Using near-term forecasts and uncertainty partitioning to inform prediction of oligotrophic lake cyanobacterial density.

Near-term ecological forecasts provide resource managers advance notice of changes in ecosystem services, such as fisheries stocks, timber yields, or water quality. Importantly, ecological forecasts can identify where there is uncertainty in the forecasting system, which is necessary to improve forecast skill and guide interpretation of forecast results. Uncertainty partitioning identifies the relative contributions to total forecast variance introduced by different sources, including specification of the model structure, errors in driver data, and estimation of current states (initial conditions). Uncertainty partitioning could be particularly useful in improving forecasts of highly variable cyanobacterial densities, which are difficult to predict and present a persistent challenge for lake managers. As cyanobacteria can produce toxic and unsightly surface scums, advance warning when cyanobacterial densities are increasing could help managers mitigate water quality issues. Here, we fit 13 Bayesian state-space models to evaluate different hypotheses about cyanobacterial densities in a low nutrient lake that experiences sporadic surface scums of the toxin-producing cyanobacterium, Gloeotrichia echinulata. We used data from several summers of weekly cyanobacteria samples to identify dominant sources of uncertainty for near-term (1- to 4-week) forecasts of G. echinulata densities. Water temperature was an important predictor of cyanobacterial densities during model fitting and at the 4-week forecast horizon. However, no physical covariates improved model performance over a simple model including the previous week's densities in 1-week-ahead forecasts. Even the best fit models exhibited large variance in forecasted cyanobacterial densities and did not capture rare peak occurrences, indicating that significant explanatory variables when fitting models to historical data are not always effective for forecasting. Uncertainty partitioning revealed that model process specification and initial conditions dominated forecast uncertainty. These findings indicate that long-term studies of different cyanobacterial life stages and movement in the water column as well as measurements of drivers relevant to different life stages could improve model process representation of cyanobacteria abundance. In addition, improved observation protocols could better define initial conditions and reduce spatial misalignment of environmental data and cyanobacteria observations. Our results emphasize the importance of ecological forecasting principles and uncertainty partitioning to refine and understand predictive capacity across ecosystems.

Assessing the effectiveness of Landsat 8 chlorophyll a retrieval algorithms for regional freshwater monitoring.

Predicting algal blooms has become a priority for scientists, municipalities, businesses, and citizens. Remote sensing offers solutions to the spatial and temporal challenges facing existing lake research and monitoring programs that rely primarily on high-investment, in situ measurements. Techniques to remotely measure chlorophyll a (chl a) as a proxy for algal biomass have been limited to specific large water bodies in particular seasons and narrow chl a ranges. Thus, a first step toward prediction of algal blooms is generating regionally robust algorithms using in situ and remote sensing data. This study explores the relationship between in-lake measured chl a data from Maine and New Hampshire, USA lakes and remotely sensed chl a retrieval algorithm outputs. Landsat 8 images were obtained and then processed after required atmospheric and radiometric corrections. Six previously developed algorithms were tested on a regional scale on 11 scenes from 2013 to 2015 covering 192 lakes. The best performing algorithm across data from both states had a 0.16 correlation coefficient (R2 ) and P ≤ 0.05 when Landsat 8 images within 5 d, and improved to R2 of 0.25 when data from Maine only were used. The strength of the correlation varied with the specificity of the time window in relation to the in-situ sampling date, explaining up to 27% of the variation in the data across several scenes. Two previously published algorithms using Landsat 8's Bands 1-4 were best correlated with chl a, and for particular late-summer scenes, they accounted for up to 69% of the variation in in-situ measurements. A sensitivity analysis revealed that a longer time difference between in situ measurements and the satellite image increased uncertainty in the models, and an effect of the time of year on several indices was demonstrated. A regional model based on the best performing remote sensing algorithm was developed and was validated using independent in situ measurements and satellite images. These results suggest that, despite challenges including seasonal effects and low chl a thresholds, remote sensing could be an effective and accessible regional-scale tool for chl a monitoring programs in lakes.

The value of hydrologic information for watershed management programs: The case of Camboriú, Brazil.

Investments in watershed services programs hold the promise to protect and restore ecosystems and water resources. The design and implementation of such programs is often accompanied by hydrologic modeling and monitoring, although the role of hydrologic information in meeting the needs of program managers remains unclear. In the Camboriú watershed, Brazil, we explored the value of hydrologic modeling and monitoring with respect to two dimensions: scientific credibility and use of generated knowledge in the design, implementation, and evaluation of the watershed management program. We used a combination of semi-structured interviews, focus groups, and hydrologic modeling under various levels of data availability to examine when improved models and data availability might build credibility and provide more useful information for decision makers. We found that hydrologic information was not actually used for the detailed design, but rather contributed to broad-scale support of the program by increasing scientific credibility. Model sophistication and data availability improved the credibility of hydrologic information but did not affect actual decisions related to program design. Hydrologic monitoring data were critical for model calibration, and high-resolution land use and land cover data, obtained via remote sensing, affected some model outputs which were not used to design the program. Our study suggests that identifying how hydrologic data will inform decision making should guide the level of effort used in hydrologic modeling and monitoring.