Balancing Monitoring and Management in the Adaptive Management of an Invasive Species

Abstract

Efficient allocation of managers' limited resources is necessary to effectively control invasive species, but determining how to allocate effort between monitoring and management over space and time remains a challenge. In an adaptive management context, monitoring data are key for gaining knowledge and iteratively improving management, but monitoring costs money. Community science or other opportunistic monitoring data present an opportunity for managers to gain critical knowledge without a substantial reduction in management funds. We designed a management strategy evaluation to investigate optimal spatial allocation of resources to monitoring and management, while also exploring the potential for community science data to improve decision-making, using adaptive management of invasive flowering rush (Butomus umbellatus) in the Columbia River, USA, as a case study. We evaluated management and monitoring alternatives under two invasion conditions, a well-established invasion and an emerging invasion, for both risk-neutral and risk-averse decision makers. Simulations revealed that regardless of invasion condition or managers' risk tolerance, allocating effort outward from the estimated center of invasion (Epicenter prioritization) resulted in the lowest overall level of infestation at the end of management. This allocation outperformed alternatives in which management occurred in fixed areas (Linear prioritization) and alternatives that targeted patchily distributed areas with the highest estimated infestation level of the invasive species (High invasion prioritization). Additionally, management outcomes improved when more resources were allocated toward removal effort than monitoring effort, and the addition of community science data improved outcomes only under certain scenarios. Finally, actions that led to the best outcomes often did not produce the most accurate and precise estimates of parameters describing system function, emphasizing the importance of using value of information principles to guide monitoring. Our adaptive management approach is adaptable to many invasive species management contexts in which ongoing monitoring allows management strategies to be updated over time.

Keywords: adaptive management; community science data; invasive species; management strategy evaluation; monitoring.

FIGURE 1

Map of the mainstem of the Columbia River, in light blue, and location of the study area, in dark blue and indicated with a box. The study area is a subset of the Columbia River that spans 40 miles (65 km) downstream of the McNary Dam.

FIGURE 2

Depiction of our management strategy evaluation (MSE) approach to adaptive management: 1. Under multiple alternative management strategies, simulate realistic population dynamics and management of the species using an operating model based on distributions of parameter values representing the current state of knowledge; 2. Gather simulated monitoring data from the operating model; 3. Analyze the monitoring data using an estimation model and potentially use informed priors generated through Bayesian updating, then predict population dynamics; 4. Based on the estimated model predictions, update the allocation of management effort, which subsequently influences the realistic population dynamics of the species in the operating model. The details in italics are specific to this study. Figure was adapted from Siple et al. (2021).

FIGURE 3

Illustration of the three spatial priority actions using a simplified river of five total segments. The arrows depict the flow of the river from upstream to downstream. Prior to management, the invasion state of each segment is estimated: Empty (E), low severity (L), or high severity (H). The colors depict a possible outcome of this estimation process for the 5‐segment example. If the Linear spatial priority action was selected, the order of segment visitation would be: 1, 2, 3, 4, 5. If the high invasion spatial priority action was selected, the order of segment visitation would be: 3, 4, 5, 1, 2. If the epicenter spatial priority action was selected, the order of segment visitation would be: 3, 4, 5, 2, 1.

FIGURE 4

Two initial invasion conditions of flowering rush simulated across the 40‐segment state space for the case study, including (A) initial invasion states by segment under condition 1, an established invasion, and (B) initial invasion states by segment under condition 2, an emerging invasion. The colors depict the initial true invasion state at each segment with orange depicting E, empty, red depicting L, low severity, and black depicting H, high severity. Segment 1 is the most upstream segment in the region and Segment 40 is the most downstream segment.

FIGURE 5

Performance of each simulated alternative in meeting the management objective to minimize the final average invasion state of flowering rush in the Columbia River study area, displayed across the three investment levels (20, 40, and 60 h of effort per week). A final average invasion state of 0 indicates eradication while a final average invasion state of 2 indicates that all segments are in the high invasion state. In box A, outcomes are displayed for an established invasion, and in box B, outcomes are displayed for an emerging invasion. In plots A and B, subplot (i) displays the expected‐value outcome for each alternative across simulations for each investment level; the bolded text displays the alternative that performed best in terms of expected value. The expected‐value criterion is used for risk‐neutral decision makers. In plots A and B, subplot (ii) displays the maximum outcome for each alternative across simulations and for each investment level, the bolded text displays the alternative that performed best in terms of minimizing the maximum potential invasion outcome, or the “mini–max” criterion. This criterion is used for risk‐averse decision makers. The alternatives are a function of spatial priority (shown with the shapes), target detection and eradication probabilities (shown with colors), and the data used in estimation (shown with the size of the points) where “A” indicates just agency were collected, and “A + C” indicates agency and community science data were collected. In the bolded text, we indicate whether the top alternative included community science data with “A + C.”

FIGURE 6

(A) Results of average percentage of segments visited for either detection or detection and removal for each alternative and investment level and (B) average weekly distance traveled (in terms of segment level units) for an emerging invasion (condition 2) for each alternative and investment level. In both plots, the boxplots represent outcomes from each alternative, the outline color of each box plot represents the spatial priority action, and the fill color represents the target (detection and eradication) probabilities. The alternatives with target priorities: (0.5, 0.75) A + C are the alternatives with the addition of community science data. In each boxplot, the colored line represents the median value, the black line is the mean value, the box displays the interquartile range, the lines indicate variability beyond the first and third quartiles, and the points represent outliers (See Figure S5 for outcomes under an established initial invasion).

FIGURE 7

Results of relative bias estimates for detection parameters p for the high invasion prioritization under the established invasion (condition 1) for the three investment levels in terms of the first distinct year across alternatives (i.e., year 4) and final time the estimation model was run (i.e., year 10). The colors represent the different target probability pairings (p = detection, $ϵ$ = eradication). The points represent average values across parameter sets and the error bars represent the upper and lower 5% quantile values.