This case study is an example of applying Bayesian Networks to guide interdisciplinary research and decision making for land use and stream health management.
This case study describes a causal linkage model between practices on dairy farms and on-farm and in-stream values in Bog Burn, Southland.
Maintaining key values in streams and rivers in areas of intensive dairy farming often requires concerted action based on a shared understanding of the links between waterway values, farm practices and stressor mitigations. A stakeholder workshop identified the key values as farm economics, trout in Bog Burn and contact recreation in the Oreti River, and developed an initial conceptual model of system linkages. Subsequent research used a combination of published information, local data, and expert knowledge to develop this into a Bayesian Belief Network (BBN) that was used to predict the effects on the key values of mitigations (singly and in combination) under simplified conditions.
(After Quinn, et al 2010)
The BBN predicted riparian fencing and planting had the greatest single-action benefit for trout in Bog Burn, whereas deferred dairy shed effluent irrigation had the best predicted single-action benefit for contact recreation in the Oreti River. Reducing phosphorus fertiliser use to ensure economically-optimal soil Olsen P tests were maintained was predicted to have the greatest single benefit for farm economic returns, whereas converting 2.5% of the land to wetlands to treat field-tile drainage had the greatest cost. When used in combination, several mitigations had enhanced effects on the key values by (i) reducing contaminant inputs (e.g., use of wintering pads, optimising P fertiliser use and stream fencing), reducing contaminant transport to water (e.g., deferred and low rate effluent application, the use of the nitrification inhibitor DCD, and constructed wetlands on mole/tile drains), or (ii) acting in combination with other drivers of in-stream responses to inputs (e.g., riparian shade influences on nuisance plant growth in Bog Burn in response to nutrient enrichment). Focusing solely on edge-of- field mitigations (riparian management and drainage treatment wetlands) was predicted to have less benefit to both farm and waterway values than combined field and edge-of-field mitigations. The top five mitigations in terms of their overall benefit for the three key values were predicted to be optimising P fertiliser to soil Olsen P, stream fencing and planting, deferred dairy effluent irrigation and/or low rate effluent irrigation, and the use of herd shelters for wintering dairy cows.
Associated Models
Bayesian Belief Network (BBN)
A BBN is a probabilistic graphical model (a type of statistical model) that represents a set of random variables and their conditional dependencies via a directed acyclic graph.
References
Quinn, J., et al (2010). Linking Farm and Waterway Values - The Bog Burn Catchment. (In) Currie,L and Christensen (Eds) Farming's Future: Minimising Footprints and Maximising Margins. 23rd Annual FLRC Workshop. Fertiliser and Lime Research Centre