. Dingle, 2002: Holocene geochronology of a continental-shelf mudbelt off southwestern Africa. Holocene, 12, 59–67, https://doi.org/10.1191/0959683602hl521rp.
Merlivat, L., and G. Nief, 1967: Fractionnement isotopique lors des changements d’état solide-vapeur et liquide-vapeur de l’eau à des températures inférieures à 0°C. Tellus, 19, 122–127, https://doi.org/10.3402/tellusa.v19i1.9756.
Miller, C., A. Hahn, D. Liebrand, M. Zabel, and E. Schefuß, 2020: Mid- and low latitude effects on eastern South African rainfall over the Holocene. Quat. Sci. Rev., 229, 106088, https://doi.org/10.1016/j.quascirev.2019.106088.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmosphere: RRTM, a validated correlated-k model for the long-wave. J. Geophys. Res., 102, 16 663–16 682, https://doi.org/10.1029/97JD00237.
Moore, M., P. N. Blossey, A. Muhlbauer, and Z. Kuang, 2016: Microphysical controls on the isotopic composition of wintertime orographic precipitation. J. Geophys. Res. Atmos., 121, 7235–7253, https://doi.org/10.1002/2015JD023763.
Nakanishi, M., and H. Niino, 2004: An improved Mellor–Yamada level-3 model with condensation physics: Its design and verification. Bound.-Layer Meteor., 112, 1–31, https://doi.org/10.1023/B:BOUN.0000020164.04146.98.
Neumann, F. H., G. A. Botha, and L. Scott, 2014: 18,000 years of grassland evolution in the summer rainfall region of South Africa: Evidence from Mahwaqa Mountain, KwaZulu-Natal. Veg. Hist. Archaeobot., 23, 665–681, https://doi.org/10.1007/s00334-014-0445-3.
Niu, G.-Y., and Coauthors, 2011: The community Noah land surface model with multiparameterization options (Noah-MP): 1. Model description and evaluation with local-scale measurements. J. Geophys. Res., 116, D12109, https://doi.org/10.1029/2010JD015139.
Norström, E., and Coauthors, 2014: Late Quaternary vegetation dynamics and hydro-climate in the Drakensberg, South Africa. Quat. Sci. Rev., 105, 48–65, https://doi.org/10.1016/j.quascirev.2014.09.016.
Nusbaumer, J., T. E. Wong, C. Bardeen, and D. Noone, 2017: Evaluating hydrological processes in the Community Atmosphere Model version 5 (CAM5) using stable isotope ratios of water. J. Adv. Model. Earth Syst., 9, 949–977, https://doi.org/10.1002/2016MS000839.
Otto-Bliesner, B., and Coauthors, 2017: The PMIP4 contribution to CMIP6—Part 2: Two interglacials, scientific objective and experimental design for Holocene and Last Interglacial simulations. Geosci. Model Dev., 10, 3979–4003, https://doi.org/10.5194/gmd-10-3979-2017.
Pascale, S., B. Pohl, S. B. Kapnick, and H. Zhang, 2019: On the Angola low interannual variability and its role in modulating ENSO effects in southern Africa. J. Climate, 32, 4783–4803, https://doi.org/10.1175/JCLI-D-18-0745.1.
Pausata, F. S. R., D. S. Battisti, K. H. Nisancioglu, and C. M. Bitz, 2011: Chinese stalagmite δ18O controlled by changes in the Indian monsoon during a simulated Heinrich event. Nat. Geosci., 4, 474–480, https://doi.org/10.1038/ngeo1169.
Pfahl, S., H. Wernli, and K. Yoshimura, 2012: The isotopic composition of precipitation from a winter storm—A case study with the limited-area model COSMOiso. Atmos. Chem. Phys., 12, 1629–1648, https://doi.org/10.5194/acp-12-1629-2012.
Pleim, J. E., 2007: A combined local and nonlocal closure model for the atmospheric boundary layer. Part I: Model description and testing. J. Appl. Meteor. Climatol., 46, 1383–1395, https://doi.org/10.1175/JAM2539.1.
Rapolaki, R. S., E. C. Blamey, J. C. Hermes, and C. J. C. Reason, 2020: Moisture sources associated with heavy rainfall over the Limpopo River basin, southern Africa. Climate Dyn., 55, 1473–1487, https://doi.org/10.1007/s00382-020-05336-w.
Risi, C., S. Bony, and F. Vimeux, 2008: Influence of convective processes on the isotopic composition (δ18O and δD) of precipitation and water vapor in the tropics: 2. Physical interpretation of the amount effect. J. Geophys. Res., 113, D19306, https://doi.org/10.1029/2008JD009943.
Roffe, S. J., J. M. Fitchett, and C. J. Curtis, 2019: Classifying and mapping rainfall seasonality in South Africa: A review. S. Afr. Geogr. J., 101, 158–174, https://doi.org/10.1080/03736245.2019.1573151.
Sachse, D., and Coauthors, 2012: Molecular paleohydrology: Interpreting the hydrogen-isotopic composition of lipid biomarkers from photosynthesizing organisms. Annu. Rev. Earth Planet. Sci., 40, 221–249, https://doi.org/10.1146/annurev-earth-042711-105535.
Schefuß, E., S. Schouten, and R. R. Schneider, 2005: Climatic controls on central African hydrology during the past 20,000 years. Nature, 437, 1003–1006, https://doi.org/10.1038/nature03945.
Schefuß, E., H. Kuhlmann, G. Mollenhauer, M. Prange, and J. Pätzold, 2011: Forcing of wet phases in southeast Africa over the past 17,000 years. Nature, 480, 509–512, https://doi.org/10.1038/nature10685.
Schimmelmann, A., A. L. Sessions, and M. Mastalerz, 2006: Hydrogen isotopic (D/H) composition of organic matter during diagenesis and thermal maturation. Annu. Rev. Earth Planet. Sci., 34, 501–533, https://doi.org/10.1146/annurev.earth.34.031405.125011.
Schüürman, J., A. Hahn, and M. Zabel, 2019: In search of sediment deposits from the Limpopo (Delagoa Bight, southern Africa): Deciphering the catchment provenance of coastal sediments. Sediment. Geol., 380, 94–104, https://doi.org/10.1016/j.sedgeo.2018.11.012.