Combining Strategies to Select Reserves in Fragmented Landscapes

Kristina D. Rothley, Claire N. Berger, Catherine Gonzalez, Elisabeth M. Webster, Daniel I. RubensteinConservation 2004 · DOI

TL;DR

This paper tackles the challenge of selecting optimal reserve networks in fragmented landscapes where data is scarce. They propose a method that field-tests the relationships between easily measurable "selection criteria" (like patch size) and actual conservation "objectives" (like species richness), then uses multi-objective programming to find the best compromises. This approach helps conservation planners navigate conflicting goals and limited data to make more informed decisions.

What Problem Does This Paper Address?

Creating effective reserve networks is crucial for conserving biodiversity, especially in fragmented landscapes where habitats are broken into small, isolated patches. A major hurdle is the lack of detailed species-specific data, forcing planners to rely on "selection criteria" (e.g., patch size, road density) as proxies for true conservation goals. However, these proxies aren't always reliable, and different conservation objectives (like protecting rare species vs. maximizing overall biodiversity) often conflict, meaning no single reserve network can be perfect for everything. This paper addresses the need for a systematic way to evaluate these proxies, identify conflicts, and find the best possible compromises when designing reserve networks under these real-world constraints.

Methods

The core of the study involved a two-pronged approach:

  1. Field Testing of Selection Criteria: The researchers conducted field surveys on Martha's Vineyard to empirically test the relationships between nine potential selection criteria (e.g., patch area, shape metrics, road density, vegetation structure) and five specific conservation objectives (e.g., rare plant species occurrence, Northern Harrier nest sites, plant species richness, invertebrate family biodiversity, Northern Harrier fledging success). This step is crucial because it moves beyond assumptions about proxy effectiveness.
  2. Multi-Objective Integer Programming (MOIP): Once selection criteria that showed a significant correlation with conservation objectives were identified, the researchers used MOIP. This mathematical technique allowed them to:
  3. Identify reserve networks that were optimal for each objective independently.
  4. Identify reserve networks that represented the best possible "trade-offs" or compromises between multiple conflicting objectives.

This MOIP approach, applied separately to two habitat types (maritime shrublands and scrub oak barrens), generated a set of optimal reserve network alternatives, from which planners could then select a final network based on their priorities.

Datasets

The study utilized data collected from field surveys on Martha's Vineyard, Massachusetts, focusing on two specific sandplain natural community types: maritime shrublands (MS) and scrub oak barrens (SO).

  • Vegetation Data: A digital polygon map of vegetation was used, with MS and SO patches identified and redrawn based on orthographic images.
  • Field Survey Data (Summer 2000 & 2001):
    • Plant Species Richness: Data collected from 1m² quadrats within 19 MS and SO patches.
  • Invertebrate Family Richness: Data collected using pitfall traps in the same 19 patches.
  • Northern Harrier Nesting Data: Field surveys in 21 patches (2000) and 46 patches (2001) to record nest presence and fledging success.
  • Small Mammal Trapping: Data from traps in the 19 patches to estimate prey density.
  • Existing Data: Digital road maps (U.S. Census Bureau 1995) were used for road density calculations.

No new benchmark datasets were introduced; rather, existing data and field-collected data were integrated for analysis.

Key Insights and Ideas

  • Selection Criteria Aren't Always Reliable Proxies: The study highlights that assumptions about selection criteria (like patch size) being good surrogates for conservation objectives can be flawed. For instance, patch size was a good predictor for Northern Harrier nesting but not for rare plant species occurrence.
  • Conflicts are Inevitable and Manageable: The research confirms that conservation objectives often conflict. For example, plant species richness and prey density (important for Northern Harriers) were negatively correlated in MS patches. However, the MOIP framework provides a structured way to identify and quantify these trade-offs.
  • Field Testing is Crucial: Empirically testing the relationships between selection criteria and objectives, rather than relying solely on theoretical assumptions, is vital for robust reserve design.
  • Compromise is Key: The MOIP approach allows planners to move beyond single-objective optimization and explore a range of "optimal compromise" reserve networks, acknowledging that perfect solutions are rare.
  • Habitat Type Matters: The study found differences in the effectiveness of selection criteria and the nature of conflicts between the two habitat types studied (MS and SO), underscoring the need for context-specific approaches.

Key Results

  • Variable Proxy Effectiveness: Patch area was a strong predictor for Northern Harrier nest site selection in both MS and SO patches. However, other criteria showed varied effectiveness: percent salt pond was a good predictor for rare plant species richness in MS, while percent grassland/heathland was best for invertebrate family richness in SO.
  • Habitat-Specific Differences: Plant species richness was significantly lower in SO patches compared to MS patches. No rare species were found in SO patches.
  • Northern Harrier Nesting: Northern Harriers showed no significant preference for MS or SO patches for nesting, but nesting success (fledging) was influenced by factors like prey density and habitat structure.
  • MOIP Outputs: The MOIP identified numerous optimal reserve network alternatives for both MS and SO patches, demonstrating the range of possible compromises. For MS, several alternatives achieved high scores for multiple objectives simultaneously. For SO, 34 optimal compromise solutions were identified.
  • Patch Overlap: In the SO analysis, three patches appeared in over 30 out of 34 optimal alternatives, suggesting these are high-priority areas for conservation. Similarly, two patches were frequently included in the MS optimal networks.

Limitations and Future Work

  • Limited Field Data: The authors acknowledge that the invertebrate survey was limited and could be expanded. Similarly, a more robust sampling technique for rare plants is suggested.
  • Assumption of Area for Plant Richness: While area was positively correlated with plant species richness in MS, the relationship was not statistically significant, suggesting a need for more field sampling to confirm this assumption.
  • Incomplete Biotic Inventory: Without a more complete inventory of all species, it's difficult to fully evaluate the representativeness and complementarity of the proposed reserve networks.
  • Potential for Overestimation of Fledging Success: The Northern Harrier nest survey in 2000 began late in the season, potentially missing early nest failures.
  • Remotely Measured Criteria: The MOIP analysis primarily used remotely measured selection criteria. Future work could incorporate locally measured criteria (e.g., specific vegetation structure) for a more comprehensive analysis.
  • Adaptability: The authors suggest their methods can be adapted to support reserve network design in other systems, even with data limitations.

Summary updated

Abstract

In the identification of reserve networks in fragmented landscapes with limited species‐specific data at hand, one approach is to use selection criteria, such as patch size, to rank the habitat patches' conservation value and evaluate reserve‐network alternatives. These criteria are assumed to be reasonable surrogates for the true network objectives. Caution is warranted, however, because the relationships between the selection criteria and the reserve‐network objectives may be inconsistent. Conflicts are also likely to arise because no single reserve network will be optimal with respect to multiple objectives (or selection criteria) simultaneously. Instead, reserve planners must compromise between conflicting demands. We field tested the relationships between a variety of selection criteria and the objectives of a reserve network for the sandplain natural communities on Martha's Vineyard Island, Massachusetts (U.S.A.). Selection criteria that correlated with the reserve‐network objectives were used in a multi‐objective integer program to identify the 10‐patch reserve networks that were optimal with each objective independently and those that offered optimal tradeoffs between the reserve‐network objectives. From these 10‐patch networks, one can select a final reserve network that provides the preferred compromise between the objectives.
BibTeX
@inproceedings{rothley2004combining,
  author = {Kristina D. Rothley and CLAIRE N. BERGER and Catherine Gonzalez and ELISABETH M. WEBSTER and Daniel I. Rubenstein},
  title = {{Combining Strategies to Select Reserves in Fragmented Landscapes}},
  booktitle = {Conservation},
  year = {2004},
  doi = {10.1111/j.1523-1739.2004.00180.x},
}
Figure