Coupled effects of surface erosion and magma intrusion on magmatic hazards under a changing climate Authors: Riccardo Lanari, Mohammad Moumeni, Riccardo Reitano, Marco Bonini, Francesca Funiciello, Chiara Del Ventisette, Domenico Montanari Published: 2026 ABSTRACT The impact of rainfall-driven erosion on the evolution of magmatic systems and, ultimately, on volcanic activity, is an emerging topic of critical importance in the context of the ongoing climate crisis. While tectonic forces control magma intrusion through the crust, climatic factors (i.e., rainfall) may influence magma ascent and contribute to volcanic eruptions. Here, we present a set of innovative analogue models integrating uplift driven by magma intrusion, and surface processes triggered by a rainfall system. By varying magma injection rates and rainfall intensities, we examine surface morphology under different conditions. Our models reveal that the topographic elevation, a proxy for the effectiveness of surface processes, does not correlate with increasing rainfall intensity. Only under heavy rainfall rates, the system significantly erodes and relocates material from the top of the rising dome to its rim, likely modulating magma ascent. Under this condition, the resulting topographic dome shows the lowest elevation and the greatest accommodation volume. We then speculate that the erosional unloading of the overburden could increase the magmatic chamber’s force to cause uplift. Ultimately, although the present work represents an initial study, we emphasize how extreme rainfall events might drive magmatic systems toward a more critical state.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-54985-0. Introduction The accelerating climate crisis highlights the urgent need to improve our understanding of magmatic systems response. Climate change models predict an increase of up to 40% of heavy rainfall events (depending on the latitude) in response to a global temperature rise of up to 5 C°1. The consequences of this increase are dramatic, as extreme and even seasonal rainfall has already been identified worldwide as a trigger for volcanic activity and seismicity2–4. Rainfall-driven erosional unloading modulates vertical crustal stress, facilitating magma ascent to the surface5–9, while the more increasingly extreme rainfall events can modulate volcanic activity, potentially leading to unexpected eruptions10. Seasonal rainfall significantly weakened the Mount Etna edifice (southern Italy), which initiated mechanical tensile failure in the volcanic flank promoting subsequent dyke intrusion triggering an eruptive cycle11. Unexpected volcanic eruptions also occurred in Kilauea Volcano (USA) following intense rainfall, as well as at Monserrat volcano12,13. In these examples, however, rainfall acts as a trigger for magmatic processes that are already near a critical state, whereby the long-term impact of extreme rainfall on magmatic systems is still unexplored. This study provides an initial investigation into the long-term effects of rainfall intensity on magma-driven topographic uplift, based on a series of specifically designed analogue models (Fig. 1). Pioneering analogue models of magma intrusion have provided key insights into describing the topographic shape after magma intrusion14–18, while other studies examine the interplay between surface processes and mechanical uplift19,20. For the first time, our models provide insights into how surface processes affect the topographic doming caused by magma emplacement at shallow crustal levels. We tested different boundary conditions, by experimenting with three magma injection rates alongside three rainfall rates. Thus, our models analyse how topography, uplift, and volume accommodation change across a broad range of the non-dimensional ratio between rainfall rate and injection rate (Ri, see also Table 1). Model setup and strategy We used a Plexiglas box with internal dimensions of 33 × 38 × 15 cm315,16, filled with a water-saturated mixture of granular materials designed to simulate both crustal deformation and surface processes21–24. The mixture consists of 40 wt% Silica Powder (SP), 40 wt% Glass Microbeads (GM), and 20 wt% Polyvinyl Chloride powder (PVC), with 25 wt% water added during preparation22. This mixed material has a mean grain size of 83 μm, a particle density of 2190 ± 1 kg/m3, porosity of 0.32, permeability of 2.9 × 10− 13 m2, and an internal friction angle of 25–36 ° (peak) and 23–34° (stable). The cohesion of the material is ca. 1 kPa, though it is highly dependent on the water content. To simulate magmatic intrusions, we used a low-viscosity Newtonian fluid: pure vegetable polyglycerine-10 (PG10) which has a viscosity of 450 Pa s and a density of 1200 kg/m³. PG10 are inter-molecular glycerol ethers, exclusively manufactured by Spiga Nord S.p.A. through the polycondensation of natural refined glycerine. The model setup was centered on the injection point, where PG10 was introduced through a cylindrical metallic pipe (1.4 mm internal diameter) positioned at the base of the model, forced to fill an air balloon covered by an 8 cm-thick sand layer (Fig. S1). Introducing the air balloon to trap the PG10 at the bottom of the box (Fig. S1) was intended to create topography and at the same time preventing an eruption of analogue magma at the surface, a phenomenon we did not intend to investigate in the present model series. From a comparison with previous model series48-54.15,16, it is possible to evaluate how this technical expedient does not preclude the potentiality of studying magmatic systems. A total of 360 cm³ of magma-analogue PG10 was injected at a rate controlled by a stepper motor connected to a central unit, enabling real-time monitoring of the injection rate throughout the model. Analogue modelling of magma intrusions inherently requires multiple simplifications. These primarily relate to the rheological behaviour of the analogue magma and its evolution during melt cooling and crystallization, as well as the rheology and thermo-mechanical evolution of the cover rocks during magma emplacement. Since the discussed models simulate magma emplacement at very shallow crustal levels, where thermal effects impact only a a very thin layer of the cover rocks, these aspects can be considered negligible for the purposes of this study12,13. The rainfall system consists of one or more sprinklers positioned above the models. Rainfall intensity and rate are controlled by the number of sprinklers and the volume of water pumped into the system. Our models aim not to simulate specific extreme rainfall events, but to explore how varying the efficiency of surface processes relative to magma injection modifies topographic evolution and its effect on the evolution of the magmatic system. Even if rainfall intensity does not precisely scale to meteorological values, its geomorphic influence on evolving structures remains dynamically comparable to natural systems, as discussed by Paola et al. (2009). We then tested three rainfall rates (620, 1088, and 2664 cm³ h⁻¹), in combination with three magma injection rates (29, 78, and 327 cm³ h⁻¹). The range of rainfall and injection rates used in this study is constrained by the technical characteristics of the model apparatus yet remains within a reasonable and experimentally tested interval capable of reproducing realistic interactions between uplift and surface processes, producing results that are both comparable and scalable to natural systems201921,22,25,26. To investigate the response of magmatic dome growth and morphology to variations in rainfall, we used the dimensionless number Ri resulting from the ratio between rainfall and magma injection rates, which ranges between ~ 2 and ~ 35. This approach allows us to investigate the trade-off between rainfall and injection rates in the development of forced folds morphology. After every 72 cm³ of injected material, the surface was scanned using a high-resolution laser scanner (Artec Leo 3D Scanner, resolution of 0.1 mm), generating digital elevation models (DEMs) and pictures (Fig. 1) that are subsequently analyzed using GIS and MATLAB software (see supplementary material). These scans provided data on the evolving dome morphology, from which we extracted parameters such: Topographic Peak (TP), which is the maximum elevation the models reach at same injected material; and the Topographic Median (TM), which is the mode of the topographic elevation the models achieve at same injected material. These topographic vertical variations represent the commonly called ‘surface-uplift’ since account for the erosion and sedimentation. Since isolating each signal from the other is challenging, to gain insight for the rock-uplift, we extract the Volume Accommodation (VA), which is the surface positive volume accumulated by the model topography at every magma injection. From the acquired DEMs we subtracted the initial topography from subsequent stages, quantifying the volume increase associated with dome formation due to the accommodation space created by magma intrusion. To ensure accurate VA measurements and mass conservation, the model design prevented material from exiting the model, effectively creating a closed system. This containment enables reliable volume balance calculations since the granular material eroded and transported was sedimented within the box, thus included in the topographic scan and accounted to the volume calculation. Due to the laser resolution, error bars are within the symbols in Fig. 2. Scaling of the models This study focuses on general processes, and so our models were not intended to investigate specific regional settings. Replicating the specific original crustal architecture or pre-existing deformation or topography, for example, would be both challenging and possibly an overparameterization of the issue. Nevertheless, the reliability of models depends on the ability to scale data to natural prototypes. Models must then be scaled in terms of geometry, kinematics, dynamics, and rheology to allow meaningful comparison with nature27–30. In geomorphic models, rigorous scaling is rarely achieved, particularly with respect to fluid dynamics (e.g., Reynolds and Froude numbers31. Nevertheless, similarity theories allow us to infer data that can be extrapolated to natural systems31,32. We assumed a length scaling factor\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{L}^{*}=\frac{{L}_{m}}{{L}_{n}}={10}^{-5}$$\end{document} so that 1 cm in the model corresponds to 1 km in nature. Considering an average density of 2700 kg m⁻³ for crustal rocks and 1700 kg m⁻³ for the analogue brittle material, the density scaling factor is\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\rho\:}^{*}\approx\:0.6.$$\end{document} The stress scaling factor can then be derived as\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}^{*}={\rho\:}^{*}{g}^{*}{L}^{*},$$\end{document} and because models evolve under natural gravity \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{g}^{*}=1,$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}^{*}\approx\:6\times\:{10}^{-6}.$$\end{document} Constraining model time scaling remains particularly challenging in geomorphic models, so various semi-empirical methods have been applied21–25,32–36. As this study is primarily concerned with the surface response to magma emplacement and its feedback, rather than with magma-related structures and dynamics, we define the time-scaling factor as follows:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}^{*}=\frac{{L}^{*}}{{v}^{*}}.$$\end{document} Erosion rates v in models typically range from 1 to 10 mm h⁻¹22,23,32,35, whereas natural prototypes exhibit erosion rates of about 0.1–1 mm yr⁻¹37,38. This implies \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{v}^{*}$$\end{document} ≈ 10⁵ and therefore \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}^{*}$$\end{document} is of the order of 10⁻⁹–10⁻¹⁰, meaning that 1 min in the model corresponds to roughly 2–20 kyr in nature. For subsequent calculations, we adopt \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{t}^{*}={10}^{-9}$$\end{document}. Topographic response at mid-low rainfall/injection ratios Considering that rainfall triggers erosion, we hypothesize that intense rainfall should enhance surface processes. Nevertheless, the actual efficiency of erosion remains debated, as it is likely controlled not only by rainfall intensity but also by its interaction with topographic gradient, which depends on the rate of uplift24,39,40. To gain insights into the linkage between rainfall/uplift and erosion, we consider topographic elevation (both peak and median) as a proxy for assessing the efficiency of surface processes; higher topography (especially the median) would imply lower erosion. In our models, topographic growth against the absolute injected material, in both TP and TM follow a logarithmic trend, with a more rapid increase during the early stages (Fig. 2a, c). This pattern remains overall consistent up to Ri exceeding ~ 21.2 (Fig. 2a and c; red symbols). This logarithmic trend shows that, in the mature stage, the system evolves towards a topographic steady state where erosion and uplift are quasi-balanced41, in agreement with previous studies20. When TP and TM are plotted against Ri (Fig. 2b, d), each evolutionary magma injection step exhibits an overall negative second-degree polynomial trend (see supplementary material). Indeed, if focussing on Ri up to ~ 21.2, both TP and TM gently increase (or remains relatively stable) at increasing Ri. According to our models, at least up to Ri of ~ 21.2, the increasing rainfall rate relatively to magma injection rates does not enhance the efficiency of surface processes. This behaviour highlights the importance of the coupling between rainfall intensity and slope gradient. In particular, the topographic median is generally slightly lower at low Ri values, when rainfall is not exceeding the injection rate. Under these conditions, erosion is more effective since it is influenced by the relative high injection and uplift rates (Fig. 2b, d). As expected, this effect is most evident during the early stages of the models (Figs. 2 and 3), when differing injection (and uplift) rates produce markedly different topographic gradients. This pattern highlights the temporal variability in the efficiency of surface processes during the same models as the landscape evolves toward a more stable configuration (Figs. 2 and 3). Conversely, VA increases linearly with the injected volumes and, when plotted against Ri, it exhibits an overall positive second-degree polynomial trend (see Supplementary material). Focussing on Ri values up to ~ 21.2, VA gently decreases as Ri increases, displaying an opposite trend to the topographic metric. We interpret this behaviour as the brittle material’s response to the imposed injection (and uplift) rates combined with the absolute duration of the models. At low Ri, the relatively high injection rate allows the model to finish in about one hour, whereas lower injection rates require more than four, or even twelve hours, to inject the full 360 cm³ of magma-analogue. At least up to Ri of ~ 21.2, we consider this timing crucial in limiting the granular material’s ability to compact by expelling excess water, which likely accounts for the modest variation in VA that we observe. Overbalance ratio between rainfall and injection rates: potential implication for magmatic systems The analogue models developed in this study show that rainfall rates can exceed magma injection rates by up to ~ 21.2 without substantially altering the overall pattern of landscape evolution (Fig. 4). Only at very high Ri (34; Fig. 2; red symbols) surface processes become predominantly governed by rainfall (Figs. 2 and 3). In this regime, the erosive power of rainfall overwhelms other concurrent processes, leading to a pronounced reduction in topographic elevation in both TP and TM (Figs. 2 and 3). The pronounced effectiveness of rainfall-driven surface processes is further evidenced by the increased areal extent of the resulting dome (Figs. 1 and 4 and Supplementary material). Under these exceptionally wet conditions, VA reaches its highest values (Fig. 2e and f) which we exclude from being attributed to the greater amount of less compacted material eroded, transported, and resedimented at the base of the topographic dome relative to the uneroded material (due to oversaturation with water). The resulting volume differences are on the order of only tens of mm³, significantly smaller than the approximately 100 cm³ discrepancy observed in models at lower Ri. The strategy employed in performing parametric studies, such as the one presented here, is to isolate the individual contribution of a single parameter in controlling the observed variations, while keeping the other parameters fixed. This applies to this model with high erosive power of rainfall, which produces a dome with the lowest elevation and highest volume of accommodation. Based on this model observation, we document potential positive feedback between the very intense rainfall rate and the long-term evolution of the magmatic system, as indicated by the highest accommodation volumes. Understanding the physical process underlying this interaction requires additional, targeted studies. Nonetheless, we endeavour to provide an initial interpretation of this behaviour, because of the enhanced erosion and sediment redistribution from the summit of the dome to its margins (Fig. 4). Such substantial redistribution of overburden material likely reduces the lithostatic pressure acting on the analogue magmatic chamber, thereby favouring rock uplift and potentially explaining the unexpected increase in VA at high Ri values (Fig. 4). Conclusions Our models suggest that climate-driven surface processes, particularly under progressively intensified rainfall, may exert an unforeseen influence on long-term subsurface magmatic dynamics, which has been previously recognized only at shorter temporal scales6,42,43. The innovative set-up presented in this study may serve as an initial step toward understanding how extreme rainfall and the associated erosion might drive magmatic systems toward shallow levels, inferenced by the dome with the lowest topographic elevation and highest accommodation volume. However, it is important to note that our models are not designed to reproduce natural systems on a one-to-one basis; rather, they suggest a potential ongoing process. Firstly, although rainfall is not scaled to natural rainfall rates or discrete events22,25,36,44,45 it acts primarily as a trigger mechanism for surface processes, thereby integrating climatic effects over long time periods. Additionally, the mechanical framework investigated in this study corresponds to a constant long-term cold–brittle crustal regime (Fig. 4). As such, our experimental conditions do not consider other plausible rheologies that are dependent on thermal evolution associated with sustained magma intrusion (e.g.46,47). Nonetheless, our results can be interpreted as representative of the cold–brittle endmember of crustal behaviour, where intrusion-induced pressure is effectively transmitted to the surface resulting in a direct uplift. Under these conditions, the stress transferred between the magma body and the overburden may remain mechanically effective during erosional unloading, potentially modifying the vertical stress field above the intrusion (Fig. 4). Ultimately, as extreme weather events are expected to become more frequent in a warming climate, incorporating these feedbacks and additional analyses into volcanic hazard models will be essential to enhance long-term risk assessments and strengthen preparedness for future climate–volcano interactions1. Table 1Details of the rainfall and injection rates and synthesis of the Ri values. For details, see the supplementary information.Volume Injected (cm3)Effective time of experiment (min)Vol/h (cm3)HighInjection36066327MidInjection36027678LowInjection36074029Rainfall g/15 minutesVol/h (cm3)LowRain155620MidRain2721088HighRain6662664NameRiLowRain_HighInjection1.9MidRain_HighInjection3.3LowRain_MidInjection7.9HighRain_HighInjection8.1MidRain_MidInjection13.9LowRain_LowInjection21.2HighRain_MidInjection34.0 Table 1Details of the rainfall and injection rates and synthesis of the Ri values. For details, see the supplementary information.Volume Injected (cm3)Effective time of experiment (min)Vol/h (cm3)HighInjection36066327MidInjection36027678LowInjection36074029Rainfall g/15 minutesVol/h (cm3)LowRain155620MidRain2721088HighRain6662664NameRiLowRain_HighInjection1.9MidRain_HighInjection3.3LowRain_MidInjection7.9HighRain_HighInjection8.1MidRain_MidInjection13.9LowRain_LowInjection21.2HighRain_MidInjection34.0 Details of the rainfall and injection rates and synthesis of the Ri values. For details, see the supplementary information. Fig. 1Images of the seven models. Columns represent individual models arranged by increasing rainfall to injection ratios. Colors represent discrete Ri intervals (see main text for details). Rows show successive model stages, defined by the percentage of material injected, from 20% to 100%. See Table 1 and supplementary for details on rainfall and injection rates. Images of the seven models. Columns represent individual models arranged by increasing rainfall to injection ratios. Colors represent discrete Ri intervals (see main text for details). Rows show successive model stages, defined by the percentage of material injected, from 20% to 100%. See Table 1 and supplementary for details on rainfall and injection rates. Fig. 2Left panels: injected volume plotted against (a) Topographic Peak, (c) Topographic Median, and (e) Volumes Accommodation. Right panels: ratio between rainfall rates and injection rates (Ri) plotted against (b) Topographic Peak, (d) Topographic Median, and (f) Volumes Accommodation. Symbols represent the measured variables: squares for Topographic Peak, triangles for Topographic Median, and circles for Volumes Accommodation. Colors indicate discrete Ri intervals (see Table 1 for details), and symbol outlines correspond to different model stages. Left panels: injected volume plotted against (a) Topographic Peak, (c) Topographic Median, and (e) Volumes Accommodation. Right panels: ratio between rainfall rates and injection rates (Ri) plotted against (b) Topographic Peak, (d) Topographic Median, and (f) Volumes Accommodation. Symbols represent the measured variables: squares for Topographic Peak, triangles for Topographic Median, and circles for Volumes Accommodation. Colors indicate discrete Ri intervals (see Table 1 for details), and symbol outlines correspond to different model stages. Fig. 3Topographic Peak, Topographic Median and Volume Accommodation plotted as a function of the boundary conditions applied, for every timestep of injection. See Table 1 and supplementary details about rates for both rainfall and injection. Colored border indicates the different Ri shown in Figs. 1 and 2. Topographic Peak, Topographic Median and Volume Accommodation plotted as a function of the boundary conditions applied, for every timestep of injection. See Table 1 and supplementary details about rates for both rainfall and injection. Colored border indicates the different Ri shown in Figs. 1 and 2. Fig. 4Conceptual model showing three models (columns) at increasing Ri (from left, relatively low, medium and high rainfall); and at two different model stages (rows): early stage with 20% of injected material at the bottom; and mature stage with 100% injected material at the top. The red magma chamber is schematic and not scaled. Conceptual model showing three models (columns) at increasing Ri (from left, relatively low, medium and high rainfall); and at two different model stages (rows): early stage with 20% of injected material at the bottom; and mature stage with 100% injected material at the top. The red magma chamber is schematic and not scaled. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2