A region where the geodetic block model suggests substantial deformation is at the northern edge of the Quito-Latacunga microblock near the border between Ecuador and Colombia (Figs.  and ). The Quito-Latacunga microblock is located at a large (∼ 250 km N–S) left-stepping section of the right-lateral eastern boundary of the NAS Fig. ,. The contractional left-stepping geometry results in ∼2 to 3.3 mm yr⁻¹ of W–E compression along the more northerly trending eastern and western boundaries of the microblock and ∼ 3 to 6 mm yr⁻¹ right-lateral strain along the more NE-trending southern and northern boundaries e.g.,. Along the western boundary, the Quito and the Latacunga thrust fault systems have been shown to accommodate much of the predicted compression . On the southeast boundary, 0.45–6.6 mm yr⁻¹ of active right-lateral slip has been observed along the Pallatanga fault and faults along strike to the northeast that traverse the eastern Cordillera of the Andes . At the eastern boundary, ∼ 10 mm yr⁻¹ of right-lateral slip has been observed along the CAS (a part of the CCPP), and ∼ 4 mm yr⁻¹ of W–E shortening has been observed on thrust faults near Reventador Volcano . Along the northern boundary of the microblock, where the ∼ 3 mm yr⁻¹ of predicted right-lateral strain is accommodated is not clear. NE-striking right-lateral Quaternary faults have been documented near the city of Pasto and on the flanks of the Galeras volcano in Colombia . Geologic and geomorphic studies of two of these structures, the Buesaco and Aranda faults, show 0.7–1.6 and 1.1–2.6 mm yr⁻¹ of right-lateral reverse slip, respectively . However, these faults are >45 km north of the northern boundary of the Quito-Latacunga microblock Fig. 2,. Historic and instrumental seismicity suggest active faulting further south in Ecuador as well, suggesting a wide zone of deformation distributed across several fault zones.

A southward decrease in the eastward component of the GNSS velocities across the northern boundary of the Quito-Latacunga microblock is consistent with the right-lateral strain predicted on ENE-WSW striking planes in the block model. GNSS stations near Tulcan, and ∼ 5 km north of the boundary observe 7.1 and 10.6 mm yr⁻¹, respectively, of ENE motion relative to stable South America, while a station in Pasto, Colombia ∼ 40 km to the north indicates 8.4 mm yr⁻¹ of NE motion (Fig. ). Although no stations are located directly south of the boundary, stations ∼ 25 km SSE of the boundary in Ibarra and Pimampiro show 6.7 and 6.3 mm yr⁻¹, respectively, of ENE motion with respect to stable South America. Thus, given the wide spacing of the GNSS stations, the location of the northern boundary of the Quito-Latacunga microblock and whether strain here is localized on a single structure, or is distributed along several structures or as off-fault deformation, cannot be discerned from the GNSS data alone. This distinction is important for seismic hazard because if all of the strain is concentrated along a single fault, this fault will likely be longer, more mature, have a higher slip rate, and be capable of producing larger earthquakes with shorter recurrence intervals. Additionally, GNSS velocities may capture magmatic processes related to recent volcanic activity at the Chiles-Cerro Negro volcanic complex and Galeras volcano.

Recent analysis of the interseismic InSAR velocity field in northern Ecuador and southern Colombia shows a better spatial resolution of deformation across the northern boundary of the Quito-Latacunga microblock, and supports a hypothesis that right-lateral deformation is distributed Fig. ,. The eastward velocity field here shows no sharp gradient in velocities across the geodetic block boundary. Instead there is a N-to-S gentle reduction in eastward velocities from ∼ 1.3 to ∼ 0.8 mm yr⁻¹ over ∼ 60 km (Fig. b). These estimates, however, may be influenced by ongoing inflation at Chiles-Cerro Negro and could be capturing transient volcanic deformation rather than being representative of long-term tectonic deformation e.g.,. This local deformation is illustrated by the substantial increase in the eastward velocity field across an axis that extends to the SSE from Chiles-Cerro Negro (Fig. a) along with a maximum uplift rate of 2.9 cm yr⁻¹ southeast of the volcano. Another InSAR analysis by shows even greater eastward velocities east of Chiles-Cerro Negro (6 mm yr⁻¹), and uplift rates of 15 mm yr⁻¹.

Crustal seismicity at the northern boundary of the Quito-Latacunga microblock (1993 to 2016, IG-EPN) also shows diffuse strike-slip and compressive strain. Small (M2.0) to moderate (M5.0) earthquakes are dispersed throughout the region and are not clustered along the boundary defined by the block model. Instead, clusters are located near the Chile-Cerro Negro and Galeras volcanic systems, and near the city of Ibarra (Fig. ). These earthquake swarms have been attributed to volcanic processes at Chiles-Cerro Negro in 2014 and Galeras in 1989 , while the swarms near Galeras in 1993 and 1995 have been suggested to have at least a partially tectonic origin . Focal mechanism analyses of this crustal seismicity indicates a S-to-N transition approximately at the Ecuador-Colombia border from compressive to strike-slip faulting with a relatively consistent W–E shortening direction . Focal mechanisms in this region of earthquakes with M>4.5 from the ISC catalogue Fig. a show predominantly right-lateral slip on W–E trending faults, however some show a substantial non-double component indicating a volcanic source e.g.,.

Instrumentally recorded and historical earthquakes with M>5.0, including the recent Mw5.6 event that is analyzed in this study, have also occurred along the modeled northern boundary of the Quito-Latacunga block. In 1868, a sequence of two large M∼6.6 and M∼7.25 earthquakes separated by 10 h occurred in northern Ecuador . Damage intensity data for the first M∼6.6 earthquake suggests an epicentre in a region between El Angel and the Chiles-Cerro Negro volcanic system (Fig. a). The subsequent M7.25 earthquake, which destroyed the city of Ibarra, is thought to have occurred either south or west of Ibarra and along the western boundary of the microblock. In Colombia, eight M∼5.6 to 6.3 earthquakes occurred between 1923 and 1947, causing substantial damage to communities such as Cumbal, Pasto and Túquerres (https://sish.sgc.gov.co/visor/, last access: August 2025). More recently, at the end of the 2014 earthquake swarm a M5.6 right-lateral earthquake occurred on a moderately-dipping SW-striking fault plane ∼ 15 to 20 km south of Chiles-Cerro Negro. InSAR and Coulomb stress modeling suggest that this earthquake could be attributed to a pore fluid pressure increase resulting from volcanic inflation. However, its focal mechanism is consistent with tectonic stress and volcanism could potentially be just a trigger of the earthquake . Finally, on 25 July 2022 a M5.6 earthquake occurred 9 km north of San Gabriel, causing damage to this and several other nearby cities such as Tulcan and El Angel.

The northern boundary of the Quito-Latacunga microblock is located within the Inter-Andean Valley, a large depression filled with Pliocene to Quaternary volcanic and volcanosedimentary rocks, which overlie and obscure a major terrane boundary between oceanic terranes to the west and metamorphic basement rocks to the east (Fig. a). The metamorphosed sediments to the east are believed to represent the ancestral western margin of South America , while the Jurassic to Cretaceous plutons that intrude into these sediments may represent a continental arc e.g.,, or intrusions in an intracontinental rift e.g.,.

Three oceanic plateaus, which subsquently accreted to the ancient margin from late Cretaceous through to the Paleocene, comprise the basement rocks west of the terrane boundary in the Inter-Andean Valley. The San Juan terrane, which accreted to the South American margin at ∼75 Ma, and the Guaranda Terrane, which then accreted to the new margin outboard the San Juan terrane at ∼68 Ma, are exposed in our study area . These basement rocks are overlain by siliciclastic sediments likely derived from exhumation during accretion , and by younger volcanic rocks related to the modern arc . The two oceanic terranes are bounded by major, steeply-dipping, SW to NE-trending suture zones Fig. a,. These suture zones have been thought to have been reactivated and may accommodate the right-lateral shear and compression resulting from the NE motion of the NAS. have shown that many of these SSW to NNE trending structures in central Ecuador have been abandoned and are now cross-cut by more recent faulting. However, in southern Colombia near Pasto, the SW–NE trending Romeral shear zone separating the accreted terranes from the continental paleomargin and parallel related structures have been shown to be reactivated as active right-lateral reverse faults .

The basement lithologies and associated suture zones are overlain by thick deposits of Miocene to Quaternary volcanic and associated volcano-sedimentary rocks deposited by the numerous volcanic centers in the study area (Fig. ). Several of these volcanic centers are clustered around the active Chiles-Cerro Negro complex, which hosted minor eruptive activity in the 19th century . Two other active volcanoes are located to the north in Colombia near the cities of Tuquerres and Pasto (Fig. ). The Azufral volcano near Tuquerres has had 6 periods of eruptive activity over the last 220 thousand years, with the most recent being ∼ 280 years BP. The Galeras volano is located 9 km west of Pasto, Colombia, and has seen continuous activity since the 16th century . This volcanic activity at the northern edge of the Quito-Latacunga microblock could play a role in the location of deformation and in the onset of earthquakes e.g.,.

Along with recent volcanism, Quaternary glaciations have shaped much of the high-elevation landscape in the Andes of northern Ecuador and southern Colombia. Glaciations in this region occurred several times throughout the Pleistocene (>130 ka BP) with a maximum extent of ice that may have reached down to 2700 m . The last glacial maximum (LGM) occurred in this region 20–12 ka BP, resulting in ice caps and valley glaciers extending down to elevations between 3000 and 3800 m a.s.l. . On the plateau surrounding the Chiles-Cerro Negro volcanic complex, U-shaped valleys, cirques, and terminal moraines that are indicative of recent glaciation e.g., are observed at elevations just below 3500 m a.s.l. Simlarly, recognize glacial landforms on the flanks of the Galeras Volcano in southern Colombia that terminate elevations just below 3400 m a.s.l. Although no ice caps remain on either of these volcanoes, the landscape suggests substantial glacial influence up to the beginning of the Holocene.

To analyze the surface deformation associated with the 25 July 2022 earthquake, we computed coseismic SAR interferograms near El Angel using both Sentinel-1 (C-band) and ALOS-2 ScanSAR (L-band) SAR data. Two Sentinel-1 interferograms, from descending track 142 (17–29 July 2022; Fig. a) and ascending track 120 (15–27 July 2022; Fig. b) were produced using the ForM@Ter GDM-SAR-In service (https://www.poleterresolide.fr/le-service-gdm-sar-in/, last access: June 2024) based on the NSBAS processing chain . One ALOS-2 ScanSAR interferogram, from descending track 139 (10 July–21 August 2022; Fig. c) was also computed using the GMTSAR software . Note that an ionospheric phase screen obtained using the split-spectrum technique was removed from the ALOS-2 interferogram . All interferograms were filtered using a sliding window removing the phase gradient before filtering and reintroducing it afterwards e.g.,. The ALOS-2 interferogram being very coherent and containing few fringes caused by coseismic displacements, could be unwrapped easily even close to the rupture using a region growing algorithm Fig. d,. An unwrapped Sentinel-1 interferogram is also provided in the data repository Fig. S2,.

Reconnaissance geomorphic mapping to map potentially active faults was conducted using GoogleEarth Pro (version: 7.3.6.10201), a hillshaded 4 m-resolution lidar- and photogrammetry-derived digital terrain model (DTM) of selected areas in Ecuador from SigTierras of the Ecuadorian Ministry of Agriculture, Quito (http://ide.sigtierras.gob.ec/geoportal/, last access: July 2023), and a 5 m resolution GeoSAR acquired DTM of southern Colombia from the Geological Survey of Colmbia GeoSAR project (2020) that is available in a data repository . Potentially active faults were mapped using QGIS software v. 3.22 , by identifying continuous lineaments in the landscape with horizontal or vertical offsets of Quaternary-aged geomorphic features (e.g. glacial moraines and channels). Often, the expected offsets of slowly-slipping faults in young recently-glaciated landscapes would be expected to be lower than the resolution of the DTMs we used e.g.,. Therefore, we also considered lineaments where clear offsets were not visible but there was evidence of topographic benches and associated landslides seen in satellite imagery.

To further assess the potentially active faults we mapped using the lower-resolution DTMs, we conducted detailed geomorphic/geologic mapping using higher resolution Pleiades satellite derived DTMs and field studies. High-resolution (<2 m) DTMs were constructed using three Pleiades 0.5 m-resolution tri-stereo panchromatic tiff images acquired northwest of El Angel, Ecuador in August 2023 (https://dinamis.data-terra.org/en/imagery/, last access: August 2023). The images and their corresponding Rational Polynomial Coefficient (RPC) models were processed using the open-source NASA Ames Stereo Pipeline (ASP) software, which uses stereogrammetry to derive topography . We first projected the three images onto an existing Copernicus 30 m-resolution DEM (https://dataspace.copernicus.eu/), and then generated three point clouds based on the three possible different stereo pairs of the photos. We then transformed these point clouds to DTMs and created a mosaic of those DTMs for final model. The code we used for the generation of our DTM with AMES and our DTM are available in the data repository .

Using the Pleiades DTM, we mapped fault scarps, and offset lateral moraines and landslides along these scarps. In particular, lateral moraine crests and one channel were delineated using breaks in slope identified in 5 and 10 m contour maps extracted from our Pleaides DTM. We estimated horizontal offset of the lateral moraine ridges by estimating where the projections of the undeformed moraine crests and channels intersect the fault plane, and measuring the horizontal distance between those projections e.g.,. Multiple projection interpretations were possible, thus we used reasonable projections with the most, and least amount of offset, to determine a range of horizontal offsets of the lateral moraines and channel, and establish uncertainty. A shapefile containing our moraine crest and glacial stream projections are located in the data repository . We attempted to use the LaDiCaoz MatLab code to backslip and match moraine ridge and stream topography across a fault to better assess slip. However, the subtle topography and the variability of the topographic features perpendicular to the fault zones (e.g., ridge width, channel morphology) led to extremely variable offset estimations that were unrealistic.

We conducted a field investigation in October and November 2022, to ground-truth and describe geomorphic features and evidence of active faulting. Where accessible, we surveyed the 25 July 2022 earthquake surface rupture that was delineated by InSAR, noting any surface deformation. We surveyed scarps, sag ponds, and landslides along other potentially active faults identified using remote sensing. We also located roadcuts that exposed active faults. To examine the earthquake history of these structures, we conducted a detailed analysis of the fault and sediments in the exposures. After cleaning and cutting the exposures to produce a vertical face, we used Agisoft Metashape software to create orthophotomosaics from overlapping digital photos of the exposure e.g.,. These photomosaics were used to map the exposure on a digital tablet. A very high-resolution photomosaic and trench log of the primary exposure is provided in a dataset .

In order to date offset glacial landforms, we collected samples of large (>2 m diameter) andesitic boulders, embedded in glacial till, that were exposed on the crest of glacial ridges/lateral moraines for cosmogenic ³He exposure dating e.g.,. By dating the exposure time of these boulders, we are estimating the time elapsed since they were transported and deposited by a glacier, assuming no exposure before this transport . The selected boulders were fresh, with little to no visible alteration near the surface, therefore we assume there is negligible erosion of the boulders.

Samples of 2 cm-thickness were collected at the top of flat exposed surfaces of three different andesite and basalt boulders, whose coordinates and elevations were precisely noted and are available in the data repository Table S1,. Given their location on the ridge crests, the samples experienced negligible mask shielding. Samples were crushed and sieved, and pure pyroxene grains of 100–500 µm were extracted by hand at the Institut des Sciences de la Terre (ISTerre) at the Université Grenoble Alpes in France. ³He and ⁴He concentrations in the pyroxene grains were then measured at the Centre de Recherches Pétrographiques et Géochimiques (CRPG) in Nancy, France. Pyroxenes were fused in high vacuum at 1500 °C using a home designed induction furnace . After gas purification, helium isotope abundances were measured with a tuned Split Flight Tube mass spectrometer, following methods described in . ⁴He furnace blanks were (4.2 ± 0.2) × 10⁻¹⁵ mol and represented from 3 % to 15 % of the sample signal, while ³He blanks were undetectable. Table S1 in the data repository shows the He isotope abundances. During the same analytical session, we analyzed a CRONUS-P standard material, that yielded an ³He concentration compatible with the certified value, within analytical uncertainties .

Given that these volcanic rocks have a Pleistocene age, we assume that both the nucleogenic ³He and the radiogenic ⁴He build ups are negligible. We considered that the measured ⁴He is magmatic and used a 3He/4He ratio of 5 Ra (Ra = 1.384 ± 10⁻⁶ being the atmospheric ratio) to correct the measured ³He concentrations . This magmatic ³He correction represents less than 10 % of the total ³He measured in the three samples. Ages were then calculated using the online CREp calculator incorporating various ³He production rate models and scaling factors. Such approach permits to encompass all the sources of uncertainties associated with this cosmogenic ³He exposure dating.

We collected 14 bulk sediment samples for radiocarbon dating to constrain the sediment deposition and earthquake event ages exposed in the roadcuts. The majority of our samples were collected from Páramo soils (Andisol), which have very high organic carbon content. Their low bulk density, high porosity, and humic/dark nature make them comparable to peat-like soils or organic-rich soils. Because of these properties, small-volume, focused sampling can retrieve sufficient organic carbon for radiocarbon dating, reducing the risk of time-averaging or mixing compared to bulk sampling in low-organic, mineral sediments. The high organic content and relative stability of organic matter in Andisol under cold, humid conditions at ∼ 4000 m reduces decomposition and thus preserves carbon, making them good candidates for radiocarbon dating. These types of samples have been shown to provide reliable radiocarbon ages in previous paleoseismic studies in similar environments along the Pallatanga fault zone and the Billecocha fault system . An identical sampling strategy resulted in precise historical ages matching with the 1797 earthquake in Pallatanga area . Nevertheless, we acknowledge that post-depositional processes such as bioturbation and root penetration could introduce additional uncertainty that is not fully captured by the laboratory-reported analytical errors e.g.,. Additionally, our samples could be detrital giving ages that are older than the unit they are located within. This is especially the case for samples from colluvial wedges e.g.,. Therefore, we assessed each sample and date and discounted dates that were likely to be affected by bioturbation and re-sedimentation.

Radiocarbon samples were dated at the Artemis AMS-French National facility CEA Saclay, LMC14;. We used OxCal version 4.4.4 for radiocarbon calibration with the IntCal20 ¹⁴C production curve and report our results as calendar calibrated ages before 1950 (cal BP) or thousands of calendar calibrated years before 1950 (ka). We then used OxCal to incorporate all available chronological constraints into a model that uses Monte Carlo routines and Bayesian statistics to determine the probability distribution function (PDF) of geologic unit and earthquake event ages e.g.,. OxCal codes are available in an archived dataset .

On 25 July 2022 a Mw5.6 earthquake occurred ∼ 10 km northeast of El Angel and ∼ 15 km southeast of the Chiles-Cerro Negro volcanic complex near 0.743° N, -77.844° E (Fig. ). This earthquake was shallow, with a hypocentral depth of <5 km (IG-EPN). The focal mechanism computed by IG-EPN indicates almost purely strike-slip kinematics on a sub-vertical fault, with either right-lateral slip on a 250°-striking fault or left-lateral slip on a 345°-striking fault.

The Sentinel-1 and ALOS-2 coseismic SAR interferograms computed for this earthquake show a clear surface deformation pattern near the epicenter of the 2022 earthquake (Fig. ). A comparative interpretation of these complementary (ascending and descending, C-band and L-band) datasets thus allows us to map precisely the coseismic surface rupture. Overall, the rupture is oriented 80° and produced right-lateral slip. The opposed signs of the Line-of-Sight (LOS) deformation lobes between ascending and descending interferograms indicate that the earthquake produced mostly horizontal motion, consistent with the almost purely strike-slip focal mechanism. In the eastern part of the rupture, close to the epicenter, a distinct surface rupture over 5 km length can be observed (solid black line in Fig. ), with up to 8 cm of localized deformation in the LOS and up to 13 cm of total surface deformation in the LOS. This 13 cm of total surface deformation (profile H-H', Fig. h) includes any deformation along the fault that did not rupture the surface, and/or inelastic, off-fault, deformation in the area around the fault e.g.,. In contrast to the eastern part of the rupture, the western part shows more off-fault deformation over about 5 more kilometers. There, the rupture divides into two strands, each of them accommodating about 5 cm of LOS deformation without clear localized offset (profile FF'). Further west, the rupture splits into more en-echelon segments with a decreasing amount of total surface deformation. Therefore, InSAR shows that the surface rupture produced by this earthquake is complex and characterized by a large amount of distributed deformation across the fault zone, which we term the 25 July fault (JF, Fig. ).

Field reconnaissance was carried out three months after the earthquake, in October 2022, aiming to study the surface rupture using InSAR maps as guides. However, the area turned out to be difficult to explore owing to the dense vegetation cover in the El Angel reserve, together with a large number of landslides closing the main road crossing the reserve. The only possible evidence of surface rupture that could be observed are sub-vertical cracks on the walls of a man-made irrigation channel, displaying a few centimeters of opening (Fig. e), and located exactly on the surface rupture inferred from InSAR (yellow star in Fig. d). However, no horizontal offsets could be observed on this channel. The organic rich Andisol soil covering the El Angel reserve (e.g., Fig. e), which may behave more ductile, coupled with the low magnitude of the 25 July earthquake may explain the low amount of localized deformation during this earthquake.

Lower resolution 4 and 5 m hillshaded DEMs and satellite imagery show a number of SW to NE lineations within a ∼ 70 km-wide zone across the northern boundary of the Quito-Latacunga microblock (Fig. a). These lineations are generally continuous between 5 and 25 km and are oblique to the previously mapped more northerly trending active faults . The lineations are formed at all elevations in the map area and cut recent volcanics, glacial sediments, and bedrock. Two lineations located north of the city of El Angel, which we have termed the Reservoir and Polylepis faults, show the clearest signs of active deformation (RF and PF in Fig. a).

Pleiades-derived DTMs show two parallel fault segments offsetting glacial moraines ∼20 km west of the epicenter of the 25 July earthquake, and ∼12 km southwest of the Chiles-Cerro Negro volcanic complex (Fig. ). Both faults are subvertical, and the northern fault strikes ∼060 to 070° while the southern fault strikes ∼070°. The surface expression of the northern fault is relatively continuous and crosses the entire width of the Pleiades DTM (10.5 km). It also may continue up to 15 km to the northeast where the low-resolution Copernicus DEM shows NE-striking lineaments along the southern flank of the Chiles volcano. The visible surface trace of the second, southern fault is less continuous, consisting of up to ∼2 km-long segments that are likely connected at depth. Topographic lineaments visible in the low-resolution Copernicus DEM suggest this southern fault continues to the northeast for up to 10 km from the edge of Fig. b. To the southwest, the projection of the more W-striking southern fault merges with the northern fault.

Along both fault traces, field reconnaissance and the Pleiades DTMs show up to 3 m-high N- and S-facing scarps and benches, along with sag ponds and marshes where streams crossed and were dammed by the fault (Figs. ; ). Small (<30 m-wide) slumps and landslides were recognized in the field and two ∼100 m-wide landslides are visible in the DTM in Fig. a, b, c, where the southern fault cross-cuts lateral moraines. Additionally, both faults offset lateral glacial moraines. The northern fault right-laterally offsets a pronounced ridge above SW-facing lateral moraine by 16±11 m, and it may offset the two adjacent lateral moraines to the east. However, where these offsets are projected, less pronounced ridge crests along the top of the moraine, or a change in strike of the moraine do not allow accurate estimations of the displaced landforms. The northern strand also offsets a creek at the base of the glacial valley in the center of Fig. a, b by 29±8 m right-laterally. A change in strike of the ridgeline where the northern fault crosses the road in Fig. d, does not allow for an accurate offset estimation, however a sag pond between the main fault branch and right-stepping branch ∼ 50 m to the south indicates extension along a right-lateral fault. The southern fault displaces both the E- and W-facing lateral moraines of a glacial valley in the center of Fig. a, by 26±11 and 21±8 m, respectively, and then displaces the stream along the basal moraine of the glacial valley by 22±10 m. Where the southern fault offsets the E-facing moraine, we have identified two fault strands on the northern and southern side of a 100 m-wide landslide and used offset measurements on both strands to determine the overall right-lateral displacement (Fig. c).

All of the scarps have minor vertical offsets along them, however, the change from N- to S-facing scarps along the fault suggest largely strike-slip deformation. The surface trace of the fault segments across the glacial valleys suggest sub-vertical to steeply N-dipping faults, also suggesting strike-slip faulting. The undulating terrain and relatively minor vertical offsets preclude viable estimates of vertical separation.

Our analyses of the 25 July 2022 Mw5.6 earthquake and fault, geomorphic mapping, and paleoseismic surveys indicate that there are several SW–NE striking active faults that accommodate right-lateral slip south of the Chiles-Cerro Negro volcanic complex. The Reservoir fault and the InSAR analyses of the earthquake on the 25 July fault both show clear evidence of right-lateral Holocene fault ruptures. The best explanation for the linear scarps, landslides, and sag ponds along the Polylepis fault is also Holocene fault rupture, however further detailed studies are required to prove this hypothesis. Nonetheless, right-lateral slip on these three parallel faults is consistent with the right-lateral strain at the northern boundary of the Quito-Latacunga microblock that is predicted by the geodetic block model (Figs. ; b). Additionally, distributed strain across these three parallel faults, which are spread over ∼ 10 km, is consistent with the lack of a sharp gradient in eastward velocities across the block boundary seen in InSAR Fig.  and the distributed instrumental seismicity (Fig. b).

Based on our geomorphic mapping, a substantial portion of the geodetically modeled 3 mm yr⁻¹ of right-lateral strain at the northern edge of the Quito-Latacunga microblock is taken up by slip on the Reservoir fault. Considering the range of offsets of the glacial moraines and creeks (Fig. ), the range of ages of these features from our cosmogenic ³He dating, and published ages we can estimate a slip rate on both the northern and southern strands of the fault. Assuming a maximum of 37 m of slip on both the northern and southern strands (26+11, 29+8 m, Fig. b), and a minimum age of 11.9 ka of the moraines (Table , Model 3, ANG-50), we calculate a maximum slip rate of 3.1 mm yr⁻¹ on both of these strands or a combined rate of 6.2 mm yr⁻¹ for both strands. Given a minimum offset of 5 m on the northern strand (16–11 m, Fig. b), 13 m on the southern strand (21–8 m, Fig. b), and a maximum deglaciation age of 20.4 ka (Table , Model 2, ANG-51), we estimate a minimum slip rate of 0.2 and 0.6 mm yr⁻¹ on the northern and southern strands respectively. The combined minimum slip rate of both fault strands is 0.8 mm yr⁻¹. Using a mean offset of 23 m for both the northern and southern strands, and an ice-free age of 15 ka , we obtain slip rates of 1.5 mm yr⁻¹ for both strands and 3 mm yr⁻¹ for the combined faults. Given these rates, slip on just the Reservoir fault could account for between one third to over double the total 3 mm yr⁻¹ deformation rate observed in the block model . Because we observe active deformation on the 25 July fault, and most likely on the Polylepis fault, we argue that deformation is distributed across several faults and that slip on the Reservoir fault takes up substantially less than 100 % of the right-lateral strain at the north edge of the Quito-Latacunga block. This argument entails that either the minimum combined slip rate of the Reservoir fault (0.8 mm yr⁻¹) is more accurate than the higher rates (>6 mm yr⁻¹), or that the geodetic instantaneous deformation rate is lesser than the geologic slip rate.

The minimum and mean slip rates we determine for the Reservoir fault are similar to right lateral slip rates determined on the NE to ENE-trending Buesaco (0.65 to 1.6 mm yr⁻¹) and Aranda (1.11 to 2.6 mm yr⁻¹) faults near Pasto, Colombia (Fig. ). These active structures, which are located >50 km north of the block model boundary, show that the deformation zone may be even more spread out and distributed (e.g., ∼ 70 km), or that it steps northward towards the east. Regional scale mapping e.g., also suggest numerous more northerly trending faults with a greater dip-slip component within this region of distributed deformation. If these structures can be shown to be active, the N-stepping strike-slip faults shown in this study may act as tear faults in a eastward stepping northerly-trending thrust belt system (Fig. ). However, field studies of these structures are required to determine if they are active and how they relate to the strike-slip faults mapped here. The Buesaco and Aranda faults are also proximal to the active Galeras volcano, similar to our study area's proximity to the Chiles-Cerro Negro volcano complex (Fig. ), suggesting that the volcanic centers may be influencing where deformation is focused see below for more discussion, e.g.,.

The proximity of our study area to recent inflation at the Chiles-Cerro Negro volcanic complex suggests that volcanism may not only localize deformation, but it also could account for a difference between geodetic and geologic slip rates. Shallow seismicity , uplift, and westward and eastward expansion away from a N–S axis centered on Chiles-Cerro Negro visible in InSAR (Fig. a), document volcanic inflation that has been undergoing since 2014. The 25 July 2022 right-lateral M5.6 rupture along a ENE-striking fault we document with InSAR (Fig. ) could be triggered by this inflation, similar to inflation-triggered strike-slip earthquakes at Makushkin Volcano in Alaska . We suggest this trigger because there is a N-to-S reduction across the 25 July fault in the eastward interseismic InSAR velocities resulting from inflation (Fig. a). Inflation northwest of the fault should induce a horizontal stress in a NE–SE direction and induce right-lateral shear on the 25 July fault, which is compatible with the observed right-lateral slip in the 25 July earthquake. This hypothesis needs further testing, however, potentially with a model calculating Coulomb stress changes resulting from inflation e.g.,. We speculate that the larger M6.6 1868 earthquake could have also been related to volcanism at Chiles-Cerro Negro as volcanic activity was reported around that time .

The influence that the Chiles-Cerro Negro volcano has on faults in the area may explain why combined right-lateral slip rates of the Reservoir, Polylepis, and 25 July faults could potentially be higher than geodetic slip rates at the northern boundary of the Quito-Latacunga microblock. The calculated slip rate using mean offsets along the Reservoir fault alone is ∼ 3 mm yr⁻¹, and given the evidence for recent activity on the other two faults it is very plausible that the combined geologic slip rate of these three faults is greater than the 3 mm yr⁻¹ predicted in the block model. The geodetic block model uses GNSS velocities from a wide area across northwestern South America and southern Central America. A model that is consistent with all of these velocities may, by definition, ignore local high strain gradients like the one caused by inflation of the Chiles-Cerro Negro volcanic complex. Additionally, volcanic unrest could trigger ruptures on critically stressed faults via pore fluid pressure increase e.g.,. Pore-fluid pressure increase could be a better explanation for triggering earthquakes on the Reservoir and Polylepis faults as they are positioned different relatively to the zone of recent inflation than the 25 July fault (Figs. , ). A period of high volcanic activity could decrease the average earthquake recurrence interval and increase the short-term slip rate. Therefore, the faults in our study area could have greater earthquake rates and slip rates over the Holocene, resulting in part from volcanic inflation, than the geodetic block model predicts. The long-term slip rate are probably not affected by changes in volcanic activity and more likely controlled by the larger scale tectonics forces. But, the dynamic stresses at the volcanic centers (Chiles-Cerro Negro and Galeras) could be triggering ruptures and localizing deformation around them. Therefore, they could be a partial control on the location of the northern edge of the Quito-Latacunga microblock.

The Polylepis and 25 July faults, and most of the lineations mapped in this study, are oriented oblique (more W–E) to the bedrock structures near Chiles-Cerro-Negro (Fig. b). The Reservoir fault, which trends northeast, however, may be more well aligned with some of these faults, for example the SW–NE trending structure immediately east of Cerro-Negro. Thick deposits of Quaternary volcanic and volcaniclastic rocks in the study area prevented us from mapping of older bedrock faults in the immediate vicinity of these faults. But, based on the orientation of previously mapped bedrock geology maps and the orientation of most of our mapped active faults, bedrock structures south of Chiles-Cerro Negro are mostly not reactivated and are therefore probably not favorably aligned with the crustal stress conditions. This result is similar to studies at the southern edge of the Quito Latacunga microblock where show that ancient terrane bounding faults are not being reactivated by more recent and active faults. In contrast, along the eastern boundary of the microblock, active faults are parallel with and likely reactivate bedrock structures . As mentioned previously, inflation of Chiles-Cerro Negro may change stress conditions at the northern edge of the Quito-Latacunga microblock, aligning them in a way that favors new fault formation. At the larger scale, the northern edge of the block is aligned with the northern side of a large constraining step-over of the entire eastern boundary of the NAS. This major change in geometry may indicate that principal stress orientations are rotating and the overall tectonic regime may also plays a role in changing stress conditions at this location.

To the north in Colombia near the Galeras volcano, the Romeral Shear zone is being reactivated along the Buesaco and Aranda faults . Our preliminary mapping shows lineaments southwest of Galeras that also may be aligned with the Cauca-Patía fault system Fig. ;. These bedrock structures are oriented slightly more W–E than the bedrock faults mapped immediately south of Chiles-Cerro Negro, and therefore they also could be more favorably aligned for reactivation.

A slip rate, and earthquake recurrence and magnitudes are required to include crustal faults as seismic sources in probabilistic seismic hazard assessment (PSHA) models. For the Reservoir fault, we have determined slip rates above. Based on the earthquake history observed in the exposures of the Reservoir fault (Figs.  and ) we can also estimate the recurrence interval for large surface rupturing earthquakes. Smaller earthquakes often do not rupture to surface, or may cause only small offset that may not be discernible in the stratigraphy e.g.,. For example, we did not observe surface offsets from the 25 July earthquake in this study, even though there was up to 8 cm of discrete offset observed with InSAR (Fig. ). Even if there was surface offset, an offset of this magnitude in a paramo environment would likely only offset Andisol against Andisol, making it impossible to perceive in trench or roadcut exposures. Therefore, we may only be observing the largest earthquakes in the exposures e.g., M>6.5,. The average time interval between earthquakes (or between the latest earthquake and present) we observe in the Reservoir fault outcrop (Fig. c) is 2.3 ± 0.5 ka. However, given that the one of these intervals was calculated using the present-day as a minimum interval, the uncertainty is greater than we calculate.

Using fault scaling relations between overall rupture length and magnitude we can estimate magnitudes of the observed paleo-earthquakes. Considering ruptures of the entire estimated length of the northern branch of the Reservoir fault of 10.5 to 25.5 km, the length scaling relation suggests a magnitude between M6.3 and 6.7. Using a similar scaling relation, we estimate paleo-earthquake magnitudes between M5.9 and M 6.5. A rupture of just the ∼ 10 to 15 km long southern branch would also fall within this magnitude range. Given the observation of parallel coeval fault ruptures during the 25 July earthquake (Fig. d), and during other earthquakes e.g., 2016 M7.8 Kaikoura,, we must consider that both north and south segments of the Reservoir fault have ruptured during a single event. If these faults were not connected at depth we would consider them as separate faults and consider their total summed length (20 to 50 km) in a magnitude estimation. These lengths would predict paleo-earthquake magnitudes between M6.6 and M7.0 using the relation and between M6.3 and M6.9 using the relation.

We can also roughly estimate paleo-earthquake magnitudes using the offsets along the northern branch of the Reservoir fault in Fig.  assuming the earthquakes observed in Fig.  account for the majority of the displacement. If there have only been three large earthquakes since the LGM, and using a total offset between 5 and 37 m of the glacial stream and lateral moraine, we estimate 1.7 to 12.3 m of horizontal slip during each earthquake. With the relation between displacement and rupture length, rupture lengths would be between 28 and 205 km corresponding to magnitudes of M6.8 and M7.7. Using the relation we estimate rupture lengths between 75 and 830 km corresponding to magnitudes of M7.2 and M8.8. Only the 28 km rupture length estimation using the minimum average slip of 1.7 m and the scaling relation falls within the range of what we observe in the field (10 to 50 km). We therefore exclude any estimates of magnitude using the relation. We also assume that if the total summed slip is greater than 5 m (three ∼ 1.7 m earthquakes), it must represent earthquakes we do not observe in the fault exposure, or fault creep. These earthquakes may be older than the oldest sediments offset in the exposures (8.54 to 8.39 ka), which is likely given the long time between deglaciation and the oldest event. The unaccounted for earthquakes could also have ruptured other fault segments (e.g., just the southern branch), not ruptured to surface, or have been smaller and not observable. In conclusion, based on the assumption of ∼ 1.7 m of horizontal slip per earthquake, or 10.5 to 50 km of rupture length, and using the scaling relation, we estimate that the paleo-earthquake magnitudes were likely between M6.3 and M7.0 (for either a single fault strand rupture or a coeval rupture of both strands).

We note there may be additional uncertainty resulting from shallow earthquake depths in the study area. The 2014 M5.6 and the 2022 M5.6 earthquake both have depths <5 km, and therefore may have a lesser rupture area for a given rupture length. The smaller rupture area would result in a lower magnitude and the magnitudes we determined above may be overestimated.

Our estimated paleo-earthquake magnitudes overlap with the proposed magnitude of the 15 August 1868 M6.4 to M6.8 earthquake that was determined to have a likely epicenter between El Angel and Chiles-Cerro Negro and close to the Reservoir, Polylepis, and 25 July faults Fig. a,. We do not see evidence for a large earthquake as recent as 1868 in the exposure of the northern branch of the Reservoir fault, but this earthquake may have ruptured the southern branch of the Reservoir fault, or another structure in the area such as the Polylepis fault. The clear surface expression and recent landslides along the Polylepis fault (Fig. ) leads us to speculate that it may have been the host of the 1868 earthquake, but detailed paleoseismic studies of this fault are needed to test this hypothesis.