Wastewater surveillance for bacterial targets: current challenges and future goals.

Wastewater-based epidemiology (WBE) expanded rapidly in response to the COVID-19 pandemic. As the public health emergency has ended, researchers and practitioners are looking to shift the focus of existing wastewater surveillance programs to other targets, including bacteria. Bacterial targets may pose some unique challenges for WBE applications. To explore the current state of the field, the National Science Foundation-funded Research Coordination Network (RCN) on Wastewater Based Epidemiology for SARS-CoV-2 and Emerging Public Health Threats held a workshop in April 2023 to discuss the challenges and needs for wastewater bacterial surveillance. The targets and methods used in existing programs were diverse, with twelve different targets and nine different methods listed. Discussions during the workshop highlighted the challenges in adapting existing programs and identified research gaps in four key areas: choosing new targets, relating bacterial wastewater data to human disease incidence and prevalence, developing methods, and normalizing results. To help with these challenges and research gaps, the authors identified steps the larger community can take to improve bacteria wastewater surveillance. This includes developing data reporting standards and method optimization and validation for bacterial programs. Additionally, more work is needed to understand shedding patterns for potential bacterial targets to better relate wastewater data to human infections. Wastewater surveillance for bacteria can help provide insight into the underlying prevalence in communities, but much work is needed to establish these methods.IMPORTANCEWastewater surveillance was a useful tool to elucidate the burden and spread of SARS-CoV-2 during the pandemic. Public health officials and researchers are interested in expanding these surveillance programs to include bacterial targets, but many questions remain. The NSF-funded Research Coordination Network for Wastewater Surveillance of SARS-CoV-2 and Emerging Public Health Threats held a workshop to identify barriers and research gaps to implementing bacterial wastewater surveillance programs.

A novel climate and health decision support platform: Approach, outputs, and policy considerations.

BACKGROUND: The adverse health impacts of climate change are increasingly apparent and the need for adaptation activities is pressing. Risks, drivers, and decision contexts vary significantly by location, and high-resolution, place-based information is needed to support decision analysis and risk reduction efforts at scale. METHODS: Using the Intergovernmental Panel on Climate Change (IPCC) risk framework, we developed a causal pathway linking heat with a composite outcome of heat-related morbidity and mortality. We used an existing systematic literature review to identify variables for inclusion and the authors' expert judgment to determine variable combinations in a hierarchical model. We parameterized the model for Washington state using observational (1991-2020 and June 2021 extreme heat event) and scenario-driven temperature projections (2036-2065), compared outputs against relevant existing indices, and analyzed sensitivity to model structure and variable parameterization. We used descriptive statistics, maps, visualizations and correlation analyses to present results. RESULTS: The Climate and Health Risk Tool (CHaRT) heat risk model contains 25 primary hazard, exposure, and vulnerability variables and multiple levels of variable combinations. The model estimates population-weighted and unweighted heat health risk for selected periods and displays estimates on an online visualization platform. Population-weighted risk is historically moderate and primarily limited by hazard, increasing significantly during extreme heat events. Unweighted risk is helpful in identifying lower population areas that have high vulnerability and hazard. Model vulnerability correlate well with existing vulnerability and environmental justice indices. DISCUSSION: The tool provides location-specific insights into risk drivers and prioritization of risk reduction interventions including population-specific behavioral interventions and built environment modifications. Insights from causal pathways linking climate-sensitive hazards and adverse health impacts can be used to generate hazard-specific models to support adaptation planning.

Urban Stormwater Characterization, Control and Treatment.

A review of literature related to stormwater runoff characterization and its subsequent management and treatment from 2017 was conducted. The 250 articles summarized herein are organized along three central themes: (1) stormwater quality and quantity characteristics, (2) site-scale stormwater management practices, with a focus on low impact development (LID) and green infrastructure (GI), and (3) watershed-scale performance of stormwater controls. Within each section, common research themes and future work are highlighted.

Urban Stormwater Characterization, Control, and Treatment.

A summary of 246 studies published in 2016 on topics related to the characterization and management of urban stormwater runoff is presented in the following review. The review is structured along three major topical areas: (1) general characterization of stormwater quantity and quality; (2) engineered systems for stormwater control and treatment, including erosion and sediment control practices, constructed stormwater ponds and wetlands, bioretention, permeable pavement, greenroofs, and rainwater harvesting and (3) watershedscale application of stormwater treatment and control practices. Common research themes and needs highlighted throughout this review include efforts to better understand stormwater transport and treatment mechanisms and their representation in models, advancements to optimize the design of stormwater control measures to meet specific hydrologic and/or water quality targets, and increasing understanding of the biophysical and social factors that influence watershed-scale implementation of low impact development and other stormwater control measures.

Tree mortality from an exceptional drought spanning mesic to semiarid ecoregions.

Significant areas of the southern USA periodically experience intense drought that can lead to episodic tree mortality events. Because drought tolerance varies among species and size of trees, such events can alter the structure and function of terrestrial ecosystem in ways that are difficult to detect with local data sets or solely with remote-sensing platforms. We investigated a widespread tree mortality event that resulted from the worst 1-year drought on record for the state of Texas, USA. The drought affected ecoregions spanning mesic to semiarid climate zones and provided a unique opportunity to test hypotheses related to how trees of varying genus and size were affected. The study was based on an extensive set of 599 distributed plots, each 0.16 ha, surveyed in the summer following the drought. In each plot, dead trees larger than 12.7 cm in diameter were counted, sized, and identified to the genus level. Estimates of total mortality were obtained for each of 10 regions using a combination of design-based estimators and calibrated remote sensing using MODIS 1-yr change in normalized difference vegetation index products developed by the U.S. Forest Service. As compared with most of the publicized extreme die-off events, this study documents relatively low rates of mortality occurring over a very large area. However, statewide, regional tree mortality was massive, with an estimated 6.2% of the live trees perishing, nearly nine times greater than normal annual mortality. Dead tree diameters averaged larger than the live trees for most ecoregions, and this trend was most pronounced in the wetter climate zones, suggesting a potential re-ordering of species dominance and downward trend in tree size that was specific to climatic regions. The net effect on carbon storage was estimated to be a redistribution of 24-30 Tg C from the live tree to dead tree carbon pool. The dead tree survey documented drought mortality in more than 29 genera across all regions, and surprisingly, drought resistant and sensitive species fared similarly in some regions. Both angiosperms and gymnosperms were affected. These results highlight that drought-driven mortality alters forest structure differently across climatic regions and genera.

Urban Stormwater Characterization, Control, and Treatment.

The following review presents a synthesis of 181 journal articles published during 2015 that represent progress toward better characterizing, controlling, and treating urban stormwater runoff. The review is structured by general topical areas related to (1) stormwater quality and quantity characterization; (2) engineered stormwater control and treatment practices, including erosion and sediment control, stormwater ponds, constructed stormwater wetlands, bioretention, permeable pavement, greenroofs, and rainwater harvesting systems; and (3) watershed-scale modeling and optimization of stormwater control and treatment practices. Common research themes emerging from this collection of studies include potential to enhance hydrologic and pollutant treatment performance of stormwater practices via media amendments and the use of innovative outlet control structures, as well as development of a more mechanistic understanding of hydrologic and water-quality functions to inform modeling and performance predictions. These studies serve to expand the field's knowledge base and will inform future efforts to further improve stormwater control and treatment at various spatial and temporal scales.

The effect of permafrost thaw on old carbon release and net carbon exchange from tundra.

Permafrost soils in boreal and Arctic ecosystems store almost twice as much carbon as is currently present in the atmosphere. Permafrost thaw and the microbial decomposition of previously frozen organic carbon is considered one of the most likely positive climate feedbacks from terrestrial ecosystems to the atmosphere in a warmer world. The rate of carbon release from permafrost soils is highly uncertain, but it is crucial for predicting the strength and timing of this carbon-cycle feedback effect, and thus how important permafrost thaw will be for climate change this century and beyond. Sustained transfers of carbon to the atmosphere that could cause a significant positive feedback to climate change must come from old carbon, which forms the bulk of the permafrost carbon pool that accumulated over thousands of years. Here we measure net ecosystem carbon exchange and the radiocarbon age of ecosystem respiration in a tundra landscape undergoing permafrost thaw to determine the influence of old carbon loss on ecosystem carbon balance. We find that areas that thawed over the past 15 years had 40 per cent more annual losses of old carbon than minimally thawed areas, but had overall net ecosystem carbon uptake as increased plant growth offset these losses. In contrast, areas that thawed decades earlier lost even more old carbon, a 78 per cent increase over minimally thawed areas; this old carbon loss contributed to overall net ecosystem carbon release despite increased plant growth. Our data document significant losses of soil carbon with permafrost thaw that, over decadal timescales, overwhelms increased plant carbon uptake at rates that could make permafrost a large biospheric carbon source in a warmer world.

Critical Review of Technical Questions Facing Low Impact Development and Green Infrastructure: A Perspective from the Great Plains.

Since its inception, Low Impact Development (LID) has become part of urban stormwater management across the United States, marking progress in the gradual transition from centralized to distributed runoff management infrastructure. The ultimate goal of LID is full, cost-effective implementation to maximize watershed-scale ecosystem services and enhance resilience. To reach that goal in the Great Plains, the multi-disciplinary author team presents this critical review based on thirteen technical questions within the context of regional climate and socioeconomics across increasing complexities in scale and function. Although some progress has been made, much remains to be done including continued basic and applied research, development of local LID design specifications, local demonstrations, and identifying funding mechanisms for these solutions. Within the Great Plains and beyond, by addressing these technical questions within a local context, the goal of widespread acceptance of LID can be achieved, resulting in more effective and resilient stormwater management.

Real-Time Monitoring of SARS-CoV-2 Variants in Oklahoma Wastewater through Allele-Specific RT-qPCR.

During the COVID-19 pandemic, wastewater surveillance was used to monitor community transmission of SARS-CoV-2. As new genetic variants emerged, the need for timely identification of these variants in wastewater became an important focus. In response to increased reports of Omicron transmission across the United States, the Oklahoma Wastewater Surveillance team utilized allele-specific RT-qPCR assays to detect and differentiate variants, such as Omicron, from other variants found in wastewater in Oklahoma. The PCR assays showed presence of the Omicron variant in Oklahoma on average two weeks before official reports, which was confirmed through genomic sequencing of selected wastewater samples. Through continued surveillance from November 2021 to January 2022, we also demonstrated the transition from prevalence of the Delta variant to prevalence of the Omicron variant in local communities. We further assessed how this transition correlated with certain demographic factors characterizing each community. Our results highlight RT-qPCR assays as a rapid, simple, and cost-effective method for monitoring the community spread of SARS-CoV-2 genetic variants in wastewater. Additionally, they demonstrate that specific demographic factors such as ethnic composition and household income can correlate with the timing of SARS-CoV-2 variant introduction and spread.

Molecular detection of Campylobacter spp. and fecal indicator bacteria during the northern migration of sandhill cranes (Grus canadensis) at the central Platte River.

The risk to human health of the annual sandhill crane (Grus canadensis) migration through Nebraska, which is thought to be a major source of fecal pollution of the central Platte River, is unknown. To better understand potential risks, the presence of Campylobacter species and three fecal indicator bacterial groups (Enterococcus spp., Escherichia coli, and Bacteroidetes) was assayed by PCR from crane excreta and water samples collected during their stopover at the Platte River, Nebraska, in 2010. Genus-specific PCR assays and sequence analyses identified Campylobacter jejuni as the predominant Campylobacter species in sandhill crane excreta. Campylobacter spp. were detected in 48% of crane excreta, 24% of water samples, and 11% of sediment samples. The estimated densities of Enterococcus spp. were highest in excreta samples (mean, 4.6 × 10(8) cell equivalents [CE]/g), while water samples contained higher levels of Bacteroidetes (mean, 5.1 × 10(5) CE/100 ml). Enterococcus spp., E. coli, and Campylobacter spp. were significantly increased in river water and sediments during the crane migration period, with Enterococcus sp. densities (~3.3 × 10(5) CE/g) 2 to 4 orders of magnitude higher than those of Bacteroidetes (4.9 × 10(3) CE/g), E. coli (2.2 × 10(3) CE/g), and Campylobacter spp. (37 CE/g). Sequencing data for the 16S rRNA gene and Campylobacter species-specific PCR assays indicated that C. jejuni was the major Campylobacter species present in water, sediments, and crane excreta. Overall, migration appeared to result in a significant, but temporary, change in water quality in spring, when there may be a C. jejuni health hazard associated with water and crops visited by the migrating birds.

The effect of Paenibacillus on IDEXX Enterolert results from freshwater stream environments.

Enterolert, a fluorogenic substrate test, is used as a quantitative method for determining freshwater concentrations of Enterococcus for water quality indicators. However, there is some evidence from recent studies suggesting that Enterolert may not suppress false positives due to pollution sources in waterbodies. In this study, we evaluated this method by analyzing field water and sediment samples from four freshwater streams. We also performed a laboratory microcosm study from two of the stream sediments. The Enterolert method was investigated by phenotypic and genomic analyses for accuracy of isolating and quantifying Enterococcus and/or Streptococcus. Additionally, we tested isolates from Enterolert panels for antibiotic resistance. Results from the field and microcosm studies from initial to final time points indicated that false positives were predominantly Paenibacillus spp. and other non-fecal indicator bacteria. Furthermore, the microcosm study indicated shifts from lactic acid to non-lactic acid bacteria between initial to final time points, but Enterococcus concentrations from Enterolert panels remained stable for the duration of the study for both stream sediments. Antibiotic resistance indicated no distinct pattern of resistance or susceptibility to a suite of antibiotics. However, all isolates tested were resistant to bacitracin and nalidixic acid. In conclusion, we found that Enterolert was not exclusively selective for Enterococcus from freshwater environments and that sediment and polluted waterbodies have the potential to skew the presumed concentrations. More research is needed to evaluate the effectiveness and selectivity of the medium used for the fluorogenic substrate test for Enterococcus enumeration.

Using Wastewater Surveillance to Monitor Gastrointestinal Pathogen Infections in the State of Oklahoma.

During the COVID-19 pandemic, wastewater surveillance was widely used to monitor temporal and geographical infection trends. Using this as a foundation, a statewide program for routine wastewater monitoring of gastrointestinal pathogens was established in Oklahoma. The results from 18 months of surveillance showed that wastewater concentrations of Salmonella, Campylobacter, and norovirus exhibit similar seasonal patterns to those observed in reported human cases (F = 4-29, p < 0.05) and that wastewater can serve as an early warning tool for increases in cases, offering between one- and two-weeks lead time. Approximately one third of outbreak alerts in wastewater correlated in time with confirmed outbreaks of Salmonella or Campylobacter and our results further indicated that several outbreaks are likely to go undetected through the traditional surveillance approach currently in place. Better understanding of the true distribution and burden of gastrointestinal infections ultimately facilitates better disease prevention and control and reduces the overall socioeconomic and healthcare related impact of these pathogens. In this respect, wastewater represents a unique opportunity for monitoring infections in real-time, without the need for individual human testing. With increasing demands for sustainable and low-cost disease surveillance, the usefulness of wastewater as a long-term method for tracking infectious disease transmission is likely to become even more pronounced.

Development and evaluation of a quantitative PCR assay targeting sandhill crane (Grus canadensis) fecal pollution.

While the microbial water quality in the Platte River is seasonally impacted by excreta from migrating cranes, there are no methods available to study crane fecal contamination. Here we characterized microbial populations in crane feces using phylogenetic analysis of 16S rRNA gene fecal clone libraries. Using these sequences, a novel crane quantitative PCR (Crane1) assay was developed, and its applicability as a microbial source tracking (MST) assay was evaluated by determining its host specificity and detection ability in environmental waters. Bacteria from crane excreta were dominated by bacilli and proteobacteria, with a notable paucity of sequences homologous to Bacteroidetes and Clostridia. The Crane1 marker targeted a dominant clade of unclassified Lactobacillales sequences closely related to Catellicoccus marimammalium. The host distribution of the Crane1 marker was relatively high, being positive for 69% (66/96) of the crane excreta samples tested. The assay also showed high host specificity, with 95% of the nontarget fecal samples (i.e., n = 553; 20 different free-range hosts) being negative. Of the presumed crane-impacted water samples (n = 16), 88% were positive for the Crane1 assay, whereas none of the water samples not impacted by cranes were positive (n = 165). Bayesian statistical models of the Crane1 MST marker demonstrated high confidence in detecting true-positive signals and a low probability of false-negative signals from environmental water samples. Altogether, these data suggest that the newly developed marker could be used in environmental monitoring studies to study crane fecal pollution dynamics.

Supportive breeding boosts natural population abundance with minimal negative impacts on fitness of a wild population of Chinook salmon.

While supportive breeding programmes strive to minimize negative genetic impacts to populations, case studies have found evidence for reduced fitness of artificially produced individuals when they reproduce in the wild. Pedigrees of two complete generations were tracked with molecular markers to investigate differences in reproductive success (RS) of wild and hatchery-reared Chinook salmon spawning in the natural environment to address questions regarding the demographic and genetic impacts of supplementation to a natural population. Results show a demographic boost to the population from supplementation. On average, fish taken into the hatchery produced 4.7 times more adult offspring, and 1.3 times more adult grand-offspring than naturally reproducing fish. Of the wild and hatchery fish that successfully reproduced, we found no significant differences in RS between any comparisons, but hatchery-reared males typically had lower RS values than wild males. Mean relative reproductive success (RRS) for hatchery F(1) females and males was 1.11 (P = 0.84) and 0.89 (P = 0.56), respectively. RRS of hatchery-reared fish (H) that mated in the wild with either hatchery or wild-origin (W) fish was generally equivalent to W × W matings. Mean RRS of H × W and H × H matings was 1.07 (P = 0.92) and 0.94 (P = 0.95), respectively. We conclude that fish chosen for hatchery rearing did not have a detectable negative impact on the fitness of wild fish by mating with them for a single generation. Results suggest that supplementation following similar management practices (e.g. 100% local, wild-origin brood stock) can successfully boost population size with minimal impacts on the fitness of salmon in the wild.

Predicting COVID-19 cases in diverse population groups using SARS-CoV-2 wastewater monitoring across Oklahoma City.

SARS-CoV-2 was discovered among humans in late 2019 and rapidly spread across the world. Although the virus is transmitted by respiratory droplets, most infected persons also excrete viral particles in their feces. This fact prompted a range of studies assessing the usefulness of wastewater surveillance to determine levels of infection and transmission and produce early warnings of outbreaks in local communities, independently of human testing. In this study, we collected samples of wastewater from 13 locations across Oklahoma City, representing different population types, twice per week from November 2020 to end of March 2021. Wastewater samples were collected and analyzed for the presence and concentration of SARS-CoV-2 RNA using RT-qPCR. The concentration of SARS-CoV-2 in the wastewater showed notable peaks, preceding the number of reported COVID-19 cases by an average of one week (ranging between 4 and 10 days). The early warning lead-time for an outbreak or increase in cases was significantly higher in areas with larger Hispanic populations and lower in areas with a higher household income or higher proportion of persons aged 65 years or older. Using this relationship, we predicted the number of cases with an accuracy of 81-92% compared to reported cases. These results confirm the validity and timeliness of using wastewater surveillance for monitoring local disease transmission and highlight the importance of differences in population structures when interpreting surveillance outputs and planning preventive action.

Spatial variability in tree-ring carbon isotope discrimination in response to local drought across the entire loblolly pine natural range.

Considering the temporal responses of carbon isotope discrimination (Δ13C) to local water availability in the spatial analysis of Δ13C is essential for evaluating the contribution of environmental and genetic facets of plant Δ13C. Using tree-ring Δ13C from years with contrasting water availability at 76 locations across the natural range of loblolly pine, we decomposed site-level Δ13C signals to maximum Δ13C in well-watered conditions (Δ13Cmax) and isotopic drought sensitivity (m) as a change in Δ13C per unit change of Palmer's Drought Severity Index (PDSI). Site water status, especially the tree lifetime average PDSI, was the primary factor affecting Δ13Cmax. The strong spatial correlation exhibited by m was related to both genetic and environmental factors. The long-term average water availability during the period relevant to trees as indicated by lifetime average PDSI correlated with Δ13Cmax, suggesting acclimation in tree gas-exchange traits, independent of incident water availability. The positive correlation between lifetime average PDSI and m indicated that loblolly pines were more sensitive to drought at mesic than xeric sites. The m was found to relate to a plant's stomatal control and may be employed as a genetic indicator of efficient water use strategies. Partitioning Δ13C to Δ13Cmax and m provided a new angle for understanding sources of variation in plant Δ13C, with several fundamental and applied implications.

Long-term effects of catastrophic wind on southern US coastal forests: Lessons from a major hurricane.

Threats posed by windstorms are an increasing concern to forest managers in the southern United States (US). Studies suggest that the southern US will experience an increase in the occurrence as well as the intensity of windstorms, such as hurricanes, in the future. However, forest managers may have difficulty preparing for this future because there is limited understanding of how windstorms affect the structure and composition of forests over the long term. In this study, we evaluated the impacts of Hurricane Ivan, which made landfall in September 2004 near Gulf Shore, Alabama, impacting forests in the western Florida Panhandle and southwestern Alabama. We acquired the United States Department of Agriculture Forest Inventory and Analysis (FIA) plot data available for the period from 2002 to 2018 for the Ivan-affected area and classified the plots into 4 categories: (1). ND (No Disturbance), (2). NDBH (No Disturbance but Harvested), (3). ID (Disturbance caused by Hurricane Ivan), and (4). IDAH (Disturbance caused by Hurricane Ivan and Harvested). The plots that were damaged by Hurricane Ivan (ID and IDAH plots) had significantly (α = 0.05) (1) higher basal area, (2) higher quadratic mean diameter and height, (3) more diverse tree species composition (species richness and Shannon diversity index), (4) denser stocking of seedling and saplings, (5) lower proportion of dead trees or saplings, and (6) higher live aboveground biomass than the plots that were not damaged by the hurricane (ND and NDBH plots). Diverse stands were not necessarily more windstorm resistant. Species diversity in the overstory may not improve forest resistance to hurricane damage but may improve its resilience following the hurricane. The study suggests that managing stand structure through density management and stand improvement could be critical to windstorm resilience and resistance in the southern US forests.

Field studies of pollutant removal from nursery and greenhouse runoff by constructed wetlands.

Plant nursery runoff commonly contains pesticides and nutrients that often threaten aquatic ecosystems. Constructed wetlands could be a tool to remove pesticides and nutrients from nursery runoff but have not been extensively studied in this setting. Two field-scale constructed wetlands (one subsurface-flow constructed wetland [SFCW] and one free-surface constructed wetland [FSCW]) were implemented and monitored for water quality improvement. The SFCW demonstrated significant mass reduction of 78% or greater for nitrate, orthophosphate, total nitrogen, total phosphorus, and total suspended solids. The SFCW also demonstrated significant mass reduction of 79% or greater for 10 of the 12 pesticide compounds detected in over half of the collected samples. The FSCW demonstrated significant mass reduction of 46% or greater for all nonpesticide analytes except total nitrogen. Loading rate and actual storage volume compared with inflow volume likely affected performance. Reduced size and increased loading rate of the FSCW likely reduced its ability to effectively reduce pesticides. Results from this study indicate that constructed wetlands are likely an effective tool for nursery runoff management. When designing and implementing constructed wetlands, it is important for practitioners to consider the tradeoff between system size (additional cost and land otherwise dedicated to production) and performance.

Urban stormwater characterization, control, and treatment.

This review summarizes over 280 studies published in 2019 related to the characterization, control, and management of urban stormwater runoff. A summary of quantity and quality concerns is provided in the first section of the review, serving as the foundation for the following sections which focus on the control and treatment of stormwater runoff. Finally, the impact of stormwater control devices at the watershed scale is discussed. Each section provides a self-contained overview of the 2019 literature, common themes, and future work. Several themes emerged from the 2019 literature including exploration of substrate amendments for improved water quality effluent from stormwater controls, the continued study of the role of vegetation in green infrastructure practices, and a call to action for the development of new models which generate reliable, computationally efficient results under the physical, chemical, biological, and social complexity of stormwater management. PRACTITIONER POINTS: Over 280 studies were published in 2019 related to the characterization, control, and treatment of urban stormwater. Studies on bioretention and general stormwater characteristics represented the two most common subtopics in 2019. Trends in 2019 included novel substrate amendments, studies on the role of vegetation, and advancements in computational models.

Accounting for two-billion tons of stabilized soil carbon.

The pedosphere is the largest terrestrial reservoir of organic carbon, yet soil-carbon variability and its representation in Earth system models is a large source of uncertainty for carbon-cycle science and climate projections. Much of this uncertainty is attributed to local and regional-scale variability, and predicting this variation can be challenging if variable selection is based solely on a priori assumptions due to the scale-dependent nature of environmental determinants. Data mining can optimize predictive modeling by allowing machine-learning algorithms to learn from and discover complex patterns in large datasets that may have otherwise gone unnoticed, thus increasing the potential for knowledge discovery. In this analysis, we identify important, regional-scale determinants for top- and subsoil-carbon stabilization in production forestland across the southeastern US. Specifically, we apply recursive feature elimination to a large suite of socio-environmental data to strategically select a parsimonious, yet highly predictive covariate set. This is achieved by recursively considering smaller and smaller covariate sets-or features-by first training the estimator on the full set to obtain feature importance. The least important features are pruned, and the procedure is recursively repeated until a desired number of covariates is identified. We show that although carbon ranges from 0.3 to 8.2 kg m-2 in the topsoil (0 to 20 cm), and from 0.4 to 17.6 kg m-2 in the subsoil (20 to 100 cm), this variability is predictably distributed with precipitation, soil moisture, nitrogen and sand content, gamma ray emissions, mean annual minimum temperature, and elevation. From our spatial predictions, we estimate that 2.6 Pg of soil carbon is currently stabilized in the upper 100 cm of production forestland, which covers 34.7 million ha in the southeastern US.