TL;DR
This paper introduces a novel animal home range estimation method (DFHRE) that accounts for terrain and obstacles by using cost-distance instead of Euclidean distance. This approach better models animal movement anisotropy and produces more realistic home range estimates, outperforming existing methods like KDE and LoCoH in validation tests.
What Problem Does This Paper Address?
Estimating an animal's home range (HR) is crucial for understanding its behavior, resource use, and ecological role. Traditional methods often struggle to accurately represent HRs because they don't adequately consider the influence of the environment, specifically terrain and physical obstacles like rivers or mountains. Animals don't move randomly; their movement is directional and influenced by these factors, leading to anisotropic behavior. Existing methods that rely on simple Euclidean distances fail to capture this, potentially leading to HR estimates that incorrectly include impassable areas or exclude important foraging grounds. This paper aims to bridge this gap by developing an HR estimator that explicitly incorporates these environmental complexities.
Methods
The core innovation of this paper is the Density-based Fuzzy Home Range Estimator (DFHRE), which leverages the Active Learning Method (ALM). The key methodological shift is the replacement of standard Euclidean distance with a cost distance calculation. This cost distance is derived using a geodesic distance transformation that accounts for terrain (elevation changes) and obstacles.
Here's a breakdown of the DFHRE's approach:
- Cost Distance Calculation: Instead of a straight-line distance, the algorithm calculates the "effort" or "cost" required to travel between points, considering elevation changes and the presence of obstacles. This is achieved using a map algebra-distance transformation with obstacles (MA-DTO) algorithm.
- Fuzzy Membership Grade: The DFHRE uses a membership grade to represent the "possibility" or "importance" of different locations within the animal's range. This allows for fuzzy boundaries, reflecting the gradual change in an animal's preference for different areas, rather than sharp, defined edges.
- Active Learning Method (ALM) Integration: The ALM framework is used to fuzzify the data and generate these membership grades. The "ink drop spread" (IDS) operator, a core component of ALMs, is adapted to work with the cost distance, ensuring that the spread of influence from an animal's location is anisotropic and respects environmental barriers.
- Parameter Optimization: The influence range (D), analogous to bandwidth in kernel density estimation, is determined using an "ad hoc smoothing" method. The topographic influence parameter (k) is optimized using a resource selection method that considers relief amplitude.
Datasets
The proposed DFHRE method was validated using three distinct datasets:
- Oriental White Stork (Ciconia boyciana): Real-world GPS tracking data from a rescued stork in the Beidagang Reservoir area. This dataset was used to test the method in a relatively flat environment with no significant obstacles.
- Simulated Sika Deer (Cervus nippon): Simulated data generated using a digital elevation model (DEM) and remote sensing imagery. This dataset incorporated various obstacles like highways, water bodies, and villages, and was used to test the method's ability to handle terrain and obstacles.
- African Buffalo (Cilla): Real-world GPS tracking data from an individual buffalo in the Kruger National Park. This dataset included a major river (Sabie River) and mountainous terrain as significant obstacles.
Key Insights and Ideas
- Anisotropy is Key: The most significant insight is that animal movement is rarely isotropic. Incorporating terrain and obstacles through cost distance is crucial for accurately modeling this anisotropic behavior.
- Fuzzy Boundaries are Realistic: Animals don't have strict boundaries for their home ranges. The DFHRE's use of fuzzy membership grades provides a more nuanced and realistic representation of an animal's probability of being in different locations.
- Cost Distance Outperforms Euclidean: The direct comparison with Euclidean distance-based methods clearly shows that cost distance is superior when environmental factors are at play, leading to more biologically plausible HR estimates.
- Obstacle Avoidance is Crucial: The DFHRE's ability to effectively "bypass" obstacles like rivers and mountains is a major advantage over methods that might incorrectly include these areas within an animal's home range.
Key Results
The paper demonstrates that the DFHRE method generally outperforms Kernel Density Estimation (KDE) and Local Convex Hull (LoCoH) methods across the three datasets:
- Dataset 1 (Stork): DFHRE produced the smallest HR, suggesting a more precise estimation in a relatively obstacle-free environment.
- Dataset 2 (Sika Deer): DFHRE's estimates were more reasonable, avoiding obstacles like highways and water bodies, unlike KDE methods which overlapped with them. The HRs estimated by DFHRE were also smaller than those from KDE.
- Dataset 3 (Buffalo): DFHRE successfully avoided overlapping with the Sabie River and showed minimal overlap with mountains, whereas KDE methods did overlap significantly. This indicates better handling of major environmental barriers.
- Cross-Validation: A leave-one-out cross-validation test showed that the DFHRE results were more stable and reasonable, with less area change when sample data was reduced, compared to MKDE and k-LoCoH, which exhibited more erratic behavior.
Limitations and Future Work
The authors acknowledge a few limitations and suggest future research directions:
- Obstacle Assumption: The current model assumes animals cannot reach or cross obstacles. Future work could explore more nuanced models where certain obstacles (like streams) might be traversable.
- Land Cover Influence: While terrain and discrete obstacles are considered, the paper suggests that quantitatively determining the influence of different land cover types (e.g., forest, grassland) on animal activity is an important area for future exploration, especially when land cover itself is fuzzy.
- Spatial Relationships: Further investigation into the spatial relationships between the oriental white stork's HR and wetlands, including topological and directional relationships, could provide deeper ecological insights.
Summary updated