Spatio‐Temporal Variations in Sediment Delivery as a Response to Rapid Quaternary Climate Change in the Lake Malawi Rift, East Africa

The interplay of rapid climate change and tectonics drives landscape development, sediment routing, and deposition in early‐stage continental rift systems. The Lake Malawi Rift, in the Western Branch of the East African Rift, is an archetype of a juvenile rift and an ideal natural laboratory for evaluating lacustrine source‐to‐sink systems on orbital or shorter timescales. We examine the interplay of these processes over the past 140 kyr using observations from nested seismic reflection data sets tied to scientific drill cores, which calibrate numerical forward models of this closed sedimentary system. Fault slip rates measured from seismic data drive tectonic displacements in the model. Satellite‐derived precipitation maps constrain modern precipitation and are scaled to previous hydrologic balance studies to reconstruct past climates. Our model reproduces known sediment thicknesses across the rift and accounts for 96% of the estimated siliciclastic sediment deposited over the past 140 kyr. The results demonstrate that the onset of arid climate conditions (140–95 kyr BP) causes extreme drainage adjustments downstream and the formation of mega‐catchments that flow axially into a shallow restricted paleo‐lake. Sedimentation rates during this time are twice the present values due to increased sediment focusing via these axial systems into a much smaller, hydrologically closed lake. As the climate became wetter (95–50 kyr BP), the lake rapidly expanded, decreasing both erosion and sedimentation rates across the rift. This closed‐loop approach allows us to evaluate the role of high‐frequency climate change in modulating basin physiography as well as sediment fluxes in juvenile rift systems.


The Malawi Rift
The Malawi (Nyasa) Rift, situated at the southern end of the WB of the EARS (Figure 1), is a Late Cenozoic extensional system (Macgregor, 2015;Rosendahl, 1987;Scholz et al., 2020).It is composed of three border fault-bounded half-graben basins (North, Central, and South, Figure 1b) that are between 130 and 200 km in length and linked by accommodation or transfer zones, which partition strain along strike (Figure 1; Scholz et al., 2020).Border fault offsets, sediment thicknesses, and measurements of extension from balanced cross sections are at a maximum in the North and Central Basins and decrease markedly in the South Basin, suggesting that the northern and central segments of the rift may have experienced more extension and may be older than their southern counterpart (Scholz et al., 2020); this hypothesis is consistent with the observation that the deepest seismic sequence observed in the North and Central Basins is missing in the South Basin (Flannery & Rosendahl, 1990;Scholz et al., 2020;Specht & Rosendahl, 1989).Scholz et al., 2020), overlain on the regional geology the rift transects (Fritz et al., 2013).Black lines represent fault heave polygons from basement-involved faults that offset the basement horizon.Regional shear zones in red.Note the spot measurements of elevation in panel (a).

10.1029/2022JF007027 4 of 25
The rift zone is positioned at the confluence of several ancient geological terranes and is aligned with, or crosscuts important continental scale shear zones (Daly, 1988;Daly et al., 1989;Ebinger et al., 1984;Fritz et al., 2013;Scholz et al., 2020;Specht & Rosendahl, 1989;Versfelt & Rosendahl, 1989).Within the lake catchment as well as the rift, most of the outcropping bedrock is Proterozoic in age and is related to a series of Pan-African orogenic and collisional events associated with the East African Orogen, and as such contacts between terranes are steeply dipping to near vertical (Figure 1, Daly et al., 1989;Fritz et al., 2013).To the west are the greenschist facies of the Mesoproterozoic Irumide Belt.In the north and northeast are the Paleoproterozoic rocks of the Ubendian Belt, which consist of gneisses, schists, granulites, granites, and gabbros.In the southwest and west are the granulite to amphibolite facies metamorphic rocks of the Southern Irumide belt, which are of Mesoproterozoic to Neoproterozoic age.To the southeast in Mozambique lies the Unango Complex in Mozambique, which consists of granulite facies metamorphic rocks, largely orthogneisses which are Meso-Neoproterozoic in age.In the east are Neoproterozoic Tixitonga Group outcrops, a volcano-sedimentary complex with minor rhyolite flows.
To the east of the rift are the Maniamba and Ruhuhu troughs, both Permo-Triassic sedimentary basins affiliated with Karoo rift events (i.e., Kreuser, 1995), both of which are bounded by modest escarpments (∼400 m) and lie at lower elevations relative to the adjacent crystalline basement (Figure 1).Across the rift from the Ruhuhu trough to the west are additional outcroppings of Karoo sedimentary rocks consisting of sandstones, conglomerates and shales with abundant coal seams (Dixey, 1927).Cretaceous sedimentary units ("Dinosaur Beds" Dixey, 1937aDixey, , 1937b) also outcrop on the northwestern margin of the lake (Figure 1), and may extend northwards into Tanzania.To the north of the rift valley is the Rungwe Volcanic Province (RVP).The RVP is a late Cenozoic extrusive system composed of basalt, trachyte, and phonolite flows, as well as tuffs (Fontijn et al., 2010) that includes the Rungwe volcano, which erupted most recently in the Holocene (Biggs et al., 2021).

Catchment and Hydroclimate
Rift-induced active subsidence, combined with a favorable hydroclimate result in a current overfilling of the rift valley with present day Lake Malawi, one of the world's largest lakes at 560 km long and with a maximum depth of more than 700 m.The total catchment area of Lake Malawi covers an area of 95,700 km 2 and several drainage systems enter the lake across different structural settings (Figure 1).The structural setting combined with pre-rift topography is the primary control on catchment size, drainage patterns, and delta morphology (Crossley, 1984;Lyons, Scholz, et al., 2011;Scholz, 1995;Wells et al., 1999).The main fluvial inputs and the lithologies they crosscut are summarized in Figure 1 and Table 1.
The climate of the Lake Malawi catchment is dominated by annual changes in precipitation rather than temperature due to the migration of the tropical rain belt, also known as the Inter-Tropical Convergence Zone (Nicholson, 2000(Nicholson, , 2017)).Annual precipitation varies spatially along the strike of the rift, ranging from 800 mm/a at the southern tip to 2,400 mm/a in the mountainous areas around the North Basin due to convergence and orographic effects (Malawi Department of Surveys, 1983).Lake level has varied cyclically as a function of hydroclimate, with evidence of 20 lowstands greater than 200 m below present lake level identified over the past 1.3 myr (Ivory et al., 2016;Lyons, Scholz, et al., 2011;Lyons et al., 2015).Indeed, low-stand deltas and deep water unconformities, plus total organic carbon/total inorganic carbon, saturated bulk density, and faunal data yield a coherent record of extremely low lake levels (−550 to 600 m relative to modern) from ∼135 to 95 kyr, with lake levels rising in a stepped fashion between 95 and 50 kyr BP and reaching near modern highstand conditions ∼50 kyr to present (Cohen et al., 2007;Scholz et al., 2007;Lyons, Kroll, & Scholz, 2011;Lyons, Scholz, et al., 2011;Lyons et al., 2015).This period of extreme climate variability is consistent with precessional forcing, although over longer time frames there is evidence of 100 kyr periodicity (Johnson et al., 2016;Lyons et al., 2015;Scholz et al., 2007); moreover, data from cores in Lake Malawi show that during the most extreme arid intervals (∼135-95 kyr BP), Lake Malawi's watershed was converted into a semi-desert environment with <400 mm/a of precipitation (Cohen et al., 2007).
These severe lowstands produced hydrologically closed, shallow, mildly saline, and alkaline lakes (Cohen et al., 2007;Ivory et al., 2016;Lyons, Scholz, et al., 2011).Hydrological modeling of the catchment water budget assuming modern bathymetry (Lyons, Kroll, & Scholz, 2011) indicates that precipitation rates of 63%-100% of modern sustain highstand equilibrium lake levels experienced between 50 kyr to present; 53%-58% of modern precipitation is required to sustain the stepped lake level rises between 95 and 50 kyr BP, and rates of 45% of modern values are required to sustain the lowstand equilibrium lake between 135 and 95 kyr BP (Lyons et al., 2015).

Methods
Different landscape evolution models have been developed to improve our understanding of landscape dynamics over spatial dimensions ranging between single catchments and entire orogens, and over temporal ranges from thousands to millions of years (e.g., Armitage et al., 2011;Cowie et al., 2000;Densmore et al., 2007;Fernandes et al., 2019;Forzoni et al., 2014;Pechlivanidou et al., 2019;Salles & Hardiman, 2016;Tucker & Hancock, 2010;Whipple & Tucker, 2002;Zhang et al., 2020).Here we use the landscape evolution model pyBadlands (Salles et al., 2018), an open source basin and landscape evolution model designed to simulate both erosion and deposition, and to investigate drainage evolution over the rift scale, through time scales of thousands to tens of millions of years.The model integrates hillslope diffusion and river incision by means of a modified stream power law.This approach allows for balanced sediment transport and deposition, varying climate and tectonic forcings, and for the thorough investigation of the interplay between uplift, erosion, precipitation, and deposition (Figure 2).

Governing Equations
The key inputs and outputs of pyBadlands are shown in Figure 2 and below we briefly present the constitutive equations (see Salles et al. (2018) for details).Continuity of mass is defined by the standard equation: where u in myr −1 represents tectonic uplift or subsidence, and q s is the depth integrated, bulk volumetric flux of sediment per unit width (m 2 yr −1 ).The downhill sediment transport rate includes transport by both channel flow, which is described by the stream power law, and long-term slope-driven diffusive processes, defined by non-linear creep (q r and q d respectively, Chen et al. (2014)).
Fluvial incision and the evolution of river networks are simulated through a single-flow-direction algorithm where water is routed over the land surface following the steepest direction of descent (Bahadori et al., 2022;Chen et al., 2014;Salles et al., 2018).For channel flow, we assume that fluvial erosion and transport are detachment limited (i.e., Howard, 1994) and thus transport by channel flow, q r , is modeled using the conventional stream power law equation, defined as a function of net topographic gradient ∇z, and surface water discharge: Uplift/subsidence (from fault movement and regional), landscape Erosion, offshore Deposition, and Lake Level change, all of which may vary through time and space.General rates and ranges of these processes are shown for the Lake Malawi Rift.In a sedimentologically closed system such as the Lake Malawi Rift, the volume of sediment eroded from the landscape E, is equal to the volume of deposited sediment, D, as there is no sediment outflow from the system.
The expression for surface water discharge relates net precipitation P, which can vary spatio-temporally (Figure 2) and drainage area A. The erodibility coefficient, κ b , is a measure of incision efficiency, and m and n are both positive and indicate how the incision rate scales with bed shear stress.The ratio of m/n within our simulations was set at 0.5 (Chen et al., 2014;Tucker & Hancock, 2010), as is routinely employed in similar studies (e.g., Bahadori et al., 2022;Pechlivanidou et al., 2019;Salles et al., 2017).We acknowledge that this is a simplification and that empirical studies have found a range of n values across various mountain landscapes (Gailleton et al., 2021).
The surficial layer (regolith) long-term transport processes (i.e., hillslope processes) are defined through a non-linear diffusion law which assumes that flux rates are proportional to the gradient of topography and increase to infinity if slope values approach a critical slope (Andrews & Bucknam, 1987;Pechlivanidou et al., 2019;Roering et al., 2001;Salles et al., 2018): where κ hl (m 2 /yr) is the diffusion coefficient with different values for the terrestrial and marine realms, and S c is the critical slope.The coefficients κ b and κ hl encompass the influence of climate, lithology, vegetation cover, and channel hydraulics, and are scale dependent (Tucker & Hancock, 2010;Whipple & Tucker, 1999).All precipitation on the landscape does work against the landscape in pyBadlands, and while the model does not explicitly account for other climate processes such as evaporation, evapotranspiration, ground water flow, or the effects of changing vegetation cover, these are implicit in κ b and κ hl .
Within pyBadlands, the source-to-sink processes assume the conservation of mass such that the amount of mass eroded equals that deposited, plus any mass outflow from the system (Bahadori et al., 2022;Chang & Liu, 2019;Salles et al., 2018).In closed systems such as the Malawi Rift, sedimentological outflow is negligible (Figure 2).
Once the sediment reaches the shoreline, fluvial transport stops and sediment accumulation is a function of three conditions: (a) the existence of a depression area identified from Planchon and Darboux (2002) pit-filling algorithms, (b) a sub-aerial land surface, and (c) a local elevation less than the aggregational slope (Bahadori et al., 2022;Salles et al., 2018).Using a pit filling algorithm, pyBadlands estimates the volume of sediment needed to fill the depression area (Salles et al., 2018), and then, based upon the amount of sediment transported by rivers to the shoreline and the maximum water depth available in a timestep, sedimentation starts.Sediments are then diffused through the lacustrine realm (Equation 3).
In pyBadlands, all clastic sediments are subject to compaction and thus to a reduction in porosity.The decrease in porosity with depth is modeled as a negative exponential function (Athy, 1930;Baldwin & Butler, 1985;Sclater & Christie, 1980): where ϕ the present-day porosity at depth, ϕ 0 is the porosity at the surface, c is the porosity depth coefficient (m −1 ), and y is the depth (m).Changes induced by sediment compaction are used to adjust both underlying basins' sedimentary thickness and surface elevation.We calibrated our model using data for the past 140 kyr of rift evolution within the Lake Malawi Rift.For this period, the history of the basin is particularly well constrained.
Values for the parameters used in the governing equations are shown in Table S1 in Supporting Information S1.

Model Setup and Boundary Conditions
pyBadlands simulates the erosion, transport, and deposition of sediment (i.e., the landscape evolution) based upon a series of inputs that we describe in detail in the following sections.These include the topography/bathymetry, precipitation, tectonic displacements, and the erodibility of the rocks over the study area, all of which can vary through space and time.Further information on model set up and boundary conditions is provided in Text S3 in Supporting Information S1.

Model Precipitation and Lake Level Evolution
We reconstruct the climate forcings from the known lake-level history (Lyons, Scholz, et al., 2011;Scholz et al., 2007) and the fraction of modern precipitation required to sustain these lake levels in the Lake Malawi Rift over the past 140 kyr using information from hydrological modeling studies (Lyons, Kroll, & Scholz, 2011).
To approximate the modern-day spatial variations in climate (precipitation) across the Lake Malawi Rift, we use monthly average Global Precipitation Measurement satellite derived estimates of precipitation at a 0.1 degree resolution (∼11,100 m, Huffman et al., 2019), averaged over the past 21 years (the available record length) (Figure 3).We compared this by the amount of modern precipitation from Lyons, Kroll, and Scholz (2011) required to sustain the key lake-level stages identified throughout the past 140 kyr in the Lake Malawi Rift (Lyons, Scholz, et al., 2011;Scholz et al., 2007).Implicit in this method is the assumption that the relative proportion of rainfall in adjacent catchments is constant through time, and only the total volume of input precipitation changes (i.e., the spatial pattern of precipitation is constant over 140 kyr).Furthermore, given that the spatial pattern of precipitation is ratioed to a constant, it is possible that upland precipitation is underestimated during intervals of extreme aridity.While we acknowledge that this approach does not capture local inter-catchment precipitation changes that may be occurring, it does reflect the known regional scale hydroclimate variability across East Africa over the past 140 kyr linked to orbital forcings (Lyons et al., 2015;Schaebitz et al., 2021;Scholz et al., 2007), and is preferable to driving the reference model with either a spatially constant precipitation value, or a model of orographic precipitation.See Text S3.1 in Supporting Information S1 for further details on precipitation input tests and grid creation.

Paleotopography Construction
We assume that the present-day topography has not changed markedly over the past 140 kyr.Indeed, this assumption is valid given that the major faults that form the shoreline and generate the local subsidence and uplift formed their lengths early on in the rift history (e.g., Scholz et al., 2020;Shillington et al., 2020).Therefore, we use the present-day surface topography for our model runs, sampled at the model resolution from the 30 m resolution Shuttle Radar Topography Mission DEM data (Farr & Kobrick, 2000).As the 140-kyr lowstand surface is mapped at a variety of scales across the rift that depend on the line spacing of the seismic reflection data, restoring the lake bottom to this surface might introduce artifacts in the sink geometry.Instead, we assume that the regional structural configuration of Lake Malawi has not changed significantly over the past 140 kyr, as evidenced by the lack of newly formed faults or sub-basins (McCartney & Scholz, 2016;Wright et al., 2023) and merge the DEM with gridded bathymetry data (Lyons, Scholz, et al., 2011;Scholz & Lyons, 2022a).We set the geodetic datum of the completed model topography to the present-day lake level.

Quantifying Tectonic Displacements
To account for tectonic uplift and subsidence in pyBadlands, we compute a total vertical surface displacement map for the Lake Malawi Rift (Figure 4 and Text S3(b) in Supporting Information S1) using the linear elastic dislocation model Coulomb 3.4 (Lin & Stein, 2004;Toda, 2005).For fault geometry we use the mapped surface fault traces from the Malawi Active Fault Database (Williams et al., 2022b), a constant dip of 65°, and a fault root depth of 30 km, as evidenced by the steep nodal planes and deep earthquake focal mechanisms recorded by the SeGMENT array (Ebinger et al., 2019).We use the cumulative fault slip values from Wright et al. (2023) where available, and where faults are not wholly offshore, we use the Malawi Seismogenic Source Model (Williams et al., 2022a).By integrating intra-and border fault displacements as well as border faults within our vertical surface displacement map, we can accurately constrain sediment pathways and routing systems within the rift basin.Figure 4b shows the resulting map of the total vertical displacement (m) over 140 kyr.The total vertical displacement map is divided equally over the run time (140 kyr) to calculate the uplift rates (m/a) for each cell.
As such, a model cell experiences a linear distribution of vertical displacement.

Quantifying Rock Erodibility
A bedrock erodibility (κ b ) map is generated using the empirical approach of Pechlivanidou et al. (2019) from the mapped terranes across the rift (Figure 1) (Fritz et al., 2013).Mapped lithologies and terranes across the rift are clustered based on their elevation and category (Figure 5), assuming that at the regional scale harder and thus less erodible rocks are more prominent, except for Quaternary volcanics, which we assume are more erodible than their elevation suggests.Sediments (Quaternary through Karoo) are classified as highly erodible.The Unango complex, Irumide and Southern Irumide Belts, the Tixitonga group, as well as the RVP are classified as moderately erodible.The Ubendian belt rocks typically occur at higher elevations and are therefore presumed relatively harder and assigned a low erodibility.We then assign values as follows: Low (κ b = 5 × 10 −7 m (1-2m) yr −1 ), medium (κ b = 1.5 × 10 −6 m (1-2m) yr −1 ), high (κ b = 2.5 × 10 −6 m (1-2m) yr −1 ) and group them into a single layer map of uniform thickness (nominally 3 km).This is justified given that the contacts between terranes and the shear zones within the region are considered steeply dipping to sub-vertical (Daly et al., 1989;Fritz et al., 2013).The inferred κ b values are calibrated by comparing modeled and estimated sediment volumes from seismic reflection data (see below).In addition, we test the sensitivity of the model to a range of spatially constant κ b values (Text S4 in Supporting Information S1).

Sediment Volumes From Seismic Reflection Data
Within both the multi-channel and single-channel seismic data (MCS and SCS respectively, Figure S1 in Supporting Information S1), we observe a shallow, high amplitude and laterally continuous reflection that occurs across the rift and is typically overlain by conformable, low amplitude and acoustically transparent reflections (Figures S2 and S3 in Supporting Information S1).This shallow lake-wide reflector marks the most recent lowstand depositional sequence boundary (∼200-550 m below present lake level) from the latest lake-level cycle (Lyons, Scholz, et al., 2011;Scholz & Rosendahl, 1988;Scholz et al., 2007).The observed seismic reflection is tied back to hole 1C from the Lake Malawi Scientific Drilling Program in the Central Basin of the rift, which is dated at 140.6 ka from the age model of Lyons et al. (2015) at this location (Figure S3 in Supporting Information S1).In this deep part of the Central Basin, the sequence boundary is marked by erosional truncations on the eastern flexural margin, and onlapping sediments in the west (Figure S3 in Supporting Information S1), and thus transitions from an unconformity to correlative conformity across the rift.The low seismic amplitudes above the sequence boundary are correlated with hole 1C to hemipelagic, organic-rich muds with low density and high water content, whereas the high amplitude reflector that delineates the sequence boundary has been correlated with high density carbonate-rich mud and sands with low water and low total organic carbon contents (Lyons, Scholz, et al., 2011;Scholz et al., 2007Scholz et al., , 2011a)).In many localities, this sequence boundary is marked by erosional truncation of underlying reflections, as well as downlap of lowstand delta foreset strata (Lyons, Scholz, et al., 2011;Scholz & Rosendahl, 1988).Deltaic deposits at 200 and 350 m below modern lake level have been observed within the depositional sequence overlying this lowstand sequence boundary in high-resolution SCS data (Scholz, 1995), and these are interpreted as stillstands during the overall long-term transgression (Lyons, Scholz, et al., 2011).These stillstands are not resolved in the MCS data; however, in the SCS data, it is observed that these stillstands are compressed within the lowstand sequence boundary to form a single reflector at progressively shallower depths.This sequence boundary has previously been referred to as the Songwe sequence boundary (Flannery & Rosendahl, 1990;Scholz et al., 1989), Sequence 1 (Lyons, Scholz, et al., 2011), or the "Megadrought" horizon (Shillington et al., 2020).Thus, the age of the mapped reflector varies with lake depth (i.To calculate the volume of sediments deposited over the last 140 kyr, we subtract this surface from the lake floor bathymetry mapped across the rift (Scholz & Lyons, 2022b).This sedimentary package is no greater than 200 ms TWTT in thickness; thus, we use a constant velocity of 1,480 m/s that is consistent with P-wave core measurements (Scholz et al., 2011b) and stacking velocity data (Lyons, Scholz, et al., 2011;McCartney & Scholz, 2016, Figure S3 in Supporting Information S1) to convert the TWTT isopach to depth.The fraction of pore space is calculated at each pixel using the depth to the center of the cell and the porosity depth relationship defined in Text S2 in Supporting Information S1.This pore space is removed and the isopach is then summed over its area to calculate the solid rock volume.
From the isopach data we estimate a total solid rock volume of 462 km 3 has been eroded from the catchment and deposited within the confines of the Lake Malawi Rift over the past 140 kyr.The 2D seismic line locations and the limits of seismic data acquisition in nearshore environments necessitate correcting this volume for no-data areas that are adjacent to the lakeshore (Figures S1 and S4 in Supporting Information S1) and this missing area is equivalent to 8,111 km 2 (∼27% of present-day Lake Malawi's aerial extent).In deep water nearshore areas that are both adjacent to border faults and covered by the seismic reflection data, sediment thickness range between ∼70 and 200 m, whereas on the shallower flexural margins where data exist, sediment thickness varies between 2 and 22 m.Therefore, in these nearshore areas, we make a conservative simplifying assumption that on average the thickness of sediment deposited in this section over the 140-kyr interval is no greater than 40 m, suggesting an additional 324 km 3 of sediment is present in these areas.Therefore, a total offshore sediment volume of ∼786 km 3 is estimated for the past 140 kyr.
While sediments of siliciclastic origin are a major component of the deposited material over the past 140 kyr (Scholz et al., 2011b), it is important, and necessary to remove the additional biogenic, carbonate, organic, and diagenetic constituents from the total measured sediment volume, as our model does not account for this autochthonous material.Approximately 20% of the material from the upper 200 m of the GLAD-MAL05-1B sediment core is autochthonous in origin (Figure S4 in Supporting Information S1); therefore, of the total 740 km 3 of sediment deposited over the past 140 kyr, ∼629 km 3 can be attributed to sediments of siliciclastic/allochthonous origin.

Landscape and Fluvial Networks
Modeled basin topography, and drainage evolution over the 140 kyr model run is in close agreement with the observed changes in the Lake Malawi Rift (Figure 6).From 135-95 kyr BP, with only 45% of the modern-day precipitation (hereafter lowstand climate), the prescribed lake level drops rapidly with the formation of a shallow lake in the Central Basin of the rift that is ∼50 m deep, 100 km long, and 50 km wide (Figure 6).This drop in the prescribed lake level causes a marked drainage reorganization within the rift valley.In the North Basin, the Songwe, Lufira, N. Rukuru, N. Rumphi, and Ruhuhu rivers that are separate drainages at present organize into a single axial system flowing through the desiccated North Basin and routing sediments into the shallow lake in the Central Basin (Figure 6).A similar pattern is observed in the model with the South Basin, whereby the Bua, Dwangwa, and Linthipe rivers organize into a single axial system that flows northwards through the desiccated South Basin into the Central Basin.Within the North Basin, we observe that the path of these systems appears to be structurally controlled by intra-rift faults, whereas structural control seems to play less of a role in the South Basin, except that axial drainage develops parallel and close to the eastern border fault (Figure 6).The S. Rukuru and N. Rumphi drainages continue to drain the footwall of the border fault in the Central Basin.
As the prescribed lake level increases within the model in a stepwise fashion from 95-50 kyr BP (Figure 6), precipitation increases from 53% to 58% of modern and we refer to this as the stillstand climate phase.With increasing precipitation, the extent of the lake rapidly expands to ∼75% of its present-day extent, forming a lake 250 km long, 60 km wide, and ∼400 m deep that occupies the entire Central Basin, and most of the North and South Basins by 50 kyr.The combined river systems that flowed axially into the Central Basin during the lowstand climate phase described above have been divided into their constituent drainages, except for the Linthipe in the South Basin that flows axially into the lake during this period, and the Songwe in the North Basin.The Songwe river is combined with the present day Kiwira river just to the North, forming the Kiwira-Songwe system (Tan & Scholz, 2021).
From 50 kyr to present, precipitation increases from 63% to 100% of modern values (Figure 3).During this period, the prescribed lake level increases and stabilizes at its modern dimensions (Figure 6), increasing its area by ∼25% and deepening by ∼200 m.Most of this expansion occurs in the South Basin, where the lake expands southwards as slopes in this region are shallower and the lake floor less subsided than in the north.We refer to this as the highstand climate phase.Relative to the stillstand phase described above, the Songwe-Kiwira system has split into its constituent rivers in the North Basin, and the Linthipe river now enters the lake on the flexural margin of the South Basin rather than axially.

Comparison to Seismic Reflection Data
To test the veracity of the pyBadlands model it is useful to compare the volume of sediment produced over the 140 kyr model run to that measured from seismic reflection data, and on a more granular basin-by-basin scale to localized high-resolution seismic data (Figure 7).From isopach data, approximately 629 km 3 of allochthonous sediment has been deposited in the Lake Malawi Rift over the past 140 kyr.We compare this to our reference model (which only accounts for the siliciclastic fraction) where ∼606 km 3 of sediment is deposited within the Lake Malawi Rift, reproducing 96% of the observed siliciclastic sediment volume.
To study within-rift sediment flux and accumulation variability, it is useful to compare the sediment accumulations, and smaller sedimentary features basin by basin over the past 140 kyr between the model and observations from transects of high-resolution single channel reflection seismic data (Figure 7, Table 2).We choose endmember locations on our 2D data that are representative of different depositional conditions and tectonic settings and compare these spot locations in the model.Overall, we note an excellent comparison between modeled and observed sediment thickness across all basins.In the North Basin (Profile A, Figure 7), we note the presence of a present-day deep-water fault-controlled axial channel on this profile, and a suite of buried axial channels close to the border fault.This fault control of channel location is also observed within our reference model.In the north of the Central Basin (Profile B Figure 7), we also note the presence of buried channels on the relay ramp and border fault fans in the SCS data, both of which are reproduced in the reference model (Figures 6 and 7).In the South Basin (Profile D, Figure 7), within the SCS data, we observe some high amplitude reflections in the hangingwall adjacent to the border fault, and no deeply incised channels are imaged.This is likely due to a combination of data resolution, local slope, and the resulting facies.The average gradient in the South Basin (0.1%) is much shallower than in the North and Central Basins (0.6% and 0.9% respectively), and thus rather than one or two large, incised channels as seen in the North and Central Basins (Profile A and B, Figure 7), we may expect many smaller, less incised channels that are at or near the resolution of our data.Furthermore, the observed acoustic response likely indicates a sandier facies that may represent the axial braided river system (i.e., Allen, 1983) that is observed in the model in the South Basin (Profile D, Figure 7).

Sensitivity Analysis
We conducted a sensitivity analysis (Table 3 and Text S4 in Supporting Information S1) to investigate the sensitivity of total sediment volumes and catchment-average sediment flux to geologic processes.Overall, the analysis suggests that erodibility has the most significant impact on total sediment volume, followed by precipitation and tectonic forcings (Table 3).This is to be expected given the known variability in these parameters both spatially (erodibility, precipitation, tectonics) and temporally in the case of precipitation across the Lake Malawi Rift, which is not captured in a constant value.In terms of erodibility, while more groups can certainly be defined, we suggest that three broad groups are appropriate given the regional nature of this study, model grid size, and scale of current geologic mapping.As explicit calibration points exist for the precipitation (e.g., Huffman et al., 2019;Lyons, Kroll, & Scholz, 2011) and tectonic (e.g., Williams et al., 2022a) forcings, we attempt to accurately capture this variability within the model.Generally, the sediment flux is as sensitive to precipitation inputs at the catchment scale, as it is at the rift regional scale.

Modeled Sediment Flux Over the Past 140 kyr
From the reference model, we calculate the thickness of lacustrine sediment deposited over each of the three key climate phases identified over the past 140 kyr (Figure 8) and thus also the average sedimentation rates (Figure 8).During the arid lowstand climate phase (140-95 kyr BP) where a shallow restricted lake is observed in the Central Basin and river systems that presently drain into the North Basin combine into one mega-system (Section 3.1), we observe that a maximum of 80 m of sediment is deposited, focused within the Central Basin of the rift, within the extents of the shallow lake (Figure 8), and adjacent to the main border fault.We also observe ∼50 m of sediment deposited as a fan at the mouth of the S. Rukuru river as it cuts down through the relay ramp.During this time, limited fluvial-shallow lacustrine deposition (∼15 m) is observed in the south of the North Basin, as this environment fluctuates between subaerial and subaqueous conditions, as evidenced by the very small, isolated ponds in Figure 8, adjacent to the axial river system.During this arid period there is very little sedimentation recorded in the South Basin (<5 m).
From 95 to 50 kyr BP (stillstand climate period Figure 8), local sedimentation rates decrease as the lake expands and sediments are dispersed across the basin.On average, 20 m of sediment is deposited in the North Basin, and 30 m is deposited across most of the Central Basin, except at the S. Rukuru fan where ∼45 m is deposited.In the South Basin, we observed sediment accumulation between 10 and 30 m of sediment accumulation in localized depocenters, mostly adjacent to the Bua and Dwangwa rivers (Figure 8).
As the climate becomes wetter and lake levels stabilize at high stand conditions (50 kyr to present), we observe an increase in the sedimentary thickness across all basins of the rift within the reference model (Figure 8).In the North Basin, a maximum of 30-40 m of sediment is deposited adjacent to the border fault, decreasing toward the flexural margin.In the Central Basin ∼40 m of sediment is deposited, and in the South Basin 20-40 m of sediment is deposited adjacent to the border fault, thinning toward the flexural margin.It should also be noted that during this time period in the model we observe sediments filling the eroded channels from the Songwe, Ruhuhu, S. Rukuru, Bua, and Dwangwa rivers formed by incision during the lowstand phase (Figure 8).During this highstand climate phase, we also notice that a non-trivial amount of sediment is stored outside the rift proper in the S. Rukuru (∼20 m) and Ruhuhu (∼15 m) catchments.

Sedimentation Rates From Seismic Reflection Data Over the Past 140 kyr
We calculate average sedimentation rates in the Central Basin of the rift from a combination of drill core data and pseudowells created from high-resolution SCS data for the three key climate phases identified over the past 140 kyr (Figure 9, Table 4).From the 2005 Lake Malawi Drilling Project core hole in the Central Basin (GLAD05-MAL07-1, Lyons, Scholz, et al., 2011;Scholz et al., 2011aScholz et al., , 2011b) ) tied to the seismic using the age model and synthetic seismogram of Lyons et al. (2015), we identify the tops and reflections corresponding to the lowstand sequence boundary (140 kyr), end of the lowstand lake (95 kyr), and the start of highstand conditions (50 kyr) (Figure 9).These are then mapped from the core location along strike and into the deep basin to two pseudo-well locations that are presumed to represent progressively deeper paleo-water-depths (Figure 9).A total vertical thickness is calculated from the tops at the drill core site and from the seismic reflection data at the pseudo-wells, assuming a constant seismic velocity of 1,480 ms −1 .The thickness is divided by the period over which it accumulated to give the average sedimentation rates for the time interval.
During the lowstand phase (140-95 kyr BP), sedimentation rates vary from 0.44 mm/a at the core site to 1.13 and 1.67 mm/a at pseudowells one and two respectively.During the stillstand phase (95-50 kyr BP), we observed sedimentation rates at the core site of 0.72 mm/a and a gradual increase to 0.74 and 0.82 mm/a at the pseudowells (Figure 9).During the highstand phase (50 kyr to present) we recorded sedimentation rates of 0.74 mm/a at the core site, and rates of 0.82, and 0.89 mm/a at the pseudowells.

Catchment-Average Erosion Rates
Catchment-average erosion rates, calculated from the total volume of sediment removed from each catchment during each of the key climate stages and for the entire model run, are presented in Figure 10 and Table 5, and an interpretation of these results is presented in Section 4.4.From 135 to 95 kyr BP during the lowstand climate phase, the northern axial mega-catchment has an erosion rate of 0.041 mm/a, while the N. Rumphi catchment has the highest erosion rate of 0.077 mm/a.Across the South Basin of the rift, the large axial mega-catchment has an average erosion rate of 0.021 mm/a.Detailed erosion rate information for each of the major rift catchments is presented in Table 5.
From 95 to 50 kyr BP during the gradual lake-level rise (i.e., the stillstand climate phase), catchment-average erosion rates increase southward from 0.035 mm/a in the Songwe River to 0.050 mm/a in the N. Rukuru river.In the Central Basin the N. Rumphi catchment shows the highest erosion rate across the rift of 0.076 mm/a.In the South Basin, catchment-average erosion rates vary from 0.011 to 0.017 mm/a (Figure 10).
From 50 kyr to present (i.e., highstand climate phase), catchment-average erosion rates increase southward across the North Basin from 0.034 to 0.05 mm/a from the Songwe to N. Rukuru catchments (Figure 10).The Ruhuhu shows an erosion rate of 0.0375 mm/a during this period.In the Central Basin, the N. Rumphi and S. Rukuru have catchment-average erosion rates of 0.079 and 0.027 mm/a respectively.Across the South Basin the Bua, Dwangwa, and Linthipe have catchment-average erosion rates of ∼0.018 to 0.020 mm/a (Figure 10).

Implications for Juvenile Rift Systems
The Malawi Rift provides exceptional constraints on paleoclimate and basin morphology over the past 140 kyr, and offers a unique insight into a juvenile terrestrial rift environment before the connection to the global ocean system (e.g., Gulf of Corinth, Gulf of California, passive rifted margins).Our new modeling results tied to observational data illustrate the significant impact of spatio-temporal variations in regional climate at orbital timescales (i.e.,     (Ben-Avraham et al., 2008;Lu et al., 2020) systems.Therefore, this work builds on the accepted tectonic model for facies development in lacustrine rift basins (e.g., Lambiase, 1990;Lambiase & Bosworth, 1995), adding the effects of high-frequency climate forcings to early syn-rift fills.The observed patterns in shoreline location, controlled by climate-induced base-level falls, have implications for the distribution and quality of potential reservoir, source, and seal rocks within the early syn-rift phase of rifted margins.First, the (cyclical) exposure of organic rich sediments to sub-aerial conditions during arid intervals will likely degrade source and seal rock quality (Kelts, 1988;Talbot, 1988).This is because sediment build up affects the burial history, preservation and thermal history of deep rifted margin sediments and organic matter.Second, it is likely that both sediment source and sink areas change as both axial and flexural margin catchments reorganize into the observed mega-systems and refocus sediment in response to the base-level fall.For example, in Lake Malawi, the axial river systems flowing through the South and North Basins during lowstands provide a fairway for focusing on sands.This has implications for the location, presence, and effectiveness of reservoir facies within juvenile rifts.
The fluctuations observed in basin physiography, and sediment focusing have important implications for how tectonics and climate interact to control the syn-rift stratigraphy in ancient rift systems worldwide (Balázs et al., 2017;McNeill et al., 2019;Thompson et al., 2021).Surface processes can impact lithospheric extensional processes (Olive et al., 2014(Olive et al., , 2022;;Wolf et al., 2022), and in particular the redistribution of mass into the deep basin may promote the greater accommodation of extension on rift faults before they are abandoned (e.g., Olive et al., 2014).Thicker sediments resulting from cyclical sediment focusing may also reduce buoyancy forces from lithospheric thinning (Bialas et al., 2010;Buck, 1991Buck, , 2004;;McKenzie, 1978), and thus promote/ entrench narrow rifting modes (Bialas & Buck, 2009;Buck, 1991), prolonging the onset of continental rupture (e.g., Martín-Barajas et al., 2013).These inferences are largely based on models or larger rift structures lacking temporal constraints.Our high-resolution spatio-temporal constraints on erosion and sediment flux within the rift suggest that spatio-temporal changes in basin physiography onshore and deposition offshore could have the potential to modulate fault evolution in early rift systems.Further application of this reference model to the full suite of geophysical and core data from the Malawi Rift will allow us to test the feedback between high-frequency climate forcings and tectonic evolution of the rift valley.
The surface processes and changes in basin physiography we observed within our reference model may also impact the spatio-temporal evolution of the sedimentary fill and pore-pressure.Within the Central Basin, a suite of mud diapirs are observed in the seismic reflection data (Scholz et al., 2020), although their origin is enigmatic.Cyclical sediment focusing and changes in sedimentation rate within the restricted central basin depocenter during arid intervals may promote the necessary conditions for mud diapir formation through differential compaction and sediment loading inducing sediment overpressure (Huuse et al., 2010;Jackson & Vendeville, 1994;Kopf, 2002;Van Rensbergen et al., 2003).

Landscape Evolution as a Function of High-Frequency Climate Change
Previous observations of lowstand channel complexes and delta deposits have been confined to localized high-resolution seismic grids, and thus spatially limited our interpretations of basin-wide paleoenvironments and basin morphology (Lyons, Scholz, et al., 2011;Scholz, 1995;Soreghan et al., 1999).Our results show that high-frequency climate change as observed across the Lake Malawi Rift over the past 140 kyr (e.g., Cohen et al., 2007;Lyons, Kroll, & Scholz, 2011;Lyons, Scholz, et al., 2011;Lyons et al., 2015;Scholz et al., 2007) has significant impacts on basin morphology and catchment reorganization (Figure 6) not captured in the previous source to sink studies (e.g., Tan & Scholz, 2021;Zhang et al., 2020).
Arid intervals (e.g., 140-95 kyr BP) cause a significant reduction in the levels of Lake Malawi, causing the near-complete desiccation of the North and South Basins and catchment amalgamation in these areas that then focus sediment into the remaining shallow lake (Figures 6 and 9).In the North Basin, and northern parts of the Central Basin where subsidence rates are highest, buried channel complexes have been observed in high-resolution SCS reflection data (e.g., Lyons, Scholz, et al., 2011;Soreghan et al., 1999) and these are likely the preserved expression of the northern axial mega-system in the model.In the South Basin of the rift there are no observations of the modeled axial mega-system to date and this may be a consequence of a variety of factors: (a) High-resolution SCS data is relatively lacking in this part of the rift, and the channel complexes are likely below the resolution of the legacy regional MCS data in the region (Scholz et al., 2020;Specht & Rosendahl, 1989).(b) Lower rates of subsidence and lower slopes in this basin, its shallower nature compared to its northern counterparts, its complete desiccation during the lowstand and therefore the significantly lower rates of sedimentation here make preserving this axial system challenging.Given that our model uses a simplified fault map for tectonic displacements, especially in the South Basin, it is also likely that this system may have been more limited than our models predict, perhaps with several smaller axial systems.Nevertheless, further high-resolution SCS data are needed to confirm these hypotheses in the South Basin.
Increasing precipitation (e.g., 95-50 kyr BP) causes a large increase in lake extents (Figure 6), and these axial systems largely disappear as catchments reorganize into their current configurations (e.g., Tan & Scholz, 2021).During this period, as lake level increases, significant stillstand delta formation is observed across all basins of the rift (e.g., Lyons, Scholz, et al., 2011;Scholz, 1995), with a deep lake occupying the North and Central Basins, and a relatively shallower one in the South Basin.As the precipitation continues to increase (50 kyr to present), lake levels stabilize at their highstand extents, and catchments that formed the axial mega-systems continue to lose area as the lake expands to current conditions.

Influences on Sediment Flux Into the Rift
Over the past 140 kyr, we observe a distinct and consistent spatiotemporal pattern in the evolution of sediment flux across both our observed data and the model.During the lowstand climate period (140-95 kyr BP), the high sedimentation rates (∼1.7 mm/a) in the relatively small restricted shallow lake are twice as high as those during the stillstand and highstand climate periods (∼0.85 mm/a).Then, during the stillstand period (95-50 kyr BP), we observed a decrease in sedimentation rates, followed by an increase during the past 50 kyr at the pseudowells (Figure 9).Rates at the two pseudowells that are located at progressively deeper paleowater depths than the core hole show marked increases in sedimentation rates, especially during the lowstand climate phase, highlighting the role of sediment focusing into the shallow lake during this period (Figures 9 and 10).These sedimentation rates are consistent with sedimentation rates observed in other WB EARS basins (e.g., Tanganyika, Albert) typically 0.3-2 mm/a (Cohen, 1989;Cohen et al., 1993;Simon et al., 2017), albeit lower than rates from Lake Turkana in the EB, typically 3-5 mm/a (Johnson et al., 1987).Moreover, these values are also broadly consistent with sedimentation rates (typically 0.1-3 mm/a) for other lacustrine sections including in the Sea of Marmara (Turkey, McHugh et al., 2008), Lake Baikal (Russia, Edgington et al., 1991), Salton Trough (California, Dorsey, 2010), and the Gulf of Corinth (Greece, McNeill et al., 2019).
The rate and loci of sediment accumulation within the Lake Malawi Rift, and rift systems in general are affected by the spatio-temporal interplay of tectonic, lithologic, and climate factors, and the unique closed-loop modeling approach taken here allows us to evaluate their interplay.Although over the history of rifting (c.a., 8.6 ma, Ebinger et al., 1993), tectonic uplift and subsidence clearly control the larger scale development of the Lake Malawi Rift (Figure 1b), and tectonic rates would not be expected to fluctuate cyclically on the timescales of 50,000 years as sedimentation rates fluctuate here.Therefore, we propose that the spatiotemporal patterns in basin sedimentation are a function of one or more of the following: (a) Climate (i.e., temperature and precipitation) controlling the magnitude of base-level fall, erosion rates, and sediment flux, (b) Vegetation type and cover changing through space and time, therefore changing sediment retention/erosion rates within the catchments surrounding the rift (e.g., Thompson et al., 2021).
Our models show that varying the precipitation rate (i.e., climate change) over the past 140 kyr strongly influenced basin morphology (Figures 6 and 9, and Figure S7 in Supporting Information S1).The arid climate (140-95 kyr BP) results in a marked base-level fall, leading to a major downstream drainage adjustment and focusing of sediment into a shallow lake, increasing local sedimentation rates within the lake, and eroding the fluvial/deltaic material exposed on the paleo-lake floor.This material is more easily erodible compared to the surrounding terrains (Figure 5) and may also be more prone to slump failure.Increases in precipitation (95-50 kyr BP) caused a rapid expansion of the restricted lake to 75% of its modern dimensions (Figure 6), and this expansion in lake area caused sedimentation to spread over a wide area.However, the still relatively low precipitation (53%-58% of modern) caused an overall net decrease in sedimentation rates (Figure 8).As the lake level stabilized under highstand conditions (50 kyr to present), there was limited expansion of lake extent and precipitation continued to increase to modern values, thus producing an overall increase in sedimentation rates (Figure 8).These results are consistent with lake core records (Ivory et al., 2016) and measurements of alluvial fan expansion from OSL data (Thompson et al., 2021) that suggest higher sedimentation rates accompanied by increased terrigenous input at ∼130 and 85 kyr.

High Catchment-Average Erosion Rates During Severe Arid Periods
Catchment-average erosion rates, as with sediment flux, are sensitive to changes in climate (i.e., precipitation) and tectonics, both of which drive changes in the relief, slope, and area of catchments (Equations 1 and 2, Adams et al., 2020;Forte et al., 2016;Olive et al., 2014;Syvitski & Milliman, 2007;Whipple & Meade, 2006;Wolf et al., 2022).Our closed-modeling approach makes it possible to evaluate the controls on catchment-average erosion rate across the region; however, to avoid over-interpreting the modeling results, we restrict our explanations to regional patterns (Figure 10), addressing the questions: (a) Why are catchment-average erosion rates lower across the flexural margin of the South Basin?(b) Why are erosion rates highest in the N. Rumphi and N. Rukuru catchments?(c) Why are erosion rates higher during arid intervals?
Modern precipitation rate varies spatially across the Lake Malawi rift (Figure 3), whereby the highest precipitation rates are in the North Basin (∼1.6 m/a) and the lowest rates are across the flexural margin catchments in the South Basin (∼0.6 m/a); therefore, less precipitation is available to work against the landscape across the flexural margin catchments and this shift is enhanced only with the onset of arid conditions.Furthermore, these catchments have lower average slopes, ∼500 m less relief, and ∼2-3 times more area (∼10,000 km 2 ) than catchments farther north (Crossley, 1984;Scholz, 1995;Tan & Scholz, 2021).We therefore suggest that it is the interplay of precipitation availability and catchment morphology that controls the erosion rate, leading to average rates of erosion in the flexural margin catchments during both arid and wet climate intervals compared to those farther north (Figure 10).
Farther north, the N. Rumphi and N. Rukuru catchments occupy a wetter environment, receiving ∼1 m/a of present-day precipitation (Figure 3), and therefore more precipitation is available to work against the landscape compared to southern catchments.However, precipitation is not the only driver of the elevated erosion rates, as the greatest precipitation rates occur farther north across the Songwe catchment, which has lower average erosion rates than the N. Rumphi and N. Rukuru.Both these catchments instead drain the Nyika plateau, a 3,700 km 2 highly elevated (∼2,500 m) region which formed as a response to Karoo rifting during the Late-Paleozoic (Dixey, 1937a(Dixey, , 1937c;;McMillan et al., 2022).This elevated region of topography (Figures 1 and 6) results in greater relief, higher average slopes and smaller catchment area (∼800-2,000 km 2 ) compared to the other catchments across the rift.We therefore suggest that it is this smaller catchment area, combined with the high relief and high slopes that drive the higher catchment-average erosion rate values here.
The onset of arid conditions (135-95 kyr BP) causes a rapid base-level fall.The 550 m drop in base-level increases maximum catchment relief (Figure 6), while simultaneously exposing softer, easily eroded lithologies previously covered by the expanse of the lake (Figure 5).Catchment amalgamation into the mega-catchments across the North and South Basins causes a three-to-six-fold increase in catchment area (Figure S5 in Supporting Information S1), and both the increase in catchment relief and area have the potential to increase erosion rates (Adams et al., 2020;Syvitski & Milliman, 2007), despite the 65% reduction in precipitation.As the climate becomes wetter (95 kyr to present), the contribution of increased precipitation is generally offset by the decreased relief and catchment area owing to the rapid lake transgression (Figures 6 and 9), thus decreasing erosion (i.e., sediment production, Zhang et al., 2020;Tan & Scholz, 2021).We hypothesize that temporal changes in catchment-average erosion rates in the Lake Malawi Rift are driven by the interplay of climate induced-base-level fall.Within our models, we note that trends in catchment-average erosion rates appear coupled to the climatically induced changes in sediment flux.This coupling may be a result of the transient nature of the landscape and relative sensitivity of catchments to the rapid changes in base-level caused by climate change-induced lake-level drop.We also note that our model does not account for the vegetation change through time or the action of fire across the landscape.Paleoenvironmental analysis before and after 85 kyr (Thompson et al., 2021) from pollen, charcoal, and archeological sites suggests (a) the expansion of fire tolerant vegetation over the past 92 kyr and (b) burning of uplands caused alluvial fan expansion, both of which will affect and buffer the rate of sediment production, transport and deposition.

Conclusions
Through a closed-loop numerical forward modeling approach, we show that rapid climate change in the Lake Malawi Rift over the past 140 kyr causes extreme basin physiographic changes and drainage reorganization.The onset of arid climate conditions (140-95 kyr BP) and the associated drop in lake level cause the formation of mega-catchments in the North and South basins of the Lake Malawi Rift that flow axially into a shallow restricted paleo-lake in the Central Basin.Increasing precipitation availability (95-50 kyr BP) causes a large increase in lake extents (Figure 6), and these axial systems largely disappear as catchments reorganize into their modern configurations.Although initial basin physiography was tectonically created, variations in the resulting sediment flux and erosion are controlled by the high-frequency climate perturbations inducing rapid variations in base-level over the past 140 kyr.This results in a twofold increase in sedimentation rates during arid intervals compared to modern times, whereby sediment is focused into a shallow and restricted paleo-lake in the Central Basin.The observed high-frequency fluctuations in basin physiography and sediment focusing within our models provide unique insights into how tectonics and climate interact to control the syn-rift stratigraphy in ancient rift systems.Moreover, the framework of the modeling approach outlined in this paper has the potential to be applied to less well-constrained systems to quantify tectono-climatic interactions.The redistribution of mass into deep basins during arid intervals has the potential to promote both narrow rifting modes and prolong fault lifespan before abandonment, and although not the focus of this contribution, it will be the focus of future research efforts.
Our sensitivity analysis shows that the model is most sensitive to appropriate bedrock erodibility values, and the application of this closed loop modeling approach to the full suite of geophysical and core data from the Malawi Rift will allow us to further calibrate these values and define more lithologic groups within the model.Additionally, it will allow the investigation of feedback between high-frequency climate forcings and tectonic evolution of juvenile rift systems.Precipitation data used in this study were accessed with the Giovanni online data system developed and maintained by the NASA GES DISC.Seismic processing and analysis were carried out using Landmark Graphics Corporation's SeisSpace, DecisionSpace, and Syntool software, provided on a software grant to CAS.LJMW is supported by a Syracuse University Research Excellence Graduate School Fellowship and from research grants from Chevron and Petrobras to CAS.CAS is supported by a grant from the U.S. National Science Foundation (NSF-EAR-2116017).We thank P. Cattaneo for data processing assistance.We are particularly grateful for support and assistance from the Malawi Geological Survey and the Geological Survey of Tanzania and for research permissions from the governments of Malawi and Tanzania.LJMW is currently employed by Chevron U.S.A.However, this study and the initial manuscript drafts were developed and completed during his PhD research at Syracuse University.The authors extend their sincere thanks to Amy East, Editor-in-Chief for editorial handling, as well as reviews by Luca Malatesta, Drake Singleton, and an anonymous reviewer that greatly improved the manuscript.

Figure 1 .
Figure 1.Overview of the Lake Malawi Rift.(a) Topographic/Bathymetric map of the Lake Malawi (Nyasa) Rift used as the input landscape in this study.Present day watershed outlined in black which defines the area of interest, with major rivers annotated and shown in blue.Topography derived from 90 m SRTM Digital Elevation Model, Bathymetry derived from Scholz and Lyons (2022b) See inset map for regional location.WB, Western Branch; EB, Eastern Branch; MER; Main Ethiopian Rift.(b) Time-Structure map of the Lake Malawi (Nyasa) Rift showing TWTT to the base of the syn-rift section (modified fromScholz et al., 2020), overlain on the regional geology the rift transects(Fritz et al., 2013).Black lines represent fault heave polygons from basement-involved faults that offset the basement horizon.Regional shear zones in red.Note the spot measurements of elevation in panel (a).

Figure 2 .
Figure 2. Schematic cartoon of the key elements controlling landscape evolution and sediment flux to a basin.P, U, E, D, LL refer to Precipitation, Uplift/subsidence (from fault movement and regional), landscape Erosion, offshore Deposition, and Lake Level change, all of which may vary through time and space.General rates and ranges of these processes are shown for the Lake Malawi Rift.In a sedimentologically closed system such as the Lake Malawi Rift, the volume of sediment eroded from the landscape E, is equal to the volume of deposited sediment, D, as there is no sediment outflow from the system.

Figure 3 .
Figure 3. (a) Simplified lake-level curve over the past 140 kyr modified fromLyons, Kroll, and Scholz (2011) showing the most recent lake-level cycle, and the maps of precipitation (b-f) with mean values required to sustain those lake levels fromLyons, Kroll, and Scholz (2011).Panel (b) is derived from modern satellite data, see text for methods.Panels (c-f) are derived by rationing (b) by the percentage required to sustain the lake level from the hydrologic model ofLyons, Scholz,  et al. (2011).Mean values of the resulting map are inset, as well as the percentage of modern precipitation.The present lake and catchment outlines are shown as white solid and dashed lines respectively.Paleo-lake extents for each timestep are shown as black solid lines in panels (c-f).

Figure 4 .
Figure 4. Inputs and resulting in a vertical displacement map of the Lake Malawi Rift over the past 140 kyr.(a) Maximum vertical offsets on the basement-involved normal faults over the late-Quaternary used as inputs into coulomb 3.4 (Toda, 2005).Gray circles represent the location of maximum offset on a fault from Wright et al. (2023) in the Central and South Basins, and Shillington et al. (2020) in the North Basin.Border fault displacements in the Central and North Basins are from the MSSD (Williams et al., 2022a).(b) Total vertical displacement over the past 140 kyr input into pyBadlands, where blues are subsiding areas and reds are uplifting areas.Maximum subsidence is ∼200 m, and maximum uplift is ∼65 m.Model faults are shown by red lines, and the contour interval is 25 m.The present day lake outline is shown in gray, and the catchment boundary in black dashes.

Figure 5 .
Figure5.Box and whisker distribution of regional geology (Figure1b) according to elevation (Figure1a), and derived bedrock erodibility map (Kb) for the Lake Malawi Rift used in the reference model run based on the elevation groupings.Catchment outline shown in dashes, and present-day extent of Lake Malawi shown in solid lines.
e., from 0 to 200 m it represents the 200 m stillstand).In water depths deeper than 350 m, this surface represents the 550 m low stand sequence boundary.Finally, between 200-350 m and 0-200 m it is interpreted to represent the 350 and 200 m stillstands.

Figure 6 .
Figure 6.Modeled basin topography and drainage evolution at key time periods over the past 140 kyr.Topography is shown as meters above lake level.Drainage networks are shown in shades of blue representing modeled downstream discharge; model precipitation is shown in the bottom left of each panel relative to modern day precipitation values.Faults are shown in red lines.Key rivers and their catchments labeled in the right panel: 1. Songwe, 2. Lufira, 3. N. Rukuru, 4. N. Rumphi, 5. S. Rukuru, 6. Bua, 7. Dwangwa, 8. Linthipe, 9. Ruhuhu.NB = North Basin, CB = Central Basin, SB = South Basin.Note the desiccation of the North and South Basins, the shallow (50 m deep) lake in the Central Basin, and drainage reorganization in desiccated areas with the formation of two major axial systems flowing into this shallow lake during the Lowstand Climate period.Catchments that form part of these major axial systems are colored through time, whereby purple represents the south axial system, orange represents the north axial system and no color fill indicates no presence in an axial system.

Figure 7 .
Figure 7.Comparison between high-resolution single-channel reflection seismic data (a-e) and the modeled sedimentary package over the past 140 kyr in the model (f).Red line in panels (a-e) shows the interpreted 140 kyr lowstand surface, the lake floor shown as blue line.Faults in black.Values at arrows show the true vertical thickness of the sediment accumulated over 140 kyr.(f) Total modeled erosion (blues) and deposition (reds) in the Lake Malawi Rift over the 140 kyr model run.Lake shore in black, key rivers annotated.See text for comparison of sediment volumes.Co-location of spot measurements indicated by dots.

Figure 8 .
Figure 8. Overview of modeled rift basin response to rapid climate change in the Lake Malawi Rift from the calibrated reference model.Lake outline shown in Black, colored surface represents total erosion and deposition (blues and reds) over the three key time periods.Overlain is the river network in blues.Note how arid conditions cause the basinward migration of sediment source areas for all key rivers.

Figure 9 .
Figure 9. Observed sediment accumulations between key stages of interest in the Central Basin.(a) Axial high-resolution single channel seismic reflection profile showing the horizon interpretation of the three key climatic stages and the location of GLAD7-MAL05-1 and pseudo well 1.Insets show graphs of calculated sedimentation versus age for the three stages, as well as the location of the seismic lines and wells colored by their relative paleo-water depth.(b) Dip high-resolution single channel profile showing the interpreted key stages of interest, pseudo well 2 and the calculated sedimentation rates.Note the expansion of the lowstand sequence in the pseudo well compared to (a), highlighting the role of sediment focusing.LFZ = Lipichilli Fault Zone, the intrabasinal high separating the well and pseudowells.

Figure 10 .
Figure 10.Modeled catchment-average erosion rates over the past 140 kyr for key catchments along the length of the Lake Malawi Rift from the calibrated reference model divided by the time periods of interest as well as the average over the 140 kyr model run.
10 to 100 s of kyr) on rift basin physiography and thus sediment sourcing and supply within early rift systems.We would expect to see similar patterns occurring in other early rift systems before they are fully connected to the open ocean, for example, the proto-Atlantic, Gulf ofCorinth (McNeill et al., 2019), or Dead Sea Transform

Table 1
Summary Statistics of the Fluvial Inputs Into the Lake Malawi Rift

Table 2
Comparison of Modeled to Observed Sediment Thicknesses From Seismic Reflection Data

Table 3
Sensitivity of Modeled Sediment Volumes to Various Input Parameters

Table 4
Age, Depth, and Thickness of Key Horizons and Intervals From Well and Seismic in the Central Basin

Table 5
Modeled Catchment-Average Erosion Rates Across the Lake Malawi Rift During the Past 140 kyr