The Department of Defense (DoD) is increasingly engaging landowners outside installation boundaries to help protect biodiversity at a landscape scale. Such activities include the acquisition of recovery habitat for species at risk. SERDP Exploratory Development (SEED) project RC-1469 evaluated how such habitat allocations would affect population viability for the red-cockaded woodpecker (Picoides borealis, RCW) on Marine Corps Base Camp Lejeune, North Carolina. As an extension of that research, this project evaluated the contribution different data types make to reducing uncertainty in individual-based, spatially explicit population models (IB-SEPMs) and, given this uncertainty, what is the most cost-effective allocation of habitat. In contrast to the SEED project, this project increased the spatial scale of analysis for RCWs across the greater Onslow Bight landscape, while incorporating data from conservation partners to help reduce uncertainty in RCW population dynamics. This project also illustrated the transferability of the techniques to lesser known species by applying them to the gopher tortoise (Gopherus polyphemus, GT) at Fort Benning (FB), Georgia, including associated Army Compatible Use Buffer (ACUB) lands.

Technical Approach

Habitat management often entails tradeoffs among habitats at different locations and at different times. Both the spatial location and timing of land use change can affect the realized conservation benefit associated with such decisions. IB-SEPMs have a proven ability to predict biological patterns in dynamic landscapes. Military installations will often wish to change habitat allocations prior to reaching a scientific consensus regarding the behavior and demography of at-risk species; therefore, Decision Analysis, a structured approach for contrasting alternative habitat allocations given uncertainty in system dynamics (e.g., dispersal and/or reproduction), is applied. This requires applying two techniques to IB-SEPMs. The first is Pattern Oriented Modeling (POM), a retrospective analysis that evaluates uncertainty in IB-SEPMs. The second simulation modeling technique is Landscape Equivalency Analysis (LEA), which estimates credits and debits that reflect changes in abundance and genetic variation toward sustainability criteria. By incorporating genetic criteria, LEA recognizes that the amount of migration required to minimize the influence of inbreeding, genetic drift, and local extinction will vary as landscape patterns change over time.

Decision Analysis for RCWs was applied using an IB-SEPM for the Coastal NC Recovery population, which includes Holly Shelter State Game Lands (HS), Croatan National Forest (CNF), and Marine Corps Base Camp Lejeune (CL). A habitat model was constructed for the one million hectare landscape, and RCW monitoring data were collected from all stakeholders. POM was applied to contrast the ability of monitoring data to reduce uncertainty in the dispersal submodel of the IB-SEPM. The five most plausible dispersal parameterizations were then used to evaluate six alternative habitat allocations for the Onslow Bight landscape using LEA.

To illustrate the transferability of the POM/LEA technologies to a lesser known species, an IB-SEPM was constructed for the GT. This was accomplished by performing an extensive literature review, constructing a habitat model, and conducting a landscape genetic study at FB. All available monitoring data were collected through collaboration with FB, The Nature Conservancy (TNC), and Auburn University. Therefore, the research team was able to apply POM to estimate the relationship between different types of monitoring data and uncertain aspects of the species’ natural history. Then, given the uncertainty remaining after POM, LEA was applied to assign conservation value to four alternative habitat allocations.


Application of Decision Analysis for RCWs indicated that the best dispersal parameterizations derived from POM all agreed regarding which allocation of habitat would lead to the greatest number of LEA credits for the number of Potential Breeding Groups (i.e., abundance; PBG). However, the parameterizations disagreed regarding which habitat allocation would minimize habitat fragmentation effects, estimated as the departure of genetic variation farther away from a pre-settlement condition compared to that expected under the current recovery plan. Based on approximate costs associated with alternative habitat allocations, the cost-effective level of investment in future dispersal research, which may lead to a habitat allocation with greater conservation benefits at a lower cost, was estimated at $1,036,679. Research in the Onslow Bight for RCWs also indicated that building landscape-scale models using less detailed RCW data from non-DoD partners did help to reduce uncertainty and improved management decisions for Encroachment Partnering (EP) programs.

Application of Decision Analysis to the GT indicated that the best emigration parameterizations derived from POM disagreed regarding which habitat allocation would provide the greatest conservation benefit for both abundance and genetic variation. Therefore, future habitat allocation decisions would benefit from either further model development and/or data collection. The habitat allocation that maximized abundance was different than the allocation that minimized the erosion of genetic variation; therefore, multiple services aid in habitat allocation decisions. Simulation results also indicated that GTs may be more sensitive to the fragmentation of habitat independent of the effects of habitat loss. Application of POM to develop and verify the IB-SEPM illustrated the relationship between model uncertainty and data commonly collected in the field. Results indicated that the mating system assumptions can have a large impact on the ability of the model to approximate data collected in the field. Extension of new techniques that integrate point pattern analysis with the landscape genetic data collected at FB indicated a mating system characterized by female philopatry and male biased dispersal. These genetic patterns, along with several demographic patterns collected in the field, indicated that strength of habitat preferences during dispersal for juvenile and subadult male dispersal also significantly impacted model fit. Certainly, much more work could be done to test the IB-SEPM, but this study has created a valuable framework with which to test hypotheses regarding GT natural history.


By integrating POM and LEA, using Decision Analysis, the most cost-effective habitat allocation and/or translocation program can be identified given the existing knowledge about system dynamics. POM provides a method for constructing and verifying IB-SEPMs by including observed data that characterize the transient dynamics of the system. LEA provides a generally applicable accounting system that allows for substitutions between habitat connectivity and area while maintaining equivalent ecological function at the landscape scale.