The terrestrial biosphere is one of the most volatile components of the global carbon cycle. In some years, forests and grasslands act as massive carbon sinks, absorbing billions of tons of anthropogenic $\text{CO}_2$; in others, droughts, heatwaves, and fires turn them into net carbon sources. To predict how these ecosystems will respond to future climate change, we cannot rely solely on statistical extrapolations of past trends. Instead, we must simulate the underlying physical, chemical, and biological mechanisms directly.
This is the core philosophy behind process-based ecosystem modeling, often realized through Dynamic Global Vegetation Models (DGVMs) and Terrestrial Biosphere Models (TBMs). By translating ecological first principles into systems of differential equations, these models allow scientists to simulate how carbon, water, and nutrients flow through plants and soils across the globe.
Simulating Ecosystems from First Principles
Traditional empirical models rely on static relationships—such as multiplying a given land area by a fixed emission factor to estimate gas fluxes. While simple, this approach fails when environmental conditions push ecosystems past historical baselines. Process-based models solve this by simulating the actual mechanisms driving ecosystem dynamics at fine temporal resolutions.
Consider how a model tracks carbon exchange. Instead of guessing net uptake, biophysical submodels simulate photosynthesis (Gross Primary Productivity, or GPP) and respiration based on hourly or daily meteorological data, as seen in work by Auke van der Woude et al. in Temperature extremes of 2022 reduced carbon uptake by forests in Europe. These models calculate stomatal conductance—the opening and closing of leaf pores—by balancing the plant's need for $\text{CO}_2$ against the risk of water loss via transpiration. As explained by Kai Wang et al. in Regional and seasonal partitioning of water and temperature controls on global land carbon uptake variability, this mechanistic foundation allows models to capture how environmental factors interact non-linearly to govern carbon uptake and release.
This principle extends beneath the surface into the soil. Rather than treating soil decomposition as a uniform decay rate, advanced models simulate microbial processes within discrete depth layers. For example, Shree R. S. Dangal et al. in Global Nitrous Oxide Emissions From Pasturelands and Rangelands: Magnitude, Spatiotemporal Patterns, and Attribution dynamically simulate microbial nitrification and denitrification. Because these biological pathways are highly sensitive to daily fluctuations in soil moisture, temperature, pH, and substrate availability, the process-based approach successfully captures the sudden, non-linear emission pulses of greenhouse gases like nitrous oxide ($\text{N}_2\text{O}$) that empirical models routinely miss.
Disentangling Confounding Drivers via Factorial Simulations
In the real world, ecosystems are bombarded by multiple shifting variables simultaneously: temperatures rise, atmospheric $\text{CO}_2$ concentrations climb, rainfall patterns shift, and humans alter land cover. If a forest's carbon sink strengthens, is it because of $\text{CO}_2$ fertilization, or because of a temporary shift in regional precipitation?
Process-based models excel at answering these "what-if" questions through a technique called factorial simulations. Because the model's equations explicitly separate the influence of each driver, researchers can run parallel virtual experiments:
- Control Run: All environmental drivers (climate, $\text{CO}_2$, land-use) vary historically.
- $\text{CO}_2$-Only Run: Atmospheric $\text{CO}_2$ increases historically, but climate and land-use are held constant.
- Climate-Only Run: Climate variables change historically, but $\text{CO}_2$ is kept at a pre-industrial baseline.
By subtracting the outputs of these runs from one another, researchers can isolate the exact contribution of each driver. Using this approach, Shilong Piao et al. in Spatiotemporal patterns of terrestrial carbon cycle during the 20th century disentangled the 20th-century terrestrial carbon cycle, demonstrating that while land-use change acted as a major carbon source, $\text{CO}_2$ fertilization and climate shifts ultimately drove the biosphere to be a net sink. Similarly, Hui Yang et al. in Strong direct and indirect influences of climate change on water yield confirmed by the Budyko framework utilized factorial simulations within the Budyko hydrological framework to confirm that global runoff changes are primarily dictated by direct and indirect climate changes rather than independent vegetation shifts.
Closing the Global Carbon Ledger
Process-based models serve as an indispensable tool for tracking the global carbon budget. The global carbon ledger must balance: emissions from fossil fuels and land-use change must equal the accumulation of $\text{CO}_2$ in the atmosphere plus the amounts absorbed by the oceans and the land. While atmospheric accumulation can be measured directly with high precision, the land sink is notoriously difficult to observe at a global scale.
In the annual global carbon budget syntheses, such as those led by Pierre Friedlingstein et al. in Global Carbon Budget 2022 and Corinne Le Quéré et al. in Global carbon budget 2014, ensembles of DGVMs are deployed to simulate the land $\text{CO}_2$ sink from the bottom up. These simulations provide a crucial cross-validation against "top-down" atmospheric inversion models and budget residuals.
When extreme climate anomalies occur, process-based models can be forced with near-real-time climate reanalysis data to provide rapid, low-latency insights. For instance, when the global land sink experienced a dramatic decline, low-latency simulations by Piyu Ke et al. in Low latency carbon budget analysis reveals a large decline of the land carbon sink in 2023 and Philippe Ciais et al. in Low latency global carbon budget indicates reduced land carbon sink in the year 2024 quickly identified that widespread extreme heat and drought were transitioning key terrestrial regions from carbon sinks into net sources.
Representing Ecological Memory and Feedbacks
Ecosystems are not passive, instantaneous responders to weather; they possess "biological memory" and exhibit complex internal feedbacks. Modern process-based modeling focuses heavily on capturing these lagged and interactive effects.
One such mechanism is the interplay between disturbance and recovery. When a wildfire burns a boreal forest, it causes an immediate pulse of carbon emissions. However, it also initiates a decades-long process of vegetation regrowth. Chao Yue et al. in How have past fire disturbances contributed to the current carbon balance of boreal ecosystems? designed models to track these parallel dynamics, revealing that while recent fires are a net carbon source, the legacy of past fires creates a long-term carbon sink through post-fire regrowth.
Similarly, models must account for chemical feedbacks in the soil. The phenomenon of "priming" occurs when the addition of fresh organic matter (like falling leaves) accelerates the decomposition of older, stable soil carbon by stimulating microbial activity. By integrating an explicit priming submodel into the ORCHIDEE land biosphere framework, Bertrand Guenet et al. in Towards a representation of priming on soil carbon decomposition in the global land biosphere model ORCHIDEE (version 1.9.5.2) demonstrated how process models can more accurately simulate long-term soil carbon stability.
Finally, these models can be coupled directly to atmospheric climate models to capture biophysical feedbacks. As shown by Lili Liang et al. in Response of terrestrial ecosystems carbon budget to large-scale direct CO2 removal using Community Earth System Model, large-scale changes in atmospheric $\text{CO}_2$ alter plant physiology, which shifts soil respiration and water-use efficiency, ultimately feeding back into regional climate dynamics. By simulating these interconnected loops from the leaf to the globe, process-based models provide the mechanistic clarity needed to navigate our changing planet.
Key Foundational Work
Spatiotemporal patterns of terrestrial carbon cycle during the 20th century (2009) — contributed 1 concept elements to this direction.
Variability and recent trends in the African terrestrial carbon balance (2009) — contributed 1 concept elements to this direction.
Effect of climate and CO2 changes on the greening of the Northern Hemisphere over the past two decades (2006) — contributed 1 concept elements to this direction.