The research, led by Wenhang Zhang of Nanjing University of Information Science and Technology together with David J. P. Moore of the University of Arizona and Ye Li, with colleagues, examined trends in the leaf area index, or LAI, across global drylands from 1982 to 2020. LAI is a fundamental measure of vegetation activity, describing the total one-sided area of leaves per unit of ground surface, and it serves as one of the most reliable proxies scientists have for tracking photosynthesis, biomass and ecosystem productivity from space. By analyzing both long-term trends and year-to-year fluctuations in LAI, the team was able to reconstruct a far more complete picture of dryland dynamics than earlier studies that focused on average greening alone.

What they found is persistent and widespread greening. Across global drylands, LAI has increased steadily throughout the nearly four-decade study period, and the rate of greening is higher than what current dynamic global vegetation models predict. The results also directly contradict several recent reports suggesting that the greening of drylands has slowed or stalled in recent years. Rising atmospheric carbon dioxide, which allows plants to photosynthesize more efficiently and lose less water in the process, along with shifts in rainfall patterns and land-use change, are among the mechanisms commonly invoked to explain the observed greening, although the study’s central concern lies not with the greening itself but with what is happening to the variability around it.

Because average trends can conceal profound underlying instability, the researchers paid close attention to interannual variability, the degree to which vegetation activity swings from one year to the next. Their analysis revealed that variability in dryland LAI has been rising sharply, and that this increase is not scattered or localized. Instead, it is spatially coherent, covering approximately 82 percent of the global dryland area. That near-universal signal is a statistical red flag: when most of a biome simultaneously begins to fluctuate more strongly, it suggests a fundamental change in how the ecosystem behaves, not a collection of isolated local stories.

The structure of this rising variability is particularly revealing. The researchers found that it is characterized by two simultaneous changes: leaf area maxima are increasing, meaning that in good years vegetation grows more luxuriantly than before, while leaf area minima are declining, meaning that in bad years vegetation drops to lower levels than in the past. The gap between the best and worst years is widening from both ends. Healthy ecosystems typically buffer against environmental swings, damping fluctuations rather than amplifying them. A system in which peaks and troughs are diverging in this way is one whose internal stabilizing mechanisms appear to be eroding, a pattern ecologists associate with ecosystems approaching critical thresholds or tipping points.

To identify what was driving the increase in variability, the team examined a suite of potential climatic and environmental factors. The dominant driver, they found, is increasing rainfall sensitivity, meaning that dryland vegetation has become progressively more responsive to fluctuations in water availability. In other words, the same amount of rainfall variability now produces a larger swing in vegetation activity than it did decades ago. Two secondary drivers reinforce this effect: variability in rainfall between years, and variability in rainfall within the growing season itself. As rainfall delivery becomes more erratic in both timing and amount, the vegetation of already water-stressed landscapes reacts with increasingly volatile swings in leaf growth and loss.

Perhaps the most sobering result concerns the models that scientists rely on to project the future of the global carbon cycle. When the researchers compared their observations with a suite of dynamic global vegetation models, the models failed to reproduce the observed increase in interannual variability and produced divergent LAI trajectories that did not match what satellites had recorded. This is a serious problem, because drylands contain large stores of soil organic carbon, and fluctuations in their vegetation determine whether that carbon is locked away in soils or released back into the atmosphere. If models cannot capture the growing volatility of dryland ecosystems, their projections of future carbon uptake and release carry uncertainties that could cascade into global climate projections.

The study’s headline finding is therefore a story of two opposing trends wrapped into a single observation. On the surface, drylands appear to be thriving: leaf area has grown consistently since 1982, and models have, if anything, underestimated this productivity gain. But beneath that greening lies an increasingly unstable system, one whose resilience appears to be wearing thin. The authors describe this dynamic as greening masking a loss of stability, a phrase that captures the paradox at the heart of the research. The very metric that suggests drylands are doing well, rising LAI, is being propped up by increasingly extreme oscillations between lush years and depleted ones.

The implications extend well beyond academic carbon accounting. Dryland ecosystems deliver services that billions of people depend upon, including forage for livestock, soil retention, water regulation and food production in some of the world’s most vulnerable regions. Rising variability in vegetation activity means that drought years will hit harder and good years cannot be counted on to compensate, complicating everything from grazing management to national food security planning. In ecological terms, the widening swings in leaf area may indicate imminent state transitions, the abrupt and often irreversible shifts by which ecosystems cross a threshold and reorganize into a new, frequently degraded configuration. Desertified landscapes, once lost, can take centuries to recover their productivity, and the carbon released during such transitions would amplify warming in a feedback loop.

The study also underscores the importance of looking beyond averages when monitoring the planet. Satellite records of vegetation have occasionally produced apparently contradictory conclusions, with some analyses reporting greening slowdowns and others continuing gains, in part because different datasets, time windows and methodological choices yield different trend estimates. By explicitly combining trend analysis with variability analysis, and by focusing on the full 1982 to 2020 period, this work suggests that the apparent contradictions may partly reflect the fact that a greening signal and an instability signal are moving in opposite directions within the same data. Averages that looked reassuring were, in effect, summarizing a system whose year-to-year behavior had become less and less predictable.

For climate scientists, the message is that dryland monitoring and modeling must evolve. Capturing the full behavior of these ecosystems requires models that can represent shifting rainfall sensitivity, the amplifying effects of both interannual and intraannual rainfall variability, and the nonlinear dynamics that precede state transitions. For the rest of the world, the message is a warning wrapped in green: the semi-arid belts that ring the planet’s deserts are changing faster and more erratically than most projections have assumed, and the apparent health of their vegetation should not be mistaken for genuine resilience. What looks like a planet getting greener may, in its driest regions, be a planet getting closer to the edge.