A Bayesian network model for predicting macroinvertebrate community diversity in the lower Loddon RiverReport 3
Land & Water Australia. 2008. A Bayesian network model for predicting macroinvertebrate community diversity in the lower Loddon River. [Online] (Updated June 6th, 2012)
Available at: http://lwa.gov.au/node/2666 [Accessed Tuesday 12th of March 2013 02:22:22 PM ].
This report is the third in a series of five produced by NPSI project UMO45 Delivering Sustainability through Risk Management. Ecological Risk Assessment (ERA) is a formal process for determining the risk posed by hazards (stressors, threats) to the health of ecosystems.
ERA evolved from the need to develop processes that better deal with the complexity of aquatic ecosystems. That is, the difficulties in assessing multiple stressors for a wide range of species within inherently variable ecosystems. Such assessments provide an explicit and transparent process for making management decisions for complex ecosystems that may not always be fully understood.
The ERA conducted in this study was focused on the assessment of the health of the Lower Loddon River. The aim of the risk assessment is to provide local resource managers with a better understanding of risks to the Lower Loddon River and the effectiveness of different management actions in protecting and rehabilitating the River. This primarily involved the development of a Bayesian network to assess the health of the macroinvertebrate community in the Lower Loddon catchment. A Bayesian decision network model for predicting macroinvertebrate community diversity in the Lower Loddon River was developed and full details are provided in the report. Sensitivity analysis showed that habitat variables (e.g. in-stream habitat, food availability, in-stream vegetation, turbidity, sedimentation, riparian vegetation, woody debris and roots, bank erosion) had the greatest influence on the predicted macroinvertebrate community diversity.
The Bayesian network predicted that the macroinvertebrate community diversity in all six ISC reaches in the Lower Loddon River would be poor. This certainly agrees with the small amount of field data available. The Bayesian network has also been used to predict the effect on the macroinvertebrate communities of three levels of stock access (low, moderate, high) to the riparian zone and the channel. Reducing stock access significantly improved the macroinvertebrate community diversity in good to very good condition from 21% for high access to around 80% for low access. These results support the current Loddon catchment management plan, where major on-ground fencing works are being implemented to reduce stock access to the riparian zone and the river.
Westbury et al