Uncertainty analysis of the performance of a management system for achieving phosphorus load reduction to surface waters

Transferability of a Bayesian Belief Network across diverse agricultural catchments using high-frequency hydrochemistry and land management data

Biogeochemical catchment models are often developed for a single catchment and, as a result, often generalize poorly beyond this. Evaluating their transferability is an important step in improving their predictive power and application range. We assess the transferability of a recently developed Bayesian Belief Network (BBN) that simulated monthly stream phosphorus (P) concentrations in a poorly-drained grassland catchment through application to three further catchments with different hydrological regimes and agricultural land uses. In all catchments, flow and turbidity were measured sub-hourly from 2009 to 2016 and supplemented with 400-500 soil P test measurements. In addition to a previously parameterized BBN, five further model structures were implemented to incorporate in a stepwise way: in-stream P removal using expert elicitation, additional groundwater P stores and delivery, and the presence or absence of septic tank treatment, and, in one case, Sewage Treatment Works. Model performance was tested through comparison of predicted and observed total reactive P (TRP) concentrations and percentage bias (PBIAS). The original BBN accurately simulated the absolute values of observed flow and TRP concentrations in the poorly and moderately drained catchments (albeit with poor apparent percentage bias scores; 76 % ≤ PBIAS≤94 %) irrespective of the dominant land use, but performed less well in the groundwater-dominated catchments. However, including groundwater total dissolved P (TDP) and Sewage Treatment Works (STWs) inputs, and in-stream P uptake improved model performance (-5 % ≤ PBIAS≤18 %). A sensitivity analysis identified redundant variables further helping to streamline the model applications. An enhanced BBN model capable for wider application and generalisation resulted.

Monitoring to detect changes in water quality to meet policy objectives

Detecting change in water quality is key to providing evidence of progress towards meeting water quality objectives. A key measure for detecting change is statistical power. Here we calculate statistical power for all regularly (monthly) monitored streams in New Zealand to test the effectiveness of monitoring for policy that aims to decrease contaminant (phosphorus and nitrogen species, E. coli and visual clarity) concentrations to threshold levels in 5 or 20 years. While > 95% of all monitored sites had sufficient power and samples to detect change in nutrients and clarity over 20 years, on average, sampling frequency would have to double to detect changes in E. coli. Furthermore, to detect changes in 5 years, sampling for clarity, dissolved reactive phosphorus and E. coli would have to increase up to fivefold. The cost of sampling was predicted to increase 5.3 and 4.1 times for 5 and 20 years, respectively. A national model of statistical power was used to demonstrate that a similar number of samples (and cost) would be required for any new monitoring sites. Our work suggests that demonstrating the outcomes of implementing policy for water quality improvement may not occur without a step change in investment into monitoring systems. Emerging sampling technologies have potential to reduce the cost, but existing monitoring networks may also have to be rationalised to provide evidence that water quality is meeting objectives. Our study has important implications for investment decisions involving balancing the need for intensively sampled sites where changes in water quality occur rapidly versus other sites which provide long-term time series.

Optimization and multi-uncertainty analysis of best management practices at the watershed scale: A reliability-level based bayesian network approach

Best management practices (BMPs) have been widely adopted to mitigate diffuse source pollutants, and the simulated processes of its pollutant reduction effectiveness suffer from manifold uncertainties, such as watershed model parameters and climate change. We presented a novel Bayesian modeling framework for BMPs planning, integrating process-based watershed modeling and Bayesian optimization algorithm to reveal the impact of multiple uncertainties. The proposed framework was applied to a BMPs planning case study in the Erhai watershed, the seventh-largest freshwater lake in China. Firstly, priority management areas (PMAs) were identified for BMPs siting using a simulation-optimization approach. Bayesian networks were subsequently embedded to reveal the multiple uncertainty sources in the optimal planning and the reliability level (RL) is introduced to represent the probability to meet the water quality target with BMPs implementation. The results suggest that E_NS of discharge and nutrients concentration simulation by LSPC are both greater than 0.5, which displays satisfactory performance. The identified PMAs account for 0.8% of the total watershed areas while contribute to more than 15% of nutrient loadings reduction. The analysis of multiple uncertainty sources reveals that precipitation is the most influential source of uncertainties in BMP effectiveness. The construction of hedgerows plays an important role in the nutrient reduction. With the improvement of the reliability levels, the cost increases sharply, indicating that the implementation of BMPs has a marginal utility. The study addressed the urgent need for effective and efficient BMPs planning by identifying PMAs and addressing multi-source uncertainties.

Harmonizing science and management options to reduce risks of cyanobacteria

Management of cyanobacteria has become an increasingly complex venture. Cyanobacteria risks have amplified as society moves forward in an era of accelerated global changes. The cyanobacteria management "pendulum" has progressively shifted from prevention to mitigation, with management considerations often put forth after bloom formation. A universal system (i.e., one-size-fits-all management) fails to provide a management path forward due to the inherent complexities of each lake. A tailored management plan is needed: the right species at the right time in the right place (i.e., the three Rs). The three Rs represent a customizable management strategy that is flexible and informed by advances in scientific understanding to lower cyanobacteria-associated risks. Identifying thresholds in risk tolerance, where thresholds are defined by community collectives, is essential to frame cyanobacteria management targets and to decide on what management interventions are warranted.