Observed and projected functional reorganization of riverine fish assemblages from global change.

Climate and land-use/land-cover change ("global change") are restructuring biodiversity, globally. Broadly, environmental conditions are expected to become warmer, potentially drier (particularly in arid regions), and more anthropogenically developed in the future, with spatiotemporally complex effects on ecological communities. We used functional traits to inform Chesapeake Bay Watershed fish responses to future climate and land-use scenarios (2030, 2060, and 2090). We modeled the future habitat suitability of focal species representative of key trait axes (substrate, flow, temperature, reproduction, and trophic) and used functional and phylogenetic metrics to assess variable assemblage responses across physiographic regions and habitat sizes (headwaters through large rivers). Our focal species analysis projected future habitat suitability gains for carnivorous species with preferences for warm water, pool habitats, and fine or vegetated substrates. At the assemblage level, models projected decreasing habitat suitability for cold-water, rheophilic, and lithophilic individuals but increasing suitability for carnivores in the future across all regions. Projected responses of functional and phylogenetic diversity and redundancy differed among regions. Lowland regions were projected to become less functionally and phylogenetically diverse and more redundant while upland regions (and smaller habitat sizes) were projected to become more diverse and less redundant. Next, we assessed how these model-projected assemblage changes 2005-2030 related to observed time-series trends (1999-2016). Halfway through the initial projecting period (2005-2030), we found observed trends broadly followed modeled patterns of increasing proportions of carnivorous and lithophilic individuals in lowland regions but showed opposing patterns for functional and phylogenetic metrics. Leveraging observed and predicted analyses simultaneously helps elucidate the instances and causes of discrepancies between model predictions and ongoing observed changes. Collectively, results highlight the complexity of global change impacts across broad landscapes that likely relate to differences in assemblages' intrinsic sensitivities and external exposure to stressors.

Explainable machine learning improves interpretability in the predictive modeling of biological stream conditions in the Chesapeake Bay Watershed, USA.

Anthropogenic alterations have resulted in widespread degradation of stream conditions. To aid in stream restoration and management, baseline estimates of conditions and improved explanation of factors driving their degradation are needed. We used random forests to model biological conditions using a benthic macroinvertebrate index of biotic integrity for small, non-tidal streams (upstream area ≤200 km2) in the Chesapeake Bay watershed (CBW) of the mid-Atlantic coast of North America. We utilized several global and local model interpretation tools to improve average and site-specific model inferences, respectively. The model was used to predict condition for 95,867 individual catchments for eight periods (2001, 2004, 2006, 2008, 2011, 2013, 2016, 2019). Predicted conditions were classified as Poor, FairGood, or Uncertain to align with management needs and individual reach lengths and catchment areas were summed by condition class for the CBW for each period. Global permutation and local Shapley importance values indicated percent of forest, development, and agriculture in upstream catchments had strong impacts on predictions. Development and agriculture negatively influenced stream condition for model average (partial dependence [PD] and accumulated local effect [ALE] plots) and local (individual condition expectation and Shapley value plots) levels. Friedman's H-statistic indicated large overall interactions for these three land covers, and bivariate global plots (PD and ALE) supported interactions among agriculture and development. Total stream length and catchment area predicted in FairGood conditions decreased then increased over the 19-years (length/area: 66.6/65.4% in 2001, 66.3/65.2% in 2011, and 66.6/65.4% in 2019). Examination of individual catchment predictions between 2001 and 2019 showed those predicted to have the largest decreases in condition had large increases in development; whereas catchments predicted to exhibit the largest increases in condition showed moderate increases in forest cover. Use of global and local interpretative methods together with watershed-wide and individual catchment predictions support conservation practitioners that need to identify widespread and localized patterns, especially acknowledging that management actions typically take place at individual-reach scales.

Time marches on, but do the causal pathways driving instream habitat and biology remain consistent?

Stream ecosystems are complex networks of interacting terrestrial and aquatic drivers. To untangle these ecological networks, efforts evaluating the direct and indirect effects of landscape, climate, and instream predictors on biological condition through time are needed. We used structural equation modeling and leveraged a stream survey program to identify and compare important predictors driving condition of benthic macroinvertebrate and fish assemblages. We used data resampled 14 years apart at 252 locations across Maryland, USA. Sample locations covered a wide range of conditions that varied spatiotemporally. Overall, the relationship directions were consistent between sample periods, but their relative strength varied temporally. For benthic macroinvertebrates, we found that the total effect of natural landscape (e.g., elevation, longitude, latitude, geology) and land use (i.e., forest, development, agriculture) predictors was 1.4 and 1.5 times greater in the late 2010s compared to the 2000s. Moreover, the total effect of water quality (e.g., total nitrogen and conductivity) and habitat (e.g., embeddedness, riffle quality) was 1.2 and 4.8 times lower in the 2010s, respectively. For fish assemblage condition, the total effect of land use-land cover predictors was 2.3 times greater in the 2010s compared to the 2000s, while the total effect of local habitat was 1.4 times lower in the 2010s, respectively. As expected, we found biological assemblages in catchments with more agriculture and urban development were generally comprised of tolerant, generalist species, while assemblages in catchments with greater forest cover had more-specialized, less-tolerant species (e.g., Ephemeroptera, Plecoptera, and Trichoptera taxa, clingers, benthic and lithophilic spawning fishes). Changes in the relative importance of landscape and land-use predictors suggest other correlated, yet unmeasured, proximal factors became more important over time. By untangling these ecological networks, stakeholders can gain a better understanding of the spatiotemporal relationships driving biological condition to implement management practices aimed at improving stream condition.

Using expert knowledge to support Endangered Species Act decision-making for data-deficient species.

Many questions relevant to conservation decision-making are characterized by extreme uncertainty due to lack of empirical data and complexity of the underlying ecologic processes, leading to a rapid increase in the use of structured protocols to elicit expert knowledge. Published ecologic applications often employ a modified Delphi method, where experts provide judgments anonymously and mathematical aggregation techniques are used to combine judgments. The Sheffield elicitation framework (SHELF) differs in its behavioral approach to synthesizing individual judgments into a fully specified probability distribution for an unknown quantity. We used the SHELF protocol remotely to assess extinction risk of three subterranean aquatic species that are being considered for listing under the U.S. Endangered Species Act. We provided experts an empirical threat assessment for each known locality over a video conference and recorded judgments on the probability of population persistence over four generations with online submission forms and R-shiny apps available through the SHELF package. Despite large uncertainty for all populations, there were key differences between species' risk of extirpation based on spatial variation in dominant threats, local land use and management practices, and species' microhabitat. The resulting probability distributions provided decision makers with a full picture of uncertainty that was consistent with the probabilistic nature of risk assessments. Discussion among experts during SHELF's behavioral aggregation stage clearly documented dominant threats (e.g., development, timber harvest, animal agriculture, and cave visitation) and their interactions with local cave geology and species' habitat. Our virtual implementation of the SHELF protocol demonstrated the flexibility of the approach for conservation applications operating on budgets and time lines that can limit in-person meetings of geographically dispersed experts.

Uso del Conocimiento Experto para Respaldar la Toma de Decisiones del Acta de Especies en Peligro para Especies con Información Deficiente Resumen Muchas preguntas relevantes para la toma de decisiones de conservación se caracterizan por una incertidumbre extrema causada por la falta de información empírica y por la complejidad de los procesos ecológicos subyacentes. Esto lleva a un rápido incremento en el uso de protocolos estructurados para obtener conocimiento de los expertos en el tema. Las aplicaciones ecológicas publicadas con frecuencia emplean un método Delphi modificado, en el cual los expertos proporcionan dictámenes anónimamente y luego se usan técnicas de agregación matemática para combinar estos dictámenes. El marco de trabajo de obtención Sheffield (SHELF) difiere en su enfoque conductual para sintetizar los dictámenes individuales en una distribución de probabilidad completamente especificada para una cantidad desconocida. Usamos el protocolo SHELF remotamente para evaluar el riesgo de extinción de tres especies acuáticas subterráneas que están siendo consideradas para ser incluidas en el Acta de Especies en Peligro de los E.U.A. Les proporcionamos a los expertos una evaluación empírica de la amenaza para cada localidad conocida durante una videoconferencia y registramos los dictámenes sobre la probabilidad de la persistencia poblacional durante cuatro generaciones por medio de formularios enviados en línea y las apps R-shiny disponibles a través del paquete SHELF. A pesar de la gran incertidumbre para todas las poblaciones, hubo diferencias importantes entre el riesgo de extirpación de las especies con base en la variación espacial en las amenazas dominantes, el uso del suelo local y las prácticas de manejo, y el microhábitat de las especies. Las distribuciones resultantes de la probabilidad proporcionaron al órgano decisorio un cuadro completo de la incertidumbre que fue consistente con la naturaleza probabilística de las evaluaciones de riesgo. Las discusiones entre los expertos durante la fase de agregación conductual de SHELF documentaron claramente las amenazas dominantes (p. ej.: desarrollo, extracción de madera, agricultura animal y visitas a las cuevas) y sus interacciones con la geología de las cuevas locales y el hábitat de la especie. Nuestra implementación virtual del protocolo SHELF demostró la flexibilidad del enfoque para las aplicaciones de la conservación que operan con presupuestos y líneas de tiempo que pueden limitar las reuniones en persona de expertos dispersados geográficamente.

Techniques to improve ecological interpretability of black-box machine learning models.

Statistical modeling of ecological data is often faced with a large number of variables as well as possible nonlinear relationships and higher-order interaction effects. Gradient boosted trees (GBT) have been successful in addressing these issues and have shown a good predictive performance in modeling nonlinear relationships, in particular in classification settings with a categorical response variable. They also tend to be robust against outliers. However, their black-box nature makes it difficult to interpret these models. We introduce several recently developed statistical tools to the environmental research community in order to advance interpretation of these black-box models. To analyze the properties of the tools, we applied gradient boosted trees to investigate biological health of streams within the contiguous U.S., as measured by a benthic macroinvertebrate biotic index. Based on these data and a simulation study, we demonstrate the advantages and limitations of partial dependence plots (PDP), individual conditional expectation (ICE) curves and accumulated local effects (ALE) in their ability to identify covariate-response relationships. Additionally interaction effects were quantified according to interaction strength (IAS) and Friedman's H2 statistic. Interpretable machine learning techniques are useful tools to open the black-box of gradient boosted trees in the environmental sciences. This finding is supported by our case study on the effect of impervious surface on the benthic condition, which agrees with previous results in the literature. Overall the most important variables were ecoregion, bed stability, watershed area, riparian vegetation and catchment slope. These variables were also present in most identified interaction effects. In conclusion, graphical tools (PDP, ICE, ALE) enable visualization and easier interpretation of GBT but should be supported by analytical statistical measures. Future methodological research is needed to investigate the properties of interaction tests.

Shale gas development has limited effects on stream biology and geochemistry in a gradient-based, multiparameter study in Pennsylvania.

The number of horizontally drilled shale oil and gas wells in the United States has increased from nearly 28,000 in 2007 to nearly 127,000 in 2017, and research has suggested the potential for the development of shale resources to affect nearby stream ecosystems. However, the ability to generalize current studies is limited by the small geographic scope as well as limited breadth and integration of measured chemical and biological indicators parameters. This study tested the hypothesis that a quantifiable, significant relationship exists between the density of oil and gas (OG) development, increasing stream water concentrations of known geochemical tracers of OG extraction, and the composition of benthic macroinvertebrate and microbial communities. Twenty-five headwater streams that drain lands across a gradient of shale gas development intensity were sampled. Our strategy included comprehensive measurements across multiple seasons of sampling to account for temporal variability of geochemical parameters, including known shale OG geochemical tracers, and microbial and benthic macroinvertebrate communities. No significant relationships were found between the intensity of OG development, shale OG geochemical tracers, or benthic macroinvertebrate or microbial community composition, whereas significant seasonal differences in stream chemistry were observed. These results highlight the importance of considering spatial and temporal variability in stream chemistry and biota and not only the presence of anthropogenic activities in a watershed. This comprehensive, integrated study of geochemical and biological variability of headwater streams in watersheds undergoing OG development provides a robust framework for examining the effects of energy development at a regional scale.

Disentangling the potential effects of land-use and climate change on stream conditions.

Land-use and climate change are significantly affecting stream ecosystems, yet understanding of their long-term impacts is hindered by the few studies that have simultaneously investigated their interaction and high variability among future projections. We modeled possible effects of a suite of 2030, 2060, and 2090 land-use and climate scenarios on the condition of 70,772 small streams in the Chesapeake Bay watershed, United States. The Chesapeake Basin-wide Index of Biotic Integrity, a benthic macroinvertebrate multimetric index, was used to represent stream condition. Land-use scenarios included four Special Report on Emissions Scenarios (A1B, A2, B1, and B2) representing a range of potential landscape futures. Future climate scenarios included quartiles of future climate changes from downscaled Coupled Model Intercomparison Project - Phase 5 (CMIP5) and a watershed-wide uniform scenario (Lynch2016). We employed random forests analysis to model individual and combined effects of land-use and climate change on stream conditions. Individual scenarios suggest that by 2090, watershed-wide conditions may exhibit anywhere from large degradations (e.g., scenarios A1B, A2, and the CMIP5 25th percentile) to small degradations (e.g., scenarios B1, B2, and Lynch2016). Combined land-use and climate change scenarios highlighted their interaction and predicted, by 2090, watershed-wide degradation in 16.2% (A2 CMIP5 25th percentile) to 1.0% (B2 Lynch2016) of stream kilometers. A goal for the Chesapeake Bay watershed is to restore 10% of stream kilometers over a 2008 baseline; our results suggest meeting and sustaining this goal until 2090 may require improvement in 11.0%-26.2% of stream kilometers, dependent on land-use and climate scenario. These results highlight inherent variability among scenarios and the resultant uncertainty of predicted conditions, which reinforces the need to incorporate multiple scenarios of both land-use (e.g., development, agriculture, etc.) and climate change in future studies to encapsulate the range of potential future conditions.

A Random Forest Approach for Bounded Outcome Variables.

Random forests have become an established tool for classification and regression, in particular in high-dimensional settings and in the presence of non-additive predictor-response relationships. For bounded outcome variables restricted to the unit interval, however, classical modeling approaches based on mean squared error loss may severely suffer as they do not account for heteroscedasticity in the data. To address this issue, we propose a random forest approach for relating a beta dis-tributed outcome to a set of explanatory variables. Our approach explicitly makes use of the likelihood function of the beta distribution for the selection of splits dur-ing the tree-building procedure. In each iteration of the tree-building algorithm it chooses one explanatory variable in combination with a split point that maximizes the log-likelihood function of the beta distribution with the parameter estimates de-rived from the nodes of the currently built tree. Results of several simulation studies and an application using data from the U.S.A. National Lakes Assessment Survey demonstrate the properties and usefulness of the method, in particular when compared to random forest approaches based on mean squared error loss and parametric regression models.

Canopy volume removal from oil and gas development activity in the upper Susquehanna River basin in Pennsylvania and New York (USA): An assessment using lidar data.

Oil and gas development is changing the landscape in many regions of the United States and globally. However, the nature, extent, and magnitude of landscape change and development, and precisely how this development compares to other ongoing land conversion (e.g. urban/sub-urban development, timber harvest) is not well understood. In this study, we examine land conversion from oil and gas infrastructure development in the upper Susquehanna River basin in Pennsylvania and New York, an area that has experienced much oil and gas development over the past 10 years. We quantified land conversion in terms of forest canopy geometric volume loss in contrast to previous studies that considered only areal impacts. For the first time in a study of this type, we use fine-scale lidar forest canopy geometric models to assess the volumetric change due to forest clearing from oil and gas development and contrast this land change to clear cut forest harvesting, and urban and suburban development. Results show that oil and gas infrastructure development removed a large volume of forest canopy from 2006 to 2013, and this removal spread over a large portion of the study area. Timber operations (clear cutting) on Pennsylvania State Forest lands removed a larger total volume of forest canopy during the same time period, but this canopy removal was concentrated in a smaller area. Results of our study point to the need to consider volumetric impacts of oil and gas development on ecosystems, and to place potential impacts in context with other ongoing land conversions.

Brook trout distributional response to unconventional oil and gas development: Landscape context matters.

We conducted a large-scale assessment of unconventional oil and gas (UOG) development effects on brook trout (Salvelinus fontinalis) distribution. We compiled 2231 brook trout collection records from the Upper Susquehanna River Watershed, USA. We used boosted regression tree (BRT) analysis to predict occurrence probability at the 1:24,000 stream-segment scale as a function of natural and anthropogenic landscape and climatic attributes. We then evaluated the importance of landscape context (i.e., pre-existing natural habitat quality and anthropogenic degradation) in modulating the effects of UOG on brook trout distribution under UOG development scenarios. BRT made use of 5 anthropogenic (28% relative influence) and 7 natural (72% relative influence) variables to model occurrence with a high degree of accuracy [Area Under the Receiver Operating Curve (AUC)=0.85 and cross-validated AUC=0.81]. UOG development impacted 11% (n=2784) of streams and resulted in a loss of predicted occurrence in 126 (4%). Most streams impacted by UOG had unsuitable underlying natural habitat quality (n=1220; 44%). Brook trout were predicted to be absent from an additional 26% (n=733) of streams due to pre-existing non-UOG land uses (i.e., agriculture, residential and commercial development, or historic mining). Streams with a predicted and observed (via existing pre- and post-disturbance fish sampling records) loss of occurrence due to UOG tended to have intermediate natural habitat quality and/or intermediate levels of non-UOG stress. Simulated development of permitted but undeveloped UOG wells (n=943) resulted in a loss of predicted occurrence in 27 additional streams. Loss of occurrence was strongly dependent upon landscape context, suggesting effects of current and future UOG development are likely most relevant in streams near the probability threshold due to pre-existing habitat degradation.

A detailed risk assessment of shale gas development on headwater streams in the Pennsylvania portion of the Upper Susquehanna River Basin, U.S.A.

The development of unconventional oil and gas (UOG) involves infrastructure development (well pads, roads and pipelines), well drilling and stimulation (hydraulic fracturing), and production; all of which have the potential to affect stream ecosystems. Here, we developed a fine-scaled (1:24,000) catchment-level disturbance intensity index (DII) that included 17 measures of UOG capturing all steps in the development process (infrastructure, water withdrawals, probabilistic spills) that could affect headwater streams (<200km2 in upstream catchment) in the Upper Susquehanna River Basin in Pennsylvania, U.S.A. The DII ranged from 0 (no UOG disturbance) to 100 (the catchment with the highest UOG disturbance in the study area) and it was most sensitive to removal of pipeline cover, road cover and well pad cover metrics. We related this DII to three measures of high quality streams: Pennsylvania State Exceptional Value (EV) streams, Class A brook trout streams and Eastern Brook Trout Joint Venture brook trout patches. Overall only 3.8% of all catchments and 2.7% of EV stream length, 1.9% of Class A streams and 1.2% of patches were classified as having medium to high level DII scores (>50). Well density, often used as a proxy for development, only correlated strongly with well pad coverage and produced materials, and therefore may miss potential effects associated with roads and pipelines, water withdrawals and spills. When analyzed with a future development scenario, 91.1% of EV stream length, 68.7% of Class A streams and 80.0% of patches were in catchments with a moderate to high probability of development. Our method incorporated the cumulative effects of UOG on streams and can be used to identify catchments and reaches at risk to existing stressors or future development.

Unconventional Oil and Gas Spills: Risks, Mitigation Priorities, and State Reporting Requirements.

Rapid growth in unconventional oil and gas (UOG) has produced jobs, revenue, and energy, but also concerns over spills and environmental risks. We assessed spill data from 2005 to 2014 at 31 481 UOG wells in Colorado, New Mexico, North Dakota, and Pennsylvania. We found 2-16% of wells reported a spill each year. Median spill volumes ranged from 0.5 m3 in Pennsylvania to 4.9 m3 in New Mexico; the largest spills exceeded 100 m3. Seventy-five to 94% of spills occurred within the first three years of well life when wells were drilled, completed, and had their largest production volumes. Across all four states, 50% of spills were related to storage and moving fluids via flowlines. Reporting rates varied by state, affecting spill rates and requiring extensive time and effort getting data into a usable format. Enhanced and standardized regulatory requirements for reporting spills could improve the accuracy and speed of analyses to identify and prevent spill risks and mitigate potential environmental damage. Transparency for data sharing and analysis will be increasingly important as UOG development expands. We designed an interactive spills data visualization tool ( http://snappartnership.net/groups/hydraulic-fracturing/webapp/spills.html ) to illustrate the value of having standardized, public data.

Unconventional oil and gas spills: Materials, volumes, and risks to surface waters in four states of the U.S.

Extraction of oil and gas from unconventional sources, such as shale, has dramatically increased over the past ten years, raising the potential for spills or releases of chemicals, waste materials, and oil and gas. We analyzed spill data associated with unconventional wells from Colorado, New Mexico, North Dakota and Pennsylvania from 2005 to 2014, where we defined unconventional wells as horizontally drilled into an unconventional formation. We identified materials spilled by state and for each material we summarized frequency, volumes and spill rates. We evaluated the environmental risk of spills by calculating distance to the nearest stream and compared these distances to existing setback regulations. Finally, we summarized relative importance to drinking water in watersheds where spills occurred. Across all four states, we identified 21,300 unconventional wells and 6622 reported spills. The number of horizontal well bores increased sharply beginning in the late 2000s; spill rates also increased for all states except PA where the rate initially increased, reached a maximum in 2009 and then decreased. Wastewater, crude oil, drilling waste, and hydraulic fracturing fluid were the materials most often spilled; spilled volumes of these materials largely ranged from 100 to 10,000L. Across all states, the average distance of spills to a stream was highest in New Mexico (1379m), followed by Colorado (747m), North Dakota (598m) and then Pennsylvania (268m), and 7.0, 13.3, and 20.4% of spills occurred within existing surface water setback regulations of 30.5, 61.0, and 91.4m, respectively. Pennsylvania spills occurred in watersheds with a higher relative importance to drinking water than the other three states. Results from this study can inform risk assessments by providing improved input parameters on volume and rates of materials spilled, and guide regulations and the management policy of spills.

Transport of hydraulic fracturing waste from Pennsylvania wells: A county-level analysis of road use and associated road repair costs.

Pennsylvania's rapid unconventional oil and gas (UOG) development-from a single well in 2004 to more than 6700 wells in 2013-has dramatically increased UOG waste transport by heavy trucks. This study quantified the amount of UOG waste and the distance it traveled between wells and disposal facilities on each type of road in each county between July 2010 and December 2013. In addition, the study estimated the associated financial costs to each county's road infrastructure over that period. We found that UOG wells produced a median wastewater volume of 1294 m(3) and a median of 89,267 kg of solid waste. The median number of waste-transport truck trips per well was 122. UOG wells existed in 38 Pennsylvania counties, but we estimated trucks transporting well waste traveled through 132 counties, including counties in West Virginia, Ohio, and New York. Median travel distance varied by disposal type, from 106 km to centralized treatment facilities up to 237 km to injection wells. Local roads experienced the greatest amount of truck traffic and associated costs ($1.1-6.5 M) and interstates, the least ($0.3-1.6 M). Counties with oil and gas development experienced the most truck traffic and incurred the highest associated roadway costs. However, many counties outside the active development area also incurred roadway repair costs, highlighting the extension of UOG development's spatial footprint beyond the active development area. An online data visualization tool is available here: www.nicholasinstitute.duke.edu/transportation-of-hydraulic-fracturing-waste.

Stream Vulnerability to Widespread and Emergent Stressors: A Focus on Unconventional Oil and Gas.

Multiple stressors threaten stream physical and biological quality, including elevated nutrients and other contaminants, riparian and in-stream habitat degradation and altered natural flow regime. Unconventional oil and gas (UOG) development is one emerging stressor that spans the U.S. UOG development could alter stream sedimentation, riparian extent and composition, in-stream flow, and water quality. We developed indices to describe the watershed sensitivity and exposure to natural and anthropogenic disturbances and computed a vulnerability index from these two scores across stream catchments in six productive shale plays. We predicted that catchment vulnerability scores would vary across plays due to climatic, geologic and anthropogenic differences. Across-shale averages supported this prediction revealing differences in catchment sensitivity, exposure, and vulnerability scores that resulted from different natural and anthropogenic environmental conditions. For example, semi-arid Western shale play catchments (Mowry, Hilliard, and Bakken) tended to be more sensitive to stressors due to low annual average precipitation and extensive grassland. Catchments in the Barnett and Marcellus-Utica were naturally sensitive from more erosive soils and steeper catchment slopes, but these catchments also experienced areas with greater UOG densities and urbanization. Our analysis suggested Fayetteville and Barnett catchments were vulnerable due to existing anthropogenic exposure. However, all shale plays had catchments that spanned a wide vulnerability gradient. Our results identify vulnerable catchments that can help prioritize stream protection and monitoring efforts. Resource managers can also use these findings to guide local development activities to help reduce possible environmental effects.

Ecological risks of shale oil and gas development to wildlife, aquatic resources and their habitats.

Technological advances in hydraulic fracturing and horizontal drilling have led to the exploration and exploitation of shale oil and gas both nationally and internationally. Extensive development of shale resources has occurred within the United States over the past decade, yet full build out is not expected to occur for years. Moreover, countries across the globe have large shale resources and are beginning to explore extraction of these resources. Extraction of shale resources is a multistep process that includes site identification, well pad and infrastructure development, well drilling, high-volume hydraulic fracturing and production; each with its own propensity to affect associated ecosystems. Some potential effects, for example from well pad, road and pipeline development, will likely be similar to other anthropogenic activities like conventional gas drilling, land clearing, exurban and agricultural development and surface mining (e.g., habitat fragmentation and sedimentation). Therefore, we can use the large body of literature available on the ecological effects of these activities to estimate potential effects from shale development on nearby ecosystems. However, other effects, such as accidental release of wastewaters, are novel to the shale gas extraction process making it harder to predict potential outcomes. Here, we review current knowledge of the effects of high-volume hydraulic fracturing coupled with horizontal drilling on terrestrial and aquatic ecosystems in the contiguous United States, an area that includes 20 shale plays many of which have experienced extensive development over the past decade. We conclude that species and habitats most at risk are ones where there is an extensive overlap between a species range or habitat type and one of the shale plays (leading to high vulnerability) coupled with intrinsic characteristics such as limited range, small population size, specialized habitat requirements, and high sensitivity to disturbance. Examples include core forest habitat and forest specialists, sagebrush habitat and specialists, vernal pond inhabitants and stream biota. We suggest five general areas of research and monitoring that could aid in development of effective guidelines and policies to minimize negative impacts and protect vulnerable species and ecosystems: (1) spatial analyses, (2) species-based modeling, (3) vulnerability assessments, (4) ecoregional assessments, and (5) threshold and toxicity evaluations.

Boosted beta regression.

Regression analysis with a bounded outcome is a common problem in applied statistics. Typical examples include regression models for percentage outcomes and the analysis of ratings that are measured on a bounded scale. In this paper, we consider beta regression, which is a generalization of logit models to situations where the response is continuous on the interval (0,1). Consequently, beta regression is a convenient tool for analyzing percentage responses. The classical approach to fit a beta regression model is to use maximum likelihood estimation with subsequent AIC-based variable selection. As an alternative to this established - yet unstable - approach, we propose a new estimation technique called boosted beta regression. With boosted beta regression estimation and variable selection can be carried out simultaneously in a highly efficient way. Additionally, both the mean and the variance of a percentage response can be modeled using flexible nonlinear covariate effects. As a consequence, the new method accounts for common problems such as overdispersion and non-binomial variance structures.

Upland disturbance affects headwater stream nutrients and suspended sediments during baseflow and stormflow.

Because catchment characteristics determine sediment and nutrient inputs to streams, upland disturbance can affect stream chemistry. Catchments at the Fort Benning Military Installation (near Columbus, Georgia) experience a range of upland disturbance intensities due to spatial variability in the intensity of military training. We used this disturbance gradient to investigate the effects of upland soil and vegetation disturbance on stream chemistry. During baseflow, mean total suspended sediment (TSS) concentration and mean inorganic suspended sediment (ISS) concentration increased with catchment disturbance intensity (TSS: R2= 0.7, p = 0.005, range = 4.0-10.1 mg L(-1); ISS: R2= 0.71, p = 0.004, range = 2.04-7.3 mg L(-1)); dissolved organic carbon (DOC) concentration (R2= 0.79, p = 0.001, range = 1.5-4.1 mg L(-1)) and soluble reactive phosphorus (SRP) concentration (R2= 0.75, p = 0.008, range = 1.9-6.2 microg L(-1)) decreased with increasing disturbance intensity; and ammonia (NH4+), nitrate (NO3-), and dissolved inorganic nitrogen (DIN) concentrations were unrelated to disturbance intensity. The increase in TSS and ISS during storms was positively correlated with disturbance (R2= 0.78 and 0.78, p = 0.01 and 0.01, respectively); mean maximum change in SRP during storms increased with disturbance (r = 0.7, p = 0.04); and mean maximum change in NO3- during storms was marginally correlated with disturbance (r = 0.58, p = 0.06). Soil characteristics were significant predictors of baseflow DOC, SRP, and Ca2+, but were not correlated with suspended sediment fractions, any nitrogen species, or pH. Despite the largely intact riparian zones of these headwater streams, upland soil and vegetation disturbances had clear effects on stream chemistry during baseflow and stormflow conditions.

Influence of catchment-scale military land use on stream physical and organic matter variables in small southeastern plains catchments (USA).

We conducted a 3-year study designed to examine the relationship between disturbance from military land use and stream physical and organic matter variables within 12 small (<5.5 km2) Southeastern Plains catchments at the Fort Benning Military Installation, Georgia, USA. Primary land-use categories were based on percentages of bare ground and road cover and nonforested land (grasslands, sparse vegetation, shrublands, fields) in catchments and natural catchments features, including soils (% sandy soils) and catchment size (area). We quantified stream flashiness (determined by slope of recession limbs of storm hydrographs), streambed instability (measured by relative changes in bed height over time), organic matter storage [coarse wood debris (CWD) relative abundance, benthic particulate organic matter (BPOM)] and stream-water dissolved organic carbon concentration (DOC). Stream flashiness was positively correlated with average storm magnitude and percent of the catchment with sandy soil, whereas streambed instability was related to percent of the catchment containing nonforested (disturbed) land. The proportions of in-stream CWD and sediment BPOM, and stream-water DOC were negatively related to the percent of bare ground and road cover in catchments. Collectively, our results suggest that the amount of catchment disturbance causing denuded vegetation and exposed, mobile soil is (1) a key terrestrial influence on stream geomorphology and hydrology and (2) a greater determinant of in-stream organic matter conditions than is natural geomorphic or topographic variation (catchment size, soil type) in these systems.