Two Hidden Model Settings Steer the Fate of Global Monsoon Rainfall

Monsoons deliver the water that billions of people depend on, yet climate models still disagree sharply about how the rains of the future will unfold. A new study published in Climate Dynamics by Huiting Zhang of Nanjing University and colleagues has taken aim at that disagreement, and the result is a strikingly concrete answer: much of the uncertainty in global monsoon precipitation projections can be traced back to just two tunable settings inside the models’ cloud and convection schemes. Rather than pointing vaguely at model imperfections, the team identified the specific parameters that matter, the physical pathways through which they act, and the regions of the monsoon world where each one dominates.

The study’s power comes from its experimental design. Instead of comparing many different models built by different groups, which makes it hard to isolate causes, the researchers used a single coupled atmosphere-ocean model and ran an ensemble of 200-year-long simulations in which sixteen cloud and convection parameters were simultaneously perturbed across plausible ranges. This perturbed parameter ensemble approach, a technique with a long pedigree in climate and atmospheric science, allows each simulation to differ from the others only in the values of these carefully chosen knobs. Any spread in the projected behavior of the monsoon can therefore be attributed directly to the uncertain physics encoded in those parameters, rather than to differences in model structure, resolution, or initialization.

The spread the team found was substantial. Projections of both mean and extreme monsoon precipitation varied widely across the ensemble, and the uncertainty was particularly large over the oceanic portions of the monsoon systems. This matters because the monsoon is a coupled land-ocean-atmosphere phenomenon: the great seasonal reversals of wind that draw moist air from the tropical oceans onto the continents are fueled by sea surface temperature patterns and by the convective storms that grow over warm water. If the models cannot agree on how oceanic convection will respond to warming, the entire supply chain of monsoon moisture becomes uncertain, with direct consequences for projections of both floods and droughts over the adjacent land.

Within this sea of uncertainty, the study uncovered a clean organizing principle. Globally, projected monsoon precipitation does tend to rise with the amplitude of global-mean surface warming, consistent with the familiar thermodynamic expectation that a warmer atmosphere holds more water vapor. But when the researchers looked at local precipitation changes across the ensemble members, the controlling variable was not how much the planet warmed. Instead, it was a quantity the authors call precipitation sensitivity to surface warming: how strongly a given configuration of the model converts each degree of warming into additional rainfall in a particular region. The one clear exception was the North American Monsoon, where the warming amplitude itself played the leading role, a distinction that turned out to unlock the physical story behind the ensemble’s most concerning result.

Two parameters emerged as the dominant levers. The first, named cf_rhminl1, is the relative humidity threshold below which low clouds cannot form in the model. Because low marine clouds reflect sunlight and cool the underlying ocean, the value of this threshold shapes how much the southeastern Pacific warms and, through that, the global-mean warming amplitude itself. The second, zm_ke, is the evaporation rate of convective precipitation, which determines how much rainfall falls as raindrops evaporate on their way down rather than reaching the surface. Strikingly, the study found that these two parameters influence the monsoon through entirely different channels: cf_rhminl1 acts mainly by changing the amplitude of surface warming, while zm_ke acts mainly by changing the local precipitation sensitivity to that warming.

The North American Monsoon finding deserves particular attention, because it links a small parameter choice to a large-scale drying signal. In ensemble members with a higher value of cf_rhminl1, the projected reduction in North American Monsoon precipitation was stronger. The authors traced this to enhanced warming over the southeastern Pacific, which can drive a divergence of moisture flux away from the monsoon region, effectively robbing it of its water supply before it arrives. The chain of causation runs deeper still: the southeastern Pacific warming response is tied to the cloud feedback effect, because the downward transport of dry air into that cloud deck depends on the baseline temperature of the historical climate, and that baseline temperature is itself strongly regulated by the cf_rhminl1 threshold. A single calibration choice about low clouds, made years before the simulation is run, cascades through cloud feedbacks, regional warming patterns, moisture transport, and finally into the projection of summer rain over northwestern Mexico and the southwestern United States.

The role of zm_ke is arguably even more consequential for most of the monsoon world. This parameter has limited influence on how much the planet warms, yet it exerts a dominant control on projected precipitation across many monsoon regions. The reason lies in elementary thermodynamics: the evaporation of falling raindrops depends on the saturation vapor pressure of the air, which is a steep, exponential function of temperature. Because the parameter governs a process whose efficiency changes systematically as the globe warms, its fingerprints on future precipitation turn out to be broadly consistent with its fingerprints in the historical climate. In other words, the way a model configuration behaves in observations of the past twentieth century is a meaningful guide to how it will behave in a future warming world, at least for this particular process. That continuity is a genuine gift for the field, because it means the historical record can, in principle, be used to constrain the most influential uncertainty.

The broader implication of the study is that monsoon projection uncertainty is not an amorphous cloud of model disagreement but a traceable quantity with identifiable origins. The finding that cloud and convection parameters dominate echoes a well-known result in climate sensitivity research, where the spread across models has been traced to the representation of atmospheric convective mixing. Here the same philosophy, applied to a coupled model with a purpose-built parameter ensemble, shows that the monsoon problem is tractable: by targeting observations at the processes governed by cf_rhminl1 and zm_ke, the scientific community could shrink the projection spread for the rainfall systems on which a large fraction of humanity depends. Observational campaigns measuring low-cloud formation conditions over the eastern subtropical oceans and the microphysics of evaporating rain in tropical convective storms would directly address the two most influential uncertainties.

There is also a caution embedded in these results. The exceptional behavior of the North American Monsoon, where warming amplitude rather than precipitation sensitivity controls the outcome, shows that a single constraint strategy will not work everywhere. Different monsoon regions respond to different aspects of the model physics, so efforts to narrow uncertainty must be tailored region by region. For the many monsoon regions governed by precipitation sensitivity, constraining zm_ke through historical observations holds real promise. For the North American Monsoon, the relevant constraint concerns cloud feedbacks and the pattern of southeastern Pacific warming. The study, published on 23 September 2026 as volume 64, article 434 of Climate Dynamics, thus does more than quantify uncertainty; it provides a roadmap for reducing it, parameter by parameter, region by region.

For the public, the takeaway is vivid. The fate of the world’s monsoons, from the rains that feed the Ganges to those that fill reservoirs in the American Southwest, hangs partly on hidden numerical thresholds inside climate models, choices about how low clouds form and how fast raindrops evaporate. The work of Zhang, Yang, Wu, Dai, Fu, Zhang, and Guo demonstrates that these choices can be identified, their physical consequences mapped, and their influence on human-scale water security quantified. That is a crucial step toward climate projections that water managers, farmers, and governments can actually plan around, and a reminder that the path from a line of Fortran code to the future of a continent’s rainfall is shorter than most people imagine.

Subject of Research: Parametric uncertainty from cloud and convection parameters in climate model projections of global monsoon precipitation

Article Title: Parametric uncertainty in future projection of global monsoon precipitation: roles of surface warming amplitude versus precipitation sensitivity to warming

Article References: Zhang, H., Yang, B., Wu, W., Dai, G., Fu, H., Zhang, Y., & Guo, Z. (2026). Parametric uncertainty in future projection of global monsoon precipitation: roles of surface warming amplitude versus precipitation sensitivity to warming. Climate Dynamics, 64(10), Article 434. https://doi.org/10.1007/s00382-026-08380-0

Image Credits: AI Generated

DOI: 10.1007/s00382-026-08380-0

Keywords: global monsoon, precipitation projection, climate model uncertainty, perturbed parameter ensemble, cloud feedback, convection parameterization, low clouds, precipitation sensitivity, surface warming, North American Monsoon, southeastern Pacific, Climate Dynamics