Rainfall-runoff modelling in karst catchments is challenging due to complex hydrogeological and climatic conditions. Conventional hydrological models can struggle to simulate the nonlinear dynamics. To address this challenge, this study proposes a multiple hybrid modelling approach to enhance daily rainfall-runoff simulations in the karst Ljubljanica River catchment in Slovenia. This approach integrates the Technische Universitat Wien (TUW) and the Genie Rural a` 5 parame`tres Journalier (GR5J) lumped conceptual models, which consider the snow processes, with the symbolic regression- genetic programming (SR-GP) data-driven model. The hybrid model, TUW-CemaNeige GR5J-SR-GP, was fortified by grey wolf optimisation (GWO) for calibration, ensemble empirical mode decomposition (EEMD) for data decomposition, and recursive feature elimination (RFE) for feature selection. Rainfall-runoff modelling was conducted for the observed and projected datasets under the Representative Concentration Pathway 4.5 (RCP4.5) and 8.5 (RCP8.5) climate change scenarios. The hybrid model improved baseflow simulation performance by 41 % (TUW) and 36 % (CemaNeige GR5J), enhanced monthly peak discharge simulation performance by 5 % and 13 %, and yielded notable improvements in simulating low and high flows under the RCP4.5 and RCP8.5 scenarios. The algebraic equations of the SR-GP model and sensitivity analysis highlighted the influence of the slow-flow components on discharge simulations. The hybrid modelling approach is a promising alternative for compensating for the limitations of the stand-alone models in karst catchments for efficient water resources management.

Understanding the factors influencing soil erosion in agricultural catchments is crucial to reduce land degradation. This study investigated the benefit of disdrometer information beyond rain-gauge observations and the factors that explain the variability in event-scale suspended sediment load. The study seasonally assessed multi-year, high frequency observations of rain-gauge and disdrometer-based rainfall and suspended sediment load measurements in the 66 ha Hydrological Open Air Laboratory agricultural catchment in Austria. Hydrometric and land management information was combined to understand the factors controlling erosion. The results confirmed that the kinetic energy of rainfall and rainfall erosivity estimated by disdrometer and rain gauge data were comparable (r = 0.93 and 0.95, respectively). Rainfall erosivity alone did not fully explain the observed variability in SSL. Winter events with small kinetic energy caused large total event suspended sediment load because bare and saturated soils reinforced surface runoff. In the spring and summer months, the average size and velocity of the drops were the largest and events with larger number of drops in the most frequent velocity class were associated with larger total event suspended sediment load. In spring and summer, the sediment response to a given level of rainfall erosivity was modulated by land management state and antecedent soil moisture rather than by rainfall properties alone. This study demonstrated that instead of disdrometers rain gauge-based estimates can suffice to estimate the erosivity of rainfall events which can be linked to soil erosion. Still, the disdrometer data provided additional knowledge on detailed characteristics of the rainfall events.

When vegetation intercepts precipitation, the quantity of rainwater reaching the ground is affected, as it passes through the canopy, drips from it, and runs down the stem. Interception also significantly alters the characteristics of rainfall, which is among others reflected in differences in the number, size and velocity of raindrops. Throughfall drop size distribution was monitored and analysed for three vegetation types, including a single pine tree in an urban park, trees in an urban mixed forest, and a maize field in an agricultural area. Velocity-diameter diagrams were compiled for the 33 selected throughfall events and grouped into three distinct clusters based on similarity using a hierarchical clustering approach. Pine throughfall events were grouped in Cluster 1, urban mixed forest events in Cluster 2, while maize events were split between Clusters 1 (with all the pine tree events) and Cluster 3. A detailed analysis of rainfall microstructure characteristics under maize and pine canopies was conducted in relation to the rainfall event conditions and crop growing stage to evaluate why, in some cases, throughfall microstructure under maize is similar to that beneath pine (events assigned to Cluster 1), and, in other cases, it differs (events assigned to Cluster 3). Throughfall events in Cluster 3 were generally larger and more intense, showing a unimodal temporal distribution. In contrast, maize throughfall events in Cluster 1 exhibited a bimodal distribution, with two intensity peaks separated by a rainfall break. Notably, the maize leaf area index (LAI) exceeded a value of 4 during the period when the shift occurred from the events assigned in Cluster 1 to the subsequent events assigned in Cluster 3. As maize leaves mature, they become less flexible and do not bend as much under the weight of rain. Consequently, throughfall consist of more drips (larger drops) than direct rainfall (smaller drops). Further research could include additional types of vegetation, and the results could be supported by measurements over a longer period of time. These values could also be used for direct analyses of rainfall erosivity.Acknowledgment: This contribution is part of the ongoing research project entitled “Evaluation of the impact of rainfall interception on soil erosion” supported by the Slovenian Research and Innovation Agency (J2-4489) and the Austrian Science Fund (FWF) I 6254-N.

Increased flooding is becoming more prevalent under an increasingly variable future of weather extremes, highlighting the need for effective mitigation strategies. Different flood mitigation measures are available, ranging from classical structural (grey) solutions to nature-based solutions (NbS). This study assessed and compared the hydrological effectiveness, cost-effectiveness, and public perception of green (NbS), grey, and hybrid flood mitigation measures in the Gradascica River catchment, Slovenia. For the hydrological assessment, a SWAT + model simulated wetland, retention polder, and dam scenarios. Results showed that wetlands had a minimal effect on flood hazard, reducing flood peaks by up to 3 %, while retention polders and dams reduced flood peaks by 51 % and 73 % and flood volumes by 28 % and 58 %, respectively. The economic analysis found wetlands to be less cost-effective than retention polders and dams. However, it should be noted that wetlands provide additional diverse co-benefits. The public perception analysis revealed significant discrepancies in perceived effectiveness, feasibility, and acceptability of flood mitigation measures across target groups, including the general public, water engineers, researchers, and agricultural advisors. While most groups ranked dams as most effective and wetlands as least effective, aligning with hydrological findings, researchers held a directly opposing view, and the public generally overestimated the performance of green measures. By examining hydrological effectiveness, cost-effectiveness, and public perception across flood mitigation measures, the research highlights the need to integrate multidisciplinary approaches to develop robust flood management strategies - an essential lens as communities confront escalating climate-driven flood risks.

Climate change affects surface water quality parameters, including river quality. This study analyses changes in climate parameters, specifically air temperature and solar radiation, and their impact on river water temperature. It also examines how changes in river water temperature and organic matter load affect oxygen saturation levels, a key indicator of river water quality. Using water quality data, the status as well as temporal and spatial trends of the analysed parameters were assessed for the period between 2007 and 2024 on the three largest Slovenian rivers: the Drava, Mura, and Sava. Relative importance analysis of temperature and biochemical oxygen demand (BOD) using the Random Forest machine learning method showed that water temperature in the analysed rivers has an impact ranging from 51% to 66% on predicting oxygen saturation. The selected approach to analysing watercourse quality parameters enables the assessment of the impact of these parameters on river water quality. Based on these results, it will be possible to implement appropriate measures promptly to achieve sustainable river management by establishing a strategy that, under climate change conditions, safeguards water quality and maintains ecosystem protection, ensuring long-term ecological and socio-economic benefits.

Different types of vegetation shape throughfall in various ways, which affects erosion and runoff generation processes. Direct comparisons of throughfall drop size distributions (TF DSD) across different vegetation systems remain unexplored; therefore, this study quantifies TF DSD of 33 events under three distinct vegetation types (a pine tree, a mixed forest, and a maize field) within a unified framework. The amount and duration of TF was the highest in the forest, while TF under the maize exhibited the largest and the most uniform intensity regime. The diameter of the drops under the maize was the largest, resulting also in the highest proportion of drips (drops > 2 mm). In the forest, the proportion of drips remained constant between events, while under the pine tree the splashes prevailed (drops < 1 mm). Further analysis of the TF DSD with velocity-diameter diagrams, grouped according to their visual similarity with hierarchical clustering, resulted in three distinct clusters (C1-C3). Events under the pine were placed in C1 and those under the mixed forest in C2, while events under the maize were divided between C1 and C3. In comparison with C1, C3 events were characterised by increased TF intensity and larger TF drops. This indicates that there is a dynamic control of TF DSD under the maize, dependent on the interaction between the characteristics of rainfall and the canopy. Understanding these differences can support the design of vegetation cover in agricultural and urban landscapes to enhance infiltration, reduce soil erosion, and regulate runoff.

The first international summer school HydRoData for Master and PhD students was held in September 2023 at the University of Ljubljana, Faculty of Civil and Geodetic Engineering. The summer school was organised by the UNESCO Chair on Water-related Disaster Risk Reduction, University of Ljubljana, Slovenia and Slovenian national IHP programme. The focus of the summer school was data in hydrology. The programme topics included data acquisition, data manipulation and analysis, data curation, data communication, FAIR data principles, and introduction to R programming for hydrology. The teaching-learning process was structured as a combination of lectures, fieldwork, group work, ICT supported learning etc. In the scope of the summer school, participants partook the measurements of hydrological processes on several experimental plots, and visited meteorological station and radar during the field trip. To wrap everything up, the participants had the opportunity to show what they have learned in the competitive quiz in R programming. The official part of the summer school was enriched by social events, enabling the participants to network and get to know each other in more relaxed set-up. Social events included ice breaker trivia quiz pizza party, and a visit to traditional Slovenian tavern. The number of the applications exceeded the number of the available spots, and regardless on the new spots opening, a selection process was made. Finally, the 26 attending participants of 21 nationalities came from 19 universities. According to the feedback questionnaire, the participants evaluated the summer school execution with the average satisfaction grade 9.27 (out of 10). Here, the participant’s feedbacks that will assist in the improvement of the learning procedure, topic selection, schedule etc. will be presented more in detail along with the establishment and realization of the summer school. Since the first edition of the summer school showed to be successful, the second HydRoData summer school is announced, with applications already open. The HydRoData summer school 2024 will be held from 2 September to 6 September in Ljubljana, Slovenia. More information and registration form can be found: https://www.unesco-floods.eu/unesco-floods-summer-school/.

Study region: This study investigated rainfall at two locations in the Danube Basin, the Hydrological Open Air Laboratory (HOAL) agricultural catchment in Petzenkirchen, Austria and an urban experimental plot in Ljubljana, Slovenia. Study focus: The variability of rainfall characteristics and rainfall erosivity were explored using measurements of precipitation and drop size distributions in the period 2014–2018. Annual and seasonal rainfall event characteristics were analyzed for each site and comparatively assessed using hierarchical clustering. Annual and seasonal variability of rainfall erosivity was compared between the plots. New hydrological insights for the region: Despite having the same Köppen-Geiger climate classification, differences were found between the sites. The long-term average annual precipitation in Ljubljana was almost twice as high as in the HOAL. According to the clustering analysis, larger and more intense rainfall events occurred in Ljubljana than in the HOAL. The average drop sizes and velocities tended to be lower in Ljubljana but the range of drop size distributions was larger in Ljubljana than in the HOAL. The seasonalities of rainfall event characteristics and rainfall erosivity were similar at the sites. Rainfall intensities tended to peak in summer when rainfall durations were shorter, and larger and faster drops were observed. Rainfall erosivity was between 2 and 7 times larger in Ljubljana than in the HOAL because of the more intense rainfall and single faster and larger drops during events.

An extreme flood event occurred in Slovenia in August 2023. This study evaluated the influence of this extreme flood on the design discharges in Slovenia. This evaluation was based on flood frequency analysis for the data from 33 gauging stations. Analyses were conducted with and without the 2023 peak discharge, i.e., for the periods 1961–2022 and 1961–2023, using eight different theoretical distribution functions. In addition, specific discharge values for the 2023 flood event were analyzed and compared with regional envelope curves for Europe. The findings of the study indicate that the impact of a single flood event on the design discharge values can be substantial. Moreover, an analysis of the specific discharges resulting from the 2023 flood event in Slovenia reveals that the values for all gauging stations considered are below the regional envelopes. Concurrently, the analysis indicates that a flood event larger than the 2023 event may occur in the future.

The study compares changes in annual maximum (AM) discharge occurrence at 33 gauging stations in Slovenia for the period 1961-2023. The entire period was divided into two sub-periods, namely the period 1961-1990 and the period 1991-2023. The frequency of AM occurrence per day of the year was calculated for all stations under consideration, and the kernel density estimate was calculated for the two periods. The findings reveal that, except for one gauging station, the AM discharge has occurred later during the recent three decades (1991-2023) than during the first 30-year period (1961-1990). The shift ranged from one to 35 days. This shift could be mainly attributed to higher air temperatures in the spring and summer months, which intensify precipitation events. At four stations, the day of maximum density shifted from spring to autumn and, at one station, from autumn to spring. All of these stations are located in the eastern part of the country. In contrast, gauging stations in Alpine regions of the western part of the country show smaller shifts, attributed to reduced snow accumulation and earlier snowmelt.

Throughfall plays a significant role in hydrological processes, defining the effective rainfall available for soil moisture and runoff generation under vegetation. This study presents the first cross-climate comparison of the drivers of throughfall under black pine trees (Pinus nigra Arnold) in urban environments. Open rainfall and throughfall were measured at two experimental sites in the city of Ljubljana, Slovenia (temperate continental), and Sopron, Hungary (humid continental). To determine which of the climatic and canopy-related variables influence the event throughfall percentage (Tf) across three periods (i.e., whole study, growing and dormant periods), regression tree (RT) and boosted regression tree (BRT) models were used. The total rainfall and mean Tf recorded during the study period from September 2023 to September 2024 were 1591.2 mm and 45% +/- 21.5% in Slovenia, respectively. Conversely, Hungary experienced 767.2 mm and 50% +/- 27.2% . Both models confirmed rainfall (RA) as the primary driver of Tf across the three examined periods in both sites. Furthermore, the BRT model confirmed rainfall intensity as the secondary influential variable, specifically during Slovenia's growing and Hungary's dormant periods. Contrarily, the RT model showed relative humidity (RH) and leaf area index as the secondary variables defining the Tf over the whole study period at both sites, with variations across the growing and dormant periods. Under a scenario that included only the atmospheric variables, the BRT model identified RH as the most significant driver across all periods and both sites. Similarly, the RT model recognised RH as the primary variable during the three periods in Hungary and the entire study period in Slovenia. Moreover, air temperatures influenced Tf in Slovenia's growing and dormant periods. The findings indicated throughfall as a climate-sensitive parameter, emphasising the significance of these results in hydrological models susceptible to climate variability and in areas characterised by comparable climatic conditions.

Rainfall intercepted by vegetation is, in many regions, an important part of the hydrological water cycle. Part of the intercepted rainfall evaporates into the atmosphere, and throughfall and stemflow contribute to runoff generation, control soil moisture and runoff connectivity patterns and affect soil erosion. The question of how changing climate and land cover conditions impact rainfall interception, raindrop microstructure, and their erosive power still needs to be better understood.This presentation introduces the main aims of a bilateral research project between TU Wien, University of Ljubljana and the Slovenian Forestry Institute that focuses on the understanding of the effect of meteorological and vegetation characteristics on changes in raindrop microstructure and, therefore, on the erosive power of rainfall. The main idea of the research cooperation is to analyse and understand the main mechanisms of the rainfall interception process in different climate conditions and vegetation settings. The high-resolution disdrometer measurements from the experimental urban and forest plots in Slovenia and a small agricultural basin in Austria are used to determine and compare raindrop distributions and their changes. The high-resolution observations of discharge, sediment concentrations and isotope analyses contribute to the understanding of the erosion processes and sediment transport in the streams. Acknowledgment: This contribution is part of the ongoing research project entitled “Evaluation of the impact of rainfall interception on soil erosion” supported by the Slovenian Research and Innovation Agency (J2-4489) and it was funded in part by the Austrian Science Fund (FWF) I 6254-N.

In natural environments, rainfall causes soil erosion, which has a significant impact on the agricultural production and the ecological conditions of the streams. Due to different types of vegetation, their unique characteristics and seasonality, there are still a lot of open scientific questions about how rainfall interception process influences the rainfall erosivity and soil erosion. With the aim of improving knowledge about rainfall interception by different vegetation and its impact on the rainfall erosivity, an interdisciplinary and international research team (Faculty of Civil and Geodetic Engineering at the University of Ljubljana, Slovenian Forestry Institute and Technical University of Vienna) work together in the research project entitled “Evaluation of the impact of rainfall interception on soil erosion”. In the scope of the project, drop size distribution measurements above and below selected plants will be conducted in combination with classical measurements of rainfall partitioning. The measurements are ongoing in the small urban park in Ljubljana, Slovenia and in the experimental catchment with mainly agricultural land use in Lower Austria (The Hydrological Open Air Laboratory HOAL in Petzenkirchen). To evaluate the differences in rainfall characteristics for the two research plots, a comparative analysis on rainfall event properties such as rainfall amount, duration and intensity, size and velocity distribution of raindrops is performed. The aim of the presentation is to introduce the project and presents the first comparison of the rainfall characteristics at research plots in Austria and Slovenia. Acknowledgments: This contribution is part of the ongoing research project entitled “Evaluation of the impact of rainfall interception on soil erosion” supported by the Slovenian Research and Innovation Agency (project J2-4489) and the Austrian Science Fund (FWF) I 6254-N.

Trees, as a vital element of urban greening, have been increasingly recognized for their hydrologic contributions to stormwater management. However, the representation of tree canopy hydrological processes is often simplified or overlooked in existing stormwater models. This study modelled and evaluated the stormwater runoff reduction potential of birch (Betula pendula) and pine (Pinus nigra) trees in three scenarios (i.e., birch, pine, and mixed planting) on a storm event basis using the updated Storm Water Management Model (SWMM) tree canopy module. The model effectively represents the rainfall interception process of both tree species during different phenoseasons, demonstrating strong correlations between simulated and observed throughfall (r = 0.97-0.99) and interception values (r = 0.72) across all storm events. The results indicate that implementing urban trees in the study area led to an average reduction of 20-25 % in runoff volume and 16-25 % in peak flow, depending on the scenarios and phenoseasons. The most significant runoff reduction benefits were observed in a mixed-species planting scenario and during the leafed season. This interplay between species highlights the advantages of mixed-species plantings in urban environments, where diverse tree characteristics can enhance hydrological performance. However, the effectiveness of trees is limited during intense, high-volume storm events, although they still provide tangible benefits of up to 13.2 % reduction. The relative contribution of canopy interception to runoff reduction is most pronounced during the leafed season, small to moderate storm events, and when trees are situated over directly connected impervious areas. Infiltration and storage beneath tree canopies are the dominant mechanism for managing and reducing surface runoff, accounting for over 20 % of the water balance. This study demonstrates that the stormwater reduction efficiency of urban trees depends on both above- and below-canopy processes and conditions.

The process of interception, whereby vegetation partitions rainfall, largely influences natural processes such as soil erosion. The kinetic energy of rainfall plays a crucial role in evaluating this impact. In this study, we measured rainfall characteristics using three disdrometers placed above and below vegetation, specifically under two distinct tree species (birch and pine) in an urban area in Ljubljana, Slovenia. The study period extends over two years, subdivided into a dry and a wet sub-period. The investigation encompasses the effects of vegetation characteristics, raindrops characteristics and meteorological variables on the kinetic energy of throughfall. Two methods, namely boosted regression trees and random forest, were used to evaluate the influence of vegetation and meteorological variables on raindrop characteristics and their kinetic energy. The results indicate that, in general, pine reduces the kinetic energy of raindrops to a much greater extent than birch, and that birch exerts a positive effect on the reduction of kinetic energy only during the leafed period. Both applied machine learning models confirmed that the amount of throughfall has the greatest influence on kinetic energy, regardless of vegetation type. Furthermore, rainfall intensity and median-volume drop diameter exhibited a greater influence in the case of pine compared to the birch tree. Additionally, the findings indicate that the influence of the event duration on kinetic energy of throughfall differs depending on the presence of foliage on the tree canopy.

The rainfall erosivity influences the detachment of soil particles, movement and washing away the surface soil layers, which affects soil degradation and leads to various environmental problems. It depends primarily on the kinetic energy of raindrops, determined by the size and velocity of raindrops. However, rainfall microstructure (size, velocity and number of raindrops) is significantly changed during the process of rainfall interception. Precipitation, that is not intercepted by the vegetation, reaches the ground as throughfall (falling directly through the gaps in the canopy or dripping from the leaves and branches) or stemflow (flowing down the branches and stem). Therefore, the kinetic energy of throughfall under the vegetation is different than kinetic energy of open rainfall. In the urban park located in Ljubljana, Slovenia, we have monitored the rainfall microstructure in the open and underneath the deciduous (Betula pendula Roth.) and coniferous (Pinus nigra Arnold) trees between 12 July 2022 and 19 July 2023. We have analysed the differences between rainfall microstructure and kinetic energy of raindrops in the open and underneath the trees. The observed average number of raindrops per event under both trees was lower than the number of raindrops in the open. Also, the average kinetic energy of drops per event was significantly lower under the trees than in the open. Additionally, an analysis of factors influencing the kinetic energy of throughfall drops underneath the both trees was performed using the boosted regression trees and random forest models. Both models identified rainfall amount as the most influencing factor. Acknowledgments: Results are part of the research programme P2-0180 and research projects J2-4489, N2-0313 and J6-4628, financed by the Slovenian Research Agency (ARIS).

Hydrological modelling can be complex in nonhomogeneous catchments with diverse geological, climatic, and topographic conditions. In this study, an integrated conceptual model including the snow module with machine learning modelling approaches was implemented for daily rainfall -runoff modelling in mostly karst Ljubljanica catchment, Slovenia, which has heterogeneous characteristics and is potentially exposed to extreme events that make the modelling process more challenging and crucial. In this regard, the conceptual model CemaNeige Ge nie Rural a ` 6 parame `tres Journalier (CemaNeige GR6J) was combined with machine learning models, namely wavelet -based support vector regression (WSVR) and wavelet -based multivariate adaptive regression spline (WMARS) to enhance modelling performance. In this study, the performance of the models was comprehensively investigated, considering their ability to forecast daily extreme runoff. Although CemaNeige GR6J yielded a very good performance, it overestimated low flows. The WSVR and WMARS models yielded poorer performance than the conceptual and hybrid models. The hybrid model approach improved the performance of the machine learning models and the conceptual model by revealing the linkage between variables and runoff in the conceptual model, which provided more accurate results for extreme flows. Accordingly, the hybrid models improved the forecasting performance of the maximum flows up to 40 % and 61 %, and minimum flows up to 73 % and 72 % compared to the CemaNeige GR6J and stand-alone machine learning models. In this regard, the hybrid model approach can enhance the daily rainfall -runoff modelling performance in nonhomogeneous and karst catchments where the hydrological process can be more complicated.

Recently, there has been a growing recognition that the generic values of runoff coefficient (C) and curve number (CN) from standard lookup tables in the literature may not accurately capture the specific characteristics of a given catchment. This study aims to estimate and compare the event C and CN of two contrasting urban catchments in Ljubljana, Slovenia using observed rainfall-runoff data. Estimated parameter values were compared to the tabulated values from the literature (i.e., ASCE, 1992; USDA-NRCS, 2004). Seasonal changes of C and CN, along with diurnal streamflow patterns, were also analyzed to understand the eco-hydrological processes in the urban forest. The results demonstrated that the urban mixed forest exhibited high variability of Cs across all rainfall events with a mean and median of 0.11 and 0.062, respectively. Most of the C values observed in the urban area are clustered around the central tendency with a mean and median of 0.60. These mean C values in both catchments were lower than the tabulated values in the ASCE (1992) design manual. Mean and median CN values in the urban area were 95.45 and 96.81, respectively, lower than the urban mixed forest's CN values of 82.69 and 83.95. An asymptotic CN infinity infinity of 90.69 was found for the urban area and 71.71 for the urban mixed forest. Central tendency-derived CN values tend to be slightly higher than the tabulated values from USDA-NRCS (2004) and gridded values from GCN250 (Jaafar et al., 2019), while asymptotic CN infinity infinity values were lower. Using central tendency measures as the single lumped value of C and CN may provide a reasonable representation of the runoff behavior in the studied urban area. However, this may present certain challenges in the urban forest catchment. Additionally, pre-event soil moisture conditions and specific storm characteristics contributed to the observed variations in C and CN. Bi-monthly analysis showed that C and CN were high during the autumn and winter months. Diurnal streamflow pattern is most prevalent during low-flow and precipitation- free periods while exhibiting seasonal structures in terms of the shape and timing of maxima and minima. Local estimation of C and CN allows for a more tailored representation of the catchment's hydrological behavior.

Hydrological modelling, essential for water resources management, can be very complex in karst catchments with different climatic and geologic characteristics. In this study, three combined conceptual models incorporating the snow module with machine learning models were used for hourly rainfall-runoff modelling in the mostly karst Ljubljanica River catchment, Slovenia. Wavelet-based Extreme Learning Machine (WELM) and Wavelet-based Regression Tree (WRT) machine learning models were integrated into the conceptual CemaNeige Génie Rural à 4 paramètres Horaires (CemaNeige GR4H). In this regard, the performance of the hybrid models was compared with stand-alone conceptual and machine learning models. The stand-alone WELM and WRT models using only meteorological variables performed poorly for hourly runoff forecasting. The CemaNeige GR4H model as stand-alone model yielded good performance; however, it overestimated low flows. The hybrid CemaNeige GR4H-WELM and CemaNeige-WRT models provided better simulation results than the stand-alone models, especially regarding the extreme flows. The results of the study demonstrated that using different variables from the conceptual model, including the snow module, in the machine learning models as input data can significantly affect the performance of rainfall-runoff modelling. The hybrid modelling approach can potentially improve runoff simulation performance in karst catchments with diversified geological formations where the rainfall-runoff process is more complex.