High Impact Weather Events in the Andes

Owing to the extraordinary latitudinal extent, a strong orographic variability with very high mountain tops, and the presence of deep valleys and steep slopes, the Andes and the population of the region are highly prone and vulnerable to the impacts of a large suite of extreme weather events. Here we provide a review of the most salient events in terms of losses of human and animal lives, economic and monetary losses in costs and damages, and social disruption, namely: (1) extreme precipitation events and related processes (Mesoscale Convective Systems, lightning), (2) cold spells, frosts, and high winds, (3) the impacts of ENSO on extreme hydro-meteorological events, (4) floods, (5) landslides, mudslides, avalanches, and (6) droughts, heat waves and fires. For our purposes, we focus this review on three distinctive regions along the Andes: Northern tropical (north of 8°S), Southern tropical (8°S-27°S) and Extratropical Andes (south of 27°S). Research gaps are also identified and discussed at the end of this review. It is very likely that climate change will increase the vulnerability of the millions of inhabitants of the Andes, impacting their livelihoods and the sustainable development of the region into the twenty first century amidst urbanization, deforestation, air, soil and water pollution, and land use changes.


INTRODUCTION
Rapid rates of human encroachment, urbanization, deforestation and land use changes are taking place along the Andes (Figure 1). These developments make the region and their populations highly vulnerable to extreme weather events and impose a huge toll in terms of socioeconomic, environmental and ecological impacts. Such high vulnerability is associated with diverse natural hazards and risks, poor (or non-existent) preventive and adaptive capabilities and fragile governance. Figures 2A,B illustrate the impacts of weather and hydrological extreme events over the Andes. Floods and cold spells lead the rank in human lives losses in the region. Between 1985 and 2014 occurred 150 disasters in the Andes triggered by these extreme events, causing economic losses of 3,138.4 million of US dollars, killing 6,664 people and affecting more than 12 million people (Stäubli et al., 2017). It is highly likely that climate change will increase such vulnerability throughout the Andes (Magrin et al., 2014). FIGURE 1 | Location and population of cities and towns along the Andes, whose populations are greater that 10,000 inhabitants. Data source: World Cities Database (https://simplemaps.com/data/world-cities).
This paper reviews high impact weather events affecting the livelihoods of millions of people, and compromise the sustainable development of the Andean region amid climate change, urbanization, deforestation, and land use changes. It is a contribution in a series of six to this special issue of Frontiers in Earth Science focused on the Andes climate, weather, hydrology and cryosphere (Condom et al., 2020;Espinoza et al., 2020;Masiokas et al., 2020;Pabón-Caicedo et al., 2020), arising from ANDEX, a prospective Regional Hydroclimate Project (RHP) of the Global Energy and Water Cycle Experiment (GEWEX; https://www.gewex.org/). For our purposes, the Andean region is divided into three distinctive regions: Northern Tropical (north of 8 • S), Southern Tropical (27 • S-8 • S) and Extratropical Andes (south of 27 • S). The work is distributed as follows. Section Extreme Precipitation and Related Processes is focused on extreme precipitation and storm events and related processes, including mesoscale convective systems and lightning. Section Cold Spells, Frosts, and High Winds deals with cold spells, frosts, and high winds, while section ENSO Impacts on Extreme Events discusses the impacts of ENSO on extreme hydro-meteorological events. Section Floods is related to floods, while section Landslides, Mudslides, Avalanches with landslides, mudslides, and avalanches, and section Droughts, Heat Waves and Fires reviews droughts, heat waves and fires. Section Research Gaps provides a summary of main research gaps regarding the monitoring, understanding, modeling, and prediction of the studied processes and phenomena, and some final remarks are pointed out in section Final Remarks.

EXTREME PRECIPITATION AND RELATED PROCESSES Northern Tropical Andes
Very localized intense storms occur in the tropical Andes during the (annual or semi-annual) wet seasons, with a diurnal cycle shifting with the seasons (Poveda et al., 2005;Bedoya-Soto et al., 2019). Over the Andes of Colombia and Venezuela, most intense storms occur during the two maxima of the annual cycle (April-May and September-November) . The most salient feature of intense storms is a strong spatiotemporal variability, which is caused by steep orography and topographic roughness, local circulations, small and large-scale convective cloud systems acting at different timescales, local circulations and regional valley/mountain breeze systems, katabatic winds (López and Howell, 1967;Bendix et al., 2009), evapotranspiration from the bottom of intra-Andean valleys, orographically-driven moisture transport (Rollenbeck and Bendix, 2011), and anabatic circulations (Bedoya- Soto et al., 2019), which in turn are influenced by synoptic mesoscale and continental conditions. Over the eastern slopes of the Ecuadorian Andes, nocturnal convection and clouds develop in spite of unfavorable largescale conditions (Bendix et al., 2009;Trachte et al., 2010). Precipitation recycling in the tropical Andes can be 70-90% of total precipitation (Zemp et al., 2014;Bedoya-Soto et al., 2019). The strong variability of tropical Andean storms is evidenced in their fractal and multifractal properties (Poveda and Mejía, 2004;Gómez and Poveda, 2008;Hurtado and Poveda, 2009;Poveda, 2011;Mesa and Peñaranda, 2015;Posadas et al., 2015;Poveda and Salas, 2015;Yarleque et al., 2016;Duffaut-Espinosa et al., 2017).
Intense storms over the tropical Andes are also associated with Mesoscale Convective Systems (MCS) over western Colombia and the far eastern tropical Pacific (Velasco and Fritsch, 1987;Houze et al., 2015;Zuluaga and Houze, 2015). These large and intense convective systems are seen to contribute to explain the existence of one of the rainiest spots on Earth over the Pacific coast of Colombia with average rain rates of 10,000-13,000 mm per year (Poveda and Mesa, 1999;Poveda et al., 2014;Yepes et al., 2019). Over Colombia's far eastern Pacific, the cross-equatorial flow become the south-westerly CHOCO LLJ that interacts with the western Andes (Poveda and Mesa, 2000;Bedoya-Soto et al., 2019;Yepes et al., 2019), exhibiting weaker winds in February-March and stronger winds during September-November (Figure 3). Such enhanced ocean-atmosphere-land interactions by the CHOCO LLJ produce world-record breaking rainfall intensities over western Colombia inland and off-shore (Poveda andMesa, 1999, 2000;Yepes et al., 2019). Mostly focused on the Caribbean and the Orinoco LLJs, easterly trade winds prevail, and their interaction with the ITCZ produce large amounts of rainfall in the Orinoco basin, and the Upper Amazon river basins in Colombia, Venezuela and Brazil (Figure 3). Figure 4 shows the distribution of average (1998-2013) rainfall over most South America during JJA and DJF, using data from the Tropical Rainfall Measuring Mission (TRMM). Most of this rain is produced by different types of convective organization at the regional scale , which are often part of large MCSs at different stages. Such types can be classified using reflectivity echoes from the TRMM 3B42 product (Kummerow et al., 1998(Kummerow et al., , 2000Houze et al., 2007Houze et al., , 2015. Figure 5A shows the probability of TRMM Precipitation Radar (PR) observing a wide convective core (WCC), with pixels exceeding 40 dBZ reflectivity over 1,000 km 2 during JJA . The WWC tend to maximize in the Pacific coast of Colombia and Panama, in the northern region of the Andes of Colombia, as well as in the Orinoco and parts of the Upper Amazon river basin. Figures S1, S2 show the MCS that triggered the flooding of Mocoa, Colombia, in April 1st, 2017, killing at least 336 people, injuring 332, and 70 others missing.
Rainfall over north-western South America exhibits different precipitation features (PF) containing elements associated with their convective life cycle, such as those with and without ice scattering and MCSs. The former two maximize in the afternoon from 12:00 to 18:00 LT at the west of the western Colombian Andes, and the eastern slope of the eastern Andes. The former occurs by the effect of Pacific sea breeze along the valley floor and upslope to the top of the Andes in the afternoon (López and Howell, 1967;Yepes et al., 2019), but also from the Orinoco LLJ during the afternoon. On the other hand, PFs with MCSs mainly occur in the morning (00:00-06:00 LT) offshore in the Pacific at 77.5 • W (Mapes et al., 2003;Yepes et al., 2019). Large precipitation features (PFs) associated with MCSs represent <1% of PFs in the eastern Amazonian slopes of the Andes (Jaramillo et al., 2017). The most common PFs exhibit small areas and contribute about 15% to total precipitation, while MCS are observed least but produce more than 50% of the total along the eastern slope of the Central Andes below 2000 m a.s.l. (Chávez and Takahashi, 2017).
Some mesoscale convective units may become broad areas of stratiform rain (BSR; TRMM echoes covering more than 50,000 km 2 ) at the later stage of MCS (Houze, 2004) (Figure 5B for JJA). They are concentrated over the eastern Pacific near Colombia, and less in the Orinoco and Upper Amazon basin. Furthermore, very deep and intense convective cores (DCC) develop over continental regions, with maxima during JJA over the north-eastern Andes of Colombia and Venezuela. Figure 5C shows DCCs higher than 10 km, associated with severe weather, lightning and flash floods (Rasmussen and Houze, 2011;Zuluaga and Houze, 2015;Hoyos et al., 2019).
After convective systems get organized, they migrate away from the topography. Usually, the migration of organized convective systems is favored by the prevailing mid-to upper level wind. Over northern western Colombia and southern Panama, organized systems tend to move westward away from the topography toward the east Pacific Ocean by the easterly flow (Mapes et al., 2003;Zuluaga and Houze, 2015). Some authors attribute the speed of flow to a gravity wave triggered by the afternoon heating of the Andes (Mapes et al., 2003), and others to the easterly wave migration over the east Pacific (Rydbeck and Maloney, 2015).
Northern South America is also affected by synoptic-scale processes favoring convective organization over the tropical Andes such as the African Easterly Waves (AEWs). These perturbations are associated with convective activity and the occurrence of storm events in northern Colombia and Venezuela (León et al., 2001;Poveda et al., 2002;Dominguez et al., 2020). To the west of the tropical Andes (north of Peru and Ecuador), rain is associated with an onshore westerly low-level flow from the equatorial Pacific, which triggers convection forced by orographic lifting of the Andes (e.g., Takahashi, 2004).
The most active regions of lightning in the world are found in the low-lands of the tropical Andes of Colombia and Venezuela (Albrecht et al., 2016), with flash rate density (FRD) of 233 flashes per km 2 per year ( Figure S3). Lightning flash density is inversely correlated with altitude along the Andes of Colombia , causing high human and livestock mortality (Cristancho et al., 2017), with 757 deaths from 2000 to 2009, or 1.78 per million per year (Cruz et al., 2013;Navarrete-Aldana et al., 2014). Also, between 2003 and 2012, 282 casualties were reported among soldiers of the Colombian Army in 149 events with 72 deaths and 210 injuries.
Lightning also affect the power transmission systems in Colombia (Aranguren et al., 2017).

Southern Tropical Andes
Most of moisture reaching the Andes highlands come from the Atlantic Ocean and the Amazon River basin via easterly aerial rivers . Westerly winds from the Pacific are not infrequent over the Andes from Colombia to Peru, and occur following the seasonal movement of the ITCZ and the strength of the Bolivian High Sakamoto et al. (2011);Poveda et al. (2014);Makowski Giannoni et al. (2016); Trachte (2018); Espinoza et al. (2020). Boers et al. (2015aBoers et al. ( ,b, 2016 studied different types of extreme storms events over the southern tropical Andes, and found that events north of 20 • S are typically generated in the Amazon Basin, whereas storms over the foothills of the eastern Andes south of 20 • S, including the Puna High Plateau, originate over the southeastern part of the continent. Such propagation during the monsoon season (December to March) is explained by mid-latitude alternating pressure anomalies embedded in a westerly Rossby wave-train in combination with a low-pressure center located over northwestern Argentina, creating favorable conditions for intense storms (Liebmann et al., 2004;Espinoza et al., 2012a;Boers et al., 2015b;Paccini et al., 2017). Over the northern part of the south tropical Andes (10 • S-5 • N), the eastward propagation of the MJO convective core during the Austral summer (DJF), modulates sub-seasonal precipitation, producing positive rainfall anomalies up to 50% compared with the climatology (Recalde-Coronel et al., 2020).
Extreme rainfall in the valleys of the southern tropical Andes results from the interplay between the Amazon and the Pacific Ocean flows, influenced by orography (Junquas et al., , 2018Trachte, 2018). Kumar et al. (2019) show that precipitating cloud systems (PCs) in the Andes and western Amazon are largely affected by the directional surface flows. In particular, larger PCs (>2,000 km 2 ) are more frequent at the western flank of the Andes during southerly and easterly surface flows, while along the eastern flank, northerly and easterly surface winds are predominant during high and large PCs (Figures 6A-D). According to Villalobos et al. (2019), convection and cloud systems are deeper over the central Andes regions than in the Amazon-Andes transition region. The central Andes witness convective (30%) and stratiform (70%) precipitation, contributing with 63.3 and 36.7% of total rain, whereas in the Amazon-Andes transition those percentages and the contribution to cumulative rain attain similar values (31 and 69%, respectively). Over the central Peruvian Andes, and using a vertically pointed profile rain radar installed in the Mantaro Basin, Kumar et al. (2020) show that most of the bright band height (an indication for the melting zone) over this region vary between 4.3 and 4.7 km. An example of the vertical profile of radar reflectivity is shown in Figures 6E,F.
The eastern Andes foothills located at windward conditions are among the rainiest regions in the Amazon basin (4,000-7,000 mm yr −1 ; Roche et al., 1990;Espinoza et al., 2015), resulting from the interaction between large-scale humidity transport from the Atlantic Ocean, Amazonian evapotranspiration and moisture recycling, and the steep Andes topography (Killeen et al., 2007;Espinoza et al., 2009;Zemp et al., 2014). Highest values of this orographic rainfall are located along the foothills of the eastern Andes of Peru and Bolivia (e.g., Figure 6B), mainly during the austral summer Espinoza et al., 2015). Most rainfall intensification has been associated to the direction of 850 hPa winds ( Figure 6G), with excessive rainfall during a north-easterly wind regime originating in subtropical South America (Garreaud and Wallace, 1998;Laurent et al., 2002;Rickenbach et al., 2002;Espinoza et al., 2015;Chávez and Takahashi, 2017;Kumar et al., 2019;Moya-Álvarez et al., 2019). Wet episodes over these regions are frequently related to large and medium MCSs, triggered by the orographic lifting of the moist South American low-level jet (SALLJ, Bendix et al., 2009;Giovannettone and Barros, 2009;Romatschke and Houze, 2013;Junquas et al., 2018). Below 2000 m, intraseasonal rainfall variability is mainly modulated by the interplay between SALLJ, cross-equatorial low-level flow and southern winds intrusion toward low latitudes (Wang and Fu, 2002;Espinoza et al., 2015;Chávez and Takahashi, 2017). In the north-western Amazon-Andes transition region (Peru, Ecuador and Colombia) intense storms are more frequent under southerly low-level winds regime, particularly from June to November (Wang and Fu, 2002;Paccini et al., 2017). Figure 5D shows how the Andes foothills region in Central South America witness localized maxima of storms containing wide convective cores (WCC). These storms are organized at the mesoscale, producing the accumulated rainfall concentrated in the hotspots shown in Figure 5B). The region also experiences storms with extensive broad areas of stratiform rain (BSR) as shown in Figure 5E, mainly localized in the Andes foothills south of Peru and northern Bolivia. Diurnal analysis of the occurrence of these storms show how WCC peaks during the evening hours while storms containing BSRs peak in the early morning, which suggest that both are part of similar MCSs in either early or later stages of their convective life cycles . Chávez and Takahashi (2017) found that surface precipitation over these rainfall "hotspots" at the eastern flank of the Peruvian Andes is characterized by maximum values at 01-06 LT and organizes into MCSs that represent with at least 50% of daily rainfall. Few storms containing deep convective elements are found in the central Andes foothills region (Figure 5F), which suggests the orographically-driven moderate convective nature of storms. This observation agrees with the notion that convection here is highly influenced by the proximity to the convection in the Amazon basin, which tends to have less deep, more maritime-like and weaker convection .
MCSs in the region are observed in about 86% of the cases (Jaramillo et al., 2017) by the moisture transport that converges in the Andes foothills, in association with a stronger SALLJ. In the Central Andes, MCSs mainly occur from December to March, as in the northern part of the La Plata River basin (Jaramillo et al., 2017). The other 14% of the cases are associated with cold surges coming from south-eastern South America. These MCSs show a strong relationship with the diurnal cycle of rainfall in the Amazon-Andes transition region; in the upper Andes and lower Amazon precipitation tends to peak in the afternoon (Killeen et al., 2007), On the eastern slopes of the Andes (800-2,000 m a.s.l.), convection is developed and organized into MCSs during the night and a peak of precipitation is observed at 01-06 LT. These then move downslope (∼700 m) and precipitation declines during the morning (Chávez and Takahashi, 2017).

Extratropical Andes
Intense precipitation events over the extratropical Andes are different than those over the northern and central Andes, and occur both in summer and winter. Most precipitation is delivered by deep stratiform clouds rooted along frontal systems (Falvey and Garreaud, 2007;Viale and Nuñez, 2011), although deep convection associated with cut-off lows (COL) also occur. These weather systems reach the extratropical Andes year-round, and are strongly distorted by the mountains (Barrett et al., 2009). In particular, the cordillera produces a west-east precipitation gradient such that precipitation increases by a factor 2-4 over the western slope of the Andes to sharply decrease to the east of the continental divide . The upstream precipitation enhancement and downstream rain shadow across the southern Andes cordillera creates one of the most extreme precipitation gradients on Earth. In frontal cases, the amount of precipitation over the mountains and adjacent lowlands scales directly with the integrated water vapor transported by the westerly winds toward the western Andes (Falvey and Garreaud, 2007). Recent research shows that such transport maximizes in the core of atmospheric rivers (Ralph et al., 2016;Nieto et al., 2019) that landfall between 35 and 45 • along the west coast of South America 3-4 times per month (Viale et al., 2018). After making landfall, the ARs face the Andes cordillera where they deliver most of their precipitating. Thus, ARs account for up to 50% of the precipitation in the western slope of the Andes and are associated with about half of the extreme precipitation events in this region (Viale et al., 2018), thus being responsible for the majority of extreme precipitation events along the west slopes of the extratropical Andes (Valenzuela and Garreaud, 2019). A dramatic example of this situation is illustrated with the integrated water vapor and accumulated rainfall on Dec. 17, 2017, which triggered a major flooding causing more than 20 fatalities in the Santa Lucia village in southern Chile (Figure 7).
The thermal structure of weather systems is also relevant in triggering extreme hydrological events over the extratropical Andes. In most storms (at least two thirds of the events) precipitation over the central valleys begins almost simultaneously with the arrival of the cold front, so the bulk of the storm accumulation occurs under cold conditions (postfrontal precipitation) with a freezing level altitude around 2,200 m (Garreaud, 2013). This level is well below the Andes crest height between 27 and 37 • S (>5,000 m a.s.l.), contributing to the build-up of a seasonal snow pack that eventually melt during spring-summer. Indeed, many cold fronts arriving to central Chile during winter produce a minor increase in the flow of the rivers draining the Andes cordillera. A few winter storms, however, feature warm conditions causing a freezing level as high as 4,000 m a.s.l. and increasing the pluvial area up to a factor 4 relative to average conditions. This winter warm storm, often associated with an AR, can cause large floods downstream (Garreaud, 2013).
Cut-off lows (COL) can cause substantial precipitation near the western slope of the Andes, because the ascent in its leading edge is reinforced by the forced ascent over the terrain. In contrast to cold fronts, COLs can reach subtropical latitudes (northern Chile to southern Peru). An extraordinary COL situation happened in March 22-24, 2015, after warm waters established along the western coast of South America since early March, providing ample moisture in this otherwise dry region. The result was a major rainstorm over the Atacama region, the most arid desert in the world: up to 100 mm during 3 days and rainfall intensities in excess of 10 mm/h at some stations (Figure 8). The 2015 Atacama flooding episode is among the worst hydrometeorological disasters in Chile in terms of losses of human lives (more than 60 fatalities) and infrastructure damage (Bozkurt et al., 2016).
Convective rainfall events also occur during dry austral summer months in the northern sector of the extratropical Andes (30-40 • S), when liquid precipitation occur as high as 4,000 m a.s.l, triggering landslides and debris flows and damaging infrastructure and international highways (Sepúlveda and Padilla, 2008;Garreaud and Viale, 2014). The synoptic conditions for these summer events are characterized by easterly winds atop of the Andes, opposing the climatological westerly flow that transport moist air toward the mountains (Figure 9). The approach of an upper level trough (or COL) is necessary to overcome the large-scale subsidence, thus contributing to exacerbate the intensity of storms (Garreaud and Rutllant, 1997;Viale and Garreaud, 2014).
The southern region of South America experience one of the deepest, and most intense convective systems on Earth (Zipser et al., 2006;, mainly over the region of Mendoza, along the Andes of Argentina. Most of this precipitation occurs in the austral spring and summer (e.g., Figure 5B), and concentrates to the east of the Andes in the La Plata River basin (Zipser et al., 2006;Salio et al., 2007;Houze, 2011, 2016). An on-going research effort is under way to monitor and understand the life cycle of convective storms associated with high impact weather in the lee side of the Andes of Argentina (Provinces of Córdoba and Mendoza), known as RELAMPAGO (Remote sensing of Electrification, Lightning, And Mesoscale/microscale Processes with Adaptive Ground Observations; https://www.eol.ucar.edu/field_projects/ relampago). It is a joint effort between USA, Brazil and Argentina which had an Extended Observing Period from June 1st 2018 to April 30th, 2019, and an Intensive Observing Period from November 1st to December 18th, 2018. This region experiences multicellular storms which rank among the most intense convective systems in the world, exhibiting high tops, large hail, and strong lightning activity. Most of the storms initiate and develop in the early afternoon and late evening hours near and just east of the Sierras de Córdoba, and are predominant during the austral summer (December to February), with large CAPE and weak low-level vertical wind shear (Mulholland et al., 2018). The RELAMPAGO project involve collaboration with CACTI (Clouds, Aerosols, and Complex Terrain Interactions), a USA funded project to study orographic clouds and their representation in multi-scale models (Varble et al., 2019).
Farther downstream of the extratropical Andes, the lowlands of the La Plata River Basin region experience an exceptionally number of localized and widespread convective storms near the La Plata River basin, just east of the Andes ranges ( Figures 5D-F). There exists a preference for initial deep convective cells to be triggered near the foothills of the Andes close to Sierra de Cordoba in Argentina (e.g., DCCs shown in Figure 5F), under a strong topographically-driven convective initiation and maintenance. Although the Amazonian moisture source for the SALLJ does not seem to be as significant as originally thought (Yang and Dominguez, 2019), it is sometimes capped by leeside subsidence of the midlevel dry air that flows over the Andes from the eastward synoptic disturbances. The breaking of this cap triggers the release of convective instability and deep convection leading to tornadoes, hail and floods along the plains between the Andes and the Atlantic coast (Grandoso, 1966;Nascimento and Marcelino, 2005;Matsudo and Salio, 2011;Rasmussen and Houze, 2011;Mezher et al., 2012;Rasmussen et al., 2014). Associated mechanisms of hail occurrence include intense convection with sufficiently high warm and moist updrafts to increase super-cooled water content in the cloud, convergence of moist enthalpy near the ground, and the passage of cold fronts that trigger instabilities favoring upward motion on the warm side of the front, and facilitating the growth of deep convection (Teitelbaum and D'Andrea, 2015). FIGURE 9 | Schematic west-east cross section during summer storms over the extratropical Andes (30-37 • S). This is the most frequent type of storm (2/3) associated with an approaching weak trough aloft and easterly winds over the Argentinean side. Adapted from Viale and Garreaud (2014).

Northern Tropical Andes
Higher elevations in the tropical Andes pose a challenge for their communities which have to adapt to the occurrence of frosts, as is the case in the northern Andes of Colombia above 2,500 m a.s.l. Such events are typical during dry seasons due to intense radiative cooling of the ground during the night in association with very low cloud cover ( Figure S4). The occurrence of intense frosts events causes large economic losses to agricultural communities in the Andes by destroying crops and pastures (Bermúdez, 1990;Holmann et al., 2005). Metz et al. (2013) find that along the Andes corridor, extreme cold events reduce 925-hPa temperatures up to 17 • C below normal. During these events 925-hPa southerly winds are intensified from 0 to 10 m s −1 and the 925-700 hPa lapse rates is low (−3 • C km −1 ). Cold surges, also known as Friajes or Surazos in Spanish (Friagems, in Portuguese), propagate equatorward reaching the Bolivian and southern Peruvian Amazon, causing sudden and severe drops in temperature, up to 10 • C in few hours, affecting people, livestock and crops (Marengo, 1983;Ronchail, 1989a,b;Marengo et al., 1997;Garreaud and Wallace, 1998;Pezza and Ambrizzi, 2005;Quispe, 2010;Sulca et al., 2018). Incursions of low-level cold southerly winds from southern South America are also associated with rainy episodes on the Amazon-Andes transition region (Chávez and Takahashi, 2017;Paccini et al., 2017), as well as formation of thick clouds in the Zongo glacier region in Bolivia (16 • S, 5,060 m a.s.l.), particularly during the austral winter and spring (Sicart et al., 2015). Hurley et al. (2015) showed that around 70% of the total accumulated snow on the Quelccaya glacier (Peruvian Andes; 5,680 m) is associated with these southern cold air intrusions.

Southern Tropical Andes
In the central Peruvian Andes, radiative frosts are a major hazard to agriculture. They are caused by low cloud cover, surface air humidity, and soil moisture (Saavedra and Takahashi, 2017), and tend to occur at night times and early morning of the dry and cold seasons (June-August) (Saavedra and Takahashi, 2017;Sulca et al., 2018). During the austral summer, extreme cold events in the Mantaro Valley (central Peruvian Andes, around 500-5,300 m a.s.l.) are caused by intrusions of air masses along the east of the Andes through equatorward displacement of extratropical Rossby waves, negative temperature advection and southeasterly surface wind anomalies, positive OLR anomalies, and possibly influenced by the Madden-Julian Oscillation.

Extratropical Andes
Severe weather events associated with intense downslope wind episodes are observed in the eastward slope of the subtropical Andes and are commonly referred to as Zonda wind events (Norte, 2015). Zonda is a high-speed warm and very dry downslope wind occurring on the lee side of the Andes characterized by very strong wind-gusts affecting several western Argentina areas. Zonda episodes are more frequent during winter and spring. The occurrence of Zonda events is often associated with large damages on a variety of systems. These damages can be either socio-economical or ecological, such as blowing off roofs, toppling of trees, downing of power electricity supply and communication lines, among others. Because of its association with dry and very warm conditions, it also favors the occurrence of fire events and damage in crops. At high altitudes over the mountains, its occurrence affects the hydrological cycle, by accelerating snow melt and evaporation, and may trigger the occurrence of avalanches (Norte et al., 2008). The classical Zonda episode, described in Norte (2015) occurs in a synoptic environment characterized by a cold front approaching the western coast of South America and the associated migratory depression, as well as a moderate northern flow over central Argentina associated with the presence of the north-western Argentinean low and a warm front located over north-eastern Argentina. These ingredients induce strong baroclinicity, intense vertical wind shear and strong wind components that are transversal to the Andes. In addition, the presence of a warm front downstream the region where the Zonda develops favors its dynamics. Hence, the Zonda is produced by the topographically and dynamically forced ascent of humid air upslope of the Andes and a subsequent orographic descent of a prefrontal air mass which is strongly baroclinic. On the basis of the knowledge of the favorable conditions for Zonda events occurrence, Otero et al. (2018) have constructed a probability index for surface Zonda wind occurrence at the city of Mendoza (Argentina), including the estimation of the onset of the event, as a tool for predicting its probability of occurrence. However, Zonda events have been less investigated over the northern part of the extratropical Andes, where several strong episodes were identified (Otero et al., 2018).

ENSO IMPACTS ON EXTREME EVENTS
El Niño-Southern Oscillation (ENSO) is the most important modulator of the global hydro-climatology at interannual timescales. Its two phases, El Niño and La Niña, trigger a large suite of hydrological and meteorological extreme events worldwide (McPhaden et al., 2006), and particularly in South America and the Andean region (Aceituno, 1988;Waylen and Poveda, 2002;Aceituno et al., 2009;Cai et al., 2020), affecting the livelihoods of societies and imposing a huge socio-economic and environmental burden to the region.

Northern Tropical Andes
During El Niño, the Andes of Colombia experience long and intense droughts, heat waves, and forest fires, while La Niña triggers intense storms, landslides, floods, avalanches and inundations (Poveda andMesa, 1996, 1997;Poveda et al., 2001aPoveda et al., , 2006Poveda et al., , 2011Hoyos et al., 2013;Bedoya-Soto et al., 2018, 2019. Figure 10 shows the annual cycle of the probability distribution function (box-plots) of maximum intensity 1-h rainfall events in 19 rain gauges on the Colombian Andes , including monthly maximum events during both extreme ENSO phases, and also maxima events (MME). In general, maximum rainfall intensities are larger (smaller) during La Niña (El Niño), although the MME are not necessarily directly associated with either ENSO phase. Similar conclusions are obtained for storm durations from 1 to 24 h . El Niño and La Niña also affect the diurnal cycle of rainfall over the Colombian Andes, by increasing (decreasing) the amplitude during La Niña (El Niño), although the phase remains unchanged (Poveda et al., 2005).
Soil moisture over the Andes of Colombia is affected consistently by both phases of ENSO. Figure 11 shows soil moisture data gathered at nearby plots covered by shaded coffee, sunlit coffee and secondary forest. Results indicate that the annual seasonal cycle exhibits a strong coupling to interannual (ENSO) variability. Soil moisture exhibited minimum values during the two drier seasons (July through September 1997, and December 1997through March 1998, amidst the 1997-1998 El Niño event. On the other hand, the bi-modal annual cycle of soil moisture disappeared during the following 1998-1999 La Niña event, exhibiting saturation values throughout the rest of that year. Sunlit coffee plants exhibited stronger water deficits than forest and shaded coffee plants, thus pointing out the relevance of vegetation and land use in controlling water stress (Poveda et al., 2001a. Atmospheric teleconnections are part of the physical mechanisms that explain the establishment and persistence of drought in the northern Andes . For instance, droughts in the Andes of Ecuador during El Niño (La Niña) are associated with clear-cut westerly (easterly) surface wind anomalies over the central tropical Pacific and the reverse flow in the higher troposphere (Wang, 2002;Vicente-Serrano et al., 2017). The significant negative correlation between SST anomalies in the central tropical Pacific and precipitation anomalies over the glaciers of Ecuador showcase their strong dependence to both phases of ENSO Francou et al., 2004;Vuille et al., 2008;Rabatel et al., 2013).

Southern Tropical Andes
The general effect of ENSO on precipitation in the Southern Tropical Andes tends to be the opposite to the Northern Tropical Andes, with positive (negative) anomalies during El Niño (La Niña). Different ENSO types involve a large range of impacts worldwide through atmospheric teleconnections, in addition to land-atmosphere feedbacks (e.g., Larkin and Harrison, 2005;Ashok et al., 2007;Sun et al., 2013;Bedoya-Soto et al., 2018). In Peru, rainfall anomalies are associated with different types of ENSO, usually denoted as with the EP (Eastern Pacific) and CP (Central Pacific) indices (Yu and Kao, 2007;Kao and Yu, 2009;Takahashi et al., 2011;Lavado-Casimiro and Espinoza, 2014;Sulca et al., 2017;Takahashi and Martínez, 2019). SST warming in the eastern (central) equatorial Pacific is frequently associated with enhanced (reduced) coastal (Andean and Amazonian) rainfall, which can occur, with varying relative magnitudes, during El Niño or, with opposite sign, during La Niña (Lagos et al., 2008;Lavado-Casimiro et al., 2013;Lavado-Casimiro and Espinoza, 2014;Rau et al., 2016;Sulca et al., 2017;Imfeld et al., 2019). Both EP and CP indices are associated with weakened upper-level easterly flow over Peru, more so over the central/southern Peruvian Andes (Sulca et al., 2017). Moreover, during EP El Niño events extreme droughts in the southern Peruvian Andes are associated with a stronger warming in the troposphere above the central Pacific Ocean (Imfeld et al., 2019). However, not all extreme droughts over the region are explained by El Niño. The Bolivian High and moisture advection from the Amazon play an important role (e.g., Segura et al., 2016Segura et al., , 2020Imfeld et al., 2019). Over the Peruvian coast, El Niño is the main modulator of rainfall variability at interannual scale. The strong 1982-83 and 1997-98 El Niño events were characterized by an enhancement of convective activity in the eastern Pacific Takahashi and Dewitte, 2016;Jauregui and Takahashi, 2017), and intense storms leading to floods in the central and northern Peruvian coast (e.g., Woodman, 1985Woodman, , 1999Lavado-Casimiro et al., 2013;Bourrel et al., 2015;Rau et al., 2016). Rainfall and streamflow records in the Piura River basin (5.2 • S, 80.6 • W, 40 m a.s.l.; Figure 12A) characterize El Niño impacts (e.g., Deser and Wallace, 1987;Takahashi and Martínez, 2019). For instance, while the mean annual discharge is around 28 m 3 /s (Lavado et al., 2012), it reached values of 442 m 3 /s and 364 m 3 /s during the 1998 and 1983 El Niños, respectively (Takahashi and Martínez, 2019). Accordingly, during extreme El Niño events suspended sediment yield in the Peruvian coast increases by 3 to 60 times compared to the 1968-2012 multiyear average ( Figure 12B; Morera et al., 2017). In turn, Rodriguez-Morata et al. (2018) estimated that 21% of the total hydro-geomorphic disasters reported in Peru during the 1970-2013 period (including generic rainfall-related damage such as muddy areas, structural damage, holes, etc.) were related to the 1972-1973, 1982-1983El Niño events. The 1997 El Niño-driven flood damages were estimated to be around USD 1 billion, primarily along the northern coast of Peru (OPS, 2000).
A complex relationship is observed between ENSO variability and rainfall over the Peruvian coast (Bourrel et al., 2015). The so-called "coastal El Niño" strongly impacts the hydrology of Peru, such as in 1891, 1925, and 2017, characterized  Over the Altiplano a dominant negative (positive) relationship between El Niño (La Niña) and summer (DJF) rainfall has been identified (Vuille et al., 2000;Garreaud et al., 2009). This signal becomes rather weak to the north of the Altiplano and to the sub-Andean valleys of the southwestern Amazon below 1,500 m a.s.l. (Ronchail and Gallaire, 2006). The ENSO-Altiplano rainfall emerges from the changes in the zonal wind atop of the Andes during austral summer. During El Niño years the subtropical jet stream is enhanced thus reducing the transport of moist air from the continent toward the mountains (Garreaud, 1999). The opposite situation occurs during La Niña years, when weaker westerlies in the mid troposphere foster the intrusion of moist air from the eastern lowlands toward the Altiplano. In a recent study, a second mechanism associated with upward motion over the western Amazon, has been proposed to explain the interannual precipitation variability over the Altiplano (Segura et al., 2020). The authors show a slight increase in rainfall over the past two decades, in association with the strengthening of upward motion over the western Amazon. Seiler et al. (2013) investigated interannual climate variability in Bolivia. Temperatures were found to be higher in the Andes during El Niño, as well as during the positive phase of PDO and the Antarctic Oscillation (AAO). During the positive phase of the PDO the total annual rainfall and the frequency of intense storms were higher, regardless of ENSO phase in the lowlands. During December-February, El Niño was associated with drier conditions in the Andes with more variable precipitation.

Extratropical Andes
ENSO is a major modulator of interannual precipitation along the extratropical Andes (25-40 • S), exhibiting positive (negative) anomalies during the occurrence of El Niño (La Niña). While most precipitation in this area occurs during austral winter, it does not coincide with the peak of ENSO. Yet, the correlation coefficient between Niño3.4 index and annual rainfall is about 0.6 across this region (Montecinos and Aceituno, 2003), which mostly relies on rainfall over the western slope of the Andes (central Chile). In this region, the probability of having an extreme precipitation event (>40 mm/day) is twice as large during El Niño events compared to neutral conditions. Contrary, the probability of extreme events does not decrease substantially during La Niña events.
On the eastern side of the Andes, El Niño (La Niña) is associated with positive (negative) anomalies in precipitation, river runoff and landslides (Compagnucci and Vargas, 1998;Moreiras, 2005) during both winter and summer months. Significant differences in the frequency of occurrence of extreme precipitation events have also been identified over western Argentina during winter and late spring, with increased (decreased) number of extreme events during El Niño (La Niña) (Grimm and Tedeschi, 2009).

FLOODS Northern Tropical Andes
Colombia ranks first in Latin America and the Caribbean in losses and damages (per 100,000 inhabitants), from natural hazards during 1999-2013 (UNISDR, 2016). Most of them are associated with massive flooding during La Niña events (Poveda andMesa, 1996, 1997;Poveda et al., 2001aPoveda et al., , 2002Poveda et al., , 2006Hoyos et al., 2013;Bedoya-Soto et al., 2018). During the 2010-11 La Niña event, widespread flooding caused damages and losses in excess of 6 Billion US dollars, or 2% of Colombia's GDP for 2011. One reason is the intensification of the CHOCO lowlevel jet during La Niña, which transport large amounts of atmospheric moisture from the far eastern Pacific onshore, thus providing extra moisture for the development of MCSs and intense storms that can lead to widespread flooding along the Andes of Colombia and Ecuador. Typically, such steep river basins witness intense storms that produce flash floods, as the Guayas River in Ecuador during El Niño years (Frappart et al., 2017). Flooding mechanisms in tropical Andean cities depend on the preceding 24-h rainfall, baseflow, urbanization patterns, soil types and saturation conditions, as in the case of Quito (Perrin et al., 2001).
On the other hand, flash flood emergencies have also been observed to occur in periods with negative monthly precipitation anomalies associated with El Niño conditions as was the case of 18 May, 2015 in La Liboriana basin in the Municipality of Salgar, Colombia, causing more than 100 casualties and significant infrastructural damage. The meteorological factors associated with the emergency showed the role of successive precipitation pulses in relatively short periods that were intensified as deep convective cores interacted with the steep topography (Hoyos et al., 2019). These results show the challenges to opportunely pinpoint convective intensification over complex mountainous terrain from observations and forecast models to improve local risk reduction strategies over vulnerable regions of the tropical Andes.

Southern Tropical Andes
Floods in the central Andes can be triggered by intense storms and/or rapid glacier melting during the spring-summer of the austral hemisphere (Huggel et al., 2015). Flash floods commonly originate in moraine-dammed lakes threatening communities located in the Cordillera Blanca of Peru (Emmer, 2017). In addition, urbanization is increasing the risks of flash floods (Bourges et al., 1995;Méndez et al., 2016). In the valleys of Urubamba and Vilcanota in the Cuzco region, extreme rainfall and floods occurred in January and February 2010 causing USD 200M in damages and losses (Lavado-Casimiro et al., 2010) See Figure S5.
In the Peruvian Coast and most of the western flank of the Peruvian Andes, extreme floods have been associated with El Niño dating back to pre-Inca societies (e.g., Takahashi, 2004;Manners et al., 2007;Lavado-Casimiro and Espinoza, 2014;Rau et al., 2016). Three distinctive flooding responses to El Niño have been found over the western Andes of the Chancay River north of Lima (Peru): coastal desert, Lomas, and mountain desert. The mountain desert and valleys of the Pacific slopes are devastated by catastrophic floods and torrential rains during El Niño. However, less abundant precipitation is observed in Lomas, favoring the development of vegetation during the austral winter (Kalicki et al., 2014).
Besides the strong impacts of El Niño on the Peruvian coast, extreme rainfall in Peru has also been reported during La Niña and neutral conditions in the equatorial Pacific (Ronchail et al., 2002;Espinoza et al., 2012b;Lavado-Casimiro and Espinoza, 2014;Sulca et al., 2017;Recalde-Coronel et al., 2020). Indeed, excessive rainfall over the central and western Amazon have been related to La Niña events, and to warm anomalies in the tropical South Atlantic Ocean (TSA), in association with extreme rainfall and floods, as reported in 2009, 2012 and 2014 (Ronchail et al., 2005;Marengo et al., 2010Marengo et al., , 2011Marengo et al., , 2013Satyamurty et al., 2013). Rodriguez-Morata et al. (2018) documented that more than 36% of the hydro-geomorphic disasters reported in Peru during the 1970-2013 period were associated with La Niña and neutral conditions in the equatorial Pacific. The very unusual wet 2014 austral-summer period originated in the eastern slope of the Peruvian and Bolivian Andes and flooded the southern Amazon Ovando et al., 2016; Figure 13). This exceptional event has been associated with warm anomalies in the western Pacific-Indian Ocean and over the subtropical South Atlantic Ocean.

Extratropical Andes
Highly intense and short storms regularly trigger severe flash floods in the Andes of Mendoza, Argentina. Although previous studies have been unsuccessful in simulating peak discharges, factors such as storms characteristics and soil variability (quasiimpervious areas) play a key role, along with vegetation, surface storage capacity, and antecedent soil moisture (Braud et al., 1999). Other studies have proposed that surface, perimeter, basin length, elevations, and slope can be useful to predict floods over that region (Perucca and Angilieri, 2011). Flash floods in the Patagonian Andes have been reconstructed from dendrochronological studies (Casteller et al., 2015).
Along the western slopes of the southern Andes (continental Chile) floods occurs mostly during winter storms with significant precipitation (>50 mm/day), and have warm characteristics (thus producing a high freezing level) or last for more than 3 days. Recent events have been observed in northern, central and southern Chile (Valenzuela and Garreaud, 2019) and they are associated either with atmospheric rivers (Viale et al., 2018) or cut-off lows (COLs), as described in section Extratropical Andes.
A trend analysis was performed on annual peak river flows between 1975 and 2008 for rivers in the semi-arid (29−32 • S) and temperate (36−38 • S) zone of the Chilean Andes (Pizarro et al., 2014). Negative trends were identified in 87% of the stations in the semi-arid zone while positive trends were found in 57% of the stations in the temperate zone.
Differences in trends sign seem to be associated with the positive trend in the altitude of the zero-degree isotherm in this region.

LANDSLIDES, MUDSLIDES, AVALANCHES
The Andes, as all high mountain regions worldwide are subject to landslides, mudslides, rockslides, debris flows, and avalanches due to the steep topography, the soil and rock characteristics, land cover/land use change, and climate and weather patterns (Hermanns et al., 2012;Stahr and Langenscheidt, 2015). Intense storms, likely exacerbated by global warming and increased climate variability are the most important cause of landslides in the Andes (Moreiras, 2006;Sepúlveda and Petley, 2015).
These risks are particularly enhanced by the combined effects of rock avalanches causing temporal lakes and the ensuing dam failures (Ferrer, 1999;Hermanns et al., 2004). Intense storms exacerbate sediment production by slope movement on hillslopes, directly affecting their transport and deposition in downstream rivers and dams, as well as producing morphological changes in stream channels. The role of vegetation is critical on the stability of runoff and sediment yield and for example deforestation of the Andes plays a deleterious role (Braud et al., 2001). The Multinational Andean Project (MAP: GAC 2002GAC -2008 have set forward regional guides for landslide studies and mapping, and have conformed the first regional group of researchers for landslide studies in the Andes (GEMMA) (Jaramillo, 2008).

Northern Tropical Andes
The combination of intense storms and steep terrain are major causes of landslides in the tropical Andes. For instance, a 3 h rainfall event (208 mm) on September 21, 1990, triggered more than 800 landslides along the Andes of Colombia (Aristizábal et al., 2015). The SHALSTAB model (Montgomery and Dietrich, 1994) was able to capture the physics involved in such landslides. López Salgado (2016) studied mass movements in the upper Coello River basin in the Central Andes of Colombia and identified that they result from the combination of high rainfall, steep slopes, and unstable volcanic ash cover.
There is also a strong connection between deforestation and land use change with erosion and landslides in the tropical Andes. Erosion rates in the Magdalena River basin, located between the central and eastern ranges of the Colombian Andes, have shown an increase from 550 ton km −2 y −1 , in the years previous to 2000, to 710 t km −2 y −1 between 2000 and 2010, as well as sediment load (average increase of 44 Mt y −1 ), coinciding with a forest loss of 40%, and an increase of 65% in areas dedicated to agriculture and pastures (Restrepo and Escobar, 2018).
Deforestation and land use change significantly alter the landslide frequency-area distributions in the Andes of Ecuador (Guns and Vanacker, 2014). A strong increase in the frequency of small landslides in anthropogenic environments explains in part the observed enhancement in landslide denudation rates in this region (Vanacker et al., 2003Molina et al., 2012;Vanacker, 2013, 2014). Marie and Veerle (2013) suggest that deforestation causes major changes in soil properties and hydrology, accelerating landslide activity. In other cases, vegetation contributes to destabilize slopes owing to its contribution to the organic layer mass, and promotes a new cycle of vegetation succession. Vanacker et al. (2007) show evidence that the steepness of topography contributes more than human activities or land use practices to the occurrence of infrequent large landslides over the Andes of Ecuador.
Although many landslides in the Andes are triggered by highly intense storms, Soto et al. (2017a) found that only 24% landslides in the Loja River basin (southern Ecuador) were triggered by intense storms whose return periods exceed 1 year, and less for the 76% of landslides. Also, rainfall events that are not exceptionally intense can reactivate these landslides (Soto et al., 2017b).
Erosion processes in the Andean river basins yield large concentrations of suspended sediment to the western part of the low-lying Amazonia. For example, as is the case of the Napo River in the Amazonian regions of Ecuador and Peru, where annual mean erosion rates are approximately 1,160 t km −2 (Laraque et al., 2009;Clark et al., 2016).
Younes-Cárdenas and Erazo-Mera (2016) developed an explanatory model to predict the degree of susceptibility of landslides along the E-20 highway in Ecuador, and concluded that the distribution of landslides was far from random, and clearly associated with several geological and environmental factors, including zones of active erosion processes, volcanic sandstones, granitic rocks and mean annual rainfall values between 1500 and 1750 mm. The non-random nature of landslides in space and time has been confirmed by Muñoz et al. (2017) who showed the multifractal space-time characteristics of landslides in the Andes of Colombia.
The main source of sediments (and nutrients) to the lower Amazon River basin is located along the Andes. Townsend-Small et al. (2008) studied the production and transport of sediments and suspended organic matter along the Chorobamba River in the central region of the Andes-Amazon transition of Peru. At least 80% of total annual suspended sediment, 74% of organic carbon, and 64% of organic nitrogen transport was concentrated in 10 days of the year, with significant seasonal changes in sources of organic matter. The latter indicates that extreme events are the main origin for this process.

Southern Tropical Andes
Global warming linked to intense storms and stronger El Niño events, along with permafrost degradation and glacier retreat promote slope instability in the Central Andes, thus increasing the frequency and intensity of landslides. These processes can modify the spatial distribution of landslides, shifting initial points of slope instability to higher altitudes, and triggering more complex and larger landslides (Moreiras and Pont, 2017). Furthermore, large mass movements that originated as rock/ice falls from the glacierized Andes can become highvolume high-velocity mud-rich debris flows, such as the 1962 and 1970 events in Nevado Huscarán (Cordillera Blanca, Peru) (Evans et al., 2009).
Landslides play an important ecological role (Restrepo et al., 2009;Smith, 2011). An inventory over the Andes of Peru indicated turn over times of 1,320 years (Clark et al., 2016), removing 264 tC km −2 yr −1 of organic carbon mostly from soils (80 %) in 25 years. A single storm event in March 2010 was responsible for 27% of the total landslide area reported, and produced 26% of the organic carbon flux associated with landslides during such 25-yr period (Huggel et al., 2015). Heavy rainfall reported during the austral summer of 2010 in Cuzco region caused economic losses estimated to be around 250 million US dollars. Landslide and flash floods destroyed around 5,000 houses, 16,300 ha of farmlands, 6.15 km of the Cusco-Machu Picchu railway line and 21 bridges (Lavado-Casimiro et al., 2010; Figure S5). Lowman and Barros (2014) investigated the links between precipitation and erosion in the central Andes, and estimated mean erosion rates of 2.1-8.5 mm per year, which happens to be 1 to 2 orders of magnitude larger than long-term estimates (1,000 to 1 million years) for the central Andes. On the other hand, erosion rates in the Eastern Cordillera (northern Bolivia) have been estimated as 5-14 mm per year (Blodgett and Isacks, 2007).
The ratio of sediment transport between hillslopes and channels in the Peruvian Andes influences geomorphological features and the morphology and dynamics along the channel network (Schneider et al., 2008). The headwaters of river channel networks exhibit rough and high-relief surfaces with deep incisions and dense drainage density. Thus, storminess influences the production and transport of sediments on hillslopes and channels, reflected into landscape morphometry.
The melting of snow accumulated in the winter conductive to saturation of soils and landslides in the following spring and summer seasons (December to February) is the major cause of slope instability in the Aconcagua Park, as evidenced in the high significant correlation between landslides and river discharges, and a weak relationship between landslide and temperature records .

DROUGHTS, HEAT WAVES, AND FIRES
ENSO is a major driver of droughts in various regions of South America (Zhou and Lau, 2001;Kane, 2006;McPhaden et al., 2010;Cai et al., 2020), in particular El Niño is associated with droughts over large parts of Colombia, Venezuela, some regions of the high Andes of Ecuador, Peru and Bolivia (Francou, 1985;Caviedes and Endlicher, 1989). At the same time, La Niña is associated with droughts in other regions of Ecuador, Bolivia, Chile and Argentina.

Northern Tropical Andes
Multiple studies have shown that droughts in the northern tropical Andes are associated with the occurrence of El Niño (section ENSO Impacts on Extreme Events, Northern Tropical Andes) through changes in the Hadley circulation Arias et al., 2015), the Walker circulation (Kousky et al., 1984;Vuille, 1999;Vuille et al., 2000;Francou et al., 2004;Poveda et al., 2006Poveda et al., , 2011Vicente-Serrano et al., 2011a, as well as to land surface-atmosphere feedbacks (Poveda and Mesa, 1997;Bedoya-Soto et al., 2018, 2019. Such droughts have deleterious impacts on water supply for humans and livestock, crops failure, and contribute to intensify forest fires, heat waves, as well as outbreaks of malaria, dengue and other vector-borne diseases, and electricity shortages owing to the importance of hydropower generation. Tropical cloud forests are facing huge threats from fires, deforestation and climate change (Mutke et al., 2017). Deforestation and land use/land change also affect the dynamics of soil moisture in the Andes of Colombia, reducing water storage capacity and leading to higher peak flows (Ramírez et al., 2017).
The frequency and duration of dry periods in Quito has been increasing since the middle of the last century, compared to the last 400 years (Domínguez-Castro et al., 2018). The strongest dry periods occurred in 1692-1701 (the severest one), 1718-723, 1976-1980, 1990-1993, and 2001-2006. The great drought that lasted from 1692 to 1701 caused generalized famine in Quito and affected the majority of population of the central Andes. Such dry periods affect water supply, agriculture and have strong ecological impacts (Chirino et al., 2017).
The unique tundra ecosystems above 3,500 m in the neotropical Andes, páramos, play a fundamental ecological and hydrological role below the snow line in the high Andes, in terms of ecosystem services, such as water retention for human and animal consumption, irrigation, and hydropower. Droughts have a strong impact on the hydrological dynamics of páramos, such that the recovery from droughts are fundamentally determined by the aridity index (Iñiguez et al., 2016).
The dynamics of fires in the Andes of Colombia is highly controlled by the occurrence of El Niño (Becerra and Poveda, 2006). On the other hand, fires in Ecuador, Peru and Bolivia (Paramo, Yungas, and Puna, respectively) do not show a significant relation to ENSO, but to periodic weather patterns of precipitation: increased rainfall the year before fire-peak seasons followed by very dry periods and unusual low temperatures during the fire-pike (Román-Cuesta et al., 2011.

Southern Tropical Andes
Over the Altiplano region century-scale dry periods have been identified using a 707-yr dendrochronological reconstruction (Morales et al., 2012). A persistent negative trend in precipitation has been identified from 1930s to date. At interannual to multidecadal time scales, those authors also identified an ENSOlike pattern. Severe drought condition in the Altiplano region were reported during 1983 and 1992 El Niño years (Garreaud et al., 2003;Segura et al., 2016). During the 2015-2016 El Niño event, extreme dry conditions were reported over the Altiplano (Marengo et al., 2017). Severe drought affected central and southern Bolivia, with deficits of rainfall surpassing 30%. Low river discharges were documented in the northern part of the Titicaca basin (Peru) (Jiménez- Muñoz et al., 2016).
During extreme drought years, impacts on the vegetation have been identified over this region, such as reduction in vegetation activity (measured using NDVI index) and forest fires during the 2005, 2010, and 2016 droughts (Fernandes et al., 2011;Espinoza et al., 2016;Jiménez-Muñoz et al., 2016). Peaks of active fires in Amazonia have been more related to extreme drought than deforestation (Aragão et al., 2018), although the intense fire 2019 summer season was indeed related with deregulation policies in the Brazilian Amazon and the concomitant deforestation (Amigo, 2000;Lovejoy and Nobre, 2019;MAAP, 2019;Wheeling, 2019). These fires and the resulting smoke and soot can reach the highest Andes, contributing to accelerate their on-going rapid melting and disappearance (Allen, 2020). Espinoza et al. (2019) documented an increase in the dry day frequency (DDF) over the central and southern Peruvian Amazon, as well as in the Bolivian Amazon, considering the 1981-2017 period. See also Espinoza et al. (2020).
Various studies document the demise and collapse of pre-Columbian civilizations due to extensive droughts in the Central Andes. For instance, during the sixth century, major cultural disturbances were reported on the northern Peruvian coast, including the Mochica culture that was conditioned by severe sixth-century droughts, spanning from 562 to 594 CE (Shimada et al., 1991). Likewise, the Tiwanaku state, in the south-central Andes, disintegrated between c. AD 1000-1100, after 700 years of growth and colonial expansion. Such droughts impacted irrigation-based agriculture in the low-altitude colonies and the groundwater-dependent raised-field systems in the Altiplano (Ortloff and Kolata, 1993).
In the central Peruvian Andes, dry spells during the wet season (December to March) are associated with a weakened convective activity over the western tropical Pacific, that is in accord with the impacts of El Niño over the region (Sulca et al., 2016). The authors also show that westerly wind anomalies over the upper troposphere reduce moisture flow from the Amazon basin, producing dry events over the Peruvian Andes. Reduced snowfall from suppressed frontal activity in the Central Andes of Argentina is another mechanism that propagates droughts into river flows, mainly driven by La Niña-like conditions in the Pacific Ocean (Rivera et al., 2017b).

Extratropical Andes
The semiarid, Mediterranean-like climate of the extratropical Andes (30-40 • S) makes this region drought-prone. Most of the annual precipitation is concentrated in few rain events. This region exhibits high year-to-year variability and is punctuated by severe droughts (30-60% deficits;Núñez-Cobo et al., 2018), lasting 1 year (winter season) and in some cases up to 2-3 years. An exceptional "megadrought" has been occurring in central Chile since 2010 (Figure 14; Garreaud et al., 2017).
Winter season droughts are accompanied by a reinforced subtropical anticyclone and deeper than average Amundsen-Bellingshausen Seas Low that conspire to weaken the westerly flow impinging in the southern Andes and restrict the number of fronts Garreaud et al., 2017). Most drought years occur under La Niña conditions (Montecinos and Aceituno, 2003), and less under ENSO neutral conditions. Indeed, the "megadrought" has occurred mostly under near normal conditions in the equatorial Pacific and it even resisted the strong 2015-16 El Niño event (Garreaud et al., 2017).
Droughts in the southern Andes affect agriculture in central Chile and central-western Argentina (Núñez-Cobo et al., 2018), but also the water supply and hydro-electricity generation, and sets the stage for forest fires Urrutia-Jalabert et al., 2018). The ecology of rivers flowing to the Pacific is also affected (Masotti et al., 2018), as exemplified in the summer of 2016 when the drought triggered the worst ever recorded harmful algal bloom in the coastal areas of Patagonia (Garreaud, 2018b). Amoroso et al. (2018) developed a 525 year-long reconstruction from the Chubut River basin in Patagonia, finding the most severe droughts in 1680-1705, 1813-1828, 1900-1920, 1993-2002, and from 2011 to the present, in association with forest fires since the 1850s with a 12 year return interval, and a lack of fire for the last 94 years.
Droughts exhibit two distinct temporal regional behaviors in terms of duration and severity between North and Central Patagonia (Rivera et al., 2018). Diverse macro-climatic phenomena such as SAM, ENSO and the Pacific Decadal Oscillation (PDO) affect the interannual and interdecadal variability of droughts in Patagonia (Rivera et al., 2018). Droughts have shown to be more persistent in Argentina, with important regional differences (Minetti et al., 2010).
Extreme droughts have been on the rise in the Temperate-Mediterranean transition (TMT) zone of the Andes (35.5 • -39.5 • S) during the last century with respect to the previous six centuries (Christie et al., 2011). The identified changes are related to the north-south oscillations in moisture conditions between the Mediterranean and temperate Andean regions, resulting from the latitudinal position of the storm tracks forced by large-scale circulation modes. In turn, moisture conditions are linked to tropical and high-latitude ocean-atmospheric forcing, with the Palmer Drought Severity Index (PDSI) positively related to Niño-3.4 SST during spring and strongly negatively correlated with the Antarctic Oscillation (AAO) during summer. Extreme dry years exhibit a strong negative PDSI vs. AAO correlation.
In addition to droughts, years of above average moisture availability precede fire years in northern Patagonia (Veblen and Kitzberger, 2002), by enhancing the growth of herbaceous plants which increases the quantity of fine fuels during the fire season a few years later. The short-term variability of available moisture leading to massive burning is linked to El Niño, due to greater moisture availability in spring, causing fires in the following years. Warmer and drier conditions during La Niña in the spring exacerbate the drying of fuels, and fire events.
Regarding heat waves, Jacques-Coper et al. (2016) identified 201 cases in south-eastern Patagonia during 1872-2010, with mean duration of 2 weeks and temperature peak of 4.3 • C on day 0 (the warmest day in the mean signal), affecting both sides of the Andes. The warming results from temperature advection and enhanced radiative heating, following a high-pressure system over southern South America, from a wave-train-like pattern along the South Pacific. Two thirds of the heat wave events are linked to enhanced ascent in the South Atlantic Convergence Zone (SACZ).
During the 2010-2015 period, the eastern flank of the southern Andes (37 • S−31 • S) experienced one of the most severe hydrological droughts on record with large socioeconomic impacts. Moreover, since snowmelt is the most important source of water, the reduced snowfall over the mountains propagated the drought signal through the streamflow in the adjacent foothills east of the Andes (Rivera et al., 2017a,b).

RESEARCH GAPS
There is a plethora of challenges toward understanding, modeling and, let alone, predicting the studied high impact weather events.
Much of the current knowledge about extreme climate and weather events in the Andes are based upon limited observational data in particular in the higher Andes. For instance, there are just eight permanent atmospheric sounding sites operating near or in the Andes: Palonegro and Bogotá located on the eastern Andes of Colombia, Alfonso Bonilla-Aragón near Cali (Colombia), Quito (Ecuador), Lima-Callao (Perú), Antofagasta (Chile), Mendoza (Argentina), and Santo Domingo (Chile). This is a meager set of atmospheric measurements to understand the spatiotemporal dynamics of extreme weather patterns in a region characterized by steep slopes, deep valleys, and strong ecological gradients from north to south and from west to east. The situation of meteorological radars is even worse, given the need of high spatiotemporal precipitation estimates in the Andes (Guallpa et al., 2019), and that just a few are operating in the region: Bogotá and Medellín (Colombia), the network of three radars in Cajas, Loja, and Guachaurco in southern Ecuador (Bendix et al., 2017), another in Piura that covers a small region of the north-western Peruvian Andes, three radars in the Province of Mendoza (San Martín, Tunuyan y San Rafael), and one in Jujuy, Argentina, which are active during the hail storm season (October to April).
Regarding the monitoring network of other hydrological, meteorological and climatic variables, the situation is better with respect to in-situ rainfall data, although large regions remain uncovered. Something similar can be said about river discharges. The network is highly deficient with respect to weather and climate data sets required to estimate evapotranspiration, and almost non-existent for soil moisture and groundwater, and for all the processes that make part of the surface energy budget. Satellite and Unmanned Aerial Vehicles (UAV) type remotely sensed observations can help improve our understanding in this region, but still require appropriate in situ evaluation. Satellite data alone is also insufficient to detect the occurrence and quantify the magnitude of extreme events given their short timescales and small spatial resolution. For instance, diverse ground-validation studies show the poor skill of the TRMM satellite to capture rainfall over the Andes (e.g., Ochoa et al., 2014;Zambrano-Bigiarini et al., 2017;Vallejo-Bernal et al., 2020). Our modeling efforts and the much-needed establishment of early warning systems and disaster prevention efforts are similarly limited by the lack of observational data and data networks. A thorough discussion of the observational and monitoring hydrometeorological network in the Andes is presented by Condom et al. (2020).
Many scientific questions arise about the spatiotemporal dynamics of the high impact weather and climate events discussed herein. For example, what are the dynamics and thermodynamics of the ocean-atmosphere-land interactions that explain the wettest spot on Earth over the Pacific coast of Colombia, in association with MCSs and other rainfallgenerating mechanisms leading to intense storms, flooding and landsliding over the Andes of Colombia.
The very topic of rainfall-runoff processes and the formation of floods along the Andes has been largely overlooked, which is more concerning given that floods constitute the most impacting events in terms of human lives and socio-economic losses in the region. The framework of statistical scaling provides a tool to understand and predict the occurrence of floods from intense storms (Gupta and Dawdy, 1995;Menabde and Sivapalan, 2001;Furey and Gupta, 2007;Mandapaka et al., 2009;Gupta et al., 2010;Ayalew et al., 2015), but also to estimate annual peak discharges of different return periods, relevant in hydrologic design (Poveda et al., 2007). Scaling properties need to be explained in terms of the physical processes and predicted by hydrologic models (Nguyen et al., 2015;Quintero et al., 2016;Wu et al., 2017;Salazar et al., 2018;Pérez et al., 2019).
Major knowledge gaps still remain regarding the non-linear interactions and synchronicity (e.g., Boers et al., 2014Boers et al., , 2016Salas et al., 2020) between the dynamics of the CHOCO, Caribbean and Orinoco LLJs, and the SALLJ with other macroclimatic phenomena and modes of variability at a wide range of temporal scales, and their association with the occurrence of extreme events, including PDO, AMO, NAO, IOD, QBO, SAM, and the Tropical Atlantic SST Dipole. Also, the interplay between highfrequency variability modes (e.g., equatorial waves, gravity waves, MJO, etc.) and ENSO, and how such interactions modulate extreme events over the Andes. There is the need to understand the role of the diurnal cycle on the formation of storms over the Andes, and its dynamics during the aforementioned diverse modes of climate variability. In that regard, what is the role of land surface-atmosphere interactions in explaining the water and energy budgets across timescales, from the interdecadal to the diurnal, and the extraordinarily high percentage (70-90%) of precipitation recycling in the Andes (Zemp et al., 2014).
Land use changes have been intense during the last decades over the Andes and the Amazon rainforest. Deforestation, agriculture, urbanization, large infrastructure development and extractive industry exacerbate these problems, such that the tropical Andes have been identified as the most critical biodiversity hotspot worldwide (Myers et al., 2000;Marchese, 2015;Hrdina and Romportl, 2017). The functioning of the coupled Andes-Amazon system (Builes-Jaramillo and Poveda, 2018) requires a thorough investigation, including how the modification in local and regional land conditions affect: (i) the frequency and magnitude of extreme weather events, owing to the alteration of the hydrological cycle and the energy budget across river basins of increasing scales, and (ii) the vulnerability of populations under risks of avalanches, flash floods, landslides, droughts, etc.
Extreme events are most often related to a local (mesoscale) intensification of extreme conditions (e.g., high rainfall rates, strong winds) embedded in broader synoptic systems (e.g., forced convection along a cold front). Most of the research in the last decades has focused on the synoptic-scale aspects of these extreme events, but examination of their mesoscale features is largely hindered by the aforementioned shortcomings of the observational network in space and time. Thus, we do not know much yet about the triggers and mechanism behind extreme precipitation events (say, brief periods with more than 20 mm/h) over the extratropical Andes.
Another limitation is the short, limited and non-systematic record of landslides, avalanches, flash floods and other groundlevel manifestations of extreme weather events. Recall that south of 20 • S the Andes is sparsely populated (most people live in the lowlands immediately to the west or east of the range). The record is composed of events that somehow reached the lowlands or interrupted some roads and railways crossing the Andes. Therefore, we may face a substantial underestimation of geo-climatic hazards in this region of extremely complex terrain.
Multiple scientific questions need to be answered regarding the dynamics and spatiotemporal persistence of droughts along the Andes, including the role of vegetation, precipitation, evapotranspiration, river flows, soil moisture and groundwater, net surface radiation and the fluxes of latent and sensible heat. This is also valid for the case of wild fires and heat waves.
Given the large socio-economic toll of high impact weather events in the Andes, the need of assessing their projected changes in future climate scenarios is evident, having in mind diverse adaptation strategies. Though there are studies focusing on the impacts of climate change in the Andes (e.g., Pabón-Caicedo et al., 2020), there is a lack of studies focusing on projected changes of high impact weather events over the region. One of the reasons for this may be the lack of reliability of climate models in reproducing the current climate conditions. Models exhibit large biases mostly in near-surface variables such as temperature, winds and precipitation and, hence, large discrepancies may be found when looking at extreme weather events. Consequently, there is a clear need on assessing the quality of climate models in reproducing high-impact events in the Andes and on assessing how these events are expected to change under future warmer conditions. The state-of-the-art climate models, either Global or Regional, operating at horizontal resolutions in a range from 100 to 20 km, are not expected to capture the high-impacts events themselves, but the large scale forcings and their interaction with the complex orography. Hence, there is also the need to evaluate the skill of convection-permitting models in terms and their capability in reproducing triggering synoptic-scale mechanisms, and the occurrence of diverse extreme events.

FINAL REMARKS
We have provided an in-depth review of the most important high impact weather events along the Andes, summarized in Table 1. It is not possible to overstate their socioeconomic, environmental and ecological deleterious impacts. Such knowledge is fundamental to assess the relevant knowledge gaps and monitoring needs, but also toward the development of associated human capabilities, the implementation of scientific networks that involve the Andean scientific community and the transfer of scientific knowledge to public policies, which are particularly necessary and particularly challenging amidst population growth and climate change, as well as deforestation and urbanization trends throughout the Andes.

AUTHOR CONTRIBUTIONS
All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

ACKNOWLEDGMENTS
We are grateful to the Global Water and Energy Exchanges Programme (GEWEX) of WCRP, in particular to Fernande Vervoort, and also to Universidad Nacional de Colombia at Medellín, and to Centro de Ciencia del Clima y la Resiliencia (CR) 2 of Universidad de Chile in Santiago (FONDAP Grant 15110009), for supporting the three regional workshops of ANDEX (https://www.gewex.org/project/andex/home/), as a prospective Regional Hydroclimate Project (RHP).