Earth’s terrestrial ecosystems act as a critical buffer against climate change, absorbing roughly a quarter of anthropogenic carbon dioxide emissions. To project how this "carbon sink" will behave in the future, scientists rely on complex Earth System Models (ESMs) and Dynamic Global Vegetation Models (DGVMs).
However, a growing body of empirical evidence reveals a systematic flaw in these simulations: biosphere models suffer from an optimism bias. They consistently overestimate the resilience of ecosystems to climate stress and mischaracterize the fundamental trade-offs plants make to survive.
To understand why these models diverge from reality, we must look at how they simulate the interactions between water, carbon, and soil.
The Water-Carbon Trade-Off: Underestimating Drought and Dryness
At the heart of plant biology lies a daily gamble. To absorb $\text{CO}_2$ for photosynthesis, plants must open microscopic pores in their leaves called stomata. However, open stomata allow water vapor to escape. When water becomes scarce, plants face a choice: keep stomata open to capture carbon and risk dehydration, or close them to conserve water and starve.
Current climate models consistently underestimate the severity of this trade-off. As shown by Peters et al., state-of-the-art models fail to replicate the dramatic increase in plant water-use efficiency observed during continental-scale droughts. Because the models do not fully capture the physiological sacrifices plants make to conserve water, they underestimate the massive drop in carbon uptake that accompanies water scarcity.
[Actual Plant Response] High VPD/Drought ──> Stomata Close Tight ──> Water Conserved, Carbon Uptake Plummets
[Model Simulation] High VPD/Drought ──> Stomata Stay Open ──> Overoptimistic Carbon Uptake Predicted
This problem is compounded by a failure to separate the impacts of dry soil from dry air. Vapor Pressure Deficit (VPD) measures the dryness of the atmosphere. Even when soil moisture is plentiful, high VPD forces plants to restrict stomatal conductance to prevent their leaves from drying out. Fu et al. demonstrated that ESMs systematically underestimate plant sensitivity to both atmospheric dryness and soil water deficits. By flattening these negative responses, models project a land surface that is far more resilient to aridification than real-world data suggests.
The Seasonal Blind Spot: Flattening the Earth's Breath
If you measure atmospheric $\text{CO}_2$ over a year, you will see it fluctuate in a distinct wave pattern. The planet "breathes in" during the spring and summer growing seasons as vegetation rapidly absorbs carbon, and "breathes out" during autumn and winter as plants senesce and microbes respire carbon back into the atmosphere.
Models struggle to capture the true amplitude of this seasonal cycle. Cadule et al. identified that shortcomings in simulating vegetation phenology and climate-driven respiration lead to a significant underestimation of seasonal $\text{CO}_2$ amplitude.
Why does this wave get flattened in simulations? A comprehensive benchmarking study by Peng et al. revealed a dual error: models frequently overestimate gross primary production (GPP) during peak seasons while simultaneously overestimating or mistiming ecosystem respiration.
This seasonal mismatch is particularly acute in rapidly warming northern permafrost regions. As the late growing season warms, respiration increases dramatically. Liu et al. found that DGVMs fail to capture this late-season respiratory surge, which acts as a major offset to early-season carbon gains. Similarly, Yin highlighted that models fail to accurately simulate how warming temperatures accelerate microbial respiration and carbon turnover. By missing these nuances, models paint an overly optimistic picture of net annual carbon retention.
Misjudging the Soil: Hyperactive Carbon Cycles
The optimism bias is not confined to living leaves; it extends deep into the dirt. Soil organic carbon (SOC) represents one of the largest active carbon reservoirs on Earth. The length of time carbon remains locked in this reservoir—its turnover time—determines how quickly carbon returns to the atmosphere.
Model Assumptions vs. Reality in Soil Carbon
┌────────────────────────────────────────────────────────────────────────┐
│ Models: Fast Turnover Time (Hyperactive cycle, rapid release simulated)│
├────────────────────────────────────────────────────────────────────────┤
│ Reality: Slow Turnover Time (Carbon stays locked in subsoil 2x longer) │
└────────────────────────────────────────────────────────────────────────┘
By evaluating global datasets, Zhang et al. discovered that global Earth System Models underestimate soil carbon turnover times by more than half, particularly in tropical and high-latitude ecosystems. Because models assume a hyperactive, rapid carbon cycle, they miss the long-term stability of deep subsoil reservoirs. This structural misrepresentation makes it incredibly difficult for models to reliably project how terrestrial ecosystems will feedback into future climate scenarios.
Tropical Phenology: Overemphasizing the Wrong Stressors
In temperate zones, seasons are defined by temperature. In the tropics, they are defined by rain and sunlight. This environmental difference causes standard vegetation models to break down entirely when applied to equatorial forests.
For years, classic land surface models assumed that tropical evergreen forests experience water stress during the dry season, leading models to predict flat or declining productivity. In reality, satellite observations show a dry-season "green-up"—a surge in canopy renewal and photosynthesis.
To fix this, Chen et al. updated the ORCHIDEE model to trigger leaf shedding and flushing based on solar radiation and VPD rather than soil moisture. They demonstrated that sunlight availability, not water limitation, drives canopy dynamics across the Amazon basin.
Furthermore, models struggle with complex, bimodal climate cycles where environmental drivers peak twice a year. As Liu et al. noted, the inability of existing models to replicate the unique bimodal phenology of Central African rainforests severely limits our capacity to predict how these vital carbon sinks will respond to shifting climate regimes.
Moving Beyond Short-Term Sensitivity
Why do these discrepancies persist? A primary reason is that models are often tuned to match short-term meteorological fluctuations, causing them to miss long-term ecological realities.
Zhang et al. compared European forest models against real-world tree ring data and found that DGVMs consistently overestimate the sensitivity of forest productivity to short-term climate variations. Crucially, the models completely miss the "biological memory"—the lagged, multi-year recovery effects—that trees exhibit after severe climatic shocks.
When models overestimate short-term growth potential, ignore atmospheric dryness constraints, flatten seasonal respiration peaks, and accelerate soil carbon turnover, their errors compound. The result is a simulated biosphere that is far more resilient, flexible, and forgiving than the fragile reality documented by field ecologists. Bridging this gap requires moving away from simplified environmental triggers and embedding true physiological trade-offs into the next generation of climate models.
Key Foundational Work
The seasonal cycle of atmospheric CO2: A study based on the NCAR Community Climate Model (CCM2) (1996) — contributed 6 concept elements to this direction.