Abstract:
Land ecosystems offer an effective nature-based solution to climate change mitigation by absorbing approximately 30% of anthropogenically emitted carbon. This estimated absorption is primarily based on constraints from atmospheric and oceanic measurements while quantification from direct studies of the land carbon cycle themselves displays great uncertainty. The latter hinders prediction of the future fate of the land carbon sink. This talk will present work done by my lab in the past two decades. Specifically, we have revealed a general dynamic pattern that the land carbon cycle changes in a direction toward a moving attractor in response to global change. This general pattern is fully captured conceptually by dynamic disequilibrium and mathematically by a matrix equation, which unifies land carbon cycle models. We have integrated the matrix equation into neural network to improve accuracy of model prediction and discover mechanisms underlying land carbon cycle dynamics. Meanwhile, we have evaluated various carbon dioxide removal strategies using the knowledge gained from our basic carbon cycle research and, thereby. identified burying woody debris from managed forests probably as the most effective, easily implemented, and highly sustainable strategy.
Knowledge gained from my lab’s research in the past decades is partially presented in the training course, New Advances in Land Carbon Cycle Modeling, and the training course textbook, Land Carbon Cycle Modeling: Matrix Approach, Data Assimilation, Ecological Forecasting, and Machine Learning. The training curse videos and textbook are freely available.
Bio:
Yiqi Luo is Liberty Hyde Bailey Professor at Cornell University, USA. He obtained his PhD degree from the University of California, Davis in 1991 and did postdoctoral research at UCLA and Stanford University from 1991 to 1994 before he worked at Desert Research Institute as Assistant and Associate Research Professor from 1994 to 1998, the University of Oklahoma as Associate, full, and George Lynn Cross Professor from 1999 to 2017, and Northern Arizona University as full and Regents Professor from 2017 to 2022. His research program has been focused on addressing three key issues: (1) how global change alters structure and functions of terrestrial ecosystems, (2) how terrestrial ecosystems feedback to regulate climate change, and (3) how ecosystem processes can be effectively manipulated to offer nature-based solutions toward carbon neutrality. To address these issues, Dr. Luo’s laboratory has conducted field global change experiments, developed terrestrial ecosystem models, synthesized extensive data sets using meta-analysis methods, integrated data and model using data assimilation techniques and knowledge-guided artificial intelligence (AI)modeling, and carried out theoretical and computational analysis. Professor Luo has published seven books (including translated and edited ones), 37 book chapters, and more than 600 papers in peer-reviewed journals. He was recognized as Highly Cited Researcher by the Web of Science Group, Clarivate Analytics in 2016-2024. He was elected fellow of American Association for the Advancement of Science (AAAS) in 2013, American Geophysical Union (AGU) in 2016, and Ecological Society of America (ESA) in 2018.
Summary:
Focus: how to bend the climate warming curve
Understanding land carbon Cycle
Carbon concentration and average global temperatures are rising
Need to Understand. Predict and Manage the global carbon cycle
Consider the equilibrium state to which natural ecosystems in each region tend to with no human intervention
Ecosystem succession: sequence of species mixtures that a region goes through over time
E.g. pine forest -> disturbance (fire) -> mixed pine broad leaf forest
The land carbon cycle has a long-term state it tends to, what is it?
Dynamics: sun+air CO2 -> C in Trees -> litter on ground (leaves, body) -> soil C (roots, dead matter)
Synthesized data on organic litter, soil C and how it depends on ecosystem state
Modeled dynamic system of carbon flow in forest
Donor pool of carbon (living on tree, refreshed annually) -> Recipient pool (dead in soil, refreshed from donor)
Modeled using differential equation
Climate cycle is changing these dynamics
External forcing: Seasonal climate, disturbances, climate change, regime, ecosystem tipping points
System equations
Response: periodicity, pulse-recovery, gradual change, disequillibrium, abrupt change
Modeling dynamics using matrix equations
System of differential equations (obey mass conservation in cycle,
Simplifies over prior work: NCAR CLM5 Matrix CN
446 pools for carbon cycle
446 pools for organic nitrogen
Non-autonomous Compartment Systems analysis
Using generalized matrix approach makes it possible to train models in data driven ways, analyze model structure and equilibrium dynamics
Predicting land carbon cycle using AI
PRODA: PROcess-guided deep learning and Data-driven modeling
Differential equation model based on meteorological, soil and vegetation data
Requires lots of data assimilation to parameterize the model
Data assimilation alone is not enough; need process modeling augmented with deep learning
BINN: Biogeochemistry-Informed Neural Network
Trained with 57k soil carbon samples globally
Used model to understand soil carbon dynamics and how it may change
Train neural network to predict parameters of matrix carbon cycle model: good accuracy in modeling evolution of soil carbon
Same approach for modeling forest growth: neural network predicts parameter of process-based model
ScIReN: Scientifically-Interpretable Reasoning Network: Kolmogorov-Arnold network since it is more interpretable
Bending the warming curve via managing land carbon cycle
Goal: remove carbon from atmosphere and store on land or in ocean
Focus: create new carbon pools on land
Need to model the land carbon cycle system so that we can modify it to accelerate carbon capture
Most effective carbon sink strategies are related to forestation
We don’t need to change the amount of carbon input into these sinks, changes in strategies primarily modify the durability of carbon in each system
Example:
Take forest debris and bury it underground
If we can store it for 100 years, we’ll remove 769 GTons of CO2, 10 GT/year