The Luangwa Rift Active Fault Database (LRAFD) contains 18 active faults (Table ; Fig. ). In Fig.  we use the Chipola Fault to illustrate how we compiled evidence of activity in the LRAFD. The 207 km long Chipola Fault is the longest fault in the Luangwa Rift and forms the western border fault of the rift basin. The Chipola escarpment and fault scarp are clear even at coarse (1:400000) scale using the DEM, slope, and hillshade maps (Fig. a–c). The northern end of the fault follows the Karoo–Basement contact (Fig. c), and a scarp that offsets Quaternary sediments correlates with the mapped faults in the geological maps. The steep scarp (26∘) occurs at the base of an abrupt elevation change (Fig. d) and has displaced Quaternary alluvial fan sediments (Fig. f). Incised rivers in the faults footwall also suggest recent uplift (Fig. e). Given this evidence, the Chipola Fault was assigned 1 for activity confidence, epistemic quality, and exposure quality.

The criteria and indicators listed in Table  and described for the Chipola Fault above were applied throughout the Luangwa Rift to create the LRAFD. Faults with a similar strength of evidence to that of the Chipola Fault are also mapped with the highest confidence (e.g. the Molaza and Kabungo faults; Figs.  and ). Scarps and escarpments are prominent on the DEM and Google Earth (e.g. Figs. a and c), and slope maps highlight steep (>20∘) fault scarps that have formed at the base of many of the escarpments (Figs. b, f, , b, and d). Figures – show slope maps of the Chipola, Mkumpa, and Molaza faults, highlighting the steep scarps, with the most prominent (steepest) scarp observed along the Molaza Fault (Fig. b). Within the Luangwa Rift, these faults show the clearest evidence of offset Quaternary sediments and sedimentary features such as alluvial fans (e.g. Fig. d–e), indicating recent fault activity and rupture events. We also observed river incision and channel steepening in the footwall of the Mukopa, Chitumbi, Kapampa, Chipola, and Molaza faults (e.g. Fig. c).

Overall, 10 faults had exposure quality scores of 1, indicating they are well exposed, whereas 8 faults scored 2, meaning they lacked a strong exposure. Epistemic quality presented 13 traces as high certainty of activity and 5 as low. Activity confidence was assigned after taking a holistic view of each trace and its likelihood of recent activity. There are six faults with the strongest confidence value (1; Chitumbi, Chipola south, Chipola, Chitembo, Kabungo, and Molaza), and three faults are assigned 4 (Chipola west, Luwi, and Mwanya), thought to have a low likelihood of activity.

We found active faults along the length of the 600 km rift, with fault lengths varying between 9 and 207 km. The faults generally trend NE–SW, with some minor faults trending N–S (Fig. ). Within the rift, the two longest faults, the Chipola Fault (207 km; Fig. ) and the Molaza Fault (142 km; Fig. ), both display evidence of well-preserved fault scarps and have the highest activity confidence (1) and exposure quality (1; Table ). These faults are located at the edge of the rift and represent the western SE-dipping (Chipola Fault) and eastern NW-dipping (Molaza Fault) border faults of the rift (Fig. ). These faults form sub-basins with the Luangwa Rift that differ in their properties. In the northern sub-basin, the Molaza Fault (Fig. ) has only one other active fault within its hanging wall. At the southern end of the Molaza Fault there are two short (9 and 16 km long) faults in a stepover geometry (LRAFD_ID: 14 and 15; Fig. a), but the rest of the fault displays a relatively simple geometry with no evidence of splays. In the southern sub-basin, where the Chipola Fault is the border fault, there are up to three faults across strike (LRAFD_ID: 8–13; Fig. a) including two intra-rift faults in the centre of the rift (LRAFD_ID 9 and 10; Fig. a). In the hanging wall of the northern end of the Chipola Fault, there are two faults (LRAFD_ID: 11 and 12; Fig. a) in a stepover geometry that are directly across strike from the stepover faults at the southern end of the Molaza Fault. This zone of distributed faulting between the northern and southern sub-basins of the Luangwa Rift is a common observation in other rift transfer sections in the EARS .

Using the fault scaling laws set out in , we derived earthquake source parameters including average fault displacement (Dav), down-dip fault rupture width (W), fault rupture depth extent (RDE), and moment magnitude (Mw; Table ). Dav varies between 0.3+0.7/-0.2 m on the Molaza-2 Fault and 4.2+9.5/-2.6 m on the Chipola Fault. Predicted W values range from 8+3/-1 to 61+27/-11 km with RDEs of 6+3/-2 to 49+28/-14 km (Table ). With these calculations, which assume a reasonable fault dip of 53∘, only the Chipola Fault produces a rupture depth extent that would exceed the crustal thickness of the region ∼45 km;, and only the Chipola and Molaza faults produce a rupture depth that exceeds the maximum depth of seismicity recorded in the region ∼ 30 km;. Potential earthquake magnitudes for whole-fault ruptures average Mw 7.0 but vary between Mw 5.8 and 8.1 (Table ; Fig. ).

Using Eqs. (5) and (6) with a logic tree approach Fig. ; adapted from, we calculated lower, intermediate, and upper earthquake recurrence intervals (R) for the Chipola and Molaza faults (individual logic trees are shown in Supplement Figs. S1 and S2). We applied the rift extension rate (V) and azimuth (ϕ) values of 0.7 ±  0.2 mm yr-1 and 108∘ from . Slip rates for the Chipola and Molaza faults are estimated at 0.4–1.7 and 0.4–1.6 mm yr-1, respectively. For whole fault ruptures along the Chipola Fault, our intermediate estimate for earthquake recurrence interval is 5000 years, with a lower estimate of 1500 years and upper estimate of 36 000 years. For the Molaza Fault the intermediate recurrence interval is estimated at 4000 years, with a lower bound of 700 years and upper bound of 28 500 years. The large uncertainties associated with these estimates represent the large epistemic uncertainties inherent when propagating uncertainties from empirical scaling relationships to fault recurrence estimates in the systems-based approach .

Example topographic profiles for the four faults with the highest confidence of activity (Chipola, Chitembo, Kabungo, and Molaza) are shown in Fig. , with the corresponding along-strike scarp height profiles in Fig. . The median scarp height of each fault ranged between 12.9 ± 0.4 (Molaza Fault) and 19.2 ± 0.9 m (Kabungo Fault; Figs.  and c). The minimum resolvable scarp heights that we are able to measure using the SRTM data are 2–3 m. However, the lower-resolution of SRTM compared with TanDEM-X; see meant that we were unable to identify clear fault segment boundaries. In the statistics quoted below, we have removed outliers from the data, which are values >2σ from the median scarp height.

The SE-dipping Chipola Fault borders the western side of the rift, with a 220 km long fault trace. The median height of the fault scarp is calculated to be 13.0 ± 0.4 m. The highest scarps are found towards the southwestern end of the fault, where the maximum scarp height was observed (28.4 ± 0.4 m). The histogram shows that most of the scarp height measurements are between 5 and 20 m (Fig. a), with the largest peak in the histogram occurring between 14 and 16 m. Smaller peaks are also present at 19 and 10 m.

The Chitembo Fault trace is 48 km long and dips SE, and it is located 20 km east of Chipola's northern tip. The median scarp height is 14.0 ± 1.0 m, with a maximum height of 39.6 ± 0.2 m (Fig. b). The histogram of scarp heights peaks between 8–10 m, although there is a long tail to the distribution, with other minor peaks observed at 18–20 and 28–30 m.

The SE-dipping 45 km long Kabungo Fault, which is 15 km east of the Chitembo Fault has the highest median scarp height of the faults that we measured: 19.2 ± 0.9 m. The maximum scarp height is found to be 36.3 ± 0.1 m. The locations where high (>30 m) scarps were observed coincided with where slopes of >25∘ were observed on the fault scarps, and where a clear change in topography was evident on Google Earth and in the DEM (Fig. ). The histogram of scarp height measurements shows two distinct peaks at 8–10 and 20–22 m (Fig. c).

The NW-dipping 142 km long Molaza Fault is the eastern border fault in the northern basin of the Luangwa Rift. The median scarp height is 12.9 ± 0.4 m, with a maximum of 30.0 ± 0.2 m found at the northern tip, where the slope map shows the most prominent scarp (Figs.  and d). Between 25 and 75 km along strike, there is a consistently preserved scarp averaging ∼ 10 m (e.g. Fig. d) that coincides with steep slope values of 19∘. The histogram of scarp height measurements displays a clear peak at ∼ 10 m with very few exceeding 25 m (Fig. d).

We found evidence for Quaternary activity on 18 faults in the Luangwa Rift Zone and developed an active fault database to systematically compile the geomorphic attributes of these faults. This builds on the previous discovery of active faulting along the Chipola South Fault (LRAFD_ID: 5) by . Within the 18 active faults in the Luangwa Rift Active Fault Database (LRAFD), 6 faults have a very high confidence of recent activity (Table ). We measured the height of the prominent scarps at the base of the footwall base of two border faults, as well as two intra-basin faults (Fig. ). Median scarp heights are between 12 and 19 m (Fig. ). Fault scaling relationships suggest that the 207 km long Chipola Fault is capable of hosting earthquakes up to Mw 8.1 with an intermediate recurrence interval estimate of 5055 years and a slip rate of 0.9 mm yr-1. The estimated potential earthquake magnitudes exceed previous recorded events in the EARS but are consistent with hazard assessments from other active fault and seismic source databases in southern and eastern Africa . The LRAFD demonstrates that the framework for the future probabilistic seismic hazard analysis in southern Africa outlined by can be successfully applied to other regions using freely available, open-access data such as SRTM. In addition, the mapped active faults provide an opportunity to analyse the seismotectonics of the Luangwa Rift and compare it to other amagmatic rifts in along the EARS, and this is the focus of this discussion.

The two main border faults in the Luangwa Rift, the Chipola and Molaza faults, both follow Karoo–Basement contacts, but our analysis shows that these faults have also offset Quaternary sedimentary deposits (Figs.  and ). These relationships suggest that these are Karoo age structures that have been reactivated during the current phase of rifting. The Quaternary fault activity adds support to the notion that the southwestern branch of the East African Rift is active and separates the Nubian plate from smaller microplates (the San, Rovuma, and possibly Angoni microplates) in southern and eastern Africa, as recently demonstrated by geodetic data .

Plate-scale modelling suggests that the extension direction across the San–Nubia plate boundary in the Luangwa Rift is 108 ± 8∘ relative to stable Nubia;. Focal mechanism inversion shows that the σ3, minimum compressive stress direction is 123∘ . These extension directions are sub-perpendicular to the fault orientation found here (mean strike: 045 ± 45∘), which follows at the kilometre scale the orientation of the Mwembeshi shear zone (Fig. ). Both geodetic and seismological data suggest a NW–SE extension direction that is orientated sub-perpendicular to the orientation of the faults in the Luangwa Rift. However, the fault orientations are still consistent with a divergent boundary as we find no geomorphic evidence of horizontal offsets, and it is not uncommon for normal faults in the EARS to reactivate at slightly oblique angles to the regional extension direction . Furthermore, the only available focal mechanism from the Luangwa Rift, from a Mb 5.7 earthquake in 1976, shows a normal faulting mechanism . Consequently, we consider that all faults in the LRAFD have pure normal kinematics but note that further work is needed to constrain the stress orientation in this region as the focal mechanism inversion of is only based on six events, and the geodetic solution of is based on a continental-scale Global Navigation Satellite System (GNSS) network with very few stations in the vicinity of the Luangwa Rift. Thus, the LRAFD demonstrates that the Luangwa Rift is an active rift system that forms the extensional boundary between the Nubia and San plates in southern Africa, with faults that have reactivated Karoo-age structures aligned with the pre-existing lithospheric-scale Mwembeshi shear zone.

Active and inactive faults are typically distinguished by the age of the most recent earthquakes . However, large-magnitude earthquakes do not always result in surface rupture, especially in regions in southern Africa where the crust can be seismogenic down to 40 km . Furthermore, we are only aware of two palaeoseismic trenches along the whole of the East African Rift that could potentially extend the record of fault ruptures beyond the small number of historically recorded tectonic earthquake surface ruptures (1910 M 7.4 Rukwa, Tanzania – ; 1928 Ms 6.9 Subukia, Kenya – ; 1966 Mw 6.8 Toro, Uganda/DRC – ; 2007 Mw 7.0 Mozambique – ; 2009 Karonga sequence, Malawi – ). Consequently, it is hard to definitively conclude that a fault is “inactive” based on the absence of direct evidence of surface rupture alone. Thus, faults that do not fulfil all active criteria are still included in the database as we applied a broad definition of active faulting to reduce the risk of excluding faults that appear to be inactive but which could rupture in a future earthquake despite displaying limited evidence of activity. In published maps of the region, no attempt was made to distinguish active and inactive faults in the Luangwa Rift . Here we classify 18 faults into varying degrees of activity confidence in a systematic active fault database.

The faults determined to have the highest confidence of activity (Chipola, Molaza, Chitumbi, Kabungo) all have prominent scarps, offset alluvial fans, and steeply incised rivers in the footwall (Figs. –). We measured the height of the prominently exposed fault scarps on each of these faults (Figs.  and ), with the median scarp height between 13–19 m (Fig. ). The Chipola Fault has a scarp height of 13.0 ± 0.4 m (Figs. a and ). It has been suggested that a ∼ 12–14 m high scarp previously detected along the Chipola South Fault formed in the last 10 kyr although no evidence was provided for this time period;, but the authors were unable to distinguish whether this resulted from a single, large-magnitude earthquake or a series of smaller events. Our measurements of scarp height exceed the average single event displacement values from the scaling relationships. Thus, our results suggest that these scarps have formed from multiple earthquakes. Along the Bilila-Mtakataka fault in Malawi, demonstrated that a 20 m high fault scarp was generated by at least two earthquakes with single event displacements possibly as high as 10–12 m, which also exceeds empirically derived single event displacement estimates. Evidence from the Hebron Fault in Namibia also suggests that normal faults that rupture thick crust may have higher single event displacement-length ratios than the dataset compiled by . Alternatively, the scarps may have formed in a single multi-fault event e.g., whose magnitude (and hence single event displacement) would be higher than indicated in the LRAFD. However, composite scarps observed on the Molaza Fault (Fig. d) and bi-modal peaks in the histograms of scarp height along the Chipola, Kabungo, and Chitembo faults (Fig. a–c) support the inference that these scarps have formed in multiple Quaternary earthquakes.

The border faults for the northern (Molaza Fault) and southern (Chipola Fault) basins of the Luangwa Rift have the highest possible values for activity confidence, exposure quality, and epistemic quality. Both faults are well exposed along their entire length (Figs.  and ), with steep fault scarps (25–30∘), and have few gaps where it was not possible to measure the scarp height (Fig. ). In contrast, there are few mapped intra-rift faults. The Luwi and Kapampa faults (LRAFD_ID: 9 and 10) are the most prominent intra-rift faults and have clear fault scarps, but they have activity confidence values of 4 and 3, respectively, and the lowest values for exposure and epistemic quality. Furthermore, these intra-rift faults are short (Luwi – 20 km; Kapampa – 40 km) compared with the two major border faults (Chipola – 207 km; Molaza – 142 km). The lack of intra-rift faulting is unlikely to be because of the inability to detect smaller scarps with SRTM data as we measured scarps as small as ∼ 3 m high, which are smaller than intra-rift faults observed in the southern Malawi Rift . Thus, deformation in the Luangwa Rift appears to be primarily accommodated across two major border faults at the edge of the rift. This differs from the Malawi Rift where deformation is equally distributed on both border and intra-rift faults , despite both rifts being magma poor and having similar extension rates ∼ 0.7 mm yr-1;, as well as similar crustal ∼ 40–45 km; and lithospheric thicknesses ∼ 150 km;. The spatial pattern of deformation in the Luangwa Rift is more similar to the Lake Tanganyika Rift, where up to 90 % of extension is accommodated on the border faults .

The localised deformation across border faults justifies our approach of using the geodetically derived regional extension rate to estimate the slip rate and earthquake recurrence interval of the border faults (Sect. 3.2.2). However, these should be treated as upper bounds on the slip rate and lower bounds on the recurrence interval. Numerical models indicate that rifts with localised deformation across border faults form in strong lithosphere, where the strength is dominated by the crust . Low Vp/Vs ratios and high horizontal shear wave velocities suggest the absence of partial melt, magmatic intrusions, or significant levels of fluid and instead imply that the crust is strong beneath the Luangwa Rift . Furthermore, although faults in the rift follow the orientation of the foliated mylonitic gneiss and eclogites within Mwembeshi shear zone or a splay of the shear zone;, experiments on similar mafic samples from the Malawi Rift suggest these rocks are unlikely to be frictionally weak . Although the high-grade metamorphic shear zones such as the Mwembeshi shear zone are more likely to be viscously weak because of grain-scale heterogeneities, it is notable that Quaternary reactivation of Karoo-age faults in southern Africa has been observed across the Lower Zambezi escarpment in Zimbabwe and in the Lower Shire Rift in Malawi , both being regions that are not underlain by lithospheric-scale shear zones. Thus, although the faults in the Luangwa Rift and other Karoo-age basins have been reactivated during the current Miocene phase of rifting in East Africa, the nature of weaknesses in the lithosphere in this region remains unclear.

The largest historical event in the Luangwa Rift was a Mw 6.7 event on 1 May 1919 (Fig. ; , NEIC earthquake catalogue; ). The ISC-GEM catalogue indicates that another Mw 6.3 event occurred in the same region in 1940 (Fig. ). The epicentres of both the 1919 and 1940 events are located close to a ∼ 50 km long section of the Molaza Fault that is exceptionally prominent, with a well-preserved, steep (20–25∘), linear fault scarp that is continuous across small stream channels (Fig. ). Empirical fault scaling laws imply that a Mw 6.7 event would cause a 30 km long rupture with an average displacement of 0.9 m, and a Mw 6.3 event would cause a 17 km long rupture . Thus, we suggest that the ∼ 50 km long exceptionally well-preserved fault scarp along the Molaza Fault was formed, in part, by the two 20th century earthquakes in the Luangwa Rift. However, earthquake location accuracy in Africa at the time of these events is low, shown by the disparity between NEIC and ISC-GEM locations (Fig. ). The 1919 earthquake recorded by ISC-GEM occurs on the same day as a Ms 6.2 event recorded on 1 May 1919, which macroseismic damage reports initially suggest was located 250 km to the north . It is unclear if these events are linked, and thus field investigations of the Molaza Fault should be a priority to establish whether this represents a rare example of a 20th century earthquake surface rupture in Africa. Nevertheless, previous seismic hazard assessment in the region by definition considers the maximum possible earthquake magnitude to be 0.5 greater than the largest recorded historical earthquake . However, this seismic hazard assessment states that the maximum-magnitude earthquake in this region is M 6.9 . Our new finding of a Mw 6.7 event on the Molaza Fault should therefore prompt a revision of the seismic hazard in the region.

Despite evidence for the activity on the Molaza Fault in the 20th century, there remains large portions of the 140 km long fault that have not ruptured recently. We estimate that the Molaza Fault is a seismic source capable of hosting an earthquake with a maximum magnitude of Mw 7.8 and a displacement of 3.1 m, which is an order of magnitude greater than earthquakes recorded in the Luangwa region and more than any event in the whole of southern/eastern Africa. The largest regional event was the 13 December 1910 Ms 7.4 Rukwa earthquake in Tanzania , and there have only been five other M≥7.0 events along the East African Rift: the 1919 Mw 7.2 Matai, Tanzania, earthquake; the 1928 Mw 7.0 Baringo, Kenya, earthquake; the two Juba, Sudan, earthquakes in 1990 Mw 7.1 and 7.2;; and the 2006 Mw 7.0 Mozambique earthquake . The LRAFD contains eight faults that exceed 50 km in length, with two faults greater than 100 km. Seismic source attributes calculated from the LRAFD indicate that there are 12 faults that have the potential to rupture in Mw≥7.0 earthquakes (Fig. ), with the 207 km long Chipola Fault capable of hosting up to a Mw 8.1 earthquake. Global compilations of continental normal faulting earthquakes suggest that they rarely exceed rupture lengths of 50 km and low Mw 7 , and there is only one event with a surface rupture length >100 km . Thus, although Mw 7+ events are likely rare, they should be considered possible in the Luangwa Rift due to the long faults that are hosted in 45 km thick crust , which has recorded seismicity to ∼ 30 km depth . Nonetheless, these large-magnitude events likely occur infrequently as >100 km long normal faults in southern Africa are often segmented , and the regional b value is ∼ 1 , implying that smaller segmented ruptures are more likely than earthquakes that rupture an entire fault.

To estimate the magnitude of smaller segmented ruptures, we attempted to identify fault segments for the four best exposed faults in the Luangwa Rift by systematically measuring along-strike fault scarp heights (Fig. ). This approach has been successful in other East African rift basins using 12.5 m resolution TanDEM-X data . Using the 30 m resolution SRTM data, we were unable to identify clear segment boundaries in either the scarp height measurements or in notable changes in fault geometry (e.g. 90∘ bends), making it challenging to assess the limits on rupture length on individual faults. The low slip rates that we calculated for these faults (0.4–1.7 mm yr-1) indicate that if large (M>7.0) events do occur, they are rare, with intermediate earthquake recurrence intervals of 5.1 kyr on the Chipola Fault and 3.4 kyr on the Molaza Fault. Thus, although we do not directly observe evidence for fault segmentation in the Luangwa Rift, the data provided here can be used to incorporate small ruptures along these faults into seismic hazard assessments by combining the provided slip rate, fault area, and magnitude estimates with a regional b value and the methodology developed by to develop continuous recurrence models for these sources see. However, it is also possible that multi-fault rupture may occur, which would increase the potential magnitude of future events. For example, multi-fault rupture of the 142 km long Molaza (LRAFD_ID: 16), 9 km long Molaza 2 (LRAFD_ID: 15), and 16 km long Molaza 3 (LRAFD_ID: 14) faults would increase the potential magnitude of the earthquake from Mw 7.8 to 8.0 compared to if the Molaza Fault ruptured on its own.

The low frequency of large earthquakes in the Luangwa Rift and the few recent destructive earthquakes means that awareness of seismic hazard and mitigation strategies in Zambia may be low. The Luangwa Valley and National Park are tourist destinations with a significant number of communities, tourists, wildlife, and economic assets exposed to seismic hazard. Kasama (200 km west) and Chipata (150 km east) are the largest nearby cities, both with populations of ∼ 90 000 . There are a combined 2.4 million people living in the Muchinga and eastern regions of Zambia, with a population growth rate of 4.3 % and 2.8 %, respectively . This means nearby populations that may be affected by seismic hazard will increase with time. Furthermore, the region has high levels of vulnerability. Recent moderate-magnitude earthquakes in Malawi led to high levels of damage and large economic losses , and research in Malawi indicates that building vulnerability in this region is higher than currently predicted by global models e.g. the USGS WHE-PAGER model;. We suggest that active fault mapping, such as has been carried out here in the LRAFD and in other active fault databases in southern Africa , provides a framework for accurate probabilistic seismic hazard assessment and thus for increasing resilience to seismic hazard throughout southern and eastern Africa.