Abstract Both variability in precipitation and rainfall extremes are key drivers of landslide activity, yet their combined influence with antecedent moisture conditions remains insufficiently quantified at regional or local scales. In this study, daily precipitation records over the past 25 years (2000–2024) were analyzed for five meteorological stations in Northern Croatia across multiple temporal scales. The aim was to investigate the impact of precipitation patterns and regime changes on landslide triggering in Hum na Sutli and the wider area. Statistical analyses (linear regression, Mann–Kendall trend assessment, and Pearson correlation) were applied, and antecedent wetness was quantified using the antecedent precipitation index (API). Results indicate weak, statistically insignificant positive trends in annual precipitation, accompanied by strong interannual variability and coherent regional behavior. Seasonal analysis reveals the dominance of warm-season precipitation with pronounced extremes, while short-duration and multi-day rainfall events exhibit high variability and clustering. The 2024 Hum na Sutli landslide coincided with elevated cumulative precipitation and sustained high API values, despite the absence of exceptionally extreme single-day rainfall events. These findings highlight the critical role of antecedent moisture accumulation combined with episodic high precipitation in slope failure. The study supports a conceptual model in which landslide triggering is governed by the interaction of preconditioning and short-term hydrometeorological factors, providing a basis for improved hazard and risk assessment. Additionally, preliminary rainfall threshold values are proposed as practical early-warning guidance for local communities in landslide-prone regions in Northern Croatia. 1. Introduction The growing challenges of water management and geohazard mitigation under changing climatic conditions require widespread scientific attention [ 1, 2, 3]. Mass movements and hydrological hazards have a strong dependency on climatic stressors, water management practices, other geohazard events (e.g., earthquakes), and anthropogenic activities [ 4, 5]. Among these factors, precipitation is recognized as a primary triggering mechanism of landslides and slope failures [ 6]. It increases pore water pressure and decreases the shear strength of slope materials [ 7]. Both short-duration, high-intensity precipitation events and prolonged periods of moderate precipitation can contribute to slope failure, depending on local hydrological and geotechnical conditions [ 8]. As a dominant landslide trigger, precipitation exhibits highly variable behavior, characterized by asymmetric distributions, temporal clustering, and irregular extremes; it is often accompanied by shifts in variability and trends over time [ 9]. Understanding these statistical properties and their temporal evolution is essential for defining critical thresholds to improve landslide hazard assessment and support the development of effective monitoring, mitigation, and early-warning systems [ 10, 11]. Northern Croatia is prone to slope failures, with specific regions characterized by high susceptibility to landslides [ 12, 13]. For example, in Zagreb City County, more than a thousand landslides have been triggered and documented over the last 25 years [ 14, 15, 16, 17, 18]. Besides geological conditions and anthropogenic modifications, slope stability is strongly interconnected with climate-driven changes in precipitation patterns [ 19, 20, 21]. Although most regional climate studies do not indicate significant long-term trends in annual precipitation since the mid-twentieth century, changes in precipitation regime are evident, including seasonal redistribution and an increase in extreme precipitation indices (e.g., the number of heavy precipitation days and maximum 1- to 5-day precipitation totals) [ 22, 23, 24, 25, 26, 27]. This study aims to assess the role of precipitation in the initiation of a landslide in the Hum na Sutli municipality and wider area in Northern Croatia. Furthermore, it aims to assess whether regional meteorological station data can also be representative for local-scale studies such as Hum na Sutli, where in situ monitoring is unavailable. The landslides investigated have a complex activation history involving anthropogenic slope loading, seismic events, and intense precipitation events as direct triggering factors [ 5]. Daily precipitation records from five nearby meteorological stations covering the period 2000–2024 were used to perform a comprehensive analysis of precipitation characteristics, including climatological variability, extreme precipitation indices, and antecedent wetness conditions [ 28]. By integrating precipitation time series analysis with a multi-hazard perspective, this study aims to (i) characterize the precipitation conditions leading up to landslide events; (ii) evaluate the extremeness of these conditions within a 25-year climatological context; (iii) reflect on historical precipitation data (1970–2000) [ 29]; and (iv) discuss indicative precipitation thresholds for slope failure in Northern Croatia. The findings contribute to a better understanding of compound landslide triggers and provide a basis for improving hazard and risk assessments and the development of early-warning systems in similar environments [ 30, 31, 32, 33]. 2. Study Area The study area is located in Hum na Sutli, a small town in Krapina-Zagorje County in northwestern Croatia ( Figure 1a). The relief is characterized by hilly and mountainous terrain, while the landscape comprises forested hills, agricultural land, scattered settlements, and river valleys. The investigated landslide is at an approximate altitude of 335 m above sea level (m a.s.l.) and is influenced by both continental and Mediterranean climate systems. According to the Köppen–Geiger classification, the local climate is Cfb type (a temperate oceanic climate), with warm summers and relatively cold winters [ 34]. Mean annual air temperature is approximately 12 °C, with mean annual precipitation of 1000–1100 mm [ 29, 34]. The wider area is mostly covered by Miocene deposits, i.e., marls, sandstone, and limestone [ 35], and the landslide area is characterized by predominantly clastic deposits, i.e., Golubovec formation, consisting of poorly sorted, unconsolidated to weakly consolidated sands, silts, and clays [ 5]. These weakly consolidated deposits are highly susceptible to slope instability and are commonly associated with landslide occurrences in Northern Croatia [ 36]. Since there are no meteorological stations in the immediate vicinity of the landslide area in Hum na Sutli, precipitation data from regional meteorological stations were utilized ( Figure 1b,c). Basic characteristics of the meteorological stations and their time series are provided in Table 1. The main meteorological station (i.e., where a full array of meteorological parameters is monitored) is located only in Krapina, while all other stations are precipitation stations; hence, only precipitation is measured. Among these, only meteorological stations in Desinić and Donji Macelj have a sufficiently long time series to enable comparison between present-day precipitation (i.e., 2000–2024) and precipitation from the reference climate period (i.e., 1970–2000 [ 29]). Landslides in Hum na Sutli have a complex history of activation spanning over a decade. Initial sliding was recorded in 2011, and the unstable slope was loaded with construction materials and debris [ 5]. In the upper section of the slope, in the immediate vicinity of a house and bakery, initial cracks appeared on the slope after the Petrinja earthquake series (December 2020–January 2021, M 6.2, 100 km distance from the epicenter) [ 38, 39]. Landowners have reported constant movements of various degrees from 2022 to the present day, culminating in massive sliding and slope failure during the spring/summer of 2024. The recent landslide is a composite type: the upper part is a rotational slide, while the lower part is a mudflow ( Figure 2a). The height of the main scarp is approximately 10 m ( Figure 2b), while the sliding plane is at an approximate depth of 5–10 m [ 5]. Since the landslide remains active, continuous erosion of material from the upper section contributes to the development of mudflow in the lower parts of the landslide, and the main scarp propagates in the NE direction ( Figure 2c). The initial deep-seated rotational failure with shallower mudflow mobilization in the lower parts of the Hum na Sutli landslide occurred simultaneously, primarily as a consequence of precipitation. 3. Materials and Methods The precipitation data used in this study were provided by the Croatian Meteorological and Hydrological Service (DHMZ). In this study, precipitation data were analyzed on annual, monthly, and daily scales. Annual precipitation data were investigated to identify the trends and deviations in the long-term climate of this area [ 29]. Monthly precipitation values were used to detect seasonal patterns and prolonged wet periods, directly influencing progressive soil saturation. Daily precipitation values were examined to detect short-term, high-intensity rainfall events that may have triggered the recent landslide movement in 2024. The combined analyses of these temporal scales enabled the evaluation of both long-term preconditioning effects and short-term triggering mechanisms associated with precipitation. In particular, linear regressions were performed to detect trends in annual precipitation time series. To assess the statistical significance of trends, a non-parametric Mann–Kendall test was used to minimize the effect of outliers and accommodate non-normal or skewed distributions. In the Mann–Kendall test, the null hypothesis assumes the absence of a monotonic trend, and trends with p-values 0.05). Strong positive correlations between station pairs ( r = 0.78–0.92) indicated a coherent regional precipitation signal, despite localized spatial variability in individual years. Seasonal precipitation exhibits two distinctive patterns across the region, with the lowest median values observed in winter (i.e., December–February) in Pregrada, Donji Macelj, and Bednja; by contrast, Krapina and Desinić had the lowest values during autumn (i.e., September–November). Median precipitation increases progressively from winter to spring and peaks during summer or autumn, indicating a pronounced warm-season precipitation regime ( Figure 4). Krapina and Desinić show peak precipitation predominantly in summer (i.e., June–August), whereas Pregrada, Donji Macelj, and Bednja exhibit the highest median values in autumn (i.e., September–November), highlighting moderate spatial variability in the timing of seasonal maxima. Variability is markedly higher in summer and autumn, as indicated by larger interquartile ranges and extended whiskers. The predominance of longer upper whiskers across seasons indicates positively skewed distributions, suggesting that extreme precipitation events contribute substantially to seasonal totals. Several outliers, particularly in spring and summer, indicate episodic extremely wet seasons (e.g., approximately 500 mm of rainfall for spring in Krapina and autumn in Donji Macelj). In contrast, colder seasons exhibit comparatively lower variability, although occasional outliers indicate intense episodic precipitation events. The temporal evolution of precipitation extremes from 2000 to 2024 reveals distinct patterns across the analyzed indices ( Figure 5). The maximum number of consecutive wet days (CWD, Figure 5a) remains relatively stable across all stations, generally ranging between 4 and 8 days, with only occasional peaks exceeding 10 days (recorded at Donji Macelj and Desinić in 2005). This indicated limited long-term variability in the duration of wet periods. Moreover, 10 consecutive wet days were recorded in Desinić in 2023, a year prior to the reactivation of the Hum na Sutli landslide. In contrast, the maximum number of consecutive dry days (CDDs, Figure 5b) exhibits pronounced interannual variability, with values fluctuating between approximately 15 and 40 days. Recurrent peaks exceeding 30 days occur throughout the time series (e.g., the mid-2000s, early to middle 2010s, and early 2020s), reflecting episodic prolonged dry periods. Short-duration extreme precipitation, represented by maximum 3-day totals (Rx3D, Figure 5c), shows substantial variability, with multiple events approaching the observed maxima of approximately 140–160 mm of rainfall, particularly in 2010 and 2013–2017. Similarly, maximum 7-day precipitation (Rx7D, Figure 5d) displays higher accumulated totals and pronounced peaks during the same intervals, indicating that extreme rainfall events often occur as part of multi-day episodes. Overall, the results highlight the stable persistence of wet seasons. This contrasts with highly variable dry spells and episodic, yet regionally consistent, multi-day extreme precipitation events. The analysis of intense rainfall events between 2000 and 2024 identified several heavy periods (i.e., 40–60 mm/day) annually across all stations. By contrast, severe rainfall (i.e., 60–80 mm/day) occurs less frequently, and remains relatively rare (i.e., >80 mm/day). This distribution highlights a marked decrease in event frequency with increasing rainfall intensity, in accordance with regional climatic characteristics and type (i.e., a humid continental climate). The maximum recorded daily rainfall was substantially higher in Donji Macelj (130 mm/day) and Krapina (109 mm/day) compared to Bednja and Desinić (80 and 86 mm/day, respectively), whereas Pregrada exhibited lower values across all intense rainfall categories, with a maximum rainfall of 69 mm/day. Despite their rarity, extreme events are typically found within years characterized by multiple heavy and severe rainfall occurrences, indicating that they seldom occur in isolation. Pronounced interannual variability is observed, with a notably lower number of intense rainfall events between 2000 and 2005 (excluding Donji Macelj and Bednja, which have no data before 2007). Krapina, Desinić, and Pregrada show similar distributions and occurrences of intense rainfall events. By contrast, Donji Macelj and Bednja show evidence of significantly higher occurrences of extreme rainfall events despite a similar distribution. Moreover, a slightly increasing trend in the occurrence of intense rainfall is evident for Krapina and Desinić. The time series is too short to support definitive conclusions regarding a local climatic shift. However, data indicate local climatic variability and recurring trends. Furthermore, the temporal evolution of daily precipitation and the antecedent precipitation index (API) highlight the critical role of cumulative moisture conditions in landslide triggering [ 51]. Given its proximity to the landslide in Hum na Sutli, Desinić station was specifically analyzed for its API ( Figure 6), using daily precipitation data for 2024. During this year, landowners reported sliding and mass movements. While most individual rainfall events (blue bars) are moderate (i.e., 80 mm/day). Higher values are highlighted in bold. Table 2. Analysis of intense rainfall events for the period 2000–2024 at the investigated meteorological stations, following the defined rainfall thresholds: heavy (40–60 mm/day), severe (60–80 mm/day), and extreme (>80 mm/day). Higher values are highlighted in bold. Station Desinić (mm/Day) Pregrada (mm/Day) Donji Macelj (mm/Day) Krapina (mm/Day) Bednja (mm/Day) Year 40–60 60–80 >80 40–60 60–80 >80 40–60 60–80 >80 40–60 60–80 >80 40–60 60–80 >80 2000 2 No data 2001 1 1 1 2002 5 1 1 2003 2 2004 1 2005 3 2 2006 1 1 2007 1 1 3 1 1 1 2008 2 1 2 1 2 2009 1 1 3 1 3 2010 1 1 1 1 3 1 2 1
