Intensification of ENSO frequency drives forest disturbance in the Andes during the Holocene

The biodiverse montane forests of the tropical Andes are today frequently disturbed by rainfall-driven mass movements which occur mostly during extreme El Ni~ no events. Over the coming decades these events are projected to double under the 1.5 C global warming scenario. The consequent increased rainfall and mass movement events likely present an elevated risk to millions of people living in the Andes. However, the impact of more frequent rainfall extremes remains unclear due to a lack of studies that directly link past changes in El Ni~ no-Southern Oscillation (ENSO) frequency to forest and landscape disturbance patterns. Here, we present the first Holocene palaeoecological record from Laguna Pallcacocha, southern Ecuador, a key site for El Ni~ no reconstructions. We demonstrate that for the past 10,000 years plant taxa indicative of recolonization e such as Alnus acuminata e covary with El Ni~ no-induced flood layers in the lake. An amplified forest disturbance pattern is observed in the late Holocene, suggesting enhanced slope instability following deforestation. The temporal pattern is not explained by tree line fluctuations or human impact, while the latter does amplify the impact of ENSO on landscape disturbance. Spatial correlations between modern ENSO and precipitation are consistent with a regional comparison of Holocene records of landscape disturbance. Our results indicate that climate extremes, such as those associated with future intensification of El Ni~ no, combined with ongoing land use change will increase the frequency of mass movements elevating risks for millions of people in the Andes. © 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
The montane forests of the western Tropical Andes are a biodiversity hotspot and contain some of the most threatened habitats on the planet. Exceptional levels of plant endemism (Myers et al., 2000;P erez-Escobar et al., 2022) combined with narrow altitudinal habitat ranges make these forests sensitive to disturbance and climate change (Foster, 2001). Small-scale yet severe mass movements, that include landslides and debris flows, occur frequently on steep slopes in tropical montane forests (Crausbay and Martin, 2016;Muenchow et al., 2012;Richter, 2009). Triggered by intense rainfall (Clark et al., 2016) and occasional earthquakes (Crausbay and Martin, 2016), mass movements create gaps in the forest canopy and remove nutrient-rich soils (Crausbay and Martin, 2016). As a result, mass movements provide niches for rare and early-successional species (Crausbay and Martin, 2016) and contribute to vegetation heterogeneity and species diversity in Andean forests (Crausbay and Martin, 2016;Ohl and Bussmann, 2004;Richter, 2009). Landslides and debris flows are also a pervasive geohazard with 52% of Andean people at risk (Comunidad Andina, 2009), and their occurrence may increase in likelihood under climate warming.
Rainfall variability in the western Tropical Andes is influenced by the El Niño-Southern Oscillation (ENSO) (Vuille et al., 2000). Every 2e7 years El Niño events amplify atmospheric convection causing regional warming and intense rainfall events along tropical western Andean slopes (Kiefer and Karamperidou, 2019;McPhaden et al., 2006;Vuille et al., 2000). High-resolution meteorology shows that rainfall events that exceed landslide threshold levels are primarily derived from the Pacific (Hagemans et al., 2021) and occur during Eastern Pacific and Coastal Pacific El Niño events (Kiefer and Karamperidou, 2019) (Fig. S3). Twenty first century climate projections estimate doubling of extreme El Niño frequency at 1.5 C climate warming and continued increase in occurrence after climate warming stabilization (Wang et al., 2017). The IPCC Sixth Assessment Report (AR6) reaffirmed the projected ENSO-tied rainfall intensity (IPCC et al., 2021), although for the projected sea surface temperature (SST) change that would drive El Niño rainfall variability the AR6 reports no model consensus. In a critical analysis of this conclusion Cai et al. (2022) demonstrate, using the Coupled Model Intercomparison Project (CIMP6) projections, a robust 20th to 21st century SST increase that sustains the earlier projections of Wang et al. (2017) with doubling of future extreme El Niño events and associated increased rainfall variability.
Already, the impact of climate warming on Andean ecosystems is evident in the upslope migration of tree species in Peru (Feeley et al., 2011) and compositional shifts and range retractions in Colombia (Duque et al., 2015). However, to what extent a warminginduced intensification in El Niño frequency will impact the (in) stability of tropical Andean slopes and composition of forests remains unknown. Monitoring of ENSO over recent decades provides detailed insights into system function under the current climate configuration, but is of insufficient length to document forest (in) stability over timescales that capture the projected future ENSO frequency change. Studying past forest (in)stability, when El Niño frequency was similar or higher than present day, can shed light on the long-term (in)stability of Andean forests for the future.
Laguna Pallcacocha (2 46 0 S, 79 14 0 W; 4050 m a.s.l.; El Cajas National Park [ECNP]) yielded the important terrestrial record of El Niño-induced rainfall variability during the past 10,000 years (Moy et al., 2002;Rodbell et al., 1999) (Fig. 1). Recent work demonstrated that Pacific-sourced rainstorm events, associated with coastal El Niños, trigger mass movements in the Pallcacocha catchment. These mass movements deposit clastic layers in the lake at frequencies that fall into the typical El Niño frequency band (Hagemans et al., 2021). This new insight answers recent questions (Schneider et al., 2018) regarding the original interpretation of the drivers of this laminae record (Moy et al., 2002;Rodbell et al., 1999). The Holocene laminae frequency pattern shows a gradual increase of El Niño events, superimposed by millennial-scale variations (Moy et al., 2002;Rodbell et al., 1999) (Fig. 2). The change from lower El Niño frequency in the early and mid-Holocene (Moy et al., 2002;Rodbell et al., 1999) to higher and more variable El Niño frequencies in the late Holocene is recognized in marine and terrestrial records globally (e.g. Donders et al., 2008;Thompson et al., 2017;White et al., 2018). The importance of original Pallcacocha laminae record warrants further investigation on the reginal consistency and consequences of the ENSO dynamics in the Andes. Here, we apply pollen and microcharcoal analyses on the organic fraction of the Pallcacocha sediments to reconstruct Holocene vegetation history and derive forest disturbance patterns and couple them to the inorganic laminae deposition evolution. Modern pollen samples from ECNP demonstrate that pollen from montane forests in a~40 km radius around Pallcacocha are recorded in the lake sediments (Hagemans et al., 2019). Based on these calibrations and ecological observations (Hagemans et al., 2019;Paolini et al., 2005), we use characteristic pollen types to determine changes in upper forest line, post-disturbance recolonization, p aramo (alpine Andean grasslands) and anthropogenic activity on decadal to centennial resolution. Our palaeoecological reconstruction couples regional-scale forest disturbance and (in)stability of the tropical Andean landscape via the iconic paleo-El Niño laminae record from Pallcacocha to El Niño-induced rainfall events.

Material and methods
A 9-m composite sediment core from Laguna Pallcacocha, retrieved in 1999 (Moy et al., 2002), was sampled at 5-cm intervals for palynological analyses and micro-charcoal processing. The 0.3 cc samples were spiked with 4 mL Lycopodium clavatum solution (i.e.,~8339 spores) and processed following standard palynological protocols (Faegri and Iversen, 1989). The sediment samples were floated over sodium polytungstate (density ¼ 2.0) to remove minerals. Pollen was counted to a minimum of 300 terrestrial pollen grains per sample with a Leica DM2500 light microscope at 400Â magnification. Pollen types were identified to the lowest taxonomic level possible following descriptions by Hooghiemstra (1984), and by comparison with the reference collections at the University of Amsterdam and Utrecht University. A. acuminata pollen, non-pollen palynomorphs (NPPs) and spores were excluded from the pollen sum shown in Fig. 2, but A. acuminata is included in the sum for the regional comparison (Fig. 4). Fungal NPPs were identified following descriptions by van Geel et al. (2011). Microcharcoal particle number and surface area and L. clavatum occurrences were recorded using digital image analysis in ImageJ (Schneider et al., 2012). Slides were digitized microscopic at 200Â optical magnification using a Leica DM6000 B automated-stage microscope. Imaging was carried out with extended focus using at 4 z-levels over 16 mm vertical distance with the Leica LAS 4.13 software equipped with Power Mosaic plus. Particle detection of the resulting composite images in ImageJ was set with a minimum threshold of 10 mm 2 and color detection ranges to; hue 90e180, brightness 0e50 and saturation 0e255. Particle circularity was set between 0.0 and 0.5. Estimates of total micro-charcoal and pollen influx were based on Lycopodium clavatum added per 1 cc of sediment. Statistically significant pollen assemblage zones were established using the CONISS (Constrained Incremental Sum of Squares) option with square-root transformation in Tilia (Grimm, 1987). A 14 C and tephra-based chronology for the core was previously established and published (Moy et al., 2002;Rodbell et al., 2002Rodbell et al., , 1999 which for reasons of consistency with the laminae frequency record we used unaltered. Regional pollen data ( Fig. 4) represent available Holocene upland records in the Neotoma Paleoecology Database (http://www.neotomadb.org) (Williams et al., 2018). Probability density functions of archaeological artefacts were based on the South-American database of archaeological 14 C dates and human demography (Goldberg et al., 2016) and calculated with the statistical package Bchron (Haslett and Parnell, 2008) in R. To estimate mass movement density under varying El Niño regimes we first defined 5 classes: 0e5, 5e10, 10e15, 15e20 and > 20 El Niño events per 100 years as recorded in the sediments from L. Pallcacocha. We then calculated the average pollen influx of A. acuminata (grains cm À3 yr À1 ) for those classes (Fig. 5). Modern day analyses in RBSF indicate a mass movement density of 19 slides km À2 per 50 years (Muenchow et al., 2012), while modern pollen samples demonstrate that in the past 50 years the average pollen influx of A. acuminata is~1088 grains cm 2 yr À1 . Projection of this data on the relation between A. acuminata influx and El Niño frequencies from Pallcacocha gives a first order estimate of mass movement density over the Holocene. All maps presented were configured with ArcMap software.

Results
The early Holocene (~10,000e7000 calibrated years before present, cal. yr BP) section of the L. Pallcacocha pollen record is characterized by assemblages indicative of moist montane forest (Fig. 2). During this period Podocarpus spp. and Moraceae reached their maximum abundance, 25% and 7% respectively. The Alnus acuminata abundance initially was relatively low and invariable. They are derived from the montane forest at lower elevation and therefore A. acuminata abundances in Fig. 2 were calculated outside the pollen percentage sum (see section 4.1). Alternatively, pollen influx data e based on counts of exotic marker grains, known sample volumes and sediment accumulation rates (grains cm 3 yr À1 ) (Stockmarr, 1971) e are unaffected by closed-sum effects and better highlight phases of enhanced pollen deposition in this setting. Low and stable influx and relative abundance values of A. acuminata pollen during the early Holocene are notable given the overall high abundance of other arboreal taxa.
The mid-Holocene (7000-4500 cal yr BP) was marked by a decrease in Podocarpus spp. to 9% and an increase in both A. acuminata influx and relative abundances, concurrent with enhanced frequencies of clastic laminae deposition (Fig. 2). Pollen of Amaranthaceae were present in very low numbers (Fig. 2) and, as expected at this elevation, the record shows little indication for extensive local on-site cultivation (e.g., Zea mays), extensive fire, or high density continuous human occupation (Goldberg et al., 2016). This suggest that the local vegetation record has little anthropogenic bias, but at a supra-regional scale humans were likely active given the consistent increases in the archaeological settlement density after 3500 cal yr BP in Ecuador (Fig. 2) (Goldberg et al., 2016).
The mid-to late Holocene (4500 -2100 cal yr BP) showed a further, sharp decline in Podocarpus spp. accompanied with an increase of local p aramo taxa (Fig. S1). Influx of A. acuminata pollen decreased between~4500 and 4000 cal yr BP, when recorded El Niño events in the flood layer record of the lake declined in frequency. Micro-charcoal became abundant~3600 cal yr BP, but declined again shortly after and remained low until~120 cal yr BP The late Holocene zone (2100 -400 cal yr BP) exhibited a highly similar pattern between high influx of A. acuminata pollen and  (Moy et al., 2002;Rodbell et al., 1999). Alnus acuminata is a pioneering taxon in montane forests and indicative for slope destabilization and forest disturbance. Pollen abundance of up to 25% for Podocarpus spp. and 7% for Moraceae indicate local montane forest near L. Pallcacocha in the early Holocene. Poaceae and Asteraceae represent the abundance of grass-p aramo vegetation which increases in the late Holocene. Polylepis present the subp aramo forest-grassland ecotone. Probability density functions (PDF) are based on archaeological 14 C dates (n ¼ 27 in a 60-km radius around L. Pallcacocha, pink curve; all Ecuadorian ages, black line, n ¼ 545) from South American human demography reconstructions (Goldberg et al., 2016). Blue symbols represent the presence of Zea mays pollen grains (<1%) in the Pallcacocha sediment record.
recorded El Niño events in the flood layer record of the lake; both reached maximum values at~1300 and~800 cal yr BP. During this period, a decline in Polylepis spp., the first occurrence of Z. mays 1500 cal yr BP and expansion of human settlement occurred in the region (Fig. S2). From 400 cal yr BP to modern times reduced influx of A. acuminata pollen mirrored the decline in laminae deposition in the lake, while Poaceae and Asteraceae expanded. The total pollen accumulation rate mirrors the high Alnus influx phases. Except for the 3500 -2500 cal yr BP period where the laminae record shows that ENSO frequencies are at intermediate levels other upper forest taxa (mostly Myrica, Podocarpus, Polylepis) show a higher influx relative to Alnus.

El Niño-driven forest disturbance
The conspicuous variability of A. acuminata at L. Pallcacocha indicates highly variable forest dynamics during the Holocene (Fig. 2) instead of upslope Andean forest migration. A. acuminata is a pioneer species in montane forests~1500e3400 m asl and is common in marshy areas, river beds and on post-mass movement soils (Marchant et al., 2002;Paolini et al., 2005;Weng et al., 2004a). The expansion of A. acuminata in the Holocene has previously been linked to an upslope migration of the montane forest in response to temperature increase (Weng et al., 2004a). However, such interpretations are generally based on surface calibrations derived from moss polsters (Olivera et al., 2009;Urrego et al., 2011b;Weng et al., 2004b), soil samples (Reese and Liu, 2005;Weng et al., 2004b) or pollen traps (Niemann et al., 2010;Olivera et al., 2009), while reconstructions of past vegetation and climate dynamics are generally based on pollen records from lake sediments (Colinvaux et al., 1997;Hansen et al., 2003;Schiferl et al., 2018;Urrego et al., 2005). Pollen rain in moss polsters and soil samples is representative of local vegetation present within a few meters, while lake sediment samples from p aramo lakes such as L. Pallcacocha represent regional vegetation of up to 40 km distance (Hagemans et al., 2019). These surface calibrations for Pallcacocha and neighbouring lakes have shown that A. acuminata is overrepresented in pollen records from p aramo lakes due to upslope wind-transport and prolific pollen production by Alnus, and therefore does not represent local vegetation around the lake. However, the same surface calibrations show that Podocarpus spp. is a (more) sensitive proxy for past forest distribution than A. acuminata in this setting. In the Tropical Andes, Podocarpus spp. are large trees that form stands in the montane forests (Marchant et al., 2002) and can reach pollen percentages above 25% in lakes surrounded by montane forest in ECNP (Hagemans et al., 2019). Such pollen percentages for Podocarpus spp. are reached in Pallcacocha in the early Holocene, indicating that the montane forest reached its highest distribution at this time, which is consistent with other terrestrial pollen records from the region (Hansen et al., 2003;Schiferl et al., 2018;Nascimento et al., 2020). Given the continued presence of local p aramo indicators like Phlegmariurus and Plantago (Fig. 2), the upper montane forest most likely did not extend above the lake elevation at 4050 m. Deltaic deposits in front of the lake (reported in Hagemans et al., 2021) contain no visible woody macrofossils or changes in peat type suggesting that, despite a higher tree line, there was no local forest presence in the Early Holocene around Pallcacocha. Such catchment changes could be significant for the likelihood of clastic laminae deposition in the lake basin, yet we do not see any evidence for this.
By the mid-Holocene (~5000 cal yr BP) Podocarpus spp. decreased to~2% in L. Pallcacocha, which is comparable to the modern day pollen abundance in the lake (Hagemans et al., 2019) and indicates a downslope migration of the upper forest line to its modern day position of 3500 m asl. Increased microcharcoal abundance regionally indicates more frequent burning of the vegetation in the region at this time, which could be sign of local settlement, and likely contributed to a downslope migration of the forest (Urrego et al., 2011a). A. acuminata reached maximum values in the late Holocene well after the mid-Holocene downslope migration of the upper forest limit. The high abundances and influx of A. acuminata in the late Holocene are therefore unrelated to a temperature driven migration of the montane forest belt and must represent another aspect of Andean forest dynamics.
We propose a new interpretation for the conspicuous variability of A. acuminata in Pallcacocha during the Holocene, based on the regional growth pattern of A. acuminata related to disturbed soils and its modern pollen distribution in ECNP (Hagemans et al., 2019). Rainfall-driven landslides play a central role in the dynamics of tropical montane forests and contribute to the maintenance of vegetation heterogeneity and species richness (Crausbay and Martin, 2016;Richter, 2009). By creating gaps in the forest canopy, landslides provide niches for the recruitment and persistence of rare and early-successional species (Crausbay and Martin, 2016). Nitrogen-fixing cryptogams are usually first to colonize the nutrient poor post-landslide soils. Mosses and lichens provide a basis for the establishment of Gleicheniaceae, Asteraceae and Poaceae (Lozano et al., 2008;Richter, 2009). Alnus acuminata is one of the first tree species to abundantly colonize post-landslides soils (Paolini et al., 2005), because of its nitrogen-fixing capabilities and light tolerance. Melastomataceae and Piperaceae characterize secondary shrub succession and are followed by tree taxa such as Rubiaceae, Lauraceae and Myrtaceae (Richter, 2009), but these taxa produce less pollen which are less taxa specific. Its abundant pollen production and regional source area is well recorded in the open landscape (Hagemans et al., 2019), which render it a primary and extra-local indicator of vegetation regeneration after episodes of slope instability in Andean sedimentary records (Brunsch€ on et al., 2010) (Fig. 5). Recent records of sedimentation and highresolution rainfall records from ECNP provide a mechanistic link between clastic sedimentation from alluvial activity in the catchment of Pallcacocha and intensive rainfall events originating from the Pacific, such as those during coastal El Niños (Hagemans et al., 2021). We submit that regional-scale landscape disturbance is reflected in the variability of A. acuminata at Pallcacocha and is the result of increased El Niño-driven mass movements. In that view, the conspicuous variability of A. acuminata at Pallcacocha indicates severe and highly variable forest disturbance during the late Holocene (Fig. 2). Our surface data indicate that the influx of A. acuminata pollen is independent of other, local vegetation elements in ECNP and can be used to document past relative levels of forest disturbance related to mass movements. The synchrony between the A. acuminata pollen influx pattern and the frequency of laminae deposition (Fig. 2) thus indicates that the dynamics of local sediment deposition in Pallcacocha parallels the developments of regional scale disturbance patterns (Fig. 4).
Comparison of the observed dynamics in pollen abundance of A. acuminata in L. Pallcacocha to other pollen records across the Tropical Andes shows that a highly similar, largely bi-modal Holocene pattern of Alnus abundance is recorded some areas (Fig. 4) that, based on spatial patterns in reanalysis data, presently experience a clear climatic impact from ENSO (Fig. 3). This pattern of bimodal Holocene expansion and retraction in A. acuminata abundance from Pallcacocha is evident in various terrestrial pollen records across the Ecuadorian Tropical Andes (Weng et al., 2004a), such as nearby lakes Laguna Llaviucu (Colinvaux et al., 1997;Nascimento et al., 2020), Laguna Chorerras (Hansen et al., 2003) and a marine record from the Pacific Guayaquil basin (Seill es et al., 2015), with a more pronounced signal in high-elevation sites (Fig. 4, Table 1). In contrast, sheltered sites situated in the inter-Andean valley (Niemann and Behling, 2008;Brunsch€ on and Behling, 2009;Villota and Behling, 2014) or further north (Graf, 1992;Gonz alez-Carranza et al., 2012, Table 1, Figs. 3 and 4) show overall much lower A. acuminata abundances and exhibit a clearly different temporal pattern with no late Holocene increase. All these sites presently show a weak or inverse correlation to ENSO-induced rainfall (Fig. 3). The Holocene Alnus patterns in the ENSO sensitive regions, while highly similar, are not identical and temporal offsets are likely related to local threshold differences in Alnus regrowth with regard to disturbance and variations in sediment accumulation rates between sites. Differences in radiocarbon calibration are unlikely to explain the site offsets as significant changes between successive radiocarbon calibration curves (Stuiver et al., 1998;Reimer et al., 2009Reimer et al., , 2013 are mostly in the pre-Holocene sections. The consistency between regional records in areas with a modern-day prominent ENSO impact provides further support for a regional factor driving A. acuminata populations in the Tropical Andes. Mass movement occurrences in the Andes have been linked to El Niño (Comunidad Andina, 2009;Keefer et al., 2003;Moreiras, 2005;Sepúlveda et al., 2006) and in the ECNP intense rainfall events are mostly triggered by coastal El Niño events (Kiefer and Karamperidou, 2019). We infer that as rainfall anomalies triggered by Eastern Pacific and Coastal Pacific El Niño events occurred more frequently in the late Holocene, it triggered mass movement activity and subsequently provided opportunities for A. acuminata to expand.
The disturbance pattern is amplified in the late Holocene as people become more active in the landscape (Fig. 2), as proposed earlier by Nascimento et al. (2020). Probability density functions (PDF) based on archaeological 14 C dates and reconstructions of human demography within~40 km radius (i.e. the pollen source area) of Pallcacocha indicate increases in human activity during the late Holocene (Fig. 2). Pollen records from the lower elevation Laguna Llaviucu (3150 m asl) in ECNP match the timing of Fig. 3. High elevation sites included in the regional analysis (details see Table 1). Overlay shows correlation between NovembereJanuary averaged precipitation and the NINO 1.2 index 1979e2019 for Ecuador and Southern Colombia. Correlation obtained with KNMI Climate Explorer (https://climexp.knmi.nl/start.cgi) and adjusted in ArcGis (Environmental Systems Research Institute, 2014).
intensification of the laminae deposition (Fig. 4), suggesting that the adoption of agricultural and pastoralist activities amplified existing ENSO signals through the increase of landscape sensitivity in the late Holocene (Nascimento et al., 2020). While the Pallcacocha data do not show clear indications for agriculture intensification (e.g., low Zea mays, Amaranthaceae) during the observed increases in late Holocene laminae and A. acuminata, more regional-scale information on the temporal patterns of habitation is required. The central Ecuadorian Azuay and Cañar provinces, in which the ECNP is located, were inhabited by the Cañari people before the Incan conquest of the region (~500 cal yr BP). The Cañari were primarily agriculturalists maintaining an intensive, irrigation-based economy in the warm low elevation valleys producing cotton and coca (Bray and Echeverría, 2018). They lived in small hamlets or individual homesteads dispersed across the landscape (Bray and Echeverría, 2018). In the ECNP and its surroundings 27 dated archaeological sites have been registered that indicate the settlement of people in the area between 3500 cal yr BP. and 500 cal yr BP (Ministerio del Ambiente del Ecuador, 2018) (Fig. 2). In Pallcacocha, the key Andean crop Z. mays first appeared in the record at 1800 cal yr BP. and remained consistently present in low abundances for 600 years until it disappeared at 1200 cal yr BP. (Fig. 2). Other anthropogenic indicators such as Amaranthaceae also remained low between 1800 cal yr BP and 1200 cal yr BP, while Fig. 4. Holocene dynamics in Alnus from L. Pallcacocha compared to available high elevation records from Ecuador and southern Colombia, as well as marine core M772056-2 from the Bay of Guayaquil. The red color intensity and derived El Niño event frequency are based on the same sediment core and chronology from L. Pallcacocha (Moy et al., 2002;Rodbell et al., 1999). All A. acuminata percentages are expressed relative to the total sum of upland herbs, shrubs and trees. See Table 1 and Fig. 3 for site details and locations relative to current ENSO climate influence. The L. Chorreras chronology has been revised based on the tephra ages from ECNP . K. Hagemans, D.H. Urrego, W.D. Gosling et al. Quaternary Science Reviews 294 (2022) 6 macro-charcoal was absent in the sediments, suggesting that the watershed of Pallcacocha was not intensively used by people, in contrast to the archaeological site density across Ecuador after 3000 cal yr BP (Fig. 2) and in broader areas in the eastern Andean flank (Sales et al., 2022). Z. mays pollen represent a predominantly local signal (Hofmann et al., 2014) and it is therefore likely that people attempted agriculture in the vicinity of Pallcacocha, but remained low intensity. Local intensive agriculture by the Cañari in the watershed would have resulted in high proportions of Caryophyllaceae, Amaranthaceae, Thalictrum sp., Poaceae, Z. mays and charcoal such as observed in the Lake Huila record in the northern Ecuadorian Andes (Loughlin et al., 2018). Nevertheless, people were active in the region which is revealed through the presence of Z. mays in Pallcacocha, Laguna Chorreras (~8 km), Laguna Llaviucu (~12 km) and archaeological records within the ECNP (Ministerio del Ambiente del Ecuador, 2018) and the surrounding region ( Fig. S2) (Goldberg et al., 2016). As agricultural activity in the region began ca. 1800 cal yr BP., the synchrony between A. acuminata influx and El Niño-induced precipitation events became more evident in the record. Early agricultural activities seem to have amplified the impact of El Niño on the landscape at levels comparable to, but not as strong as, human amplification of the El Niño signal in Lake Sauce in the western Amazonian lowlands  and Laguna Llaviucu in the ECNP (Nascimento et al., 2020).

Implications
Mass movements are a pervasive geohazard in the Andes (Fig. 5). Between 1970 and 2007 C E approximately 11,000 people were killed and 38,000 homes were destroyed by debris flows and landslides in the Andes, with peaks in occurrences in Ecuador and Peru during El Niño years (Comunidad Andina, 2009). In Ecuador, 150 casualties were linked to a single mass movement event during the 1983e83 El Niño alone (Schuster et al., 2002). To provide a first order estimate of predicted mass movement occurrence with a doubling in El Niño frequency under a 1.5 C warming scenario (Wang et al., 2017), we compared modern pollen influx of A. acuminata and mass movements in the Ecuadorian Andes between 1969 and 2000 C E. Approximately 34% of the ECNP area is susceptible to mass movements such as debris flows and landslides (Fig. 5) (Ministerio del Ambiente del Ecuador, 2018), but mass movement density and activity in the ECNP area is yet unknown. We therefore used data from the nearest reference site Reserva Biol ogica San Francisco (RBSF), 120 km south of ECNP, situated in the Ecuadorian Andes. In RBSF, a landslide/debris flow density of 12 slides km À2 was reported between 1969 and 2000 C E (Muenchow et al., 2012). Additional pollen data from sediments deposited between 1969 and 2000 C E in Laguna Pallcacocha (Hagemans et al., 2021) show an average pollen influx of~1088 grains cm 2 yr À1 for A. acuminata 1969e2000 C E (Fig. S3). Projection of this data on the historical relation between binned A. acuminata influx and El Niño frequencies from L. Pallcacocha (Fig. 6) show that a potential future El Niño frequency of~20 moderate to extreme events per 100 years compares to a 2e4 time higher levels of landscape disturbance as indicated by the A. acuminata influx rates. This projection does not take into account potential non-linearities in this relation and extension to occurrences of mass movements is therefore not yet possible with these data. Human activities in forests can further exacerbate mass movement hazards. Deforestation and conversion to pastures results in permanent reduced slope stability (Guns and Vanacker, 2013), while mass movement susceptibility significantly increases by one order of magnitude near highways (Brenning et al., 2015). Our study shows that the interaction of climatic extremes such as those associated with future El Niño intensification (Wang et al., 2017;Cai et al., 2022) and ongoing land use change (Guns and Vanacker, 2013) is likely to increase the frequency of mass movements. These aspects require further study and should consider the temporal and spatial change of human disturbance patterns and deforestation in the Andes. These combined threats not only threatens infrastructure, ecosystems and economics, but also poses Table 1 Site details for regional comparison of pollen records in Fig. 4 Site  Stuiver et al., 1998. d Weninger et al., (2008. e Reimer et al., (2009) .   Fig. 6. Binned A. acuminata influx (grains cm À2 yr À1 ) versus the number of El Niño events per 100 years based on the Laguna Pallcacocha laminae record (Moy et al., 2002;Rodbell et al., 1999) for 0e5, 5e10, 10e15, 15e20 and > 20 El Niño events per 100 years. Hatched bar represents the influx of A. acuminata pollen in the past century ( Supplementary Fig. S3) versus the frequency of modern-day strong El Niño events.
an increased risk to the millions of people living in the Andes (Schoolmeester et al., 2016).

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
i have matached my data in the supplement