scope and boundaries of CSB and to refine the underlying principles of CSB science. Moreover, a robust theoretical framework is crucial to effectively guide the search for and discern valuable insights from the vast influx of data and to moderate the increasing dependence on artificial intelligence (AI) and statistical forecasting, which we worry can lack principled approaches (Coveney et al. 2016; Enquist et al. 2024.) There is a pressing need for new initiatives to forge an integrative theory that spans the disciplines within CSB. Such a theory would not only synthesize and harmonize existing theoretical frameworks but also enhance our understanding and management of complex biological systems from genes to the biosphere.

A Grand Challenge—Forecasting the Future of the Biosphere

Box 1-2 describes as one of the characteristics of CSB that it “inherently incorporates multiple scales, from the subcellular to the global biosphere (Figure 1-1), from the local to global spatial extents, from less than a second to millennia.” As emphasized by Harrison et al. (2021), our current ability to forecast the future of the biosphere and model its dynamics is, to put it mildly, challenged. The significant shortfalls in modeling the functioning of the biosphere as a core component of the climate system highlights significant gaps in our biosphere theory. Despite successes, substantial hurdles persist, particularly in reproducing large-scale phenomena. Both Earth system models (ESMs) and dynamic global vegetation models struggle with accuracy, failing to capture the amplification of the high-latitude seasonal cycle of atmospheric CO₂ (Grave et al. 2013, Thomas et al. 2016) and the relationship between the ¹³C/¹²C stable isotope ratio of atmospheric CO₂ and global land–atmosphere carbon exchange (Peters et al. 2018).

Persistent discrepancies in models over the effects of global warming on primary production, vegetation responses to precipitation changes, and the influence of CO₂ and nutrient availability (Ciais et al. 2013, Huntzinger et al. 2017, Wieder et al. 2015) have been highlighted for decades (Friedlingstein et al. 2006, VEMAP 1995) and were notably mentioned in the IPCC Fifth Assessment Report (IPCC, 2013). Recent studies confirm these ongoing issues (Arora et al. 2020).

Significant fundamental uncertainties still plague our understanding of the biosphere’s responses to environmental changes. There remains considerable disagreement and uncertainty about how the biosphere, including its interactions with human activities, will react to increased atmospheric CO₂ levels and subsequent rises in ambient temperatures (Arora et al. 2020, Friedlingstein et al. 2006, 2014). This uncertainty extends to several critical areas: the existence and thresholds of specific ecological tipping points (Chaparro-Pedraza and de Roos 2020, Ditlevsen and Johnsen 2010, Drake et al. 2020, Dudney and Suding 2020, Lenton 2013), the actual trends in global biodiversity and whether it is truly decreasing (Dornelas et al. 2014, McGill et al. 2015), the rate and implications of species extinctions in the Anthropocene (Ceballos and Ehrlich 2018, Rothman 2017), and the long-term effects of geoengineering initiatives such as iron fertilization (Keith 2021). These failures and uncertainties indicate a pressing need to reassess and potentially overhaul the assumptions and methodologies used in current vegetation models. Developing new or significantly improved theoretical frameworks

to enhance their predictive capabilities is crucial, necessitating a multidisciplinary approach that incorporates broader ecological and evolutionary insights. Such advancements are essential to effectively tackle the interlinked challenges of climate change, biodiversity conservation, and ecosystem management, propelling CSB forward.

Background

Science seeks to enhance our comprehension of the natural world, thriving on the dynamic interplay and tension between induction and deduction, and the balance between empiricism and theoretical frameworks. Observational and experimental data offer insights into the structure and functionalities of the natural phenomena around us. Theory, at its core, reflects our attempt to understand biological and physical phenomena and is used to classify, interpret, and predict. As this chapter’s epigraph notes, a theoretical framework is more than a collection of facts or data. It is a logical framework developed for understanding and interpreting observations and facts. A previous report (NRC 2008) defines theory as “integral to each specific kind of scientific activity, including experimentation, observation, exploration, description, and technology development as well as hypothesis testing” and differentiates “facts and data,” as distinct from theory, which are the structures that explain and interpret data and facts. Theory may evolve with new data, but facts/data do not.”

Theory unifies disparate observations and empirical laws under a single conceptual umbrella, serving as a compass for scientific exploration. Indeed, as Marquet et al. (2014) note, “[t]heory reduces the apparent complexity of the natural world, because it captures essential features of a system, provides abstracted characterizations, and makes predictions for as-yet unobserved phenomena that additional data can be used to test. . . Data gathered through observation and experimentation provide clues about the structure and function of the natural world, and theory organizes existing data and new ideas into a cohesive conceptual framework to both explain existing observations and make novel predictions.”

Theory enables us to classify, interpret, and predict observations and natural phenomena. It integrates various aspects of a phenomenon, offering a coherent narrative that elucidates underlying principles. By explaining observations as parts of a greater whole, theory guides our expectations under specific conditions. It is inherently dynamic, evolving with new evidence or contradictory findings. Theories are inherently predictive, setting the stage for empirical testing and verification.

The initial focus is to develop simple, tractable, mechanistic theories with relatively few variables and parameters. These may be caricatures of the system, but they play a crucial role because they attempt to incorporate the important variables and essential features that determine the system’s organization, structure, development, and dynamics (Servedio et al. 2014). In Isaac Newton’s Untitled Treatise on Revelation (section 1.1) he states, “choose those constructions which without straining reduce things to the greatest simplicity …. Truth is ever to be found in the simplicity, and not in the multiplicity and confusion of things.” In his Mathematical Principles of Natural Philosophy (1687), his Rule I is “No more causes of natural things should be admitted than are both true and

sufficient to explain their phenomena.” Ecology has a long history of building theories (Scheiner and Willig, 2011); the power of a parsimonious approach to theory development is that it is typically falsifiable so that models can be appropriately modified when their predictions are confronted with data. As Enquist et al. (2024) note, “A well-defined theory with specific testable predictions proven wrong by confrontation with data can provide important insights for moving a field in the right direction.”

The power of a parsimonious approach to theory development is that it is typically falsifiable. The emphasis is on making these models falsifiable so that they can be appropriately modified when their predictions are confronted with data. Ecology has a long history of building theories (Scheiner and Willig, 2011); a well-defined theory with specific testable predictions proven wrong by confrontation with data can provide important insights for moving a field in the right direction.

There is a need to develop a core body of theory in CSB based on zeroth-order frameworks and first principles rooted in the foundational laws of biology, physics, and chemistry. A “zeroth-order framework” is the use of simplifying assumptions to first give a rough approximation to the solution to a problem. The most basic, essential factors are first considered and more complex or minor influences are ignored. This is typically the starting point in solving complex problems, allowing a scientist to capture the primary behavior of a system with minimal computational or analytical complexity (West 2017). Successive iterations, or orders of approximation, include increasingly more influences on the solution (e.g., more starting assumptions) that hopefully will successively better refine the approximation of the truth. Using the principle of parsimony, starting assumptions and input need to be chosen carefully. To quote the biologist J.B.S. Haldane, “In scientific thought we adopt the simplest theories which will explain all the facts under consideration and enable us to predict new facts of the same kind” (Haldane 1927).

A zeroth-order framework contains the building blocks of foundational assumptions or axioms from which a theory is built. These principles are typically so fundamental that they are generally accepted without needing empirical evidence. Building theory starting with a zeroth-order framework ensures a robust, logically coherent framework that can be universally applied and tested, and theory can be systematically built and expanded to include more complex and specific phenomena. Next, first principles are derived from zeroth-order assumptions but are often more specific and can be proved or derived through reasoning and logical deduction. “First principles are the bedrock of science—that is, quantitative law-like postulates about processes underlying a given class of phenomena in the natural world with well-established validity, both theoretical and empirical (i.e., core knowledge)” (Marquet et al. 2014, p. 703). In physics, using first principles often involves starting with basic laws like Newton’s laws of motion or the laws of thermodynamics and applying them to work out complex phenomena from these fundamental truths. In biology, first principles are fundamental concepts or foundational truths from which more complicated theories and models are derived (e.g., the principles of homeostasis, stoichiometry, evolution by natural selection, conservation of energy, trade-offs in resource allocation, principles of exponential growth and carrying capacities, and trophic structure). Each of these principles provides a foundational

framework from which more detailed and specific scientific inquiries and hypotheses can be constructed, explaining complex biological phenomena across different scales and contexts. “The ultimate goal is to develop quantitative, predictive theories grounded in underlying principles and supported by data, observation, and experimentation” (Enquist et al. 2024).

There is increasing urgency to address many significant biosphere challenges (Box 3-1) that directly threaten human well-being and socioeconomic stability (Díaz et al. 2019, McMichael 2013, Ruckelshaus et al. 2020, WMO 2021). The ability to accurately forecast biodiversity and ecosystem functioning, predict the onset and extent of future pandemics, or determine when the Amazon rainforest might hit a catastrophic tipping point demands our attention. Identifying the critical parameters and dynamics that drive these threats is essential for developing quantitative strategies to minimize adverse outcomes and mitigate potential disasters. However, the inherent complexity of the biosphere, characterized by biological processes that span a broad spectrum of spatial and temporal scales, presents a formidable challenge.

Such a coherent theoretical framework would help guide observations and experiments and will enable scientists to understand mechanisms of climate change, land use, and other significant aspects of CSB. Nature-based strategies, recent efforts to protect and improve the natural and enhanced environment by addressing biodiversity challenges and assessing climate mitigation strategies, could be improved by developing additional tools, including theories (Novick et al. 2024).

Lessons can be learned from the effectiveness of a theory-driven approach in climate and atmospheric sciences (Emanuel 2020, Enquist et al. 2024), which has proven highly effective in predicting climate patterns, understanding atmospheric processes, and forecasting future climate scenarios. Such a theory-based framework is an example that CSB could strive to emulate to predict changes in the biosphere amidst complex environmental challenges. The Earth sciences provide a compelling example, having anticipated the unfolding climate crisis by blending fundamental theories with progressively more sophisticated observations and experiments within a theory-driven simulation framework. Atmospheric and ocean science has not only accurately predicted global temperature changes but has also provided increasingly detailed projections of past, present, and future shifts in temperature and circulation patterns (Arias et al. 2021, Hausfather et al. 2020). In contrast, similar efforts to predict corresponding phenomena within the biosphere have not progressed as rapidly (e.g., compare and contrast discussions in Doak et al. 2008, Fisher et al. 2018, Moorcroft 2006). This comparison underscores the need for CSB to challenge past approaches and research investments to better adopt and adapt these successful theory-based approaches to better anticipate and respond to biospheric changes.

CHALLENGES IN DEVELOPING CSB THEORIES

The committee identified three challenges in developing CSB theories. First, a major challenge is the explosion in our capacity to amass extensive datasets of different parameters, detail life’s rich biology at multiple scales, map the diversity of life, and

reach a granularity that was once unthinkable. However, as Brenner underscores in this chapter’s epigraph, the sheer quantity of data, even if atom-by-atom, does not automatically translate to an enhanced understanding of biological phenomena. Currently, the unprecedented levels of detail that can be seen at any scale are far outpacing the derivation of theoretical frameworks that can process this information across multiple scales and glean understanding and predictive ability.

Second, multiscale biological research is challenged by the need to go from detail to insight. Much advancement has been made in our ability to map and describe the intricate details of biological and ecological processes across scales (Hampton et al. 2013). A theoretical framework is vital because it provides the necessary abstraction that allows for meaningful interpretation, prediction, and utility in scientific endeavors and applications. In essence, science is the delicate balance between abstraction and synthesis (theory) and the retention of minimal, but sufficient detail for a theory to be effective in science.

Third, the lag in developing predictive CSB science is influenced not only by the complexity of biology or data scarcity but also by the tensions between the way scientists gather and interpret data as characterized by three predominant scientific cultures within biology (Enquist et al. 2024). The “variance culture” or natural history, focuses on detailed observation and cataloging, such as bird counts and collecting flora and fauna, emphasizing the meticulous documentation of biological diversity. It is arguably the basis of modern biology, including molecular biology, genetics, and numerous fields that currently do not rely much on natural history. The emphasis is on detailed observations and focuses on differences and deviations of taxa, clades, and specific

locations or regions. It leans more toward experimental and observational methods of investigation. The “exactitude culture” advocates for highly detailed models to mimic real-world complexity, often prioritizing precision over practical scalability and emphasizes incorporating more detail, typically focusing on specific problems or phenomena and more general concepts or understanding. Approaches include detailed statistical modeling, machine learning (ML), and AI untethered to parsimony and assessing competing models based on information criteria. Models tend to be phenomenological with many parameters, often disconnected from underlying principles. Conversely, the “coarse-grained culture” prioritizes simplification and general principles, aiming to distill complex information into overarching insights. “This approach can include mathematical derivations of probabilistic outcomes or take the form of a parsimonious statistical theory” (Enquist et al. 2024). The perspectives from each scientific culture, while valuable, often operate in isolation within the biosciences. Progress in tackling complex problems and advancing CSB emerges when these diverse approaches are integrated, combining detailed empirical data with high-level theoretical synthesis to foster a comprehensive understanding essential for addressing global ecological challenges.

These three challenges are further intensified by the climate and biodiversity crises. The escalating climate crisis, the alarming rates of species extinction, and the urgent need to preserve biodiversity underscore the necessity for a predictive science of the biosphere and to help guide issues central to the UN Climate Change Conference COP25 agenda and the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services. The ability to not just collect but also to guide the collection and distillation of vast biological data, meaningfully interpret, obtain knowledge, and use this information to then make predictions is needed.

Theories are needed that not only bridge the micro and macro scales of life but also provide actionable insights into mitigating climate change impacts, conserving biodiversity, and enhancing ecosystem services. In this context, the balance between detailed data collection and theoretical abstraction is not just a scientific endeavor; it’s a requisite tool for addressing some of the most pressing ecological and societal challenges of our time. As discussed below, while several elements of theoretical strands are in place for CSB there are several notable gaps.

REQUIREMENTS OF CSB THEORIES

General theories for CSB requires bridging gaps in our current understanding of both small- and large-scale biological and ecological processes. For example, climate change affects ecosystems through altered mean conditions and increased variability (Turner et al. 2020) alongside rising atmospheric carbon dioxide levels. These changes exacerbate other ecological pressures such as habitat loss and degradation, defaunation, and fragmentation. As emphasized by Malhi et al. (2020), understanding the ecological dynamics of these impacts, pinpointing hotspots of vulnerability and resilience and identifying effective management interventions are essential for enhancing biosphere resilience (Jung et al. 2021, Molotoks et al. 2020). Moreover, ecosystems and biodiversity play a crucial role in both mitigating and adapting to climate change (Morecroft et

al. 2019). Increasingly CSB is being called to assess how ecosystem management and restoration have the potential to contribute nature-based solutions to address both the causes and consequences of climate change. However, the effectiveness, scalability, and magnitude of different nature-based strategies need to be explored, better understood, and evaluated (Bennett et al. 2016, Malhi et al. 2020). Exploring and quantifying the mechanisms, potential, and limits of nature-based solutions are vital for informed decision making and policy formulation (Malhi et al. 2020, Morecroft et al. 2019).

Processes that define CSB encompass multiple dimensions of biodiversity, including genomic, taxonomic, functional, and phylogenetic measures (Cadotte et al. 2009, Naeem et al. 2012, 2016). CSB theory needs to provide a conceptual framework upon which multiple dimensions of biodiversity can be linked to multiscale (including molecular, cellular, organismal, and ecological) data, recognizing that the framework may change as new technologies and knowledge evolve. For example, theories can be applied to a variety of questions spanning everything from research on invasive species dynamics, biodiversity responses to various drivers, mass extinction, pandemics, to the spatial variation of ecosystem stocks and flows. However, an integrative cycle between observation, data analysis, and theory development (see Box 3-2) would enable linkages between biosphere predictions and would foster the development of new approaches and technologies to address critical ecosystem and societal needs.

CSB theories, present and future, will necessarily range from targeted (e.g., trait-based theories of carbon sequestration) to global in utility and function (e.g. projecting geographic variation of extinction risks to differing climate and human drivers). Below, the committee provides a catalog of current CSB-relevant theories that embody one or more of the following core attributes for developing effective CSB theories. These core attributes are:

1. Scaling and Multiscale Application: CSB theories need to be applicable at various spatial and temporal scales. These scales encompass attributes of individuals (traits, genes), populations, and species assemblages on landscapes, to ecosystem functions and up to the entire biosphere. CSB theories can provide solutions that navigate cross-scale questions and consider all the dimensions of biodiversity, including genomic, taxonomic, functional, and phylogenetic diversity, and include ecosystem pools and fluxes.
2. Data Guidance and Technological Integration: Modern technology and informatics offer unprecedented monitoring capabilities for biosphere processes. However, integration of data across scales for theory application and development remains a challenge. Theory is needed to inform the collection and management of big data by playing a pivotal role in ensuring that data collection aligns with the needs of understanding multiscale biological and ecological processes. Theoretical frameworks for CSB need to mesh and help guide the collection of diverse datasets, technologies, and monitoring programs and will help in interpreting which of their outputs is needed.
3. Identify and Amplify Biological Linkages: CSB theories also need to forge connections across biological processes and scales from the molecular and cellular levels to populations, entire ecosystems, and the biosphere (Figure 1-1).

4. Comprehensive Scope: CSB theories need to unify disparate functional processes from microbial to ecological and physiological processes, including material flux through the biosphere.
5. Human Aspect Theories Affecting Study Sites, Experimental Design That Might Bottleneck Theory Development. Humans make decisions on experimental design, site selection, hypotheses to be examined, and how data are analyzed, which can potentially add bias, which needs to be evaluated prior to study initiation. Encompassing different perspectives, experiences, and knowledge foundations of research teams can drive innovation and analyses for all aspects of CSB.
6. Inclusion of a Metacoupling Framework in Theory Development That Synthesizes Human Social and Ecological Interactions Across Scales. Humans also affect the function of the biosphere, but too often their effects on these processes are not fully captured in experimental designs and ultimately, in theory. As described in Chapter 2, metacoupling analysis is an emerging effort to define and address connectivity between human influences and natural ecosystems near and far. However, capturing these processes and biases in theory development to inform experimental questions is a challenge.

The cyclical process of theory development is intrinsic to the progress of CSB (Box 3-2). It emphasizes that theories are never static but continually evolving entities based on hypothesis testing and new data. Assumptions and predictions are not rigid constructs but flexible tools that adapt and grow as our understanding deepens. They facilitate a dynamic dialogue between theories and empirical evidence, driving the field forward and providing a structured approach to complex biological inquiry. Theory helps determine what data are crucial, what to measure, and where to focus attention. It synthesizes data by highlighting connections and predicts unmeasured aspects.

Select CSB-Relevant Theories

In reviewing the landscape of biological theories that is relevant for CSB, the committee focused on a sample of theories that provide a framework for integration across spatial, organizational, and temporal scales. The committee also considered the theory’s ability to assist in interpreting empirical data, moving us closer to a more integrative, quantitative, and predictive understanding of biological systems.

Neutral Theory of Biodiversity

The neutral theory of biodiversity (NTB) provides a framework for understanding the spatial and temporal dynamics of biodiversity via the impact of stochastic demographic processes on community structure and dynamics across ecological to macroevolutionary timescales (Hubbell 2001, Rosindell et al. 2012). NTB generates a broad array of predictions concerning phenomena such as species abundance distributions, species–area relationships, phylogenetic tree structures, and correlations between species

richness and macroevolutionary rates of speciation and extinction. With remarkably few variables, NTB posits that niche differences among species do not significantly influence their ecological success, effectively treating species as demographically equivalent in terms of individual rates of speciation, birth, death, and dispersal (Volkov et al. 2005).

NTB’s applicability extends across various scales of biological organization, from microbial communities to global biodiversity patterns, making it particularly relevant to CSB. It offers predictions on the distribution of species commonness, rarity, and the temporal dynamics of biodiversity change (Hubbell 2001), serving as a foundational null model to establish baseline expectations. Although many biodiversity patterns diverge from these neutral expectations, deviations are analytically valuable. For instance, deviations in studies like those analyzing human microbiome datasets reveal that the host environment significantly shapes community composition and assembly (Li and Ma 2016), thus providing critical insights into the specific ecological and evolutionary mechanisms influencing biodiversity.

• Reiteration of the Cycle: The process then repeats, with the refined assumptions and predictions undergoing further development, testing, and improvement. The cycle is an ongoing, iterative process that ensures theories evolve with scientific advancements and remain relevant and robust.

Phenotypic Optimality from Ecoevolutionary Optimality

Physiological optimality models offer an alternative to leverage current ecoevolutionary optimality (EEO) theory to understand how plants optimize traits in response to environmental pressures. EEO theory posits that natural selection eliminates less competitive traits, allowing plants to adjust their physiological responses across timescales from days to millennia for optimal survival and reproduction. The theory hinges on the mechanistic links between plant functional traits, resource acquisition, and biogeochemical cycling, impacting plant competitiveness (Franklin et al. 2020, Harrison et al. 2021).

EEO models offer parsimonious, testable parameters that encapsulate critical tradeoffs, such as maximizing carbon gain while minimizing water loss. This approach has successfully been applied to predict vegetation patterns across climates, evidenced in both natural and agricultural ecosystems (Qiao et al. 2020, 2021; Yang et al. 2018). Enhanced by the wealth of data from plant trait databases and satellite remote sensing, these models can rigorously test the universal patterns and simulate ecosystem responses to environmental drivers (Wieder et al. 2015).

ability of ecosystems and their organisms. This includes linking evolutionary processes to biogeochemical cycling, organismal acclimation, and adaptation, as well as understanding the implications for human health and zoonoses. For instance, deeper insights into evolutionary mechanisms could elucidate how species adapt to climate change, resist diseases, or manage geographical expansions. Similarly, advancing theories on biogeochemical cycling could clarify the interactions between nutrient flows and ecosystem functioning under anthropogenic stress. Furthermore, theories that explore organismal acclimation and adaptation are essential for predicting responses to rapid environmental changes, which are vital for formulating effective conservation strategies and sustainability plans. Integrating these theories with studies on human health and zoonoses can also bridge crucial knowledge gaps in how environmental changes promote the emergence and spread of diseases.

Second, there is a strong need to integrate and potentially unify existing ecological and evolutionary theories to create a more cohesive framework capable of explaining a broader range of biological phenomena, thereby enhancing the predictive power of CSB. Merging concepts such as niche theory with metapopulation dynamics could shed new light on species distributions under environmental stress, while incorporating evolutionary game theory could provide deeper insights into adaptive behaviors in fluctuating ecosystems. Such theoretical integration is crucial not only for solidifying the scientific foundations of CSB but also for fostering interdisciplinary collaborations that bring together diverse fields such as ecology, evolutionary biology, climatology, and public health.

Third, advancing CSB theory requires integration of biological feedbacks and human impacts. It is essential to address the complex interactions between biological and social systems, particularly how human activities influence continental-scale biological processes. CSB must account for the role of human populations in shaping continental-scale biological processes. By integrating CSB theory with established ecological and biological theories, we enhance continuity and leverage existing knowledge, highlighting how human activities influence large-scale ecological dynamics. Furthermore, theory development needs to consider conditions not only within a place but also interactions with other places nearby and far away (Frans and Liu 2024).

Fourth, to support this collaborative and cross-scale theory development, CSB could establish dedicated platforms or working groups aimed at synthesizing and advancing ecological and evolutionary theory. These groups could operate within existing organizational structures or through newly established interdisciplinary institutes designed to encourage cross-disciplinary collaborations. Promoting regular workshops, symposia, and joint research initiatives will be crucial to encourage dialogue and the exchange of ideas. Furthermore, integrating diverse knowledge systems, including indigenous and local ecological knowledge (Kimmerer and Artelle 2024), would provide valuable perspectives and enrich the theoretical frameworks, making them more applicable to real-world scenarios. This interdisciplinary and inclusive approach is essential for addressing the complex challenges posed by global ecological changes and ensuring the development of robust, applicable, and inclusive theory within CSB.

insights in theory. In addition to FAIR principles, CARE (collective benefit, authority to control, responsibility, and ethics) principles are important to make sure that the governance and rights of the owners who provide knowledge of various environments and ecosystems that produce experimental datasets are included in discussions of results with consideration of rights and access to data (Carroll et al. 2021).

Improper Data Practices and Incomplete Representation of Functions

“As researchers increasingly utilize publicly available databases to guide research questions and conduct analyses, the abundance of currently available data will influence the trajectory of future research and data collection” (Augustine et al. 2024). Multiple issues associated with incorrect dataset uploading and inadequate curation of database data are suspected in increasing error (Augustine et al. 2024). There is also considerable bias and variation in how data are sampled. For example, in plants, most trait data are poorly sampled. In the TRY plant trait,¹ mean trait completeness is less than 1 percent, although recent efforts focused on particular traits (e.g., growth form) have reached high levels of completeness.

Biodiversity Bias

Biodiversity is a multidimensional construct, the three most common dimensions in use being taxonomic diversity (e.g., species richness), phylogenetic diversity (e.g., cumulative branch length of a phylogenetic tree encompassing species under investigation), and functional diversity (e.g., trait-based diversity metrics), although there are more (e.g., trophic, landscape, genomic, etc.) (Naeem et al. 2016). The majority of research in biodiversity across all scales is biased toward unidimensional research, with taxonomic diversity being the dominant dimension. Although studies of phylogenetic and functional diversity are on the rise, as is multidimensional biodiversity research, studies across all scales are deficient in their coverage of species. For example, plant species are not sampled representatively across the tree of life, with some clades (e.g., Poaceae) being relatively well sampled while others (e.g., Orchidaceae) are relatively poorly sampled.

Challenge of the “Siren Call” of Machine Learning and Artificial Intelligence

Navigating the expanding influence of big data, ML, and AI within the scientific community presents a substantial challenge. The rapid acceleration of science through data-derived modeling is undeniable (Krenn et al. 2022), yet it introduces a critical trade-off. Although ML and AI can deliver remarkable accuracy by leveraging existing datasets, their tendency to overfit and the lack of transparent, mechanistic models raise concerns (Mitchell 2019). These algorithms excel at generating short-term forecasts but

___________________

¹ See https://www.try-db.org/TryWeb/Home.php (accessed April 27, 2024).

often falter in more extended projections that divert focus from the pursuit of fundamental mechanistic insights in the study of complex adaptive systems.

The efficacy of integrating increased biological detail and mechanisms into forecasting models, without simultaneously broadening our theoretical foundations, remains uncertain. ML and data-derived modeling are frequently seen as sophisticated regression techniques (Mitchell 2019), sharing both their strengths and inherent limitations. Conversely, coarse-grained methods from theory offer the potential to generate reliable predictions beyond the immediate scope of training data, uncover novel simplifications, and mitigate the risk of overfitting. While regression-based methods, including ML, have their place in scenarios where theoretical underpinnings are underdeveloped or for in-sample predictions, there is much potential synergy between theory-driven and data-driven discovery. AI and ML methods should be seen as tools to do data-driven discovery more quickly. Further, the development of innovative coarse-grained ML techniques holds promise (e.g., Brunton et al. 2016, Han et al. 2018, Schmidt and Lipson 2009, Udrescu and Tegmark 2020). These emerging approaches could bridge the gap between data-driven discovery, accuracy, and the quest for mechanistic understanding, underscoring the need for a balanced integration of theory, ML, and AI in advancing the science of complex ecological systems (Han et al. 2023).

Challenges to Developing Theory Across Scales (Molecules to Biosphere) and Implementation of Theory for Informing Hypotheses and Experimentation Across Scales

Challenge of Trade-Offs

The development of theory across scales, from molecules to the biosphere, presents significant challenges, particularly in modeling and applying theory to inform hypotheses and experimentation. One major hurdle is the inherent trade-offs involved in scientific modeling, as identified half a century ago by Levins (1966). These tradeoffs often manifest in the choice between modeling a few locations or species in great detail versus many locations and species more superficially. An overly abstract or reductionistic representation may inadequately highlight pattern and identify the essence of the system. For CSB-related experimentation, this may translate as balancing model accuracy with generalizability, essential for applying theory across varied ecosystems and scales. Further, a representation of exact replication becomes cumbersome and essentially indistinguishable from the entity it seeks to describe (Enquist et al. 2024). Balancing between virtues such as accuracy, precision, realism, and generality becomes an inherent part of the endeavor of theory development. Accepting such trade-offs is not a sign of theoretical weakness. Instead, trade-offs are indicative of the nuanced choices scientists make in their journey to understand complex phenomena.

Another trade-off is navigating the integration of big data, ML, and statistical complexity. Data-driven models, while accelerating scientific progress (Krenn et al. 2022) often produce accurate but potentially overfitted predictions that lack transparency in mechanisms (Mitchell 2019, West 2013). This limitation hampers long-term forecasting

and may shift the focus away from developing deep mechanistic insights essential for understanding complex adaptive systems (Enquist et al. 2024). As we develop theory across scales—from molecules to the biosphere—it is crucial to maintain a balance between utilizing advanced computational tools and fostering robust theoretical frameworks. This approach ensures that our reliance on modern technologies complements rather than supplants the pursuit of comprehensive, mechanistically informed scientific theory, thereby enhancing both immediate data analysis and long-term predictive capabilities.

Challenge of Biodiversity Complexity

Biodiversity, as is commonly defined, includes all dimensions, from ecological to evolutionary, across all dimensions of space and time, but in practice, theoretical and empirical approaches focus on single dimensions within scales. An example of the challenge of trade-offs in the inherent tension between modeling a select number of species or locations or a specific biological process in depth versus attempting to capture a broad array of species and functions by working at higher orders, such as tree-species richness and Normalized Difference Vegetation Index across kilometers. Addressing these trade-offs, especially when constructing process-based models for numerous locations or species, among others, remains one of biodiversity science’s most formidable challenges (Figure 3-2).

The committee posits that these challenges can be effectively navigated, irrespective of the chosen study group or modeling framework, by leveraging modeling techniques that harness strength across locations and dimensions of biodiversity, by filling data voids using proxies, amalgamating varied data sources, and doing so across different scales. Opportunities exist to further develop existing theory to help guide process-based models for many sites and across multiple levels of biodiversity by using hierarchical and inverse modeling methods, to fill data gaps, integrate diverse datasets, and model across biological and spatial scales (Evans 2016, Levin 2000).

CONCLUSIONS ON DEVELOPING THEORY TO CONNECT RESEARCH ACROSS SCALES

Theoretical considerations should guide the collection and management of big data, that is, what data that we should collect and what we need to collect. In addition, theory should support our understanding of the causes and consequences of continental-scale biodiversity and resilience in the context of major global change (Figure 3-2). The committee concluded that theory is especially needed in the following three areas.

Conclusion 3-1: Theory is needed that links research at multiple organizational, spatial, and temporal scales, from the micro to meter to landscapes up to the biosphere. The multidimensional and hierarchical multiscale nature of biodiversity requires solutions that can address cross-scale questions and identify cross-scale phenomena (Isbell et al. 2017, Soranno et al. 2014). For this approach, theory is needed that meshes with

our current technologies and informatics that collectively monitor biosphere processes. Consistent with the theme in Chapter 2 about the need for integrated yet flexible frameworks for CSB, theory also needs to be based on conceptual frameworks that integrate multiscale data. These include molecular, microbial processes, genomes, environmental DNA, metagenomics, metatranscriptomics, stable isotope labeling, and metabolomics. These data sources are crucial for linking local ecological and physiological processes of organisms to broader patterns and data collection efforts such as the distribution of species, movement of individuals and species, the functioning of ecosystems, and the flux of material and matter through the biosphere at multiple scales. This integration will enable the refinement and development of CSB theory, enhancing our ability to model and manage environmental changes effectively. By incorporating larger-scale data from remote sensing, tower-based systems, global animal tracking, and sensor networks, we can enrich this framework, providing a more comprehensive understanding necessary for predictive modeling and sustainable ecosystem management.

Conclusion 3-2: Theory is needed to improve climate and global change models by including biological feedbacks. Biological processes that result in feedbacks to ecosystems and climate are a challenge to incorporate into climate and global change theories, presenting considerable uncertainty (Figure 3-2). The inclusion of biological feedback to continental-scale models of global change will enhance our ability to predict future trends and identify cross-scale solutions and will be a key component of clarifying and improving climate and global change models. Refinement of biological feedback theories into continental-scale models and extension to climate and global change theories will improve our ability to both predict future trends as well as identify solutions that cut across scales.

Conclusion 3-3: Theory is needed that incorporates the effects of human-induced environmental changes (including climate change) to predict changes within an ecosystem and to assess metacoupled cascading effects across adjacent and distant systems. Theory is needed to predict interactions among system components across all scales that impact adjacent and distant environments. The inclusion of theory that incorporates human activity will enable the prediction of synergistic, cascading, or trade-off effects on resilience and sustainability of ecosystems and the biosphere across time and space.

To summarize, theory is important to CSB because a well-defined theory with specific testable predictions that can be proven wrong by confrontation with data can provide important insights for moving a field in the right direction (Enquist et al. 2024). Core attributes of successful theories for CSB are that they need to be applicable at various biological organizational, spatial, and temporal scales; they need to be able to inform the collection and management of big data and ensure that data collection aligns with the needs of understanding large-scale biological and ecological processes; they need to unify disparate functional processes from microbial to ecological and physiological processes, including material flux through the biosphere; and they need to be transparent about the role of different perspectives, experiences, and knowledge foundations of research team members in driving innovation and analyses.