Winter_2013

The northern red-legged frog (Rana aurora draytonii) was listed as threatened in 1996. It historically occurred occur along the Pacific Coast from southwestern Vancouver Island and Sullivan Bay, British Columbia to the California coast north of Mendocino County. Populations in the Willamette Valley OR are believe to be the most at risk of extirpation due to fragmentation associated with intensive agricultural land use and presence of non-native predators, chiefly introduced Centrarchids (i.e., sunfish and black basses) and bull frogs In the Willamette Valley, northern red-legged frogs breed in winter through early spring with adults frequently undertaking long migrations from summer habitats in mesic forests and riparian areas to breeding areas. Females lay eggs in vegetated shallows with little or no current. Eggs take approximately 6 weeks to hatch and an additional 3 months to reach metamorphosis. The wetlands can be temporary or permanent but they need to hold sufficient water until the juvenile frogs metamorphose in late June-July. Metamorphosed juveniles general remain at or near natal wetlands for several weeks before dispersing to summer habitats. Larval northern red-legged frogs consume epiphytic algae and generally use areas of dense vegetative cover, presumably to minimize predation risk.

Spatial and temporal dimensions

The spatial extent of this decision is the Willamette River Valley from Eugene OR north to Portland. Land use within the watershed is primarily a mixture of irrigated row crop agriculture and with heavily forested riparian zones. The valley is highly fragmented with wetlands occurring primarily is areas adjacent to the river. Riparian buffers are highly variable throughout the basin but tended to be smallest nearest urban areas. The grain of the decision is individual wetlands.

Legal, regulatory, and institutional constraints

The management of natural resources within the state of Oregon is the responsibility of the Oregon Department of Fisheries and Wildlife (ODFW). The broad responsibility lies in managing public lands and establishing best management practices standards, and setting wildlife conservation laws. Management of the Federally-listed species is the responsibility of the US Fish and Wildlife Service (USFWS) under the framework of the endangered species act.

Stakeholders

Identifying stakeholders was not part of this exercise. However, if I were to guess, the stakeholders for the decision would include: ODFW, USFWS, and landowners and managers in the Willamette Valley.

Objectives

During class,we created a mean objectives network and identified two fundamental objectives. One of the fundamental objectives was the long term persistence of northern red-legged frog and the other was to minimize the cost of management actions. To maximize northern red-legged frog persistence, the class believed that we should maximize the distribution and abundance of northern red-legged frogs in the Willamette Valley. A quantifiable attribute of this objectives at the wetland scale (i.e., the grain) was the presence of juvenile frogs during the early to mid summer. Thus, the quantifiable attributes of our fundamental objectives were maximizing juvenile from occupancy and minimizing costs.

Decision alternatives

After a brainstorming exercise, we arrived at three decision alternatives that could be applied to individual wetlands: mow exotic vegetation, remove exotic vegetation and replant native vegetation, and control (remove) exotic bull frogs. (Real world decision alternatives would probably more complex and exhaustive, but give us a break we only had 1/2 hr to brainstorm!!)

Valuation of outcomes

We used proportional scoring to standardize the objective specific outcomes. Here, objective specific utilities are rescaled as (actual outcome - worst outcome)/(best outcome - worst outcome). This results in utilities with a maximum value of 1 and a minimum value of 0. For example, the maximum cost for an action ( worst case replanting) was $20 K and the minimum cost ( best case mowing) was $0.5 K , so if the actual cost of an action was $5 K, the cost utility was: (5 - 20)/(0.5 - 20) = 0.769. For occupancy, the best outcome was 1 (100%) and the worst outcome was 0. The two utilities were weighted and combined as: combined utility = U.cost*0.375 + U.occupancy*0.625. The weights 0.375 and 0.625 needed to sum to one and the values were based on a majority vote by the class. The weights mean that the stakeholders value frog occupancy 0.625/0.375 = 1.67 times more than cost.

Decision model overview

The wetland decision model is a stochastic model that estimates the probability of juvenile northern red-legged frog occupancy in an individual wetland in response to a management decision. It is composed of the current state of a watershed: size, distance from nearest wetland, native vegetation buffer, and the presence of bull frog, population dynamics, and management components. The model went through three iterations before the final model. Here are the versions in Netica format (Version 1, Version 2, Version 3).

The final model is graphically represented as an influence diagram (below) that consists of model components, referred to as nodes with each node consisting of environmental states that are mutually exclusive and collectively exhaustive. The directed arcs indicate casual relationships between model components with parent nodes influencing (pointing into) child nodes. For instance, future native vegetation buffer (a child node) is influenced by current native vegetation buffer (a parent node) and the management decision (also a parent node). The relationships among components were parameterized using expert opinion. Elicitation of student experts included function elicitation, frequency elicitation, and value elicitation via the 4 step process. The excel file used to conduct the expert elicitation can be found here and the R script used for the 4 step process can be found here. Below we describe the model components and the sources of information that were used to parameterize the relations among model components.