Characterizing Spatiotemporal Variability in Phosphorus Export across the United States through Bayesian Hierarchical Modeling.

Phosphorus inputs from anthropogenic activities are subject to hydrologic (riverine) export, causing water quality problems in downstream lakes and coastal systems. Nutrient budgets have been developed to quantify the amount of nutrients imported to and exported from various watersheds. However, at large spatial scales, estimates of hydrologic phosphorus export are usually unavailable. This study develops a Bayesian hierarchical model to estimate annual phosphorus export across the contiguous United States, considering agricultural inputs, urban inputs, and geogenic sources under varying precipitation conditions. The Bayesian framework allows for a systematic updating of prior information on export rates using an extensive calibration data set of riverine loadings. Furthermore, the hierarchical approach allows for spatial variation in export rates across major watersheds and ecoregions. Applying the model, we map hotspots of phosphorus loss across the United States and characterize the primary factors driving these losses. Results emphasize the importance of precipitation in determining hydrologic export rates for various anthropogenic inputs, especially agriculture. Our findings also emphasize the importance of phosphorus from geogenic sources in overall river export.

Estimating the benefits of stream water quality improvements in urbanizing watersheds: An ecological production function approach.

Streams in urbanizing watersheds are threatened by economic development that can lead to excessive sediment erosion and surface runoff. These anthropogenic stressors diminish valuable ecosystem services and result in pervasive degradation commonly referred to as "urban stream syndrome." Understanding how the public perceives and values improvements in stream conditions is necessary to support efforts to quantify the economic benefits of water quality improvements. We develop an ecological production framework that translates measurable indicators of stream water quality into ecological endpoints. Our interdisciplinary approach integrates a predictive hierarchical water quality model that is well suited for sparse data environments, an expert elicitation that translates measurable water quality indicators into ecological endpoints that focus group participants identified as most relevant, and a stated preference survey that elicits the public's willingness to pay for changes in these endpoints. To illustrate our methods, we develop an application to the Upper Neuse River Watershed located in the rapidly developing Triangle region of North Carolina (the United States). Our results suggest, for example, that residents are willing to pay roughly $127 per household and $54 million per year in aggregate (2021 US$) for water quality improvements resulting from a stylized intervention that increases stream bank canopy cover by 25% and decreases runoff from impervious surfaces, leading to improvements in water quality and ecological endpoints for local streams. Although the three components of our analysis are conducted with data from North Carolina, we discuss how our findings are generalizable to urban and urbanizing areas across the larger Piedmont ecoregion of the Eastern United States.

Advancing freshwater ecological forecasts: Harmful algal blooms in Lake Erie.

Ecological models help provide forecasts of ecosystem responses to natural and anthropogenic stresses. However, their ability to create reliable predictions requires forecasts with track records sufficiently long to build confidence, skill assessments, and treating uncertainty quantitatively. We use Lake Erie harmful algal blooms as a case study to help formalize ecological forecasting. Key challenges for models include uncertainty in the deterministic structure of the load-bloom relationship and the need to assess alternative drivers (e.g., biologically available phosphorus load, spring load, longer term cumulative load) with a larger dataset. We enhanced a Bayesian model considering new information and an expanded data set, test it through cross validation and blind forecasts, quantify and discuss its uncertainties, and apply it for assessing historical and future scenarios. Allowing a segmented relationship between bloom size and spring load indicates that loading above 0.15 Gg/month will have a substantially higher marginal impact on bloom size. The new model explains 84 % of interannual variability (9.09 Gg RMSE) when calibrated to the 19-year data set and 66 % of variability in cross validation (12.58 Gg RMSE). Blind forecasts explain 84 % of HAB variability between 2014 and 2020, which is substantially better than the actual forecast track record (R2 = 0.32) over this same period. Because of internal phosphorus recycling, represented by the long-term cumulative load, it could take over a decade for HABs to fully respond to loading reductions, depending on the pace of those reductions. Thus, the desired speed and endpoint of the lake's recovery should be considered when updating and adaptively managing load reduction targets. Results are discussed in the context of ecological forecasting best pactices: incorporate new knowledge and data in model construction; account for multiple sources of uncertainty; evaluate predictive skill through validation and hindcasting; and answer management questions related to both short-term forecasts and long-term scenarios.

Elucidating controls on cyanobacteria bloom timing and intensity via Bayesian mechanistic modeling.

The adverse impacts of harmful algal blooms (HABs) are increasing worldwide. Lake Erie is a North American Great Lake highly affected by cultural eutrophication and summer cyanobacterial HABs. While phosphorus loading is a known driver of bloom size, more nuanced yet crucial questions remain. For example, it is unclear what mechanisms are primarily responsible for initiating cyanobacterial dominance and subsequent biomass accumulation. To address these questions, we develop a mechanistic model describing June-October dynamics of chlorophyll a, nitrogen, and phosphorus near the Maumee River outlet, where blooms typically initiate and are most severe. We calibrate the model to a new, geostatistically-derived dataset of daily water quality spanning 2008-2017. A Bayesian framework enables us to embed prior knowledge on system characteristics and test alternative model formulations. Overall, the best model formulation explains 42% of the variability in chlorophyll a and 83% of nitrogen, and better captures bloom timing than previous models. Our results, supported by cross validation, show that onset of the major midsummer bloom is associated with about a month of water temperatures above 20 °C (occurring 19 July to 6 August), consistent with when cyanobacteria dominance is usually reported. Decreased phytoplankton loss rate is the main factor enabling biomass accumulation, consistent with reduced zooplankton grazing on cyanobacteria. The model also shows that phosphorus limitation is most severe in August, and nitrogen limitation tends to occur in early autumn. Our results highlight the role of temperature in regulating bloom initiation and subsequent loss rates, and suggest that a 2 °C increase could lead to blooms that start about 10 days earlier and grow 23% more intense.

Exploring nutrient and light limitation of algal production in a shallow turbid reservoir.

Harmful algal blooms are increasingly recognized as a threat to the integrity of freshwater reservoirs, which serve as water supplies, wildlife habitats, and recreational attractions. While algal growth and accumulation is controlled by many environmental factors, the relative importance of these factors is unclear, particularly for turbid eutrophic systems. Here we develop and compare two models that test the relative importance of vertical mixing, light, and nutrients for explaining chlorophyll-a variability in shallow (2-3 m) embayments of a eutrophic reservoir, Jordan Lake, North Carolina. One is a multiple linear regression (statistical) model and the other is a process-based (mechanistic) model. Both models are calibrated using a 15-year data record of chlorophyll-a concentration (2003-2018) for the seasonal period of cyanobacteria dominance (June-October). The mechanistic model includes a novel representation of vertical mixing and is calibrated in a Bayesian framework, which allows for data-driven inference of important process rates. Both models show that chlorophyll-a concentration is much more responsive to nutrient variability than mixing, light, or temperature. While both models explain approximately 60% of the variability in chlorophyll-a, the mechanistic model is more robust in cross-validation and provides a more comprehensive assessment of algal drivers. Overall, these models indicate that nutrient reductions, rather than changes in mixing or background turbidity, are critical to controlling cyanobacteria in a shallow eutrophic freshwater system.

Quantifying uncertainty cascading from climate, watershed, and lake models in harmful algal bloom predictions.

In response to increased harmful algal blooms (HABs), hypoxia, and nearshore algae growth in Lake Erie, the United States and Canada agreed to phosphorus load reduction targets. While the load targets were guided by an ensemble of models, none of them considered the effects of climate change. Some watershed models developed to guide load reduction strategies have simulated climate effects, but without extending the resulting loads or their uncertainties to HAB projections. In this study, we integrated an ensemble of four climate models, three watershed models, and four HAB models. Nutrient loads and HAB predictions were generated for historical (1985-1999), current (2002-2017), and mid-21st-century (2051-2065) periods. For the current and historical periods, modeled loads and HABs are comparable to observations but exhibit less interannual variability. Our results show that climate impacts on watershed processes are likely to lead to reductions in future loading, assuming land use and watershed management practices are unchanged. This reduction in load should help reduce the magnitude of future HABs, although increases in lake temperature could mitigate that decrease. Using Monte-Carlo analysis to attribute sources of uncertainty from this cascade of models, we show that the uncertainty associated with each model is significant, and that improvements in all three are needed to build confidence in future projections.

Fusion-Based Hypoxia Estimates: Combining Geostatistical and Mechanistic Models of Dissolved Oxygen Variability.

The need to characterize and track coastal hypoxia has led to the development of geostatistical models based on in situ observations of dissolved oxygen (DO) and mechanistic models based on a representation of biophysical processes. To integrate the benefits of these two distinct modeling approaches, we develop a space-time geostatistical framework for synthesizing DO observations with hydrodynamic-biogeochemical model simulations and meteorological time series (as covariates). This fusion-based approach is used to estimate hypoxia in the northern Gulf of Mexico across summers from 1985 to 2017. Deterministic trends with dynamic covariates explain over 35% of the variability in DO. Moreover, cross-validation results indicate that 58% of DO variability is explained when combining these trends with spatiotemporal interpolation, which is substantially better than mechanistic or conventional geostatistical hypoxia modeling alone. The fusion-based approach also reduces hypoxic area uncertainties by 11% on average and up to 40% in months with sparse sampling. Moreover, our new estimates of mean summer hypoxic area changed by >10% in a majority of years, relative to previous geostatistical estimates. These fusion-based estimates can be a valuable resource when assessing the influence of hypoxia on the coastal ecosystem.

Bayesian mechanistic modeling characterizes Gulf of Mexico hypoxia: 1968-2016 and future scenarios.

The hypoxic zone in the northern Gulf of Mexico is among the most dramatic examples of impairments to aquatic ecosystems. Despite having attracted substantial attention, management of this environmental crisis remains challenging, partially due to limited monitoring to support model development and long-term assessments. Here, we leverage new geostatistical estimates of hypoxia derived from nearly 150 monitoring cruises and a process-based model to improve characterization of controlling mechanisms, historic trends, and future responses of hypoxia while rigorously quantifying uncertainty in a Bayesian framework. We find that November-March nitrogen loads are important controls of sediment oxygen demand, which appears to be the major oxygen sink. In comparison, only ~23% of oxygen in the near-bottom region appears to be consumed by net water column respiration, which is driven by spring and summer loads. Hypoxia typically exceeds 15,600 km2 in June, peaks in July, and declines below 10,000 km2 in September. In contrast to some previous Gulf hindcasting studies, our simulations demonstrate that hypoxia was both severe and worsening prior to 1985, and has remained relatively stable since that time. Scenario analysis shows that halving nutrient loadings will reduce hypoxia by 37% with respect to 13,900 km2 (1985-2016 median), while a +2°C change in water temperature will cause a 26% hypoxic area increase due to enhanced sediment respiration and reduced oxygen solubility. These new results highlight the challenges of achieving hypoxia reduction targets, particularly under warming conditions, and should be considered in ecosystem management.

A space-time geostatistical model for probabilistic estimation of harmful algal bloom biomass and areal extent.

Harmful algal blooms (HABs) have been increasing in intensity worldwide, including the western basin of Lake Erie. Substantial efforts have been made to track these blooms using in situ sampling and remote sensing. However, such measurements do not fully capture HAB spatial and temporal dynamics due to the limitations of discrete shipboard sampling over large areas and the effects of clouds and winds on remote sensing estimates. To address these limitations, we develop a space-time geostatistical modeling framework for estimating HAB intensity and extent using chlorophyll a data sampled during the HAB season (June-October) from 2008 to 2017 by five independent monitoring programs. Based on the Bayesian information criterion for model selection, trend variables explain bloom northerly and easterly expansion from Maumee Bay, wind effects over depth, and variability among sampling methods. Cross validation results demonstrate that space-time kriging explains over half of the variability in daily, location-specific chlorophyll observations, on average. Conditional simulations provide, for the first time, comprehensive estimates of overall bloom biomass (based on depth-integrated concentrations) and surface areal extent with quantified uncertainties. These new estimates are contrasted with previous Lake Erie HAB monitoring studies, and deviations among estimates are explored and discussed. Overall, results highlight the importance of maintaining sufficient monitoring coverage to capture bloom dynamics, as well as the benefits of the proposed approach for synthesizing data from multiple monitoring programs to improve estimation accuracy while reducing uncertainty.

Space-Time Geostatistical Assessment of Hypoxia in the Northern Gulf of Mexico.

Nearly every summer, a large hypoxic zone forms in the northern Gulf of Mexico. Research on the causes and consequences of hypoxia requires reliable estimates of hypoxic extent, which can vary at submonthly time scales due to hydro-meteorological variability. Here, we use an innovative space-time geostatistical model and data collected by multiple research organizations to estimate bottom-water dissolved oxygen (BWDO) concentrations and hypoxic area across summers from 1985 to 2016. We find that 27% of variability in BWDO is explained by deterministic trends with location, depth, and date, while correlated stochasticity accounts for 62% of observational variance within a range of 185 km and 28 days. Space-time modeling reduces uncertainty in estimated hypoxic area by 30% when compared to a spatial-only model, and results provide new insights into the temporal variability of hypoxia. For years with shelf-wide cruises in multiple months, hypoxia is most severe in July in 59% of years, 29% in August, and 12% in June. Also, midsummer cruise estimates of hypoxic area are only modestly correlated with summer-wide (June-August) average estimates ( r2 = 0.5), suggesting midsummer cruises are not necessarily reflective of seasonal hypoxic severity. Furthermore, summer-wide estimates are more strongly correlated with nutrient loading than midsummer estimates.

Relating soil geochemical properties to arsenic bioaccessibility through hierarchical modeling.

Interest in improved understanding of relationships among soil properties and arsenic (As) bioaccessibility has motivated the use of regression models for As bioaccessibility prediction. However, limits in the numbers and types of soils included in previous studies restrict the usefulness of these models beyond the range of soil conditions evaluated, as evidenced by reduced predictive performance when applied to new data. In response, hierarchical models that consider variability in relationships among soil properties and As bioaccessibility across geographic locations and contaminant sources were developed to predict As bioaccessibility in 139 soils on both a mass fraction (mg/kg) and % basis. The hierarchical approach improved the estimation of As bioaccessibility in studied soils. In addition, the number of soil elements identified as statistically significant explanatory variables increased when compared to previous investigations. Specifically, total soil Fe, P, Ca, Co, and V were significant explanatory variables in both models, while total As, Cd, Cu, Ni, and Zn were also significant in the mass fraction model and Mg was significant in the % model. This developed hierarchical approach provides a novel tool to (1) explore relationships between soil properties and As bioaccessibility across a broad range of soil types and As contaminant sources encountered in the environment and (2) identify areas of future mechanistic research to better understand the complexity of interactions between soil properties and As bioaccessibility.

Relationship Between Total and Bioaccessible Lead on Children's Blood Lead Levels in Urban Residential Philadelphia Soils.

Relationships between total soil or bioaccessible lead (Pb), measured using an in vitro bioaccessibility assay, and children's blood lead levels (BLL) were investigated in an urban neighborhood in Philadelphia, PA, with a history of soil Pb contamination. Soil samples from 38 homes were analyzed to determine whether accounting for the bioaccessible Pb fraction improves statistical relationships with children's BLLs. Total soil Pb concentration ranged from 58 to 2821 mg/kg; the bioaccessible Pb concentration ranged from 47 to 2567 mg/kg. Children's BLLs ranged from 0.3 to 9.8 µg/dL. Hierarchical models were used to compare relationships between total or bioaccessible Pb in soil and children's BLLs. Total soil Pb concentration as the predictor accounted for 23% of the variability in child BLL; bioaccessible soil Pb concentration as the predictor accounted for 26% of BLL variability. A bootstrapping analysis confirmed a significant increase in R2 for the model using bioaccessible soil Pb concentration as the predictor with 99.0% of bootstraps showing a positive increase. Estimated increases of 1.3 µg/dL and 1.5 µg/dL in BLL per 1000 mg/kg Pb in soil were observed for this study area using total and bioaccessible Pb concentrations, respectively. Children's age did not contribute significantly to the prediction of BLLs.

Ensemble modeling informs hypoxia management in the northern Gulf of Mexico.

A large region of low-dissolved-oxygen bottom waters (hypoxia) forms nearly every summer in the northern Gulf of Mexico because of nutrient inputs from the Mississippi River Basin and water column stratification. Policymakers developed goals to reduce the area of hypoxic extent because of its ecological, economic, and commercial fisheries impacts. However, the goals remain elusive after 30 y of research and monitoring and 15 y of goal-setting and assessment because there has been little change in river nitrogen concentrations. An intergovernmental Task Force recently extended to 2035 the deadline for achieving the goal of a 5,000-km2 5-y average hypoxic zone and set an interim load target of a 20% reduction of the spring nitrogen loading from the Mississippi River by 2025 as part of their adaptive management process. The Task Force has asked modelers to reassess the loading reduction required to achieve the 2035 goal and to determine the effect of the 20% interim load reduction. Here, we address both questions using a probabilistic ensemble of four substantially different hypoxia models. Our results indicate that, under typical weather conditions, a 59% reduction in Mississippi River nitrogen load is required to reduce hypoxic area to 5,000 km2 The interim goal of a 20% load reduction is expected to produce an 18% reduction in hypoxic area over the long term. However, due to substantial interannual variability, a 25% load reduction is required before there is 95% certainty of observing any hypoxic area reduction between consecutive 5-y assessment periods.

Tracking cyanobacteria blooms: Do different monitoring approaches tell the same story?

Cyanobacteria blooms are a major environmental issue worldwide. Our understanding of the biophysical processes driving cyanobacterial proliferation and the ability to develop predictive models that inform resource managers and policy makers rely upon the accurate characterization of bloom dynamics. Models quantifying relationships between bloom severity and environmental drivers are often calibrated to an individual set of bloom observations, and few studies have assessed whether differences among observing platforms could lead to contrasting results in terms of relevant bloom predictors and their estimated influence on bloom severity. The aim of this study was to assess the degree of coherence of different monitoring methods in (1) capturing short- and long-term cyanobacteria bloom dynamics and (2) identifying environmental drivers associated with bloom variability. Using western Lake Erie as a case study, we applied boosted regression tree (BRT) models to long-term time series of cyanobacteria bloom estimates from multiple in-situ and remote sensing approaches to quantify the relative influence of physico-chemical and meteorological drivers on bloom variability. Results of BRT models showed remarkable consistency with known ecological requirements of cyanobacteria (e.g., nutrient loading, water temperature, and tributary discharge). However, discrepancies in inter-annual and intra-seasonal bloom dynamics across monitoring approaches led to some inconsistencies in the relative importance, shape, and sign of the modeled relationships between select environmental drivers and bloom severity. This was especially true for variables characterized by high short-term variability, such as wind forcing. These discrepancies might have implications for our understanding of the role of different environmental drivers in regulating bloom dynamics, and subsequently for the development of models capable of informing management and decision making. Our results highlight the need to develop methods to integrate multiple data sources to better characterize bloom spatio-temporal variability and improve our ability to understand and predict cyanobacteria blooms.

Assessing biophysical controls on Gulf of Mexico hypoxia through probabilistic modeling.

A mechanistic model was developed to predict midsummer bottom-water dissolved oxygen (BWDO) concentration and hypoxic area on the Louisiana shelf of the northern Gulf of Mexico, USA (1985-2011). Because of its parsimonious formulation, the model possesses many of the benefits of simpler, more empirical models, in that it is computationally efficient and can rigorously account for uncertainty through Bayesian inference. At the same time, the model incorporates important biophysical processes such that its parameterization can be informed by field-measured biological and physical rates. The model is used to explore how freshwater flow, nutrient load, benthic oxygen demand, and wind velocity affect hypoxia on the western and eastern sections of the shelf, delineated by the Atchafalaya River outfall. The model explains over 70% of the variability in BWDO on both shelf sections, and outperforms linear regression models developed from the same input variables. Model results suggest that physical factors (i.e., wind and flow) control a larger portion of the year-to-year variability in hypoxia than previously thought, especially on the western shelf, though seasonal nutrient loads remain an important driver of hypoxia, as well. Unlike several previous Gulf hypoxia modeling studies, results do not indicate a temporal shift in the system's propensity for hypoxia formation (i.e., no regime change). Results do indicate that benthic oxygen demand is a substantial BWDO sink, and a better understanding of the long-term dynamics of this sink is required to better predict how the size of the hypoxic zone will respond to proposed reductions in nutrient loading.

Independent data validation of an in vitro method for the prediction of the relative bioavailability of arsenic in contaminated soils.

In vitro bioaccessibility (IVBA) assays estimate arsenic (As) relative bioavailability (RBA) in contaminated soils to improve accuracy in human exposure assessments. Previous studies correlating soil As IVBA with RBA have been limited by the use of few soil types and sources of As, and the predictive value of As IVBA has not been validated using an independent set of As-contaminated soils. In this study, a robust linear model was developed to predict As RBA in mice using IVBA, and the predictive capability of the model was independently validated using a unique set of As-contaminated soils. Forty As-contaminated soils varying in soil type and contaminant source were included in this study, with 31 soils used for initial model development and nine soils used for independent model validation. The initial model reliably predicted As RBA values in the independent data set, with a mean As RBA prediction error of 5.4%. Following validation, 40 soils were used for final model development, resulting in a linear model with the equation RBA = 0.65 × IVBA + 7.8 and an R(2) of 0.81. The in vivo-in vitro correlation and independent data validation presented provide critical verification necessary for regulatory acceptance in human health risk assessment.

A scenario and forecast model for Gulf of Mexico hypoxic area and volume.

For almost three decades, the relative size of the hypoxic region on the Louisiana-Texas continental shelf has drawn scientific and policy attention. During that time, both simple and complex models have been used to explore hypoxia dynamics and to provide management guidance relating the size of the hypoxic zone to key drivers. Throughout much of that development, analyses had to accommodate an apparent change in hypoxic sensitivity to loads and often cull observations due to anomalous meteorological conditions. Here, we describe an adaptation of our earlier, simple biophysical model, calibrated to revised hypoxic area estimates and new hypoxic volume estimates through Bayesian estimation. This application eliminates the need to cull observations and provides revised hypoxic extent estimates with uncertainties corresponding to different nutrient loading reduction scenarios. We compare guidance from this model application, suggesting an approximately 62% nutrient loading reduction is required to reduce Gulf hypoxia to the Action Plan goal of 5000 km(2), to that of previous applications. In addition, we describe for the first time, the corresponding response of hypoxic volume. We also analyze model results to test for increasing system sensitivity to hypoxia formation, but find no strong evidence of such change.

Retrospective analysis of midsummer hypoxic area and volume in the northern Gulf of Mexico, 1985-2011.

Robust estimates of hypoxic extent (both area and volume) are important for assessing the impacts of low dissolved oxygen on aquatic ecosystems at large spatial scales. Such estimates are also important for calibrating models linking hypoxia to causal factors, such as nutrient loading and stratification, and for informing management decisions. In this study, we develop a rigorous geostatistical modeling framework to estimate the hypoxic extent in the northern Gulf of Mexico from data collected during midsummer, quasi-synoptic monitoring cruises (1985-2011). Instead of a traditional interpolation-based approach, we use a simulation-based approach that yields more robust extent estimates and quantified uncertainty. The modeling framework also makes use of covariate information (i.e., trend variables such as depth and spatial position), to reduce estimation uncertainty. Furthermore, adjustments are made to account for observational bias resulting from the use of different sampling instruments in different years. Our results suggest an increasing trend in hypoxic layer thickness (p = 0.05) from 1985 to 2011, but less than significant increases in volume (p = 0.12) and area (p = 0.42). The uncertainties in the extent estimates vary with sampling network coverage and instrument type, and generally decrease over the study period.

Spatial and temporal trends in Lake Erie hypoxia, 1987-2007.

Hypoxic conditions, defined as dissolved oxygen (DO) concentrations below 2 mg/L, are a regular summertime occurrence in Lake Erie, but the spatial extent has been poorly understood due to sparse sampling. We use geostatistical kriging and conditional realizations to provide quantitative estimates of the extent of hypoxia in the central basin of Lake Erie for August and September of 1987 to 2007, along with their associated uncertainties. The applied geostatistical approach combines the limited in situ DO measurements with auxiliary data selected using the Bayesian Information Criterion. Bathymetry and longitude are found to be highly significant in explaining the spatial distribution of DO, while satellite observations of sea surface temperature and satellite chlorophyll are not. The hypoxic extent was generally lowest in the mid-1990s, with the late 1980s (1987, 1988) and the 2000s (2003, 2005) experiencing the largest hypoxic zones. A simple exponential relationship based on the squared average measured bottom DO explains 97% of the estimated variability in the hypoxic extent. The change in the observed maximum extent between August and September is found to be sensitive to the corresponding variability in the hypolimnion thickness.

Quantifying the impacts of stratification and nutrient loading on hypoxia in the northern Gulf of Mexico.

Stratification and nutrient loading are two primary factors leading to hypoxia in coastal systems. However, where these factors are temporally correlated, it can be difficult to isolate and quantify their individual impacts. This study provides a novel solution to this problem by determining the effect of stratification based on its spatial relationship with bottom-water dissolved oxygen (BWDO) concentration using a geostatistical regression. Ten years (1998-2007) of midsummer Gulf of Mexico BWDO measurements are modeled using stratification metrics along with trends based on spatial coordinates and bathymetry, which together explain 27-61% of the spatial variability in BWDO for individual years. Because stratification effects explain only a portion of the year-to-year variability in mean BWDO; the remaining variability is explained by other factors, with May nitrate plus nitrite river concentration the most important. Overall, 82% of the year-to-year variability in mean BWDO is explained. The results suggest that while both stratification and nutrients play important roles in determining the annual extent of midsummer hypoxia, reducing nutrient inputs alone will substantially reduce the average extent.