technology to capture population-level or community dynamics; (b) remote sensing technologies that capture processes at large spatial scales; (c) long-term, ground-based experiments that manipulate variables and are replicated across space; (d) modeling approaches that connect disparate information to enable inferences from observations or predict outcomes over large spatial extents and through time; and (e) tools that support modeling efforts, such as data harmonization, machine learning (ML), and artificial intelligence (AI) approaches.

Observational Studies Using Ground-Based Technologies

Observational studies that employ measurement techniques and allow organizational, spatial, and temporal scales to be traversed or connected are critical for CSB. Some of the most valuable, economical, efficient, and widely used tools in ground-based studies are briefly described below.

ity of analyzing the expressed functions and levels of stress of organisms, and/or even the genomics of RNA viruses.

For microorganisms, molecular approaches used to understand the roles, interactions, and functional dynamics across extensive spatial and temporal scales have been fundamental to the application of so-called multi-omics approaches. One of these approaches, “amplicon sequencing,” has been used by the Earth Microbiome Project on over 200,000 samples to catalog continental-scale microbial diversity (Thompson et al. 2017). However, amplicon sequencing, as with most eDNA approaches for macrobes, provides information only on the taxonomic distribution of microorganisms.

For analysis of the distribution of microbial genomes and their functional traits we deploy metagenomics, metatranscriptomics, MetaRiboSeq, metabolomics, and metaproteomics (see Box 4.1). Those techniques provide vast data resources that can facilitate robust hypothesis testing to explore the ecology of microorganisms and their impact on hydrological dynamics, nutrient cycling, and climate active atmospheric processes. For example, ’omics can be used to study the role of soil-associated fungi and bacteria, such as Verrucomicrobia, in the uptake of carbon in grasslands (Brewer et al. 2017, Fierer et al. 2013), which, in turn, can provide insight into how plant distributions at local and regional scales influence soil microbial carbon dynamics. For aquatic ecosystems, multi-omic techniques elucidate microbial community dynamics in response to significant events such as oceanic current shifts or river diversions, supported by databases such as the Genome Resolved Open Watersheds database (Borton et al. 2023).

crop productivity and sustainability across diverse biomes. This integrative approach is crucial for addressing global challenges such as food security, climate change, and ecosystem conservation.

Some unique challenges exist with eDNA and eRNA techniques. They need to be calibrated based on body mass or genomic polyploidy to ensure accurate quantitation of the target organism. They also need to take into consideration the degradation rates of DNA and RNA in different environmental matrices and contexts. Finally, they are affected by sampling bias, with fluid matrices (e.g., streams and rivers) having a greater distribution potential for biomarkers, and spatial analysis. Most multi-omic data are currently proportional, which limits the opportunities for continental-scale integration of multiple sites or temporal scales. Because of these limitations, multi-omic data are not sufficiently integrated into models. Recent advances in quantification need to be further expanded to create quantitative multi-omic resources and fully integrate these data into models that inform policy.

Isotopes: Stable and Radioactive

Stable and radioactive isotopes are one of the most powerful tools we have to resolve biological processes across scales. For example, radioactive carbon isotopes (¹⁴C) are being used to resolve the mean age and transit times of carbon in terrestrial systems. In the mid-to-late-1950s, the prevalence of atomic bomb tests and use increased the amount of radioactive ¹⁴C in the atmosphere. This pulse of radiocarbon, along with known decay rates and discrimination processes, allowed organic material to be dated and tracked through ecosystems with relatively high precision (Hasler 2022). This tool has helped us understand that the mean age of carbon is much greater than the mean transit time of carbon in terrestrial systems, and that the mean age of soil carbon in tropical systems is an order of magnitude younger than the soil carbon in permafrost regions, both of which constrain and inform soil carbon loss and sequestration in Earth system models (Shi et al. 2020).

Combining classical metabolomics, which focuses on metabolite levels, with substrates that are labeled with stable isotopes can yield insights into metabolic flux and help resolve metabolic rate and flow (Yu et al. 2023). For example, ammonia and nitrate labeled with ¹⁵N coupled with metabolomics can quantify the rates of and pathways used by plants to assimilate the nitrogen from soil that they need for growth (Kurczy et al. 2016). As another example, stable carbon isotope labeling in trees experiencing prolonged drought can demonstrate that belowground tissues store and use carbohydrates preferentially over aboveground tissues during recovery from drought (Hagedorn et al. 2016).

Some challenges exist with isotopic approaches. Radioactive carbon isotopes (¹⁴C) face challenges of decay and dilution. Fossil fuel emissions, which produce a large amount of CO₂ with no ¹⁴C signal because fossil fuels have lost all ¹⁴C over millions of years of radioactive decay, are thus diluting the ¹⁴C tracer. Atmospheric CO₂, and therefore newly produced organic material, will appear as though it has “aged,” or lost ¹⁴C by decay. By 2050, fresh organic material could have the same ¹⁴C/¹²C ratio as

samples from 1050, and thus be indistinguishable by radiocarbon dating. As Graven (2015) notes, “Some current applications for ¹⁴C may cease to be viable, and other applications will be strongly affected.” Dating using ¹⁴C from solar proton events may offer a solution (e.g., Walker et al. 2023).

Atmospheric Gases, Flux Towers

Measures of atmospheric greenhouse gases—such as carbon dioxide, methane, and nitrous oxides—and water vapor, provide critical insights into ecosystem dynamics, carbon cycling, and climate feedback mechanisms. They enable monitoring of the

influence of biological variation, human activities, land-use change, and climate variations on the biosphere dynamics. Atmospheric gas concentrations have a long history of measurement, from flask samples to flux towers, where analyzers are used in combination with three-dimensional (3D) sonic anemometers to estimate gas flux. Flux towers facilitate the measurement and detection of gas exchanges between the biosphere and the atmosphere, serving as an essential tool for assessing how biological activities influence and are influenced by atmospheric changes. For example, towers measuring carbon and water fluxes over Northern Hemisphere forests demonstrated that these forests have increased their carbon gain per unit of water used over a 20-year period (Keenan et al. 2013). Systematic analyses of various competing hypotheses to explain this trend indicated that the observed increase is most consistent with a strong CO₂ “fertilization” effect, whereby trees are growing more due to more carbon substrate in the air.

Flux towers are important for scaling between terrestrial and atmospheric processes. For example, to gain insight into Earth’s metabolism and how it is changing in response to increasing surface temperature, networks of long-term measurements of atmospheric CO₂ concentrations have been combined with Earth system models to resolve why the Northern Hemisphere’s atmospheric (CO₂) seasonal amplitude (the difference between summer and winter CO₂ levels) has increased with surface warming. Forkel et al. (2016) found climate-warming stimulated plant carbon uptake faster than respiratory carbon release from the terrestrial biosphere, although seasonal respiration processes are influenced by a range of interacting drivers including snow, permafrost, vegetation composition and structure, drought, soil properties, and fire disturbance history (Chylek et al. 2022, Previdi et al. 2021, Rantanen et al. 2022, Treat et al. 2024). Efforts to further incorporate these process-level interactions are needed to better resolve both spatial and temporal variations in atmospheric CO₂ integrated across vast spatial scales. Other examples using flux towers combined with peripheral sensors, such as those that measure volatile organic compounds, are measuring forest stress from disturbances such as drought and insect outbreaks (Kravitz et al. 2016).

In specific cases where a flux tower or flask sampling is absent, biological samples can serve as indices for air quality. For nearly 25 years, the U.S Department of Agriculture (USDA) Forest Service’s Forest Inventory and Analysis (FIA) program collected lichens and recorded lichen richness on a subset of standardized plots throughout the United States (Jovan et al. 2021). These vouchered samples and data can serve as a proxy for air quality when they are validated or related to quantitative measures of air quality. For example, the relative dominance of lichen functional groups can be directly related to nitrogen deposition and be used to estimate an ecosystem’s critical load (Root et al. 2015).

Physical and Digital Samples and Collections That Inform Biodiversity

Biodiversity collections housed in natural history museums and herbaria provide the foundations of our knowledge and documentation of life on Earth (NASEM 2020). Preserved specimens capture species variation over time and space and are important to maintain because they allow us to study functional traits of organisms in common environments (Fontes et al. 2022, Perez et al. 2019). Increasingly, botanical gardens,

zoos, aquaria, seed collections, and insectaria are hubs of ex situ conservation efforts (Westwood et al. 2021, Wood et al. 2020). A wide range of efforts has been undertaken to enhance such collections, including digitization, aggregation of digital records into databases, enhancement of collections by citizen science, and collections of nonphysical specimens such as images or recordings.

Digitization of physical collections is an ongoing and important effort (NASEM 2020). A major effort by the National Science Foundation (NSF) to enable the digitization of specimens across the tree of life has been ongoing for the past decade.¹ These efforts are directly in line with NSF’s current emphasis on Innovative Use of Scientific Collections, which includes the Directorate of Biological Sciences (Division of Integrative Organismal Systems, Division of Biological Infrastructure, and Division of Environmental Biology), aimed at fostering innovative and diverse uses of collections and/or associated digital data for novel research and training applications. These data have provided the foundation for extended specimens and next-generation digitization efforts (i.e., digitizing additional data from original specimens). Extended specimens provide additional information derived from physical specimens that can be used to understand functional and ecological attributes of organisms. These include foliar spectral data from physical leaf samples (Kothari et al. 2023, Meireles et al. 2020), which can provide estimates of forest ecosystem function, and CT (computed tomography) scans of skeletons (Poo et al. 2022, Shi et al. 2018). These next-generation spectroscopic data (reflected or transmitted light across many wavelengths, typically 400–2,500 nm) collected from foliar samples can enable prediction of chemical and other functional traits of plants, such as leaf nutrient content predicting photosynthesis (Kothari et al. 2023, Serbin et al. 2014, Singh et al. 2015).

Digital records can then be aggregated into databases for multiple uses to connect across scales. Specimen records are incorporated by aggregator nodes across the globe that input species occurrence records and geographic coordinates and uncertainties into the Global Biodiversity Information Facility (GBIF²). In effect, these collections are a proxy for species distributions or trait information, provided the label information on the specimens and geographic data are accurate (Zizka et al. 2020). For species, researchers can then validate these proxies of species distributions with species distribution models, and merge with remotely sensed species occurrences, to improve estimates of changing population size and abundance in the face of global change (Cavender-Bares et al. 2020, Fretwell and Trathan 2021, Guzmán et al. 2023). For traits, such as plant functional trait data in databases such as the Plant Trait Database (TRY) and the Botanical Information and Ecology Network (BIEN) (Enquist et al. 2009, Kattge et al. 2020), researchers can link these to spectroscopic remote sensing (see below) to map functional trait variation in vegetation to provide critical habitat and ecosystem information. Similarly, in soil systems, microbial functional trait analysis is gaining prominence. Akin to the aggregation of plant trait data in TRY and BIEN, microbial trait data are pivotal in understanding soil microbial communities and their impact on ecosystem health (Barberán et al. 2015,

___________________

¹ See https://www.nsf.gov/awardsearch/showAward?AWD_ID=2027654 (accessed May 8, 2024).

² See gbif.org.

Buzzard et al. 2019, Wieder et al. 2015, Yang 2021). The integration of these microbial data with advanced spectroscopic techniques, which predict chemical and functional traits of plants, offers a comprehensive view of ecosystem functions and can be linked to remote sensing for habitat and ecosystem mapping.

Physical collections can be extended by citizen/community science efforts. Efforts such as iNaturalist³ enhance biological collections by providing a wealth of additional species occurrence records, greatly expanding the spatial extent and amount of available data. Despite the use of online training on data collection and protocols in community science efforts, there can be uncertainties and other problems with the collected data. Other collection approaches are coupled with digital libraries to train AI models that classify the identity of organisms with increasing accuracy (Vélez et al. 2023). These include camera trap images and acoustic recordings (e.g., Clark et al. 2023, Quinn et al. 2023).

Acoustic recording units (ARUs) facilitate the passive observation of birds and wildlife. Acoustic technologies are rapidly developing to obtain species occurrence and animal migration patterns at scales where monitoring has not previously been possible. Acoustic libraries of birds, for example, are well developed and are used to train models for identification with high accuracy (Kahl et al. 2021, Ruff et al. 2023). Similarly, camera traps provide a means to capture local occurrence and behavior records that enable population information on animals. These cameras are equipped with a motion sensor, usually a passive infrared sensor or an active infrared sensor using an infrared light beam and are automatically triggered by a change in activity in the vicinity, such as animal motion (O’Connell et al. 2011). As with ARUs, images of species obtained from camera traps can be automatically identified using ML models, facilitating efficient analysis pipelines (Tabak et al. 2019). One approach coupled with Zooniverse⁴ crowdsources the identification of images, compared to expert identification, to develop AI identification from camera images (Willi et al. 2019).

Finally, tracking of animal movement and migrations have advanced the study of a wide range of wildlife, including birds, zebras, elephants, caribou, and many other species. Collection of these data into large, organized databases such as MoveBank (Kays et al. 2022) have advanced this research substantially (Davidson et al. 2020, Tucker et al. 2018). The International Cooperation for Animal Research Using Space (ICARUS) antenna on the International Space Station received signals from tiny transmitters attached to over 800 species of animals ranging from bats to elephants to track migration patterns, but unfortunately in March 2022 was unexpectedly terminated, impacting our ability to assess the influence of biodiversity on ecosystem function. This functionality is expected to resume with the resurrection of ICARUS on small satellites (CubeSats) in 2025. Other future advances in animal tracking and expanded use of drones have potential to augment the resumption of tracking data from ICARUS.

___________________

³ See https://www.inaturalist.org/.

⁴ See https://www.zooniverse.org/.

temporal extents using airborne lidar data acquisitions, such as those systematically collected over NEON sites (Hakkenberg et al. 2023). These recent lidar-based 3D structure metrics and associated data products also provide valuable calibration information, particularly when they themselves are calibrated with field data (which is the case with GEDI lidar products), to map and model 3D structural property metrics over continental to global spatial extents (Duncanson et al. 2022, Ma et al. 2023).

Large-scale, coordinated remote sensing campaigns from government agencies, such as NASA, have been central to the vision for CSB, that is, conducting coordinated research across scales to provide novel insights that would not otherwise be possible. NASA campaigns and missions have evolved from the First International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment, which focused on the Konza Prairie in Kansas during the 1980s (Sellers et al. 1988); to the Boreal Ecosystem-Atmosphere Study (BOREAS) which targeted a gradient of boreal forest sites in central Canada during the 1990s (Sellers et al. 1995, 1997); the Large-Scale Biosphere-Atmosphere Experiment in Amazonia (LBA) which focused on Amazonian tropical forests during the 1990s and 2000s (Avissar and Nobre 2002, Avissar et al. 2002); and most recently the ongoing Arctic-Boreal Vulnerability Experiment (ABoVE) which is targeted on the tundra and taiga forest biomes of Alaska and western Canada; and the Department of Energy’s Next-Generation Ecosystem Experiments in the Arctic (NGEE Arctic) and the Tropics (NGEE Tropics). The BioSCAPE mission in South Africa and others currently being scoped will span tropical wet and dry forest regions in Central and South America, including in the Amazon and the Brazilian Cerrado, and West Africa (PAN tropical investigation of bioGeochemistry and Ecological Adaptation [PANGEA]), and separately across a range of arid land ecosystems (Actionable Science for Earth’s Changing Drylands [ARID]).

These coordinated campaigns, augmented with more localized manipulation experiments such as the Spruce and Peatland Responses Under Changing Environments (SPRUCE), provide valuable insights and serve as models for continental-scale research. The campaigns have allowed some of the best examples of CSB, including documenting how disturbances interact to produce a range of ecosystem states that persist for decades (Foster et al. 2022), how fire disturbance influences vegetation succession and associated feedbacks to climate through changing surface reflectivity (Massey et al. 2023), how the carbon budget of high latitudes has changed in recent decades (Wang et al. 2021, Watts et al. 2023), where methane sources arise across wetland landscapes and as a result of fire (Yoseph et al. 2023), how trees have been expanding their northern range limits in recent decades (Dial et al. 2022), where the boreal forest is becoming more and less productive (Berner and Goetz 2022), and how wildlife, including beavers and moose, are expanding their ranges and altering Arctic ecosystems (Tape et al. 2022), among many other examples.

Remote sensing campaigns and Earth observation data face challenges associated with continuity and storage. As technology advances and campaigns are discontinued and others launched (e.g., ICARUS), retrospective research to ensure continuity of

observations through time or that allows interoperability through statistically robust documentation of directional change (e.g., via time series analysis) will be challenging. This may require aggregating the spatial and temporal resolution of recent data to match that of data collected in prior years and decades. The sheer volume of data may also be a challenge for many users of these data due to computing demands (storage and processing speed). NSF’s CyVerse at the University of Arizona (formerly the iPlant Collaborative) was designed to create the capacity to address some of these challenges with its high-performance computing capabilities and cloud storage of large datasets.

Long-Term Experiments—Longitudinal in Time, Replicated in Space

Manipulative experiments are those that control specific variables that allow us to determine the causal factors that drive observed processes. Relevant to enabling CSB are experiments that control key variables, such as plant diversity, climate, and greenhouse gases. Experiments that control these variables have been critical in deciphering the impacts of global change factors on organism function, community assembly processes, and ecosystem function (Blondeel et al. 2024, Isbell et al. 2015, Kolton et al. 2022, Pastore et al. 2021, Pellegrini et al. 2021, Reich et al. 2018). Distributed long-term manipulative experiments that are replicated across ecosystems and biomes enable us to examine how the same causal drivers of change have consistent and heterogeneous impacts in space and time.

These distributed manipulative experiments often become organized networks (see section below), for example, Nutrient Network (NutNet) (Box 4-4), Drought Network (DroughtNet), and Tree Diversity Network (TreeDivNet) (Table 4-1). Control replicates (i.e., replicates that do not have treatments imposed) from long-term manipulative experiments by themselves often become some of the most valuable observational studies—which are located at specific sites and replicated in space and/or time. The results of these experiments can reveal global change feedbacks in changing environments on ecosystems and change our understanding of synergistic impacts of different factors (e.g., the Biodiversity, CO₂, and Nitrogen [BioCON] experiment (Reich 2009).

As with the other tools described above, long-term experiments experience myriad challenges. Investigators who measure any one variable for decades, such as air humidity, can experience changes in both theory and equipment based on that theory (Sonntag et al. 2021). For example, early measurements of air humidity at Coweeta Hydrologic Lab, a USDA Forest Service Experimental Forest that was established in the early 1930s, used hair hygrometers, while more recent measurements use electrodes and optical measurements (Miniat et al. 2021). When sensors and theory do change during the course of a long-term experiment, conducting simultaneous measurements with both sets of sensors over the full range of conditions is always desirable, but rarely available or affordable. Challenges with continuity of measurements, including human and financial resources, are thus almost always a challenge in long-term experiments.

Quantifying these mechanisms is still a challenge, with limited evidence outside of medicine for the effective use of multi-omics in this context.

Tools to Support Modeling for CSB: ML, AI, and Data Harmonization Centers

Approaches to facilitate data integration and interpretation include the application of AI techniques, for example, ML analysis to identify traits that predict variance in key parameters. These approaches can be used to identify features that can be integrated into hierarchical modeling to capture the dynamic processes that underpin emergent properties of the complex systems that make up ecosystems. Leveraging burgeoning multi-omic datasets necessitates AI and ML techniques to identify dynamics and ecosystem disturbances. Challenges in the application of AI and ML are described further in Chapter 3.

Distributed Active Archive Centers (DAACs) work to harmonize data. The Oak Ridge National Laboratory’s DAAC publishes and preserves NASA data relevant to terrestrial ecology, primarily field and airborne data. They also facilitate the use of NASA data in ways that address terrestrial ecology needs. They host over 1,700 datasets across 9 science themes and 36 missions and projects. They make large and complex remote sensing data easily accessible to researchers focusing on smaller scales of organization (e.g., organisms, plots), and allow integration of diverse and discrete ecological data at smaller scales and connecting them to large remote sensing data. One example is the soil moisture visualizer, where soil moisture datasets collected across observational networks (e.g., USDA Soil Climate Analysis Network (USDA SCAN), SNOw TELmetry Network (USDA SNOTEL), and NASA’s Soil Moisture Active Passive datasets) are aggregated and harmonized and made available to users (Shrestha and Boyer 2019). Data aggregation, harmonization, and visualization allow users, in this case, to make inferences on soil moisture availability or drought across wide expanses that eclipse the spatial extent of any one network.

Digital twin platforms, emerging technologies that integrate computer science, mathematics, engineering, and life sciences, offer transformative potential for CSB research (de Koning et al. 2023, NASEM 2024). These platforms can create high-fidelity, real-time digital replicas of biological systems, enabling researchers to conduct virtual experiments and simulations that are impractical or impossible in the real world. This capability is particularly valuable for studying large-scale biological processes and ecosystems. For example, digital twins can model complex interactions within ecosystems, providing insights into the impact of environmental changes, such as climate change or habitat destruction, on biodiversity. Researchers can use these models to test hypotheses, predict future scenarios, and develop conservation strategies. Moreover, digital twins can facilitate data integration from various sources, including satellite imagery, sensor networks, and field observations, helping researchers identify patterns and trends that might not be apparent from smaller-scale studies.

By leveraging advanced computational techniques such as ML and AI, digital twin platforms can analyze vast amounts of data, identify relationships, and predict outcomes

with high accuracy. This approach, linked with theory (Chapter 3, Figure 3-1-1), will enable the identification of gaps in knowledge and data collection, enabling hypothesis testing and refinement of models and theory that will lead to robust predictive power for ecological outcomes associated with climate change or ecosystem disturbances. This capacity is crucial for understanding the complexities of continental-scale biological phenomena, where traditional analytical methods may fall short.

NETWORKS

Scientific observational and experimental networks are an organized group of people, knowledge systems, infrastructure, and data focused on central scientific questions or goals that conduct experiments and/or observations across space and/or time (see Table 4-1 for some examples). Networks may standardize measurements across locations, or if they are not standardized but measure the same process, they can at least provide a way to scale or link different measurements or tools together to allow broad-scale inference. Successful networks establish transparent and inclusive policies for participation as a member, collaboration and publication, and the use and reuse of data (NRC 2015, Wilkinson et al. 2016). To promote collaboration and scientific advancements, networks often adopt open science and team science principles (NRC 2015). Networks involve coordination and communication across people, and although data are one of their outputs, they are not the only product (e.g., events, data visualizations, data management and oversight, AI/ML experts, natural history experts).

Many networks are established to answer a basic question or fulfill a mandate, for example: What is the U.S. forest inventory (USDA Forest Service FIA)? How and where is acid deposition affecting air and water quality (National Atmospheric Deposition Program [NADP] and the Clean Air Status and Trends NETwork [CASTNET])? What are the material biospheric fluxes of energy, carbon, and water across the U.S. terrestrial ecosystems’ boundary layer (U.S. Department of Energy’s [DOE’s] Ameriflux network)? How general is our current understanding of productivity–diversity relationships (NutNet)? How are the world’s forests changing in biodiversity and carbon (ForestGEO network)? In other words, they are question driven. Some networks, though, are not established to answer a particular question; rather, they function more as data collection or observational networks (e.g., NEON).

Similar to the varied reasons that networks are established, how networks are established also varies. Some are established in a top-down fashion; for example, networks are established by government mandate (e.g., FIA and NADP mentioned above), to accomplish a government agency’s mission (e.g., USDA Forest Service Experimental Forest and Range network, DOE’s Ameriflux, National Oceanic and Atmospheric Administration’s U.S. Climate Reference Network), or to coordinate experiments and/ or observations (e.g., NEON, Forest GEO, NutNet, DroughtNet, International Tundra EXperiment). Often, smaller networks can grow into larger and more formal networks. The Research Coordination Networks (RCNs) program under NSF has been a very successful tool in establishing networks and expanding smaller networks. For example, NutNet was initiated to address two key human influences on ecosystems—global nutrient augmentation and changes in consumer dynamics (Box 4-4). Networks can

TABLE 4-1 Scientific Observational and Experimental Networks

TABLE 4-1 Continued

TABLE 4-1 Continued

TABLE 4-1 Continued

TABLE 4-1 Continued

also be developed in response to requests for proposals, such as those from DOE. In the context of studying how physical, biological, and chemical processes change across scales in response to climate and anthropogenic drivers, RCNs can play a pivotal role in bringing together scientists to generate data, test hypotheses, and develop models.

Two networks have been instrumental in developing and enabling CSB—the Long Term Ecological Research (LTER) and NEON networks. The LTER network, founded in 1980, is a research network whereby scientists design and carry out long-term experimental and observational studies at one or more LTER sites to address ecological questions, often in a hypothesis-driven approach. The current 28 LTER sites span all major ecosystems across the United States and its territories, including marine, terrestrial,

Low-Investment, High-Impact Cross-Site Experimentation: Emphasizing the ethos of collective effort in CSB, NutNet promotes experiments that require minimal resource investment from individual investigators but collectively yield insights spanning diverse herbaceous ecosystems. Adopting an inclusive approach, NutNet’s membership is open to ecologists dedicated to its overarching goals.

NutNet has made important discoveries about how biodiversity and ecosystem function are linked in naturally assembled ecosystems, how consumers influence ecosystem productivity, and the extent of variation in drivers of species invasion.

and freshwater sites on the North American continent, South Pacific, and Antarctica. The LTER network was instrumental in informing the design of NEON, which grew out of a desire to create coordinated and systematic observations of biotic and abiotic phenomena across ecosystems at a continental scale. This systematic approach would reduce bias and uncertainty associated with individualized measurements at many different sites, from many different studies, and where data integration and synthesis are more difficult due to the variety of approaches, instruments, and survey designs deployed at each site and even within sites.

NEON has 81 sites across the United States that are highly coordinated in a hierarchical design (Figure 4-2). Measurements and observations are systematically

coordinated with standardized methods at every site. NEON is solely focused on observational approaches to measure biotic and abiotic ecological variables in a spatially nested design where plots are nested within sites, nested with 20 ecoclimate domains. All data are collected in a coordinated and standardized way within 47 terrestrial and 34 freshwater systems across the following data themes: atmosphere, biogeochemistry, ecohydrology, land cover and processes, and organisms, populations, and communities. Although separate networks, LTER and NEON are complementary; LTER’s long-term record and experimental insights can be paired with NEON’s spatially nested design of plots within sites within domains. In fact, 12 NEON sites are co-located with LTER sites, which directly leverage the long-term data and findings of the LTER sites. As with individual LTER sites and the NEON network, which is funded for 30 years, the longevity of a network can be finite, particularly when funding support or mandates disappear, or when questions are fully answered or change. A network and its products can then be repurposed to allow synthesis.

CENTERS OF SYNTHESIS

While collaborative networks provide a powerful means to bring together scientists across disciplines to generate and synthesize complex data across space and time, synthesis centers play an increasingly critical role in data integration, processing, and synthesis. Synthesis centers are places where researchers come together to synthesize large datasets, test hypotheses across systems, develop new theories, refine conceptual frameworks, or generate insights that pertain to large-scale biological processes, combining perspectives across disciplines. Such centers often provide resources, infrastructure, and collaborative environments to foster integrative research, especially in fields that deal with big data and complex systems, such as CSB. Such centers are generally focused on using existing data to address questions of critical importance.

Similar to the evolution of NASA missions and the evolution of networks, centers of synthesis have evolved in ways that support and enable CSB. Synthesis centers in the United States began in 1995 with the launch of the National Center for Ecological Analysis and Synthesis (NCEAS). Funded by NSF for 17 years, NCEAS focused on synthesis, data access, and collaboration. NCEAS is now funded by a range of foundations, institutes, individuals, and partnerships, and continues to host many synthesis and workshop events. Further, NCEAS now operates and hosts the LTER Network Office, enabling further synthesis within LTER. NCEAS also led to the development of the Knowledge Network for Biocomplexity Data Repository and the Environmental Data Initiative, which now connect and support data synthesis across NEON and LTER, as well as many other networks. As one of the first NSF-funded synthesis centers, the NCEAS model has led to over 20 other synthesis centers. NSF-funded synthesis centers in the United States include the Socio-Environmental Synthesis Center (SESYNC), the National Institute for Mathematical and Biological Synthesis, the National Evolutionary Synthesis Center, and now the Environmental Data Science Innovation & Inclusion Lab (ESIIL). SESYNC is aimed at bringing together natural and socioeconomic data to

address questions of societal importance. ESIIL is the newest NSF–funded synthesis center aimed at developing a user community to leverage the wealth of environmental data to address environmental science questions. ESIIL emphasizes inclusion and diversity in team science and specializing in collaborative cyberinfrastructure to support the use of continental-scale data including that generated by NEON. ESIIL is explicitly envisioned as a critical synthesis center for supporting large-scale research that spans levels of biological organization, that is, described herein as CSB.

ATTRIBUTES OF SUCCESSFUL TOOLS AND NETWORKS AND CHALLENGES

Tools

Successful tools for CSB should be able to help bridge or scale processes across space and time. Most of the tools presented above allow inference across space or time and scaling or integrating across levels of biological organization. The NSF Biology Integration Institutes (BII) were established by the NSF Division of Biological Infrastructure to integrate across spatial-, temporal-, and biological scale interdisciplinary science and to promote transdisciplinary team science. The BII program specifically addresses the problem of fragmentation of the biological sciences into subdisciplines and promotes bridging scales. For example, the ASCEND (Advancing Spectral biology in Change ENvironments to understand Diversity) BII uses spectroscopic reflectance data as a common data type across scales—measured from hand-held instruments on plant tissues and from airborne and spaceborne platforms on canopies, ecosystems, and landscapes—to detect changes in biological variation and its emergent properties within whole plants, ecosystems, landscapes, and at continental scale (Box 4-3). These and related efforts have used ML and statistical models from spectroscopic imagery to predict aboveground functional traits of vegetation (Asner and Martin 2008, Miraglio et al. 2023, Wang et al. 2020), plant diversity (Gholizadeh et al. 2022, Laliberte et al. 2020, Wang et al. 2018), phylogenetic composition (Griffith et al. 2023), forest productivity (Williams et al. 2020), disease impacts on trees (Guzmán et al. 2023, Sapes et al. 2024), as well as belowground properties and microbial processes based on mechanistic linkages between aboveground and belowground portions of ecosystems (Cavender-Bares et al. 2022, Sousa et al. 2021). The integration of spectral reflectance and lidar tools with experimental, ground-based observations and laboratory analyses to test hypotheses within conceptual and theoretical frameworks is a promising approach to CSB (Box 4-3).

As the realm of CSB research continues to expand, we recognize the emergent and multifaceted challenges tied to the storage and management of biological data. The rich tapestry of life on Earth, replete with its vast array of specimens, necessitates a profound evolution in our storage methodologies. Traditionally, institutions such as museums, herbariums, and biological repositories have undertaken the formidable task of physically preserving specimens. For example, the Smithsonian National Museum of

Natural History safeguards over 148 million specimens, spanning algae to mammals⁵; the Microbiota Vault (Bello et al. 2018) is storing physical materials from global ecosystems to preserve microbial diversity. The intricacies of storing such diverse specimens demand environments with meticulous control measures to deter degradation, underlined by state-of-the-art infrastructure and consistent upkeep.

The preservation paradigm has broadened to include the digital sphere. Beyond the conventional realm of photographs and descriptive annotations, there is an influx of high-resolution 3D scans, intricate genetic sequences, and expansive data encompassing an organism’s habitat and behavioral traits. Online platforms such as the Barcode of Life Data System support these diverse data types, amalgamating both molecular and morphological data, subsequently assisting researchers in their endeavors related to biodiversity and species identification. The rapid expansion of such platforms underscores an acute surge in data storage requisites. Furthermore, ensuring the accuracy of geospatial data remains paramount. Precise location details of specimens enrich biological studies, particularly those anchored in conservation, biogeography, and ecology. By facilitating the tracking of migratory patterns and assessing the repercussions of climate change on specific habitats, accurate geospatial data prove invaluable. Repositories such as the GBIF epitomize this endeavor, curating biodiversity records while preserving the integrity of geospatial information, thereby offering a comprehensive vista of biodiversity distribution. Yet, with data accumulation comes the imperative for ensuring privacy. Coupled with this is the imperative for rigorous security protocols, ensuring the safeguarding of data against potential breaches. The protocols embraced by entities such as GenBank serve as benchmarks in this domain, emphasizing data integrity and provenance and the privacy rights intrinsically linked to it. Last, the concept of “ownership” of data might prove restrictive and even inappropriate. Given the sheer scale of CSB, collaborations are inevitable. These collaborations must be based on mutual respect, trust, and a shared vision.

Networks

Attributes of networks and their “high-performing collaborative research teams” (Cheruvelil et al. 2014) that make them successful for tackling CSB include having sites within the network that represent the array of biotic and abiotic conditions across the network, and sites that have stable and adequate financial resources. Successful networks also adopt FAIR data principles, which aim to make data Findable, Accessible, Interoperable, and strive to have data Reused in other studies (Wilkinson et al. 2016). Published data that are submitted to archives, data repositories, or journals are typically assigned a Digital Object Identifier (DOI), which allows data to be found. Published data packages that contain metadata, raw (Level 0, or L0) and processed (Level 1-4, or L1-L4) data files, and any accompanying information, such as maps, documentation, or code, allow data to be reused. Interoperable data are those from

___________________

⁵ See https://naturalhistory.si.edu/about#:~:text=We%20steward%20a%20collection%20of,moments%20we%20find%20Earth’s%20story (accessed April 28, 2024).

different methods or sources that can be merged or integrated with minimal effort, for example by using standards. Efforts by iDigBio to develop a national infrastructure that oversees implementation of standards and best practices for digitization, include developing a customized cloud computing environment for collections and planning for long-term sustainability of the national digitization effort. Successful networks also adopt principles of team science, leadership, and governance (Stokols et al. 2008). They espouse inclusive and diverse membership, where members have a voice, and where team diversity (broadly defined) is effectively fostered (Cheruvelil et al. 2014). Interpersonal skills and training are taught and followed (Cheruvelil et al. 2014), and there is strong leadership from multiple people. The people within networks also share a sense of holding each other accountable to achieving the scientific goals and delivering to a diverse group of stakeholders.

Many networks formally solicit input from and transfer knowledge to their stakeholder groups. For example, the USDA Forest Service FIA program solicits and listens to user feedback through an annual user group meeting, as well as from collaborators at academic, other government, and nongovernmental organizations. FIA has adapted to changing user needs by increasing their capacity to analyze and publish data, and by expanding the scope of their data collection to include an additional suite of attributes on a subsample of plots such as soils, understory vegetation, tree crown conditions, down woody material, and invasive species.

Networks also face numerous challenges, from data inoperability to funding. Some networks are comprised of sites that measure the same process but use widely differing methods to do so. For example, the LTER network is comprised of separate sites that conduct long-term, place-based observations. Measurement approaches and frequencies are not standardized across sites. Thus, comparing or synthesizing observations across the network or subsets of sites is challenging. NEON attempted to address the limitations of the LTERs by collecting systematic measurements across sites, but NEON data are big and complex. NEON sites collect 110 TB yr^-1, and many NEON data products are not model-ready. Raw sensor or human-collected data (L0) are available upon request, but progressively processed data (e.g., L1–L4) are served on the NEON Data Portal and readily usable. One challenge for NEON is that NEON data have a high level of uncertainty at the continental scale; more data coverage across more ecosystems is needed to be of most use to CSB.

Even if network data are readily available and usable, there remains a challenge of data “unfamiliarity,” or data users who are one step removed from the physical data collection and thus unfamiliar with the data’s limitations. Historically, this challenge is overcome by involving the principal investigator (PI) in data analysis and/or publications. But, widening a research team takes time and effort to build relationships, and if involving the PI is not an option, then data usage and scientific advancements can be limited. One network is experimenting with this challenge. Data collected across the Ameriflux network previously fell under their “legacy data policy,” which mandated that the PI be a co-author. Ameriflux now has an open-data policy, where the PI does not have to be on the publication. This experiment could be risky, because it removes the site knowledge from the paper.

• ecology, engineering, atmospheric science, remote sensing, hydrology, and other disciplines such as social sciences.
• Allocate resources for next-generation digitization of biodiversity collections to enhance their utility as reference standards for CSB and to enable the development of digital ecosystem twins. This will require new bioinformatics tools to enable access to and management of preserved and living collections to facilitate their utility for interpretation of in situ and remotely sensed data.
• Develop communities that can leverage interdisciplinary data from NSF platforms and various networks, akin to the use of National Ecological Observatory Network (NEON) data by researchers funded by the previous Macrosystem Biology Program. An example would be incorporating macrosystems/synthesis research to create living data products (those that are continually updated) that inform biological processes at continental scale. If done, data from one platform could serve as calibration/validation for other data products and layers from other platforms to facilitate interpolation, extrapolation, and/or imputation. Major government assessments, such as the National Nature Assessment, could also leverage data provided by integration of these platforms.
• Explore joint support of integrative science via interagency (e.g., NASA, U.S. Department of Energy) RFPs, for example, multiscale coordinated interdisciplinary field campaigns (e.g., Arctic-Boreal Vulnerability Experiment, the Biodiversity Survey of the Cape, and the Large-scale Biosphere-Atmosphere Experiment in Amazonia).
• Support efforts to understand how to sample for continental-scale biological questions. Few spatially distributed networks with standardized sampling exist, and those that do exist require great resources. Investment in research to understand sampling theory (time and space) for capturing continental scales and cross-boundary interactions (e.g., metacoupling, as discussed in Chapter 2) is needed.
• Explore development of interagency incentives and mechanisms for public–private partnerships that can facilitate targeted private investment in data development and integration across scales and types, for example, engaging in AI-driven data analysis and data product development.

Conclusion 4-1: Development of research infrastructure for CSB would also benefit from actions by other agencies. Examples include the following:

• The Small Business Administration could develop agency-specific innovation research (Small Business Innovation Research) and technology transfer research (Small Business Technology Transfer Research) RFPs that focus on AI, machine learning, and sensor development for the biological and environmental sciences.
• NASA’s continued support for the Surface Biology and Geology (SBG) mission, the only satellite instrument dedicated to biological processes that will specifically enable CSB, is an important contribution. SBG will provide

• continuous data across continents and the globe to fill in the gaps from NEON and enable baseline information to track changes in biological processes through time. Other NASA Explorer and Incubator missions recommended by the Decadal Survey in Earth Science and Applications from Space (NASEM 2018) complement SBG via lidar and radar measurements of ecosystem three-dimensional D structure.
• Agencies engaged in these efforts could continue to support scientific assessments and action-oriented efforts that inform policies guided by or consistent with the UN Convention on Biological Diversity. These include the Global Biodiversity Information Facility, the Group on Earth Observations Biodiversity Observation Network, the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services.
• Continue support of related domestic efforts that contribute to CSB, for example, the ongoing National Nature Assessment, U.S. Geological Survey national Biodiversity and Climate Change Assessment, and the U.S. contribution to the 30×30 Conservation initiative (White House, 2021).

REFERENCES