The Problem
Traditional conservation planning often evaluates species independently. When multiple threatened species occupy overlapping landscapes, species-by-species planning leads to inefficient land allocation and unnecessary spatial footprint.
Recovery planners needed to:
- Identify the smallest land area meeting habitat requirements for all target species
- Prioritize high-quality habitat, not just representation
- Handle incompatibility constraints between species
- Evaluate tradeoffs between footprint and habitat quality
Existing tools excel at maximizing representation targets but do not directly optimize habitat quality metrics under footprint minimization.
A new approach was required.
The Solution
optimTFE implements a stochastic greedy algorithm to solve multi-species spatial prioritization problems.
Core optimization modes include:
- Minimizing total area while meeting species-specific habitat targets
- Maximizing habitat quality within a fixed area budget
- Exploring tradeoff solutions between footprint and habitat quality
Key features include:
- Species-specific habitat quality weights and minimum area thresholds
- Incompatibility constraints for mutually exclusive habitat requirements
- Protection of known populations
- Regional and subregional distribution targets
- Maximum species per planning unit constraints
- Parallel solution generation for exploring tradeoff space
For 36 at-risk native plant species in East Maui, optimTFE identified recovery solutions using 36% less area than species-by-species planning while meeting all habitat targets. Compared to representation-based approaches, optimTFE produced equal-area solutions with higher average habitat quality.
Technical Architecture
The optimization core is implemented in C++ via Rcpp for performance-critical operations. The stochastic greedy algorithm iteratively selects planning units that maximize marginal habitat quality gains while respecting species-specific constraints and incompatibility rules.
Performance features include:
- Parallel solution generation via future and furrr
- Efficient data handling with Apache Arrow
- 10,000 solutions generated in under 8 seconds on an 8-core machine
This architecture bridges algorithmic optimization with practitioner decision-making.
Outcome
optimTFE is actively used for conservation planning in Hawaiʻi and has been applied to recovery scenarios involving dozens of threatened species.
The methods and results are published in Conservation Biology (doi:10.1111/cobi.14421). The package is released as a USGS software product under the CC0 public domain dedication and the most recent development branch is available on GitHub.
The project demonstrates how performance-oriented algorithm design can directly inform conservation outcomes, reducing spatial footprint while improving habitat quality.
To support expert validation of proposed recovery areas, I developed a companion Expert Feedback Platform enabling structured review of optimization outputs.