Better weather forecasts are often presented as one of the most practical tools for adapting to climate change. If farmers and livestock owners know a difficult season is coming, the logic goes, they can make better decisions before drought turns into crisis.
New modelling of southern African rangelands suggests that information may only be part of the solution.
A study published in Nature Sustainability on 14 September 2026 found that increasingly variable rainfall could reduce pastoralists’ average economic well-being while widening wealth inequality. Weather forecasts could help people adjust herd management, but forecasts on their own were often insufficient to protect communities across the climate scenarios tested.
The strongest results emerged when forecasting was paired with risk-smoothing support, particularly supplemental fodder during poor rainfall periods. In the model, that combination reduced some of the trade-offs between household wealth, inequality, livestock numbers and vegetation condition that appeared when interventions were used separately.
A model built around seven southern African countries
Researchers Matt Clark, Cosima Fröhner and colleagues constructed a spatial agent-based model using empirical distributions of rainfall, cattle populations and community economic well-being from Botswana, Eswatini, Lesotho, Mozambique, Namibia, South Africa and Zimbabwe.
The model represented 100 communities, each containing four grazing plots. Initial herd sizes varied stochastically but generally ranged from about 15 to 75 animals per community, with a median of around 40. Communities also differed in economic resources, reflecting the substantial inequality present in the regional data used to construct the stylised landscape.
The researchers then exposed this system to three rainfall-variability scenarios linked to SSP1-2.6, SSP2-4.5 and SSP5-8.5. These represented futures ranging from lower emissions to substantially greater fossil-fuel development.
The contrast between scenarios was substantial. Maximum modelled wet-season rainfall variability under SSP2-4.5 was 1.3 times the level used for SSP1-2.6 and the pre-industrial comparison. Under SSP5-8.5 it was 3.3 times as high.
The researchers stress that the model is intended to expose system dynamics rather than predict the exact future of individual communities. That distinction matters. The results are better read as evidence about how interventions can interact under different conditions than as a forecast of a particular village’s income or herd size in 2060.
Greater rainfall variability hurt wealth and increased inequality
The first experiment asked what happened when rainfall became more variable without widespread new adaptation measures.
The answer was consistently negative. As precipitation variability increased, average livestock numbers and economic well-being fell. Wealth also became more unequal.
The mechanism behind that inequality is important. Poorer livestock owners had less capacity to buy fodder when drought reduced available grazing. They consequently faced greater livestock losses and, in some cases, exited livestock rearing altogether. Better-resourced households were more able to absorb the shock.
Vegetation cover changed less dramatically than household economic outcomes. After dry periods reduced livestock numbers, subsequent vegetation growth could temporarily face less grazing pressure. This created a buffering effect in the ecological part of the model even while households experienced economic losses.
That divergence illustrates one of the study’s central themes: an intervention can look successful against one measure while performing poorly against another.
Forecasts helped, but information alone had limits
The second experiment scaled a conservation initiative from 10% to 90% of communities. The baseline intervention combined rotational grazing with improved access to livestock markets. Researchers then tested whether outcomes changed when it was supplemented with weather forecasts, supplemental fodder, or both.
Rotational grazing and better market access alone had relatively little effect on the four main outcomes when averaged across the modelled landscape and time.
Adding adaptive weather forecasting by itself also produced limited system-wide improvements. Under the middle SSP2-4.5 scenario, scaling forecast-informed management could generate economic benefits while maintaining relatively high vegetation cover, but those benefits remained unequal.
The reason is intuitive. Knowing that a dry season is coming does not automatically give a household the resources needed to respond. A pastoralist may know that herd reductions, additional feed or other adjustments are advisable while still lacking cash, market access or practical options to make those changes.
This distinction between information and capacity is particularly relevant as climate services expand across Africa. Improving forecast accuracy and access is valuable, but the usefulness of a forecast ultimately depends on whether the recipient can act on it.
Risk smoothing changed the picture
The more promising results appeared when forecasting was paired with measures that reduced the immediate financial consequences of poor rainfall.
In the model, supplemental fodder acted as a form of risk smoothing. During periods when natural forage was insufficient, it helped prevent livestock losses that poorer households would otherwise struggle to avoid.
Used without careful adaptive management, however, fodder support can create another problem. Keeping more livestock alive can increase grazing pressure and contribute to overstocking. The researchers therefore examined the interventions together rather than assuming that either was universally beneficial.
Pairing risk smoothing with adaptive forecast use produced a better balance. Forecasts provided information for adjusting livestock decisions, while fodder gave households greater capacity to survive difficult periods. Together, the interventions mitigated more of the direct climate impacts and reduced trade-offs associated with either measure on its own.
Support also affected whether communities stayed involved
The third experiment added another layer: communities could decide over time whether to join or abandon the conservation initiative.
Those decisions were modelled through payoff-biased social learning. Communities observed outcomes around them and were more likely to adopt approaches associated with greater observed wealth.
Supplemental fodder became particularly important in this setting because it made successful participation more visible and helped sustain engagement. Its influence became stronger as rainfall variability increased.
Under the highest-variability SSP5-8.5 scenario, supplemental fodder without forecast use produced the highest overall engagement. Yet that did not necessarily generate the best combination of wealth equality and vegetation outcomes.
This is a useful warning for real programmes. High participation does not automatically mean an intervention is producing the best ecological or distributional outcome. A programme can be attractive to participants while still creating longer-term trade-offs elsewhere in the system.
Forecast precision was not the decisive factor
The researchers also tested whether their conclusions depended heavily on highly precise forecasts or on which communities received forecasting information.
They reduced forecast precision and separately targeted forecasts at the richest quartile, the poorest quartile and all communities. Neither forecast precision nor the targeting strategy produced strong changes across the modelled outcomes.
Within this particular model, the more important issue was whether forecasting was paired with resources that allowed households to respond.
That result should not be interpreted as evidence that forecast quality does not matter in the real world. Instead, it suggests that improvements in information can encounter diminishing practical value when households remain financially constrained.
Why this matters for South Africa
South Africa is directly represented in the empirical landscape used to construct the model, alongside six neighbouring or nearby southern African countries.
The study therefore speaks to a regional problem rather than importing findings from an unrelated climate or agricultural system. Livestock producers across southern Africa already operate under substantial rainfall variability, and future climate change may make the timing and severity of dry periods harder to manage using historical experience alone.
At the same time, access to satellite observations, seasonal forecasts and other climate information is improving. The temptation is to treat better information as the intervention itself.
The modelling suggests a more demanding standard. Forecasts are most useful when households have the economic and institutional capacity to respond to them.
That could include carefully designed fodder support, drought-linked cash assistance, functioning livestock markets or other forms of temporary risk smoothing. The authors note that direct monetary support during drought could potentially provide benefits similar to fodder in some settings.
Design matters because support must remain aligned with local ecology. Protecting livestock without allowing rangelands to recover can simply transfer the cost of a drought from household balance sheets to vegetation.
The model deliberately simplifies a complicated system
The researchers are explicit about important limitations.
The simulations do not fully represent biome-specific vegetation responses, differences among cattle, sheep and goats, livestock theft, conflict, changing policy, household livelihood strategies, wildlife interactions or movement of pastoralists and livestock between communities.
The model also imposes simplified relationships between grazing, vegetation and livestock. Rotational grazing, for example, can affect soil condition, carbon storage and biodiversity in ways that are not captured by a model focused largely on standing vegetation cover.
These omissions mean the numerical outputs should not be treated as literal forecasts. They are a structured way to test how behavioural, economic and ecological mechanisms interact.
That is also the study’s strength. Climate adaptation is often evaluated one intervention at a time. Real households experience several constraints simultaneously, and solving one can expose another.
Climate information works best when people can act on it
The study ultimately challenges a simple version of technology-led adaptation.
A future in which pastoralists can see difficult weather months in advance is clearly preferable to one in which they cannot. But information does not buy feed, move livestock, create market access or replace animals lost during drought.
As rainfall becomes more volatile, the households with the least financial room to manoeuvre may be precisely those least able to turn better forecasts into better outcomes.
The modelling therefore points towards a broader definition of climate resilience. Better prediction matters, but so does the capacity to respond without deepening inequality or degrading the landscapes on which livestock production depends.
Source Information
Study Title: Minimizing climate change adaptation trade-offs in African rangelands
Authors: Matt Clark, Cosima Fröhner, Andreas Christ Sølvsten Jørgensen, Thomas Pienkowski, Sibabalo Yekela, Aamirah Isaacs, Olivia Crowe, Jeffrey Andrews, Paul E. Smaldino, Iacopo Tito Gallizioli, Gina Arena and Morena Mills
Journal: Nature Sustainability
Published: 14 September 2026
DOI: 10.1038/s41893-026-01938-0
Study setting: Agent-based modelling calibrated with empirical data from Botswana, Eswatini, Lesotho, Mozambique, Namibia, South Africa and Zimbabwe
Model design: 100 communities, four grazing plots per community, three rainfall-variability scenarios, 50 repetitions per condition and 100 iterations per simulation
Main finding: Weather forecasting alone produced limited system-wide protection under greater rainfall variability. Pairing adaptive forecast use with risk-smoothing support such as supplemental fodder better reduced economic, equity and ecological trade-offs.
