Using Knowledge-Guided Machine Learning To Assess Patterns of Areal Change in Waterbodies across the Contiguous United States.

Lake and reservoir surface areas are an important proxy for freshwater availability. Advancements in machine learning (ML) techniques and increased accessibility of remote sensing data products have enabled the analysis of waterbody surface area dynamics on broad spatial scales. However, interpreting the ML results remains a challenge. While ML provides important tools for identifying patterns, the resultant models do not include mechanisms. Thus, the "black-box" nature of ML techniques often lacks ecological meaning. Using ML, we characterized temporal patterns in lake and reservoir surface area change from 1984 to 2016 for 103,930 waterbodies in the contiguous United States. We then employed knowledge-guided machine learning (KGML) to classify all waterbodies into seven ecologically interpretable groups representing distinct patterns of surface area change over time. Many waterbodies were classified as having "no change" (43%), whereas the remaining 57% of waterbodies fell into other groups representing both linear and nonlinear patterns. This analysis demonstrates the potential of KGML not only for identifying ecologically relevant patterns of change across time but also for unraveling complex processes that underpin those changes.

Anoxia decreases the magnitude of the carbon, nitrogen, and phosphorus sink in freshwaters.

Oxygen availability is decreasing in many lakes and reservoirs worldwide, raising the urgency for understanding how anoxia (low oxygen) affects coupled biogeochemical cycling, which has major implications for water quality, food webs, and ecosystem functioning. Although the increasing magnitude and prevalence of anoxia has been documented in freshwaters globally, the challenges of disentangling oxygen and temperature responses have hindered assessment of the effects of anoxia on carbon, nitrogen, and phosphorus concentrations, stoichiometry (chemical ratios), and retention in freshwaters. The consequences of anoxia are likely severe and may be irreversible, necessitating ecosystem-scale experimental investigation of decreasing freshwater oxygen availability. To address this gap, we devised and conducted REDOX (the Reservoir Ecosystem Dynamic Oxygenation eXperiment), an unprecedented, 7-year experiment in which we manipulated and modeled bottom-water (hypolimnetic) oxygen availability at the whole-ecosystem scale in a eutrophic reservoir. Seven years of data reveal that anoxia significantly increased hypolimnetic carbon, nitrogen, and phosphorus concentrations and altered elemental stoichiometry by factors of 2-5× relative to oxic periods. Importantly, prolonged summer anoxia increased nitrogen export from the reservoir by six-fold and changed the reservoir from a net sink to a net source of phosphorus and organic carbon downstream. While low oxygen in freshwaters is thought of as a response to land use and climate change, results from REDOX demonstrate that low oxygen can also be a driver of major changes to freshwater biogeochemical cycling, which may serve as an intensifying feedback that increases anoxia in downstream waterbodies. Consequently, as climate and land use change continue to increase the prevalence of anoxia in lakes and reservoirs globally, it is likely that anoxia will have major effects on freshwater carbon, nitrogen, and phosphorus budgets as well as water quality and ecosystem functioning.

Near-term phytoplankton forecasts reveal the effects of model time step and forecast horizon on predictability.

As climate and land use increase the variability of many ecosystems, forecasts of ecological variables are needed to inform management and use of ecosystem services. In particular, forecasts of phytoplankton would be especially useful for drinking water management, as phytoplankton populations are exhibiting greater fluctuations due to human activities. While phytoplankton forecasts are increasing in number, many questions remain regarding the optimal model time step (the temporal frequency of the forecast model output), time horizon (the length of time into the future a prediction is made) for maximizing forecast performance, as well as what factors contribute to uncertainty in forecasts and their scalability among sites. To answer these questions, we developed near-term, iterative forecasts of phytoplankton 1-14 days into the future using forecast models with three different time steps (daily, weekly, fortnightly), that included a full uncertainty partitioning analysis at two drinking water reservoirs. We found that forecast accuracy varies with model time step and forecast horizon, and that forecast models can outperform null estimates under most conditions. Weekly and fortnightly forecasts consistently outperformed daily forecasts at 7-day and 14-day horizons, a trend that increased up to the 14-day forecast horizon. Importantly, our work suggests that forecast accuracy can be increased by matching the forecast model time step to the forecast horizon for which predictions are needed. We found that model process uncertainty was the primary source of uncertainty in our phytoplankton forecasts over the forecast period, but parameter uncertainty increased during phytoplankton blooms and when scaling the forecast model to a new site. Overall, our scalability analysis shows promising results that simple models can be transferred to produce forecasts at additional sites. Altogether, our study advances our understanding of how forecast model time step and forecast horizon influence the forecastability of phytoplankton dynamics in aquatic systems and adds to the growing body of work regarding the predictability of ecological systems broadly.

Increased adoption of best practices in ecological forecasting enables comparisons of forecastability.

Near-term iterative forecasting is a powerful tool for ecological decision support and has the potential to transform our understanding of ecological predictability. However, to this point, there has been no cross-ecosystem analysis of near-term ecological forecasts, making it difficult to synthesize diverse research efforts and prioritize future developments for this emerging field. In this study, we analyzed 178 near-term (≤10-yr forecast horizon) ecological forecasting papers to understand the development and current state of near-term ecological forecasting literature and to compare forecast accuracy across scales and variables. Our results indicated that near-term ecological forecasting is widespread and growing: forecasts have been produced for sites on all seven continents and the rate of forecast publication is increasing over time. As forecast production has accelerated, some best practices have been proposed and application of these best practices is increasing. In particular, data publication, forecast archiving, and workflow automation have all increased significantly over time. However, adoption of proposed best practices remains low overall: for example, despite the fact that uncertainty is often cited as an essential component of an ecological forecast, only 45% of papers included uncertainty in their forecast outputs. As the use of these proposed best practices increases, near-term ecological forecasting has the potential to make significant contributions to our understanding of forecastability across scales and variables. In this study, we found that forecastability (defined here as realized forecast accuracy) decreased in predictable patterns over 1-7 d forecast horizons. Variables that were closely related (i.e., chlorophyll and phytoplankton) displayed very similar trends in forecastability, while more distantly related variables (i.e., pollen and evapotranspiration) exhibited significantly different patterns. Increasing use of proposed best practices in ecological forecasting will allow us to examine the forecastability of additional variables and timescales in the future, providing a robust analysis of the fundamental predictability of ecological variables.