The Impact of Agricultural Irrigation on Landslide Triggering: A Review from Chinese, English, and Spanish Literature

: Landslides are considered a natural process, with hundreds of events occurring every year in many regions of the world. However, human activities can signiﬁcantly affect how stable a slope or cliff is, increasing the chances of slope collapses. Moreover, agricultural irrigation has potential to saturate subsurface materials well below ground level and is known to be an important factor that can trigger landslides in many countries. A macroregional literature review of irrigation-induced landslides was developed in this investigation, considering what has been published in Chinese, English, and Spanish. A total of 115 peer-reviewed papers, books and book chapters, graduate and undergraduate theses, and government reports were found, including 82 case studies (23 in Chinese, 26 in English, and 33 in Spanish). Results from this analysis indicate that studies focusing on this important topic have increased exponentially since the 1960s, with most irrigation-induced landslides occurring in dry climates (precipitation less than 600 mm/year), with rainfall concentrated during summer months. The majority of studies have been done in the loess region of China ( Asian region ), followed by Peru ( Latin American region ), though cases were found from other macroregions ( African, Indian, Russian, Angloamerican, and Indonesian ). Based on this global review, new agricultural irrigation projects located in collapsible areas must include a landslide risk analysis. Cultivated areas can follow a series of measures to minimize the chances of triggering a landslide, which would put human lives, ecosystems, food production, and infrastructure at risk.

Landslides can also be triggered by human activities [32], with numerous cases around the world documented in the literature, including mass movements initiated by excavations [33], piping of soil [34], terracing [35,36], deforestation [37,38], and urbanization [39], among others. Moreover, agricultural irrigation can also lead to slope instability, a topic that has been studied for decades (e.g., [40]), simply because of its effects on percolation and aquifer recharge [41], decreasing effective soil cohesion in certain portions of the slope profile due to saturation [42]. Irrigation effects are even stronger when combined with other previously mentioned factors. While climate change continues to increase the likelihood of landslides occurring due to the presence of storms with higher intensities and rainfall depths [43][44][45], new agricultural projects are established every year, e.g., [46], affecting the water cycle not only by shortening the availability of the resource [47], but also by increasing the amount of water that percolates through the soil profile, increasing water table levels [48] and exacerbating the other adverse effects on slope stability.
Considering all of the above and following the global macroregional classification defined by Anděl et al. [49], herein we develop a worldwide review focused on the effects of agricultural irrigation on landslide triggering. The literature from this review was primarily from studies reported in Chinese, English, and Spanish languages (due to a lack of translation expertise, other languages such as Arabic, Japanese, Russian, etc., were not included). This extensive search was based on case studies and reviews from peerreviewed scientific literature (SciELO, Latindex, Scopus, Web of Science, etc.), books and book chapters, graduate/undergraduate theses, and reports from government agencies, all under the following keywords: Landslides; Mass movement; Irrigation; Agriculture; Crops; Landslide-induced; and Landslide triggering. The above work was developed using different methods such as universities' inter-library systems from China, Peru, and the United States, government agencies' publication databases, and Internet search engines. Only cases where agricultural irrigation, irrigation canals, and/or storage ponds were the main landslide-inducing factor were considered in this analysis (not necessarily excluding other contributing factors such as rainfall). The overall purpose was to obtain a better understanding of the impacts of agricultural irrigation on slope mass movement mechanisms and identifying the best management practices intended to prevent future episodes.

A Worldwide Review on Landslides Triggered by Agricultural Irrigation
Many irrigation-induced landslides have been reported in the Asian region, specifically in the Loess Plateau, which covers 635 thousand km 2 in China. As a result of dam construction for hydroelectric power in parts of the Yellow River (West Gansu Province, North-West China), thousands of farmers were re-settled in the Heifangtai loess platform located 42 km southwest of Lanzhou city in Yongjing County (Gansu Province), where they started irrigating intensively around 1968, resulting in geological hazards such as soil salinization, among others [50], but also increasing perched water table levels and saturating subsurface soil layers, e.g., [51]. This resulted in thousands of soil mass movement events in this area [40,[52][53][54][55][56]. These events have been widely studied by Chinese scientists to describe the process of slope collapse, e.g., [13,[57][58][59][60][61][62][63][64][65][66][67][68], and to understand why many of them transition to dangerous fluidized forms [69][70][71][72][73][74]. Furthermore, Derbyshire [75] developed a complete analysis of the risk that landslides represent in China's loess region, concluding that agricultural activities represent a clear threat.
Irrigation-induced landslides in the region are often triggered by infiltration through surface cracks and sinkholes [76][77][78][79][80]. In fact, irrigation not only triggers landslides in this area [81], but the occurrence of a landslide itself also increases the chances for new events to happen [82], as observed in the Jiaojiayatou region [83,84]. Studies suggest a noticeable relationship between groundwater table level changes (as a result of agricultural irrigation) and landslide history [85][86][87], with authors such as Zhang [88] reporting water table increases of more than 20 m over four decades in some areas of the Heifangtai loess plateau, and Xu et al. [89], who estimated an annual increase of 0.18 m in the same region. Moreover, Xu et al. [90] simulated the hydrological effects of irrigation on the stability of a loess cliff edge in Heifangtai, concluding that the main triggering cause of the ongoing landslides within this geological formation was the establishment of large agricultural fields. Also in the Heifangtai Platform, Gu et al. [91] and Wu et al. [92] investigated the effects of irrigation on the stability of the loess landform, with both studies concluding that the excess water from agricultural flooding irrigation systems was primarily responsible for recent landslides in the area, agreeing with the results obtained by Hou et al. [93] on the instability of the Heifangtai platform, and Duan et al. [94] on the instability of the Jingyang loess platform located in the Shaanxi Province (northern China). In that same region, Li and Jin [95] analyzed the initiation of deep landslides, concluding that the triggering cause was an increase in water tables created by intense irrigation activities, similar to what Gu et al. [96], Gao et al. [97], and Qui et al. [98] concluded in their analyses of landslides in the Gansu, Sichuan, and Wuhai provinces, respectively.
Early studies such as Lei [99] and Meng and Derbyshire [55] describe how seepage water from earthen irrigation canals on the upper portions of loess formations can trigger landslide events. Later on, Wen and He [100] evaluated the effects of percolated water from irrigation projects on the reactivation of landslides located in red mudstone plateaus of northern China, suggesting that agriculture represents a concern for future mass movements in the Lanzhou area, one of the largest cities in northern China. Ma et al. [101] used field investigations, geological exploration, numerical simulation, isotropically consolidated undrained triaxial tests, and ring shear tests on four landslides that occurred simultaneously in the Shaanxi Province to identify their initiation and movement mechanisms, concluding that the main cause for such events was the presence of diversion-based irrigation canals, which constantly percolate water downward. This finding agrees with Xi et al. [102], who concluded that leakage from irrigation canals was the main triggering factor in the Gaoloucun landslide (Shaanxi Province).
Wang [103] was a pioneer of irrigation-induced landslide prediction in China, anticipating the occurrence of the Huangci landslide in Yongjing county (Gansu Province). Furthermore, landslide modeling in China's loess region has also been done by authors such as Wang et al. [104], Cui et al., [105], Lian et al. [106], Xing [107], and Zhang and Wang [108], including centrifuge experiments [109,110], Visual MODFLOW [87], and FEFLOW [84] to simulate the relationship between the evolution of groundwater dynamics and landslide disasters. These studies all identified intensive irrigation as the main factor triggering mass movements. Similarly, experimental rainfall and irrigation simulations made by Cao and Yin [111] and Ding [112], as well as slope stability analyses made by Duan et al. [113], also concluded that artificially induced water percolation from agricultural activities was the main triggering factor in landslides occurring in dry climatic regions of the country.
According to Zhuang et al. [114], most landslides in the Chinese loess region are shallow, with volumes of less than 100,000 m 3 , with concave south-east-facing slope profiles (i.e., less vegetation), slope angles of 20-35 • , and occurring as long run-out events that generally transform into mudflows. In this sense, antecedent soil moisture (from rainfall or irrigation) plays an important role in triggering loess landslides.
Four types of landslides have been identified for the loess region of China [67]: irregularly moving, fast moving, slow moving, and mudflows. Lei [115] notes that human activities can cause all of these, and one initiating from plateau surfaces is often related to irrigation activities. Most irrigation-induced landslides in this region occur during the thaw period (March) and the rainy season (July) [116], and they have been carefully monitored using high-resolution satellite data [117] and mitigated using a variety of measures [50].
In Argentina, Jurio et al. [144] used over 20 years of monitoring data to characterize and analyze the gravitational movements affecting the edge of the plateau in Vista Alegre (Neuquén Province). They attributed the landslides to an irrigated area of only 12 hectares located above the collapsed volume in this arid region of the country.
Similarly, Villaseñor-Reyes et al. [145] evaluated two landslide events on volcanicderived soils through an integrated study, including detailed lithology, morpho-structural inventories, analysis of land use, and pluviometric regimes in eastern Michoacán (Mexico), attributing excessive agricultural irrigation as the main triggering factor. Also in Mexico, García [146] concluded that excessive irrigation was one of the main contributing factors leading to landslides in Metztitlán.
In Ecuador, Yadún et al. [147] documented several irrigation-triggered landslides in the Escudillas watershed, developing a risk map using GIS techniques and field visits, emphasizing the need to be aware of how irrigation canals affect soil mass movement. Similarly, Samaniego [148] evaluated past irrigation-triggered landslides in Pueblo Viejo (Ecuador), an old mining town that became farmland later on. The author developed a risk analysis for future episodes so that local authorities can make proper prevention decisions.
Rosales and Centeno [149] evaluated the susceptibility to slope failure in La Conquista (Nicaragua), attributing inefficient irrigation as the main reason why several landslide episodes have occurred (and might continue to occur) in the area, similar to what Reyes et al. [150] and Carbajal [151] concluded in their studies at the Talgua (Honduras) and Tabarcia (Costa Rica) watersheds, respectively.
As for the Indonesian region, the 2018 Palu valley landslides in Indonesia killed thousands of people and were initially attributed to a strong earthquake. However, a recent study by Bradley et al. [152] concluded that the main triggering factor in such devastating landslide was actually rice irrigation fields located uphill, a statement corroborated later on by Watkinson and Hall [153] and Cummins [154].
Evidence of irrigation-induced landslides have been also documented in the Angloamerican region. Knott [155] investigated the causes triggering two long-dormant landslides in Ventura County (California), attributing irrigation of avocado fields as the main factor. Similarly, Clague and Evans [156] evaluated a series of landslides next to the Thompson River valley (British Columbia, Canada) that have been occurring since 1880 over a 10 km long section of the riverside, concluding that most ancient events were charac-terized as slow moving, until uphill agriculture began, triggering rapid and dangerous mass movements.
After a landslide killed 300 people in Bududa (eastern Uganda, African region), Gorokhovich et al. [157] concluded that even though the event was triggered by heavy rains, a variety of factors contributed to initiation of the slope failure, including the absence of drainage systems, the local topography, as well as land use changes (Arabica coffee plantations). The abundance of rainfall in the area, however, means that irrigation is not required, except perhaps during the dry season (January), so land use and cover changes may be more important than the addition of water.
Through geophysics, Domej et al. [158] evaluated the contributions of irrigation canals on slope stability in Tajikistan (Russian region), where landslides occurred and were initially thought to be caused by either glacial retreat and/or strong earthquakes. This area of the country is known for its extreme aridity and for the presence of a permeable surface layer that covers a clay-rich horizon, making it susceptible to landslides since agricultural fields were established, according to the authors. Similarly, though triggered by an earthquake, Ishihara et al. [159] demonstrated that agricultural irrigation was a critical component for the initiation of a landslide in that same country.
Finally, in Pakistan (Indian region), Ali et al. [160] challenged the attribution of climate change as the primary factor triggering landslides, identifying the political dimension of water management that also contributed to mass movement in the Yourjogh's mountainous region. The authors concluded that while geological factors conducive to landslides are present in the area, the main cause of the events is the deterioration of the water governance system, leading to negligent overuse of the resource, which is readily available.
No irrigation-induced landslides were found for the Islamic, European, Sino-Japanese, and Australian-Oceanic regions within the Chinese, English, or Spanish scientific literature, or were not reviewed because of translation issues.

Discussion and Conclusions
A total of 115 irrigation-induced landslide investigations around the world, including 82 case studies in Chinese (23), English (26), and Spanish (33), were found, considering scientific publications, university theses, books and book chapters, and reports from government agencies. It is interesting to note that the number of worldwide studies focusing on the effects of irrigation on landslide triggering has increased exponentially during the last decades (Figure 1), with four additional studies (including this one) already been published in 2020 (Lacroix et al. [120] and Ingemmet [141] in Peru, and Yadún et al. [147] in Ecuador). This is expected, considering that increasing global agricultural demands are anticipated to double by 2050 [161]. As a consequence, new irrigation projects are established every year in areas that are susceptible irrigation-induced landslides, which are often mistakenly attributed to earthquakes, e.g., [152].
Results from this literature review indicate that irrigation-induced landslides represent a growing risk in areas where agriculture and certain climatic and geologic conditions are present. Moreover, documented landslides from around the world suggest that the majority of irrigation-induced landslides occur in dry climates (Table 1), with 80% of cases happening on areas with less than 600 mm/year of rainfall. Similarly, 95% of cases have occurred in climates where rainfall is concentrated during summer months (see Table 1), possibly because of the lower kinematic viscosity of rainwater (i.e., higher infiltration capacity) at warmer temperatures, e.g., [162], resulting in higher effective permeability of soil and higher saturation of soil masses, leading to collapse, e.g., [112,113]. In other words, wherever underground geologic layers are dry and stable under ambient conditions, but highly unstable when saturated due to percolating water from irrigation projects (and many times combined with summer rainfall), landslides are more likely to occur. Such collapses are triggered mainly due to an increase in water table levels (e.g., [99]) or even from small seepage impediments [51]. Finally, 49% of the reported irrigation-induced landslides have been documented on loess terrains, followed by sedimentary (26%), alluvial (13%), and volcanic (12%) geologic materials. Results from this literature review indicate that irrigation-induced landslides represent a growing risk in areas where agriculture and certain climatic and geologic conditions are present. Moreover, documented landslides from around the world suggest that the majority of irrigation-induced landslides occur in dry climates (Table 1), with 80% of cases happening on areas with less than 600 mm/year of rainfall. Similarly, 95% of cases have occurred in climates where rainfall is concentrated during summer months (see Table 1), possibly because of the lower kinematic viscosity of rainwater (i.e., higher infiltration capacity) at warmer temperatures, e.g., [162], resulting in higher effective permeability of soil and higher saturation of soil masses, leading to collapse, e.g., [112,113]. In other words, wherever underground geologic layers are dry and stable under ambient conditions, but highly unstable when saturated due to percolating water from irrigation projects (and many times combined with summer rainfall), landslides are more likely to occur. Such collapses are triggered mainly due to an increase in water table levels (e.g., [99]) or even from small seepage impediments [51]. Finally, 49% of the reported irrigation-induced landslides have been documented on loess terrains, followed by sedimentary (26%), alluvial (13%), and volcanic (12%) geologic materials.   In terms of global geographical distribution, most irrigation-induced landslides have been reported in the Asian Region, with the majority of cases documented in China, followed by Peru in the Latin American Region (see Figures 2 and 3, and Table 1). Moreover, the most widely studied geological formation on the planet, when it comes to irrigation-induced landslides, is the Loess Plateau of China. In fact, just like volcanic, alluvial, and sedimentary materials in general, dry loess usually has high stability, e.g., [75], so disturbances such as earthquakes and excavations are normally the main factors triggering landslides in these regions, e.g., [22]. However, loess may collapse rapidly with increasing water content from irrigation [65,163], which is not only one of the most common triggers of loess landslides in China (as discussed earlier), but it also contributes to collapsed masses travelling longer distances downwards, thereby putting local communities at risk [69]. The most direct external manifestations of the collapsibility of loess are sinkholes and surface cracks, with the former formed on the edge of the plateau, while the latter formed mostly through the upper and lower layers of hillslopes, reducing the integrity of the soil mass and affecting the stability of the plateau, e.g., [59,71]. Both phenomena represent important ways for surface storm runoff to percolate into lower layers, in addition to the constant percolation that occurs below irrigation.  Despite the above, studies have shown that irrigation-induced events can be avoided e.g., [164]. A simple geological analysis of the material located below current or future agricultural projects could lead to the prevention of future episodes, specifically looking for the presence of geological components that can be stable when dry, but very collapsi ble when saturated such as alluvial, loess, sedimentary, or volcanic units (as shown in this review), e.g., [27,[165][166][167]. Therefore, human activities (in this case, agricultural irrigation should include a landslide-triggering risk analysis, especially for projects located in dry climates with rainfall concentrated in summer months [115]. Irrigation programs should avoid the establishment of new crops above hillslopes and cliffs, or on edges of plateaus e.g., [120]. Moreover, landslide modeling, e.g., [168], as well as monitoring mass move ment through high-resolution remote sensing [169] and other methods, e.g., [170][171][172][173], are key tools for predicting future irrigation-induced landslides. However, in most cases the damage has already been done, and agricultural (irriga tion) fields have been established for decades in collapsible areas. In these cases, usefu mitigation options include minimizing irrigation water input and controlling the rise o  Despite the above, studies have shown that irrigation-induced events can be avoided, e.g., [164]. A simple geological analysis of the material located below current or future agricultural projects could lead to the prevention of future episodes, specifically looking for the presence of geological components that can be stable when dry, but very collapsible when saturated such as alluvial, loess, sedimentary, or volcanic units (as shown in this review), e.g., [27,[165][166][167]. Therefore, human activities (in this case, agricultural irrigation) should include a landslide-triggering risk analysis, especially for projects located in dry climates with rainfall concentrated in summer months [115]. Irrigation programs should avoid the establishment of new crops above hillslopes and cliffs, or on edges of plateaus, e.g., [120]. Moreover, landslide modeling, e.g., [168], as well as monitoring mass movement through high-resolution remote sensing [169] and other methods, e.g., [170][171][172][173], are key tools for predicting future irrigation-induced landslides.
However, in most cases the damage has already been done, and agricultural (irrigation) fields have been established for decades in collapsible areas. In these cases, useful mitigation options include minimizing irrigation water input and controlling the rise of water tables, with the purpose of avoiding saturation of slopes [88,99]. This can be done Despite the above, studies have shown that irrigation-induced events can be avoided, e.g., [164]. A simple geological analysis of the material located below current or future agricultural projects could lead to the prevention of future episodes, specifically looking for the presence of geological components that can be stable when dry, but very collapsible when saturated such as alluvial, loess, sedimentary, or volcanic units (as shown in this review), e.g., [27,[165][166][167]. Therefore, human activities (in this case, agricultural irrigation) should include a landslide-triggering risk analysis, especially for projects located in dry climates with rainfall concentrated in summer months [115]. Irrigation programs should avoid the establishment of new crops above hillslopes and cliffs, or on edges of plateaus, e.g., [120]. Moreover, landslide modeling, e.g., [168], as well as monitoring mass movement through high-resolution remote sensing [169] and other methods, e.g., [170][171][172][173], are key tools for predicting future irrigation-induced landslides.
However, in most cases the damage has already been done, and agricultural (irrigation) fields have been established for decades in collapsible areas. In these cases, useful mitigation options include minimizing irrigation water input and controlling the rise of water tables, with the purpose of avoiding saturation of slopes [88,99]. This can be done through significantly decreasing the size of irrigation projects, using pumping wells to prevent groundwater from flowing towards potentially unstable slopes, setting up efficient groundwater and surface drainage systems on slopes, and returning farmland to forest and grass in potentially unstable areas [174]. Other approaches include perforating impermeable layers to allow seepage water from irrigation to pass through, e.g., [175], or stabilizing landslides by constructing retaining walls or emplacing concrete and steel piles within the slope, for example [55,57].
Replacing flood-irrigation and earth irrigation canals (which result in more seepage water, soil mass saturation, and increases in water table levels) with more efficient irrigation practices can significantly reduce the probability of future landslide events. Moreover, drip irrigation systems have the advantage of minimizing the amount of water needed per unit area, simply because the technique provides moisture to the plant in a localized manner within the soil, normally directly to the root system. If applied during daylight (when plants are active), roots can immediately absorb the provided water, thus significantly reducing seepage volumes, e.g., [176].
Equally important is the application of water-retaining polymers (ideally, potassiumbased to avoid environmental impacts), which can save up to 90% of water volumes needed to irrigate a certain area and can rapidly absorb the applied water during each irrigation period, retaining the water for use by plants over time. Additionally, polymers can decrease the number of times that a certain crop needs to be irrigated (e.g., from daily to biweekly in extreme cases, depending on site conditions and species). Therefore, the application of water retention polymers allows a farmer to cultivate more hectares with the same amount of available water, but more importantly it limits seepage processes, reducing the probability of landslides initiation, e.g., [177].
Following the same approach, reducing seepage from current irrigation structures such as canals and storage ponds can minimize the risk of landslides as well. This can be accomplished using liners of low-cost geotextiles instead of expensive concrete [178]. Finally, slopes can also be stabilized by afforestation, since the roots of trees help to reinforce soil and resist sliding [179]. It should be noted that revegetation programs typically should not include trees that require irrigation to become established (i.e., native or low water demanding species, ideally with water retaining polymers), since adding more water to the slope would defeat the purpose of stabilizing the slope by revegetation.
While climate change continues to increase the risk of landslides [180,181], findings from this review suggest that agricultural irrigation can drastically change local hydrologic cycles in many regions of the world, e.g., [182], leading to the collapse of slopes that can endanger human lives, ecosystems, infrastructure, and food production, and possibly result in lawsuits, e.g., [183]. Finally, as Hermanns et al. [184] suggest, there is an urgent need to improve international collaboration for the mapping of landslides, as this review indicates that each country acts on its own. Funding: Funding for this project was provided by the Center for Mining Sustainability, a joint venture between the Universidad Nacional San Agustin (Arequipa, Peru) and Colorado School of Mines (USA).