Predictable patterns within the kelp forest can indirectly create temporary refugia from ocean acidification.

Kelps are recognized for providing many ecosystem services in coastal areas and considered in ocean acidification (OA) mitigation. However, assessing OA modification requires an understanding of the multiple parameters involved in carbonate chemistry, especially in highly dynamic systems. We studied the effects of sugar kelp (Saccharina latissima) on an experimental farm at the north end of Hood Canal, Washington-a low retentive coastal system. In this field mesocosm study, two oyster species (Magallana gigas, Ostrea lurida) were exposed at locations in the mid, edge, and outside the kelp array. The Hood Head Sugar Kelp Farm Model outputs were used to identify dominating factors in spatial and temporal kelp dynamics, while wavelet spectrum analyses helped in understanding predictability patterns. This was linked to the measured biological responses (dissolution, growth, isotopes) of the exposed organisms. Positioned in an area of high (sub)-diel tidal fluxes with low retention potential, there were no measurable alterations of the seawater pH at the study site, demonstrating that the kelp array could not induce a direct mitigating effect against OA. However, beneficial responses in calcifiers were still observed, which are linked to two causes: increased pH predictability and improved provisioning through kelp-derived particulate organic resource utilization and as such, kelp improved habitat suitability and indirectly created refugia against OA. This study can serve as an analogue for many coastal bay habitats where prevailing physical forcing drives chemical changes. Future macrophyte studies that investigate OA mitigating effects should focus also on the importance of predictability patterns, which can additionally improve the conditions for marine calcifiers and ecosystem services vulnerable to or compromised by OA, including aquaculture sustainability.

MassWateR: Improving quality control, analysis, and sharing of water quality data.

The long-term protection and restoration of aquatic resources depends on robust monitoring data; data that require systematic quality control and analysis tools. The MassWateR R package facilitates quality control, analysis, and data sharing for discrete surface water quality data collected by monitoring programs of various size and technical capacity. The tools were developed to address regional needs for programs in Massachusetts, USA, but the principles and outputs can be applicable to monitoring data collected anywhere. Users can create quality control reports, perform outlier analyses, and assess trends by season, date, and site for more than 40 parameters. Users can also prepare data for submission to the United States Environmental Protection Agency Water Quality Exchange, thus sharing data to the largest water quality database in the United States. The automated and reproducible workflow offered by MassWateR is expected to increase the quantity and quality of publicly available data to support the management of aquatic resources.

Natural Analogues in pH Variability and Predictability across the Coastal Pacific Estuaries: Extrapolation of the Increased Oyster Dissolution under Increased pH Amplitude and Low Predictability Related to Ocean Acidification.

Coastal-estuarine habitats are rapidly changing due to global climate change, with impacts influenced by the variability of carbonate chemistry conditions. However, our understanding of the responses of ecologically and economically important calcifiers to pH variability and temporal variation is limited, particularly with respect to shell-building processes. We investigated the mechanisms driving biomineralogical and physiological responses in juveniles of introduced (Pacific; Crassostrea gigas) and native (Olympia; Ostrea lurida) oysters under flow-through experimental conditions over a six-week period that simulate current and future conditions: static control and low pH (8.0 and 7.7); low pH with fluctuating (24-h) amplitude (7.7 ± 0.2 and 7.7 ± 0.5); and high-frequency (12-h) fluctuating (8.0 ± 0.2) treatment. The oysters showed physiological tolerance in vital processes, including calcification, respiration, clearance, and survival. However, shell dissolution significantly increased with larger amplitudes of pH variability compared to static pH conditions, attributable to the longer cumulative exposure to lower pH conditions, with the dissolution threshold of pH 7.7 with 0.2 amplitude. Moreover, the high-frequency treatment triggered significantly greater dissolution, likely because of the oyster's inability to respond to the unpredictable frequency of variations. The experimental findings were extrapolated to provide context for conditions existing in several Pacific coastal estuaries, with time series analyses demonstrating unique signatures of pH predictability and variability in these habitats, indicating potentially benefiting effects on fitness in these habitats. These implications are crucial for evaluating the suitability of coastal habitats for aquaculture, adaptation, and carbon dioxide removal strategies.

Initial estuarine response to inorganic nutrient inputs from a legacy mining facility adjacent to Tampa Bay, Florida.

Legacy mining facilities pose significant risks to aquatic resources. From March 30th to April 9th, 2021, 814 million liters of phosphate mining wastewater and marine dredge water from the Piney Point facility were released into lower Tampa Bay (Florida, USA). This resulted in an estimated addition of 186 metric tons of total nitrogen, exceeding typical annual external nitrogen load estimates to lower Tampa Bay in a matter of days. An initial phytoplankton bloom (non-harmful diatoms) was first observed in April. Filamentous cyanobacteria blooms (Dapis spp.) peaked in June, followed by a bloom of the red tide organism Karenia brevis. Reported fish kills tracked K. brevis concentrations, prompting cleanup of over 1600 metric tons of dead fish. Seagrasses had minimal changes over the study period. By comparing these results to baseline environmental monitoring data, we demonstrate adverse water quality changes in response to abnormally high and rapidly delivered nitrogen loads.

Multi-scale trend analysis of water quality using error propagation of generalized additive models.

Effective stewardship of ecosystems to sustain current ecological status or mitigate impacts requires nuanced understanding of how conditions have changed over time in response to anthropogenic pressures and natural variability. Detecting and appropriately characterizing changes requires accurate and flexible trend assessment methods that can be readily applied to environmental monitoring datasets. A key requirement is complete propagation of uncertainty through the analysis. However, this is difficult when there are mismatches between sampling frequency, period of record, and trends of interest. Here, we propose a novel application of generalized additive models (GAMs) for characterizing multi-decadal changes in water quality indicators and demonstrate its utility by analyzing a 30-year record of biweekly-to-monthly chlorophyll-a concentrations in the San Francisco Estuary. GAMs have shown promise in water quality trend analysis to separate long-term (i.e., annual or decadal) trends from seasonal variation. Our proposed methods estimate seasonal averages in a response variable with GAMs, extract uncertainty measures for the seasonal estimates, and then use the uncertainty measures with mixed-effects meta-analysis regression to quantify inter-annual trends that account for full propagation of error across methods. We first demonstrate that nearly identical descriptions of temporal changes can be obtained using different smoothing spline formulations of the original time series. We then extract seasonal averages and their standard errors for an a priori time period within each year from the GAM results. Finally, we demonstrate how across-year trends in seasonal averages can be modeled with mixed-effects meta-analysis regression that propagates uncertainties from the GAM fits to the across-year analysis. Overall, this approach leverages GAMs to smooth data with missing observations or varying sample effort across years to estimate seasonal averages and meta-analysis to estimate trends across years. Methods are provided in the wqtrends R package.

Multivariate Analysis of Sediment Toxicity in an Ocean Ecosystem: A Southern California Bight Case Study.

Contaminated sediments can negatively affect aquatic organisms and beneficial uses of coastal regions. Monitoring programs typically collect many indicators of sediment toxicity, yet multivariate approaches that comprehensively evaluate data across heterogeneous spatial environments are frequently not performed. In this paper, we explore a multivariate approach to show that a list of suspected drivers of sediment toxicity to native Mytilus galloprovincialis (mussel) and Eohaustorius estuarius (a marine amphipod) population can be narrowed down without excluding samples, and that redundancies in sampling sites can be identified and isolated. Using a 153 × 28 data matrix assembled from a southern California-wide bight monitoring program, we demonstrate by this approach that Port of Los Angeles (PLA) and San Diego Bay (SDB) contained the most toxic sediments in the bight in 2008, the nature of which was unique to each locality. (Note: Little toxicity was observed here in 2013 and 2018.) In PLA sediments, mussels were more affected than amphipods, with higher survivability associated with low Hg and Sn levels. Conversely, amphipods had higher mortality than mussel embryos in SDB sediments, with higher survivability associated with low Be and Co levels. Nitrogen, organic content, and finer sediment particles were not related to the survivability of these organisms.

Severe biological effects under present-day estuarine acidification in the seasonally variable Salish Sea.

Estuaries are recognized as one of the habitats most vulnerable to coastal ocean acidification due to seasonal extremes and prolonged duration of acidified conditions. This is combined with co-occurring environmental stressors such as increased temperature and low dissolved oxygen. Despite this, evidence of biological impacts of ocean acidification in estuarine habitats is largely lacking. By combining physical, biogeochemical, and biological time-series observations over relevant seasonal-to-interannual time scales, this study is the first to describe both the spatial and temporal variation of biological response in the pteropod Limacina helicina to estuarine acidification in association with other stressors. Using clustering and principal component analyses, sampling sites were grouped according to their distribution of physical and biogeochemical variables over space and time. This identified the most exposed habitats and time intervals corresponding to the most severe negative biological impacts across three seasons and three years. We developed a cumulative stress index as a means of integrating spatial-temporal OA variation over the organismal life history. Our findings show that over the 2014-2016 study period, the severity of low aragonite saturation state combined with the duration of exposure contributed to overall cumulative stress and resulted in severe shell dissolution. Seasonally-variable estuaries such as the Salish Sea (Washington, U.S.A.) predispose sensitive organisms to more severe acidified conditions than those of coastal and open-ocean habitats, yet the sensitive organisms persist. We suggest potential environmental factors and compensatory mechanisms that allow pelagic calcifiers to inhabit less favorable habitats and partially offset associated stressors, for instance through food supply, increased temperature, and adaptation of their life history. The novel metric of cumulative stress developed here can be applied to other estuarine environments with similar physical and chemical dynamics, providing a new tool for monitoring biological response in estuaries under pressure from accelerating global change.

The impact of climate change induced alterations of streamflow and stream temperature on the distribution of riparian species.

Distributions of riparian species will likely shift due to climate change induced alterations in temperature and rainfall patterns, which alter stream habitat. Spatial forecasting of suitable habitat in projected climatic conditions will inform management interventions that support wildlife. Challenges in developing forecasts include the need to consider the large number of riparian species that might respond differently to changing conditions and the need to evaluate the many different characteristics of streamflow and stream temperature that drive species-specific habitat suitability. In particular, in dynamic environments like streams, the short-term temporal resolution of species occurrence and streamflow need to be considered to identify the types of conditions that support various species. To address these challenges, we cluster species based on habitat characteristics to select habitat representatives and we evaluate regional changes in habitat suitability using short-term, temporally explicit metrics that describe the streamflow and stream temperature regime. We use stream-specific environmental predictors rather than climatic variables. Unlike other studies, the stream-specific environmental predictors are generated from the time that species were observed in a particular reach, in addition to long term trends, to evaluate habitat preferences. With species occurrence data from local monitoring surveys and streamflow and stream temperature modeled from downscaled Coupled Model Intercomparison Project - Phase 5 (CMIP5) climate projections, we predict change in habitat suitability at the end-of-century. The relative importance of hydrology and stream temperature varied by cluster. High altitudinal, cold water species' distributions contracted, while lower elevation, warm water species distributions expanded. Modeling with short-term temporally explicit environmental metrics did produce different end-of-century projections than using long-term averages for some of the representative species. These findings can help wildlife managers prioritize conservation efforts, manage streamflow, initiate monitoring of species in vulnerable clusters, and address stressors, such as passage barriers, in areas projected to be suitable in future climate conditions.

The importance of open science for biological assessment of aquatic environments.

Open science principles that seek to improve science can effectively bridge the gap between researchers and environmental managers. However, widespread adoption has yet to gain traction for the development and application of bioassessment products. At the core of this philosophy is the concept that research should be reproducible and transparent, in addition to having long-term value through effective data preservation and sharing. In this article, we review core open science concepts that have recently been adopted in the ecological sciences and emphasize how adoption can benefit the field of bioassessment for both prescriptive condition assessments and proactive applications that inform environmental management. An example from the state of California demonstrates effective adoption of open science principles through data stewardship, reproducible research, and engagement of stakeholders with multimedia applications. We also discuss technical, sociocultural, and institutional challenges for adopting open science, including practical approaches for overcoming these hurdles in bioassessment applications.

Exoskeleton dissolution with mechanoreceptor damage in larval Dungeness crab related to severity of present-day ocean acidification vertical gradients.

Ocean acidification (OA) along the US West Coast is intensifying faster than observed in the global ocean. This is particularly true in nearshore regions (<200 m) that experience a lower buffering capacity while at the same time providing important habitats for ecologically and economically significant species. While the literature on the effects of OA from laboratory experiments is voluminous, there is little understanding of present-day OA in-situ effects on marine life. Dungeness crab (Metacarcinus magister) is perennially one of the most valuable commercial and recreational fisheries. We focused on establishing OA-related vulnerability of larval crustacean based on mineralogical and elemental carapace to external and internal carapace dissolution by using a combination of different methods ranging from scanning electron microscopy, energy dispersive X-ray spectroscopy, elemental mapping and X-ray diffraction. By integrating carapace features with the chemical observations and biogeochemical model hindcast, we identify the occurrence of external carapace dissolution related to the steepest Ω calcite gradients (∆Ωcal,60) in the water column. Dissolution features are observed across the carapace, pereopods (legs), and around the calcified areas surrounding neuritic canals of mechanoreceptors. The carapace dissolution is the most extensive in the coastal habitats under prolonged (1-month) long exposure, as demonstrated by the use of the model hindcast. Such dissolution has a potential to destabilize mechanoreceptors with important sensory and behavioral functions, a pathway of sensitivity to OA. Carapace dissolution is negatively related to crab larval width, demonstrating a basis for energetic trade-offs. Using a retrospective prediction from a regression models, we estimate an 8.3% increase in external carapace dissolution over the last two decades and identified a set of affected OA-related sublethal pathways to inform future risk assessment studies of Dungeness crabs.

Assessment of the cumulative effects of restoration activities on water quality in Tampa Bay, Florida.

Habitat and water quality restoration projects are commonly used to enhance coastal resources or mitigate negative impacts of water quality stressors. Significant resources have been expended for restoration projects, yet much less attention has focused on evaluating broad regional outcomes beyond site-specific assessments. This study presents an empirical framework to evaluate multiple datasets in the Tampa Bay area (Florida, USA) to identify 1) the types of restoration projects that have produced the greatest improvements in water quality, and 2) over which time frames different projects may produce water quality benefits. Information on the location and date of completion of 887 restoration projects from 1971 to 2017 were spatially and temporally matched with water quality records at each of 45 long-term monitoring stations in Tampa Bay. The underlying assumption was that the developed framework could identify differences in water quality changes between types of restoration projects based on aggregate estimates of chlorophyll-a concentrations before and after the completion of one to many projects. Water infrastructure projects to control point source nutrient loading into the Bay were associated with the highest likelihood of chlorophyll-a reduction, particularly for projects occurring prior to 1995. Habitat restoration projects were also associated with reductions in chlorophyll-a, although the likelihood of reductions from the cumulative effects of these projects were less than those from infrastructure improvements alone. The framework is sufficiently flexible for application to different spatiotemporal contexts and could be used to develop reasonable expectations for implementation of future water quality restoration activities throughout the Gulf of Mexico.

Phytoplankton decline in the eastern North Pacific transition zone associated with atmospheric blocking.

Global climate change can significantly influence oceanic phytoplankton dynamics, and thus biogeochemical cycles and marine food webs. However, associative explanations based on the correlation between chlorophyll-a concentration (Chl-a) and climatic indices is inadequate to describe the mechanism of the connection between climate change, large-scale atmospheric dynamics, and phytoplankton variability. Here, by analyzing multiple satellite observations of Chl-a and atmospheric conditions from National Center for Environmental Prediction/National Center for Atmospheric Research reanalysis datasets, we show that high-latitude atmospheric blocking events over Alaska are the primary drivers of the recent decline of Chl-a in the eastern North Pacific transition zone. These blocking events were associated with the persistence of large-scale atmosphere pressure fields that decreased westerly winds and southward Ekman transport over the subarctic ocean gyre. Reduced southward Ekman transport leads to reductions in nutrient availability to phytoplankton in the transition zone. The findings describe a previously unidentified climatic factor that contributed to the recent decline of phytoplankton in this region and propose a mechanism of the top-down teleconnection between the high-latitude atmospheric circulation anomalies and the subtropical oceanic primary productivity. The results also highlight the importance of understanding teleconnection among atmosphere-ocean interactions as a means to anticipate future climate change impacts on oceanic primary production.

NeuralNetTools: Visualization and Analysis Tools for Neural Networks.

Supervised neural networks have been applied as a machine learning technique to identify and predict emergent patterns among multiple variables. A common criticism of these methods is the inability to characterize relationships among variables from a fitted model. Although several techniques have been proposed to "illuminate the black box", they have not been made available in an open-source programming environment. This article describes the NeuralNetTools package that can be used for the interpretation of supervised neural network models created in R. Functions in the package can be used to visualize a model using a neural network interpretation diagram, evaluate variable importance by disaggregating the model weights, and perform a sensitivity analysis of the response variables to changes in the input variables. Methods are provided for objects from many of the common neural network packages in R, including caret, neuralnet, nnet, and RSNNS. The article provides a brief overview of the theoretical foundation of neural networks, a description of the package structure and functions, and an applied example to provide a context for model development with NeuralNetTools. Overall, the package provides a toolset for neural networks that complements existing quantitative techniques for data-intensive exploration.

A novel imputation methodology for time series based on pattern sequence forecasting.

The Pattern Sequence Forecasting (PSF) algorithm is a previously described algorithm that identifies patterns in time series data and forecasts values using periodic characteristics of the observations. A new method for univariate time series is introduced that modifies the PSF algorithm to simultaneously forecast and backcast missing values for imputation. The imputePSF method extends PSF by characterizing repeating patterns of existing observations to provide a more precise estimate of missing values compared to more conventional methods, such as replacement with means or last observation carried forward. The imputation accuracy of imputePSF was evaluated by simulating varying amounts of missing observations with three univariate datasets. Comparisons of imputePSF with well-established methods using the same simulations demonstrated an overall reduction in error estimates. The imputePSF algorithm can produce more precise imputations on appropriate datasets, particularly those with periodic and repeating patterns.

Four decades of water quality change in the upper San Francisco Estuary.

Quantitative descriptions of chemical, physical, and biological characteristics of estuaries are critical for developing an ecological understanding of drivers of change. Historical trends and relationships between key species of dissolved inorganic nitrogen (ammonium, nitrate/nitrite, total) from the Delta region of the San Francisco Estuary were modeled with an estuarine adaptation of the Weighted Regressions on Time, Discharge, and Season (WRTDS). Analysis of flow-normalized data revealed trends that were different from those in the observed time-series. Flow-normalized data exhibited changes in magnitude and even reversal of trends relative to the observed data. Modelled trends demonstrated that nutrient concentrations were on average higher in the last twenty years relative to the earlier periods of observation, although concentrations have been slowly declining since the mid-1990s and early 2000s. We further describe mechanisms of change with two case studies that evaluated 1) downstream changes in nitrogen following upgrades at a wastewater treatment plant, and 2) interactions between biological invaders, chlorophyll, macro-nutrients (nitrogen and silica), and flow in Suisun Bay. WRTDS results for ammonium trends showed a distinct signal as a result of upstream wastewater treatment plant upgrades, with specific reductions observed in the winter months during low-flow conditions. Results for Suisun Bay showed that chlorophyll a production in early years was directly stimulated by flow, whereas the relationship with flow in later years was indirect and influenced by grazing pressure. Although these trends and potential causes of change have been described in the literature, results from WRTDS provided an approach to test alternative hypotheses of spatiotemporal drivers of nutrient dynamics in the Delta.

Opening the black box of neural networks: methods for interpreting neural network models in clinical applications.

Artificial neural networks (ANNs) are powerful tools for data analysis and are particularly suitable for modeling relationships between variables for best prediction of an outcome. While these models can be used to answer many important research questions, their utility has been critically limited because the interpretation of the "black box" model is difficult. Clinical investigators usually employ ANN models to predict the clinical outcomes or to make a diagnosis; the model however is difficult to interpret for clinicians. To address this important shortcoming of neural network modeling methods, we describe several methods to help subject-matter audiences (e.g., clinicians, medical policy makers) understand neural network models. Garson's algorithm describes the relative magnitude of the importance of a descriptor (predictor) in its connection with outcome variables by dissecting the model weights. The Lek's profile method explores the relationship of the outcome variable and a predictor of interest, while holding other predictors at constant values (e.g., minimum, 20th quartile, maximum). While Lek's profile was developed specifically for neural networks, partial dependence plot is a more generic version that visualize the relationship between an outcome and one or two predictors. Finally, the local interpretable model-agnostic explanations (LIME) method can show the predictions of any classification or regression, by approximating it locally with an interpretable model. R code for the implementations of these methods is shown by using example data fitted with a standard, feed-forward neural network model. We offer codes and step-by-step description on how to use these tools to facilitate better understanding of ANN.

Quantifying seagrass light requirements using an algorithm to spatially resolve depth of colonization.

Depth of colonization (Zc ) is a useful seagrass growth metric that describes seagrass response to light availability. Similarly, percent surface irradiance at Zc (% SI) is an indicator of seagrass light requirements with applications in seagrass ecology and management. Methods for estimating Zc and % SI are highly variable making meaningful comparisons difficult. A new algorithm is presented to compute maps of median and maximum Zc , Zc, med and Zc,max , respectively, for four Florida coastal areas (Big Bend, Tampa Bay, Choctawhatchee Bay, Indian River Lagoon). Maps of light attenuation (Kd ) based on MODIS satellite imagery, PAR profiles, and Secchi depth measurements were combined with seagrass growth estimates to produce maps of % SI at Zc,med and Zc,max . Among estuary segments, mean Zc,med varied from (±s.e.) 0.80±0.13 m for Old Tampa Bay to 2.33±0.26 m for Western Choctawhatchee Bay. Standard errors for Zc,med were 1-10% of the segment means. Percent SI at Zc,med averaged 18% for Indian River Lagoon (range = 9-24%), 42% for Tampa Bay (37-48%) and 58% for Choctawhatchee Bay (51-75%). Estimates of % SI were significantly lower in Indian River Lagoon than in the other estuaries, while estimates for Tampa Bay and Choctawhatchee Bay were higher than the often cited estimate of 20%. Spatial gradients in depth of colonization and % SI were apparent in all estuaries. The analytical approach could be applied easily to new data from these estuaries or to other estuaries and could be incorporated routinely in assessments of seagrass status and condition.

Seasonal oxygen dynamics in a warm temperate estuary: effects of hydrologic variability on measurements of primary production, respiration, and net metabolism.

Seasonal responses in estuarine metabolism (primary production, respiration, and net metabolism) were examined using two complementary approaches. Total ecosystem metabolism rates were calculated from dissolved oxygen time series using Odum's open water method. Water column rates were calculated from oxygen-based bottle experiments. The study was conducted over a spring-summer season in the Pensacola Bay estuary at a shallow seagrass-dominated site and a deeper bare-bottomed site. Water column integrated gross production rates more than doubled (58.7 to 130.9 mmol O2 m-2 d-1) from spring to summer, coinciding with a sharp increase in water column chlorophyll-a, and a decrease in surface salinity. As expected, ecosystem gross production rates were consistently higher than water column rates, but showed a different spring-summer pattern, decreasing at the shoal site from 197 to 168 mmol O2 m-2 d-1 and sharply increasing at the channel site from 93.4 to 197.4 mmol O2 m-2 d-1. The consistency among approaches was evaluated by calculating residual metabolism rates (ecosystem - water column). At the shoal site, residual gross production rates decreased from spring to summer from 176.8 to 99.1 mmol O2 m-2 d-1, but were generally consistent with expectations for seagrass environments, indicating that the open water method captured both water column and benthic processes. However, at the channel site, where benthic production was strongly light-limited, residual gross production varied from 15.7 mmol O2 m-2 d-1 in spring to 86.7 mmol O2 m-2 d-1 in summer. The summer rates were much higher than could be realistically attributed to benthic processes, and likely reflected a violation of the open water method due to water column stratification. While the use of sensors for estimating complex ecosystem processes holds promise for coastal monitoring programs, careful attention to the sampling design, and to the underlying assumptions of the methods, is critical for correctly interpreting the results. This study demonstrated how using a combination of approaches yielded a fuller understanding of the ecosystem response to hydrologic and seasonal variability.

R Package imputeTestbench to Compare Imputation Methods for Univariate Time Series.

Missing observations are common in time series data and several methods are available to impute these values prior to analysis. Variation in statistical characteristics of univariate time series can have a profound effect on characteristics of missing observations and, therefore, the accuracy of different imputation methods. The imputeTestbench package can be used to compare the prediction accuracy of different methods as related to the amount and type of missing data for a user-supplied dataset. Missing data are simulated by removing observations completely at random or in blocks of different sizes depending on characteristics of the data. Several imputation algorithms are included with the package that vary from simple replacement with means to more complex interpolation methods. The testbench is not limited to the default functions and users can add or remove methods as needed. Plotting functions also allow comparative visualization of the behavior and effectiveness of different algorithms. We present example applications that demonstrate how the package can be used to understand differences in prediction accuracy between methods as affected by characteristics of a dataset and the nature of missing data.

Numerical and qualitative contrasts of two statistical models for water quality change in tidal waters.

Two statistical approaches, weighted regression on time, discharge, and season (WRTDS) and generalized additive models (GAMs), have recently been used to evaluate water quality trends in estuaries. Both models have been used in similar contexts despite differences in statistical foundations and products. This study provided an empirical and qualitative comparison of both models using 29 years of data for two discrete time series of chlorophyll-a (chl-a) in the Patuxent River Estuary. Empirical descriptions of each model were based on predictive performance against the observed data, ability to reproduce flow-normalized trends with simulated data, and comparisons of performance with validation datasets. Between-model differences were apparent but minor and both models had comparable abilities to remove flow effects from simulated time series. Both models similarly predicted observations for missing data with different characteristics. Trends from each model revealed distinct mainstem influences of the Chesapeake Bay with both models predicting a roughly 65% increase in chl-a over time in the lower estuary, whereas flow-normalized predictions for the upper estuary showed a more dynamic pattern, with a nearly 100% increase in chl-a in the last 10 years. Qualitative comparisons highlighted important differences in the statistical structure, available products, and characteristics of the data and desired analysis.